Original Research

Estrogen-containing hormone therapies (HT) have been used to alleviate menopausal symptoms and to prevent chronic diseases common to postmenopausal women, including osteoporosis and cardiovascular disease.^1,2 In this analysis, we use the abbreviation “HT” to refer to postmenopausal hormone therapies, either estrogen alone or combined with progestin.

Based on the findings of the randomized, double-blind Women’s Health Initiative (WHI) study involving estrogen-progestin,³ the Food and Drug Administration (FDA) recommends using HT to treat moderate to severe symptoms of vulvar and vaginal atrophy and vasomotor symptoms associated with the menopause, and to prevent post-menopausal osteoporosis.⁴ When HT is prescribed only to prevent osteoporosis in women without menopausal symptoms, the FDA recommends that other approved, non-estrogen therapies be considered and that HT be used at the lowest dose for the shortest duration to achieve treatment goals.⁴ Many postmenopausal women have chosen to discontinue HT in light of these recommendations.⁵ However, discontinuing HT may increase bone resorption and accelerate bone loss,^6,7 which, if untreated, places women at risk for osteoporotic fractures.

The serum estrogen receptor modulator (SERM) raloxifene is not an estrogen, a progestin, or a hormone,⁸ but it binds to the estrogen receptor to exert effects in the skeletal and cardiovascular systems and in breast tissue.⁹ In the 4-year Multiple Outcomes of Raloxifene Evaluation (MORE) osteoporosis treatment trial of postmenopausal women, raloxifene 60 mg/d, the approved dose for post-menopausal osteoporosis prevention and treatment, increased bone mineral density (BMD) and significantly reduced the risk for new vertebral fractures with sustained efficacy.¹⁰ With the declining use of long-term HT,⁵ it is clinically relevant and important to determine whether a history of HT use has any influence on the effects of other antiresorptive agents, such as raloxifene, which may be subsequently used for postmenopausal osteoporosis prevention and treatment. The objective of this analysis is to determine the effects of raloxifene on BMD, and the risks of vertebral fractures, cardiovascular events, and breast cancer in post-menopausal women who did or did not use HT prior to screening for the MORE osteoporosis study.

Materials and methods

Subjects and treatment

Details on subject recruitment and follow-up, and complete inclusion and exclusion criteria, were previously described for the MORE study.¹¹ The trial examined the incidence of osteoporotic fractures as a primary endpoint and the incidence of breast cancer as a prespecified secondary endpoint, and it collected reports of cardiovascular events as a secondary safety endpoint.

Researchers enrolled 7705 women up to 80 years of age who were at least 2 years post-menopausal, and who had osteoporosis as defined by radiographically apparent vertebral fractures at baseline or BMD criteria. Women were randomly assigned to receive raloxifene 60 or 120 mg/d, or an identically appearing placebo.¹¹ All women received daily supplements of calcium (500 mg) and vitamin D (400 to 600 IU). An ethical review board at each site approved the MORE study protocol. All women gave written informed consent to participate in the study in accordance with the ethical principles stated in the Declaration of Helsinki.

At study screening, women were asked if they had ever taken HT. Women were excluded from the study if they were experiencing clinically severe menopausal symptoms at the beginning of the study that required estrogen. Women were excluded if they had been treated with therapeutic doses of androgen, calcitonin, or estrogen (>1 cycle or 28 days) alone or with progestin (>1 cycle or 28 days) within 6 months of beginning the study. Women were permitted to have used systemic (oral or transdermal) estrogen and progestin for up to 1 cycle (28 days) during the 6 months before the study. No systemic estrogen and progestin use was allowed within 2 months before study entry. Occasional use of topical estrogens (3 times per week), and oral estriol (2 mg/d) for menopausal symptoms was permitted.

A 1-year double-blind extension phase was added to the 3-year treatment phase.¹⁰ All vertebral fracture, cardiovascular, and breast cancer endpoints that occurred over the 48-month study period were included in the present analyses.

Fracture and BMD assessment

New morphometric vertebral fractures, defined using semiquantitative assessment criteria,¹² were identified by comparing spinal radiographs taken at 2, 3, and 4 years with baseline radiographs. A new vertebral fracture was defined as a vertebral fracture occurring in a vertebra that was not fractured at baseline. New clinical vertebral fractures were defined as those associated with signs or symptoms suggestive of vertebral fracture, such as back pain, reported either at an interim 6-month clinic visit or at any time between clinic visits,¹³ and which were subsequently corroborated with radiographs and adjudicated as previously described.¹⁰

Lumbar spine and femoral neck BMD were measured annually using dual-energy x-ray densitometry, as previously described.¹¹

Cardiovascular event assessment

Cardiovascular events were collected as a secondary safety endpoint in the MORE trial, as previously described in detail.¹⁴ Women were asked at each clinic visit if they had had a myocardial infarction (MI), coronary bypass surgery, percutaneous coronary revascularization, or a stroke since the previous visit, and unsolicited reports of adverse cardiovascular events were recorded. All summaries of reported cardiovascular events were reviewed and adjudicated by 1 board-certified cardiologist, contracted by the sponsor, who was not associated with the trial and was blinded to treatment assignment.

Women with 4 or more risk points, assessed using the same criteria as for enrollment in the Raloxifene Use for The Heart (RUTH) trial,¹⁵ were considered to be at high risk for cardiovascular events. Coronary events included MI, unstable angina, or coronary ischemia.¹⁴ Cerebrovascular events included stroke and transient ischemic attack.¹⁴

Breast cancer assessment

Assessment of breast cancer was a prespecified secondary endpoint of the MORE trial, and was previously described in detail.¹⁶ All diagnoses of breast cancer were adjudicated by an independent oncology review board consisting of 5 physician specialists in breast cancer, and chaired by a pharmacologic scientist, none of whom were employed by the sponsor. Previous publications of the MORE breast cancer data have reported 61 cases of invasive breast cancer.¹⁶ A subsequent review of the MORE dataset found 1 fewer case of invasive breast cancer in each of the placebo and pooled raloxifene groups, so that 59 cases of invasive breast cancer were confirmed, and this number will be used in the present analysis. The change in the number of cases of invasive breast cancer was small, and had no impact on the overall interpretation of the breast cancer results from the MORE trial.

Results

Characteristics for all subjects at baseline and for women in the placebo group

Of the 7705 women enrolled, 7682 (99.7%) reported their status of previous HT use, with 2235 women (29.1%) having used HT before participating in MORE ( W1, available at www.jfponline.com). Baseline characteristics that were significantly different between women who reported prior HT use and those who reported no prior HT use included age, BMD, and the incidences of vertebral fractures, coronary angioplasty, hypertension, hyperlipidemia, and family histories of osteoporosis or breast cancer (Table W1, available at www.jfponline.com). In the subsets of women who did and who did not use HT previously, the baseline characteristics were not significantly different between the placebo and raloxifene groups, except for diabetes (placebo, 1.9%; raloxifene 60 mg/day, 3.3%; pooled raloxifene, 2.9%; P=.02).

In the placebo group, the incidence of new vertebral fractures, cardiovascular events, and breast cancer at 4 years were not significantly different in women with prior HT use compared with women without prior HT use.

Vertebral fracture events and bone mineral density

After 4 years of treatment with raloxifene 60 mg/d, women with and without prior HT use exhibited significant reductions in new vertebral fractures compared with those taking placebo ( Figure ). Vertebral fracture risk reductions between treatment groups were also statistically significant in subgroups of women with and without prevalent vertebral fractures ( Table1 ). Raloxifene 60 mg/d also reduced the risk of new clinically-apparent vertebral fractures, compared with placebo, in women with prior HT use (absolute risk reduction (ARR)=1.8%; RR=0.52 [95% CI, 0.28–0.96]), and in women without prior HT use (ARR=1.7%; RR=0.61 [95% CI, 0.43–0.87]; interaction P=.66]. The interaction P-values remained similar after adjusting for the baseline fracture risk factors (TableW1 , at www.jfponline.com) that were significantly different between women with and women without prior HT use. Women with and without prior HT use treated with raloxifene 60 mg/d had significant increases in BMD, at the lumbar spine (2.7% and 2.5%, respectively; interaction P=.54), and femoral neck (2.6% and 1.9% respectively; interaction P=.06), compared with placebo. Similar results were observed for fracture and BMD in women treated with the pooled raloxifene doses.

TABLE 1
Absolute and relative risks of new vertebral fractures with raloxifene 60 mg/d compared with placebo^a

FIGURE
Percentage of women with and without prior HT use who experienced new vertebral fractures in the 4-year MORE study

Cardiovascular events

In women with and without prior HT use, treatment with raloxifene (pooled doses) did not result in statistically significant changes in the incidence of new cardiovascular, coronary, or cerebrovascu-lar events, compared with placebo ( Table2 ). In a subgroup of women who were at high risk of cardiovascular disease,¹⁵ prior HT use had no effect on the incidence of new cardiovascular events with raloxifene (pooled doses) treatment ( Table 2 ). The interaction P-values remained similar after adjusting for the baseline cardiovascular risk factors (Table W1, at www.jfponline.com) that were significantly different between women with and without prior HT use.

TABLE 2
Absolute and relative risks of cardiovascular events with raloxifene (pooled doses) compared with placebo^a

Breast cancer

In women with and without prior HT use, similar reductions in the incidence of breast cancer (regardless of invasiveness), invasive breast cancer, and estrogen-receptor positive invasive breast cancer, were observed after raloxifene treatment (pooled doses) compared with placebo ( Table 3 ). The interaction P-values remained similar after adjusting for the baseline breast cancer risk factors (Table W1) that were significantly different between women with and without prior HT use.

TABLE 3
Absolute and relative risks of breast cancer with raloxifene (pooled doses) compared with placebo^a

Discussion

This analysis examined the effects of raloxifene in women who reported their use of postmenopausal hormone therapies before enrolling in the MORE osteoporosis trial. Compared with placebo, women treated with raloxifene experienced significant decreases in the risks for new vertebral fractures and the incidence of breast cancer, without significant changes in the incidence of cardiovascular events, regardless of previous HT use. These analyses provide further information on the effects of raloxifene on the risks of vertebral fractures,¹⁰ cardiovascular events,¹⁴ and breast cancer¹⁶ seen in the overall MORE study population at 4 years.

This analysis found a differential reduction in vertebral fracture risk with raloxifene between women who did and did not have prior HT use, which may result from possible differences in women who chose to use HT before participating in MORE. In women with previous HT use, a greater proportion had a family history of osteo-porosis and a lower proportion had prevalent ver-tebral fractures at baseline, compared with women who had not used HT. Other unidentified confounding factors, such as the “healthy user” bias commonly associated with women who chose to use HT,¹⁷ may also contribute to the differential vertebral fracture risk reduction with raloxifene treatment. The Study of Osteoporotic Fractures showed that women with current estrogen use had significantly decreased fracture risks, but the risk reduction waned in women who discontinued estrogen.¹⁸ After HT discontinuation, BMD loss resumes at a rate similar to that seen in women shortly after menopause, suggesting that prior HT use may have limited residual effects on maintaining BMD.^7,19 Such findings raise the urgency of evaluating the risk for osteoporosis in women who discontinue HT.

Women treated with raloxifene had no significant changes in the incidence of cardiovascular events, with no differential treatment effect based on prior HT use. In the HERS²⁰ and WHI³ trials, which studied the outcomes of estrogen-progestin therapy in postmenopausal women, similar analyses did not show any significant differential effects of prior HT use on the incidence of cardiovascular events with estrogen-progestin during the respective trials.

In this analysis, women treated with raloxifene had a significantly lower incidence of breast cancer compared with those who received placebo, and this incidence was comparable between women with and without prior HT use. In contrast, women who had used HT before the WHI study had a significant increase in the risk of breast cancer with estrogen-progestin therapy during the study, compared with those who had not used HT.³

A limitation of our analysis that a history of HT use was based on participants’ self-report, which depended on their ability to recall medication they may have taken years earlier. Also, no information was obtained on therapy duration and the doses and formulations of HT. Since the MORE trial was conducted in 25 countries, the patterns and types of HT regimens are expected to be different. The strength of our analysis is that the MORE population was large enough to prospectively collect data on multiple clinical outcomes.

In summary, postmenopausal women treated with raloxifene experienced a significant risk reduction for vertebral fractures, regardless of prior HT use, but women who had used HT may exhibit greater reductions. Women who used raloxifene had no change in the incidence of cardiovascular events and a lower incidence of breast cancer, compared with placebo, regardless of their history of HT use. Since HT is becoming increasingly limited to short-term use for menopausal symptoms, women and their physicians may consider several other therapeutic options to address postmenopausal health concerns.

Acknowledgments

The authors acknowledge the contributions of Leo Plouffe Jr., MD, and Somnath Sarkar, PhD, for suggestions on manuscript content, and Sharon Xiaohan Zou, MS, for statistical programming. A complete list of all investigators in the MORE trial is found in J Clin Endocrinol Metab 2002; 87:3609–3617. Portions of this work were presented at the following meetings: Third European Symposium on Clinical and Economic Aspects of Osteoporosis and Osteoarthritis, Barcelona, Spain, 2002; International Society for Clinical Densitometry (ISCD), Los Angeles, USA,2003; European Calcified Tissue Society (ECTS), Rome, Italy, 2003; International Bone and Mineral Society (IBMS), Osaka, Japan, 2003; European Menopause and Andropause Society (EMAS), Bucharest, Romania, 2003; Ninth Bath Conference on Osteoporosis, Bath, UK, 2003. Eli Lilly and Company sponsored the Multiple Outcomes of Raloxifene Evaluation (MORE) Trial.

Corresponding author
Olof Johnell, MD, PhD, Department of Orthopedics, Universitetssjukhuset MAS, Malmo, SE-20502, Sweden. E-mail: [email protected].

Estrogen-containing hormone therapies (HT) have been used to alleviate menopausal symptoms and to prevent chronic diseases common to postmenopausal women, including osteoporosis and cardiovascular disease.^1,2 In this analysis, we use the abbreviation “HT” to refer to postmenopausal hormone therapies, either estrogen alone or combined with progestin.

Based on the findings of the randomized, double-blind Women’s Health Initiative (WHI) study involving estrogen-progestin,³ the Food and Drug Administration (FDA) recommends using HT to treat moderate to severe symptoms of vulvar and vaginal atrophy and vasomotor symptoms associated with the menopause, and to prevent post-menopausal osteoporosis.⁴ When HT is prescribed only to prevent osteoporosis in women without menopausal symptoms, the FDA recommends that other approved, non-estrogen therapies be considered and that HT be used at the lowest dose for the shortest duration to achieve treatment goals.⁴ Many postmenopausal women have chosen to discontinue HT in light of these recommendations.⁵ However, discontinuing HT may increase bone resorption and accelerate bone loss,^6,7 which, if untreated, places women at risk for osteoporotic fractures.

The serum estrogen receptor modulator (SERM) raloxifene is not an estrogen, a progestin, or a hormone,⁸ but it binds to the estrogen receptor to exert effects in the skeletal and cardiovascular systems and in breast tissue.⁹ In the 4-year Multiple Outcomes of Raloxifene Evaluation (MORE) osteoporosis treatment trial of postmenopausal women, raloxifene 60 mg/d, the approved dose for post-menopausal osteoporosis prevention and treatment, increased bone mineral density (BMD) and significantly reduced the risk for new vertebral fractures with sustained efficacy.¹⁰ With the declining use of long-term HT,⁵ it is clinically relevant and important to determine whether a history of HT use has any influence on the effects of other antiresorptive agents, such as raloxifene, which may be subsequently used for postmenopausal osteoporosis prevention and treatment. The objective of this analysis is to determine the effects of raloxifene on BMD, and the risks of vertebral fractures, cardiovascular events, and breast cancer in post-menopausal women who did or did not use HT prior to screening for the MORE osteoporosis study.

Materials and methods

Subjects and treatment

Details on subject recruitment and follow-up, and complete inclusion and exclusion criteria, were previously described for the MORE study.¹¹ The trial examined the incidence of osteoporotic fractures as a primary endpoint and the incidence of breast cancer as a prespecified secondary endpoint, and it collected reports of cardiovascular events as a secondary safety endpoint.

Researchers enrolled 7705 women up to 80 years of age who were at least 2 years post-menopausal, and who had osteoporosis as defined by radiographically apparent vertebral fractures at baseline or BMD criteria. Women were randomly assigned to receive raloxifene 60 or 120 mg/d, or an identically appearing placebo.¹¹ All women received daily supplements of calcium (500 mg) and vitamin D (400 to 600 IU). An ethical review board at each site approved the MORE study protocol. All women gave written informed consent to participate in the study in accordance with the ethical principles stated in the Declaration of Helsinki.

At study screening, women were asked if they had ever taken HT. Women were excluded from the study if they were experiencing clinically severe menopausal symptoms at the beginning of the study that required estrogen. Women were excluded if they had been treated with therapeutic doses of androgen, calcitonin, or estrogen (>1 cycle or 28 days) alone or with progestin (>1 cycle or 28 days) within 6 months of beginning the study. Women were permitted to have used systemic (oral or transdermal) estrogen and progestin for up to 1 cycle (28 days) during the 6 months before the study. No systemic estrogen and progestin use was allowed within 2 months before study entry. Occasional use of topical estrogens (3 times per week), and oral estriol (2 mg/d) for menopausal symptoms was permitted.

A 1-year double-blind extension phase was added to the 3-year treatment phase.¹⁰ All vertebral fracture, cardiovascular, and breast cancer endpoints that occurred over the 48-month study period were included in the present analyses.

Fracture and BMD assessment

New morphometric vertebral fractures, defined using semiquantitative assessment criteria,¹² were identified by comparing spinal radiographs taken at 2, 3, and 4 years with baseline radiographs. A new vertebral fracture was defined as a vertebral fracture occurring in a vertebra that was not fractured at baseline. New clinical vertebral fractures were defined as those associated with signs or symptoms suggestive of vertebral fracture, such as back pain, reported either at an interim 6-month clinic visit or at any time between clinic visits,¹³ and which were subsequently corroborated with radiographs and adjudicated as previously described.¹⁰

Lumbar spine and femoral neck BMD were measured annually using dual-energy x-ray densitometry, as previously described.¹¹

Cardiovascular event assessment

Cardiovascular events were collected as a secondary safety endpoint in the MORE trial, as previously described in detail.¹⁴ Women were asked at each clinic visit if they had had a myocardial infarction (MI), coronary bypass surgery, percutaneous coronary revascularization, or a stroke since the previous visit, and unsolicited reports of adverse cardiovascular events were recorded. All summaries of reported cardiovascular events were reviewed and adjudicated by 1 board-certified cardiologist, contracted by the sponsor, who was not associated with the trial and was blinded to treatment assignment.

Women with 4 or more risk points, assessed using the same criteria as for enrollment in the Raloxifene Use for The Heart (RUTH) trial,¹⁵ were considered to be at high risk for cardiovascular events. Coronary events included MI, unstable angina, or coronary ischemia.¹⁴ Cerebrovascular events included stroke and transient ischemic attack.¹⁴

Breast cancer assessment

Assessment of breast cancer was a prespecified secondary endpoint of the MORE trial, and was previously described in detail.¹⁶ All diagnoses of breast cancer were adjudicated by an independent oncology review board consisting of 5 physician specialists in breast cancer, and chaired by a pharmacologic scientist, none of whom were employed by the sponsor. Previous publications of the MORE breast cancer data have reported 61 cases of invasive breast cancer.¹⁶ A subsequent review of the MORE dataset found 1 fewer case of invasive breast cancer in each of the placebo and pooled raloxifene groups, so that 59 cases of invasive breast cancer were confirmed, and this number will be used in the present analysis. The change in the number of cases of invasive breast cancer was small, and had no impact on the overall interpretation of the breast cancer results from the MORE trial.

Results

Characteristics for all subjects at baseline and for women in the placebo group

Of the 7705 women enrolled, 7682 (99.7%) reported their status of previous HT use, with 2235 women (29.1%) having used HT before participating in MORE ( W1, available at www.jfponline.com). Baseline characteristics that were significantly different between women who reported prior HT use and those who reported no prior HT use included age, BMD, and the incidences of vertebral fractures, coronary angioplasty, hypertension, hyperlipidemia, and family histories of osteoporosis or breast cancer (Table W1, available at www.jfponline.com). In the subsets of women who did and who did not use HT previously, the baseline characteristics were not significantly different between the placebo and raloxifene groups, except for diabetes (placebo, 1.9%; raloxifene 60 mg/day, 3.3%; pooled raloxifene, 2.9%; P=.02).

In the placebo group, the incidence of new vertebral fractures, cardiovascular events, and breast cancer at 4 years were not significantly different in women with prior HT use compared with women without prior HT use.

Vertebral fracture events and bone mineral density

After 4 years of treatment with raloxifene 60 mg/d, women with and without prior HT use exhibited significant reductions in new vertebral fractures compared with those taking placebo ( Figure ). Vertebral fracture risk reductions between treatment groups were also statistically significant in subgroups of women with and without prevalent vertebral fractures ( Table1 ). Raloxifene 60 mg/d also reduced the risk of new clinically-apparent vertebral fractures, compared with placebo, in women with prior HT use (absolute risk reduction (ARR)=1.8%; RR=0.52 [95% CI, 0.28–0.96]), and in women without prior HT use (ARR=1.7%; RR=0.61 [95% CI, 0.43–0.87]; interaction P=.66]. The interaction P-values remained similar after adjusting for the baseline fracture risk factors (TableW1 , at www.jfponline.com) that were significantly different between women with and women without prior HT use. Women with and without prior HT use treated with raloxifene 60 mg/d had significant increases in BMD, at the lumbar spine (2.7% and 2.5%, respectively; interaction P=.54), and femoral neck (2.6% and 1.9% respectively; interaction P=.06), compared with placebo. Similar results were observed for fracture and BMD in women treated with the pooled raloxifene doses.

TABLE 1
Absolute and relative risks of new vertebral fractures with raloxifene 60 mg/d compared with placebo^a

FIGURE
Percentage of women with and without prior HT use who experienced new vertebral fractures in the 4-year MORE study

Cardiovascular events

In women with and without prior HT use, treatment with raloxifene (pooled doses) did not result in statistically significant changes in the incidence of new cardiovascular, coronary, or cerebrovascu-lar events, compared with placebo ( Table2 ). In a subgroup of women who were at high risk of cardiovascular disease,¹⁵ prior HT use had no effect on the incidence of new cardiovascular events with raloxifene (pooled doses) treatment ( Table 2 ). The interaction P-values remained similar after adjusting for the baseline cardiovascular risk factors (Table W1, at www.jfponline.com) that were significantly different between women with and without prior HT use.

TABLE 2
Absolute and relative risks of cardiovascular events with raloxifene (pooled doses) compared with placebo^a

Breast cancer

In women with and without prior HT use, similar reductions in the incidence of breast cancer (regardless of invasiveness), invasive breast cancer, and estrogen-receptor positive invasive breast cancer, were observed after raloxifene treatment (pooled doses) compared with placebo ( Table 3 ). The interaction P-values remained similar after adjusting for the baseline breast cancer risk factors (Table W1) that were significantly different between women with and without prior HT use.

TABLE 3
Absolute and relative risks of breast cancer with raloxifene (pooled doses) compared with placebo^a

Discussion

This analysis examined the effects of raloxifene in women who reported their use of postmenopausal hormone therapies before enrolling in the MORE osteoporosis trial. Compared with placebo, women treated with raloxifene experienced significant decreases in the risks for new vertebral fractures and the incidence of breast cancer, without significant changes in the incidence of cardiovascular events, regardless of previous HT use. These analyses provide further information on the effects of raloxifene on the risks of vertebral fractures,¹⁰ cardiovascular events,¹⁴ and breast cancer¹⁶ seen in the overall MORE study population at 4 years.

This analysis found a differential reduction in vertebral fracture risk with raloxifene between women who did and did not have prior HT use, which may result from possible differences in women who chose to use HT before participating in MORE. In women with previous HT use, a greater proportion had a family history of osteo-porosis and a lower proportion had prevalent ver-tebral fractures at baseline, compared with women who had not used HT. Other unidentified confounding factors, such as the “healthy user” bias commonly associated with women who chose to use HT,¹⁷ may also contribute to the differential vertebral fracture risk reduction with raloxifene treatment. The Study of Osteoporotic Fractures showed that women with current estrogen use had significantly decreased fracture risks, but the risk reduction waned in women who discontinued estrogen.¹⁸ After HT discontinuation, BMD loss resumes at a rate similar to that seen in women shortly after menopause, suggesting that prior HT use may have limited residual effects on maintaining BMD.^7,19 Such findings raise the urgency of evaluating the risk for osteoporosis in women who discontinue HT.

Women treated with raloxifene had no significant changes in the incidence of cardiovascular events, with no differential treatment effect based on prior HT use. In the HERS²⁰ and WHI³ trials, which studied the outcomes of estrogen-progestin therapy in postmenopausal women, similar analyses did not show any significant differential effects of prior HT use on the incidence of cardiovascular events with estrogen-progestin during the respective trials.

In this analysis, women treated with raloxifene had a significantly lower incidence of breast cancer compared with those who received placebo, and this incidence was comparable between women with and without prior HT use. In contrast, women who had used HT before the WHI study had a significant increase in the risk of breast cancer with estrogen-progestin therapy during the study, compared with those who had not used HT.³

A limitation of our analysis that a history of HT use was based on participants’ self-report, which depended on their ability to recall medication they may have taken years earlier. Also, no information was obtained on therapy duration and the doses and formulations of HT. Since the MORE trial was conducted in 25 countries, the patterns and types of HT regimens are expected to be different. The strength of our analysis is that the MORE population was large enough to prospectively collect data on multiple clinical outcomes.

In summary, postmenopausal women treated with raloxifene experienced a significant risk reduction for vertebral fractures, regardless of prior HT use, but women who had used HT may exhibit greater reductions. Women who used raloxifene had no change in the incidence of cardiovascular events and a lower incidence of breast cancer, compared with placebo, regardless of their history of HT use. Since HT is becoming increasingly limited to short-term use for menopausal symptoms, women and their physicians may consider several other therapeutic options to address postmenopausal health concerns.

Acknowledgments

The authors acknowledge the contributions of Leo Plouffe Jr., MD, and Somnath Sarkar, PhD, for suggestions on manuscript content, and Sharon Xiaohan Zou, MS, for statistical programming. A complete list of all investigators in the MORE trial is found in J Clin Endocrinol Metab 2002; 87:3609–3617. Portions of this work were presented at the following meetings: Third European Symposium on Clinical and Economic Aspects of Osteoporosis and Osteoarthritis, Barcelona, Spain, 2002; International Society for Clinical Densitometry (ISCD), Los Angeles, USA,2003; European Calcified Tissue Society (ECTS), Rome, Italy, 2003; International Bone and Mineral Society (IBMS), Osaka, Japan, 2003; European Menopause and Andropause Society (EMAS), Bucharest, Romania, 2003; Ninth Bath Conference on Osteoporosis, Bath, UK, 2003. Eli Lilly and Company sponsored the Multiple Outcomes of Raloxifene Evaluation (MORE) Trial.

Corresponding author
Olof Johnell, MD, PhD, Department of Orthopedics, Universitetssjukhuset MAS, Malmo, SE-20502, Sweden. E-mail: [email protected].

Estrogen-containing hormone therapies (HT) have been used to alleviate menopausal symptoms and to prevent chronic diseases common to postmenopausal women, including osteoporosis and cardiovascular disease.^1,2 In this analysis, we use the abbreviation “HT” to refer to postmenopausal hormone therapies, either estrogen alone or combined with progestin.

Based on the findings of the randomized, double-blind Women’s Health Initiative (WHI) study involving estrogen-progestin,³ the Food and Drug Administration (FDA) recommends using HT to treat moderate to severe symptoms of vulvar and vaginal atrophy and vasomotor symptoms associated with the menopause, and to prevent post-menopausal osteoporosis.⁴ When HT is prescribed only to prevent osteoporosis in women without menopausal symptoms, the FDA recommends that other approved, non-estrogen therapies be considered and that HT be used at the lowest dose for the shortest duration to achieve treatment goals.⁴ Many postmenopausal women have chosen to discontinue HT in light of these recommendations.⁵ However, discontinuing HT may increase bone resorption and accelerate bone loss,^6,7 which, if untreated, places women at risk for osteoporotic fractures.

The serum estrogen receptor modulator (SERM) raloxifene is not an estrogen, a progestin, or a hormone,⁸ but it binds to the estrogen receptor to exert effects in the skeletal and cardiovascular systems and in breast tissue.⁹ In the 4-year Multiple Outcomes of Raloxifene Evaluation (MORE) osteoporosis treatment trial of postmenopausal women, raloxifene 60 mg/d, the approved dose for post-menopausal osteoporosis prevention and treatment, increased bone mineral density (BMD) and significantly reduced the risk for new vertebral fractures with sustained efficacy.¹⁰ With the declining use of long-term HT,⁵ it is clinically relevant and important to determine whether a history of HT use has any influence on the effects of other antiresorptive agents, such as raloxifene, which may be subsequently used for postmenopausal osteoporosis prevention and treatment. The objective of this analysis is to determine the effects of raloxifene on BMD, and the risks of vertebral fractures, cardiovascular events, and breast cancer in post-menopausal women who did or did not use HT prior to screening for the MORE osteoporosis study.

Materials and methods

Subjects and treatment

Details on subject recruitment and follow-up, and complete inclusion and exclusion criteria, were previously described for the MORE study.¹¹ The trial examined the incidence of osteoporotic fractures as a primary endpoint and the incidence of breast cancer as a prespecified secondary endpoint, and it collected reports of cardiovascular events as a secondary safety endpoint.

Researchers enrolled 7705 women up to 80 years of age who were at least 2 years post-menopausal, and who had osteoporosis as defined by radiographically apparent vertebral fractures at baseline or BMD criteria. Women were randomly assigned to receive raloxifene 60 or 120 mg/d, or an identically appearing placebo.¹¹ All women received daily supplements of calcium (500 mg) and vitamin D (400 to 600 IU). An ethical review board at each site approved the MORE study protocol. All women gave written informed consent to participate in the study in accordance with the ethical principles stated in the Declaration of Helsinki.

At study screening, women were asked if they had ever taken HT. Women were excluded from the study if they were experiencing clinically severe menopausal symptoms at the beginning of the study that required estrogen. Women were excluded if they had been treated with therapeutic doses of androgen, calcitonin, or estrogen (>1 cycle or 28 days) alone or with progestin (>1 cycle or 28 days) within 6 months of beginning the study. Women were permitted to have used systemic (oral or transdermal) estrogen and progestin for up to 1 cycle (28 days) during the 6 months before the study. No systemic estrogen and progestin use was allowed within 2 months before study entry. Occasional use of topical estrogens (3 times per week), and oral estriol (2 mg/d) for menopausal symptoms was permitted.

A 1-year double-blind extension phase was added to the 3-year treatment phase.¹⁰ All vertebral fracture, cardiovascular, and breast cancer endpoints that occurred over the 48-month study period were included in the present analyses.

Fracture and BMD assessment

New morphometric vertebral fractures, defined using semiquantitative assessment criteria,¹² were identified by comparing spinal radiographs taken at 2, 3, and 4 years with baseline radiographs. A new vertebral fracture was defined as a vertebral fracture occurring in a vertebra that was not fractured at baseline. New clinical vertebral fractures were defined as those associated with signs or symptoms suggestive of vertebral fracture, such as back pain, reported either at an interim 6-month clinic visit or at any time between clinic visits,¹³ and which were subsequently corroborated with radiographs and adjudicated as previously described.¹⁰

Lumbar spine and femoral neck BMD were measured annually using dual-energy x-ray densitometry, as previously described.¹¹

Cardiovascular event assessment

Cardiovascular events were collected as a secondary safety endpoint in the MORE trial, as previously described in detail.¹⁴ Women were asked at each clinic visit if they had had a myocardial infarction (MI), coronary bypass surgery, percutaneous coronary revascularization, or a stroke since the previous visit, and unsolicited reports of adverse cardiovascular events were recorded. All summaries of reported cardiovascular events were reviewed and adjudicated by 1 board-certified cardiologist, contracted by the sponsor, who was not associated with the trial and was blinded to treatment assignment.

Women with 4 or more risk points, assessed using the same criteria as for enrollment in the Raloxifene Use for The Heart (RUTH) trial,¹⁵ were considered to be at high risk for cardiovascular events. Coronary events included MI, unstable angina, or coronary ischemia.¹⁴ Cerebrovascular events included stroke and transient ischemic attack.¹⁴

Breast cancer assessment

Assessment of breast cancer was a prespecified secondary endpoint of the MORE trial, and was previously described in detail.¹⁶ All diagnoses of breast cancer were adjudicated by an independent oncology review board consisting of 5 physician specialists in breast cancer, and chaired by a pharmacologic scientist, none of whom were employed by the sponsor. Previous publications of the MORE breast cancer data have reported 61 cases of invasive breast cancer.¹⁶ A subsequent review of the MORE dataset found 1 fewer case of invasive breast cancer in each of the placebo and pooled raloxifene groups, so that 59 cases of invasive breast cancer were confirmed, and this number will be used in the present analysis. The change in the number of cases of invasive breast cancer was small, and had no impact on the overall interpretation of the breast cancer results from the MORE trial.

Results

Characteristics for all subjects at baseline and for women in the placebo group

Of the 7705 women enrolled, 7682 (99.7%) reported their status of previous HT use, with 2235 women (29.1%) having used HT before participating in MORE ( W1, available at www.jfponline.com). Baseline characteristics that were significantly different between women who reported prior HT use and those who reported no prior HT use included age, BMD, and the incidences of vertebral fractures, coronary angioplasty, hypertension, hyperlipidemia, and family histories of osteoporosis or breast cancer (Table W1, available at www.jfponline.com). In the subsets of women who did and who did not use HT previously, the baseline characteristics were not significantly different between the placebo and raloxifene groups, except for diabetes (placebo, 1.9%; raloxifene 60 mg/day, 3.3%; pooled raloxifene, 2.9%; P=.02).

In the placebo group, the incidence of new vertebral fractures, cardiovascular events, and breast cancer at 4 years were not significantly different in women with prior HT use compared with women without prior HT use.

Vertebral fracture events and bone mineral density

After 4 years of treatment with raloxifene 60 mg/d, women with and without prior HT use exhibited significant reductions in new vertebral fractures compared with those taking placebo ( Figure ). Vertebral fracture risk reductions between treatment groups were also statistically significant in subgroups of women with and without prevalent vertebral fractures ( Table1 ). Raloxifene 60 mg/d also reduced the risk of new clinically-apparent vertebral fractures, compared with placebo, in women with prior HT use (absolute risk reduction (ARR)=1.8%; RR=0.52 [95% CI, 0.28–0.96]), and in women without prior HT use (ARR=1.7%; RR=0.61 [95% CI, 0.43–0.87]; interaction P=.66]. The interaction P-values remained similar after adjusting for the baseline fracture risk factors (TableW1 , at www.jfponline.com) that were significantly different between women with and women without prior HT use. Women with and without prior HT use treated with raloxifene 60 mg/d had significant increases in BMD, at the lumbar spine (2.7% and 2.5%, respectively; interaction P=.54), and femoral neck (2.6% and 1.9% respectively; interaction P=.06), compared with placebo. Similar results were observed for fracture and BMD in women treated with the pooled raloxifene doses.

TABLE 1
Absolute and relative risks of new vertebral fractures with raloxifene 60 mg/d compared with placebo^a

FIGURE
Percentage of women with and without prior HT use who experienced new vertebral fractures in the 4-year MORE study

Cardiovascular events

In women with and without prior HT use, treatment with raloxifene (pooled doses) did not result in statistically significant changes in the incidence of new cardiovascular, coronary, or cerebrovascu-lar events, compared with placebo ( Table2 ). In a subgroup of women who were at high risk of cardiovascular disease,¹⁵ prior HT use had no effect on the incidence of new cardiovascular events with raloxifene (pooled doses) treatment ( Table 2 ). The interaction P-values remained similar after adjusting for the baseline cardiovascular risk factors (Table W1, at www.jfponline.com) that were significantly different between women with and without prior HT use.

TABLE 2
Absolute and relative risks of cardiovascular events with raloxifene (pooled doses) compared with placebo^a

Breast cancer

In women with and without prior HT use, similar reductions in the incidence of breast cancer (regardless of invasiveness), invasive breast cancer, and estrogen-receptor positive invasive breast cancer, were observed after raloxifene treatment (pooled doses) compared with placebo ( Table 3 ). The interaction P-values remained similar after adjusting for the baseline breast cancer risk factors (Table W1) that were significantly different between women with and without prior HT use.

TABLE 3
Absolute and relative risks of breast cancer with raloxifene (pooled doses) compared with placebo^a

Discussion

This analysis examined the effects of raloxifene in women who reported their use of postmenopausal hormone therapies before enrolling in the MORE osteoporosis trial. Compared with placebo, women treated with raloxifene experienced significant decreases in the risks for new vertebral fractures and the incidence of breast cancer, without significant changes in the incidence of cardiovascular events, regardless of previous HT use. These analyses provide further information on the effects of raloxifene on the risks of vertebral fractures,¹⁰ cardiovascular events,¹⁴ and breast cancer¹⁶ seen in the overall MORE study population at 4 years.

This analysis found a differential reduction in vertebral fracture risk with raloxifene between women who did and did not have prior HT use, which may result from possible differences in women who chose to use HT before participating in MORE. In women with previous HT use, a greater proportion had a family history of osteo-porosis and a lower proportion had prevalent ver-tebral fractures at baseline, compared with women who had not used HT. Other unidentified confounding factors, such as the “healthy user” bias commonly associated with women who chose to use HT,¹⁷ may also contribute to the differential vertebral fracture risk reduction with raloxifene treatment. The Study of Osteoporotic Fractures showed that women with current estrogen use had significantly decreased fracture risks, but the risk reduction waned in women who discontinued estrogen.¹⁸ After HT discontinuation, BMD loss resumes at a rate similar to that seen in women shortly after menopause, suggesting that prior HT use may have limited residual effects on maintaining BMD.^7,19 Such findings raise the urgency of evaluating the risk for osteoporosis in women who discontinue HT.

Women treated with raloxifene had no significant changes in the incidence of cardiovascular events, with no differential treatment effect based on prior HT use. In the HERS²⁰ and WHI³ trials, which studied the outcomes of estrogen-progestin therapy in postmenopausal women, similar analyses did not show any significant differential effects of prior HT use on the incidence of cardiovascular events with estrogen-progestin during the respective trials.

In this analysis, women treated with raloxifene had a significantly lower incidence of breast cancer compared with those who received placebo, and this incidence was comparable between women with and without prior HT use. In contrast, women who had used HT before the WHI study had a significant increase in the risk of breast cancer with estrogen-progestin therapy during the study, compared with those who had not used HT.³

A limitation of our analysis that a history of HT use was based on participants’ self-report, which depended on their ability to recall medication they may have taken years earlier. Also, no information was obtained on therapy duration and the doses and formulations of HT. Since the MORE trial was conducted in 25 countries, the patterns and types of HT regimens are expected to be different. The strength of our analysis is that the MORE population was large enough to prospectively collect data on multiple clinical outcomes.

In summary, postmenopausal women treated with raloxifene experienced a significant risk reduction for vertebral fractures, regardless of prior HT use, but women who had used HT may exhibit greater reductions. Women who used raloxifene had no change in the incidence of cardiovascular events and a lower incidence of breast cancer, compared with placebo, regardless of their history of HT use. Since HT is becoming increasingly limited to short-term use for menopausal symptoms, women and their physicians may consider several other therapeutic options to address postmenopausal health concerns.

Acknowledgments

The authors acknowledge the contributions of Leo Plouffe Jr., MD, and Somnath Sarkar, PhD, for suggestions on manuscript content, and Sharon Xiaohan Zou, MS, for statistical programming. A complete list of all investigators in the MORE trial is found in J Clin Endocrinol Metab 2002; 87:3609–3617. Portions of this work were presented at the following meetings: Third European Symposium on Clinical and Economic Aspects of Osteoporosis and Osteoarthritis, Barcelona, Spain, 2002; International Society for Clinical Densitometry (ISCD), Los Angeles, USA,2003; European Calcified Tissue Society (ECTS), Rome, Italy, 2003; International Bone and Mineral Society (IBMS), Osaka, Japan, 2003; European Menopause and Andropause Society (EMAS), Bucharest, Romania, 2003; Ninth Bath Conference on Osteoporosis, Bath, UK, 2003. Eli Lilly and Company sponsored the Multiple Outcomes of Raloxifene Evaluation (MORE) Trial.

Corresponding author
Olof Johnell, MD, PhD, Department of Orthopedics, Universitetssjukhuset MAS, Malmo, SE-20502, Sweden. E-mail: [email protected].

Racial and ethnic disparities in health care have been catalogued across numerous diseases and care settings.¹ By clarifying the causes of these disparities, we can develop solutions. In a seminal study, Shulman found in patient simulations that identical presentations for heart disease received different recommendations for care based on the patient’s race and gender, thus pinpointing the patient-provider relationship as a potential source of disparities.² Other research suggests interactions with non-physician health care personnel might also be a source of negative experiences with care.³

Research is beginning to identify how the health care encounter might relate to disparities in use of services and quality of care. For example, race concordance between the physician and patient, at least for blacks, is associated with higher patient satisfaction and greater participatory decision-making. This in turn can impact compliance and possibly outcomes.^4-6 While black patients who have black physicians are more likely to report receipt of counseling about preventive care and cancer screening,⁷ race concordance does not appear to be independently associated with different patterns of utilization.⁸

Perceived discrimination has also been associated with lower levels of satisfaction with the health care system.⁹ In one survey, two thirds of respondents reported feeling discriminated against in their interactions with health care providers due to their race or socioeconomic status.¹⁰ How perceived discrimination influences quality and outcomes of care has not been fully explored.

We hypothesized that minority patients and those who do not speak English perceive negative experiences with the health care encounter more often than whites or English-speakers. We further hypothesized that patients who report such negative experiences are less likely to seek care initially or return for follow-up care. We tested these hypotheses using data from the Commonwealth Fund 2001 Health Care Quality Survey.

Methods

Sample

Respondents were from a nationally representative sample of 6722 adults, aged 18 years and older, living in the continental United States, and who speak English, Spanish, Mandarin, Cantonese, Vietnamese, or Korean.

The sampling frame was based on random-digit dialing; telephone exchanges with higher-than-average numbers of minority households were oversampled. In addition to the oversampling based on telephone exchanges, we interviewed members of 394 households identified from a nationwide demographic tracking survey as having an Asian/Asian American or African American family member. Interviews were conducted in English, Spanish, Mandarin, Cantonese, Vietnamese, or Korean, depending on the respondent preference. The response rate for the entire sample was 53.1%.

The final sample was weighted to correct for the disproportionate sample design and to ensure the sample was representative of all adults aged 18 years and older based on the March 2001 Current Population Survey (CPS). The final weighted sample is therefore reflective of the 193 million adults in the United States who have telephones. A more detailed description of the sampling and weighting methods can be found else-where.¹¹ Data were collected between April 30 and November 5, 2001.

We focus on the subset of questions addressing respondent’s experience with the health care encounter and their use of certain health care services. We used STATA Version 6.0¹² to conduct statistical analyses using the weighted sample.

Dependent variables

Negative perceptions of the patient-provider relationship. We identified factors that define the patient-provider relationship, and formulated questions based on these factors that would elicit meaningful responses. Specific questions included:

• “Did the doctor treat you with a great deal of respect and dignity, a fair amount, not too much, or none at all?” (4-point scale)
• Please tell me if you strongly agree, somewhat agree, somewhat disagree, or strongly disagree with the statement, “I often feel as if my doctor looks down on me and the way I live my life.” (4-point scale)
• “Thinking about all of the experiences you have had with health care visits in the last two years, have you ever felt that the doctor or medical staff you saw judged you unfairly or treated you with disrespect because of how well you speak English?” (yes/no)
• “Thinking about all of the experiences you have had with health care visits in the last two years, have you ever felt that the doctor or medical staff you saw judged you unfairly or treated you with disrespect because of your race or ethnic background?” (yes/no)
• “Do you think there was ever a time when you would have gotten better medical care if you had belonged to a different race or ethnic group?” (yes/no)

Those who said they did not have a doctor were unable to answer these questions and were excluded from our analysis. We combined the first 2 questions into a single dichotomous variable, characterized as “being treated with disrespect,” because both questions described negative perceptions of the health care encounter and because doing so preserved sample size for our analyses.

Utilization and optimal care

We examined self-reported use of specific services, including whether respondents had a physical exam within the past year. For cancer screening and chronic disease care, we created variables designed to represent optimal care. For example, optimal cancer screening was defined as being up to date on all tests for which the individual was eligible, based on age and gender.

These included:

• fecal occult blood testing for colon cancer screening within the prior year (both female and male respondents aged >50 years )
• cervical cancer screening within the prior 3 years (all women over the age of the 18)
• mammography within the prior year (women over the age of 50).

We excluded men younger than 50 years since colon cancer screening is not routinely recommended.

We considered respondents to have optimal chronic disease testing if they reported receiving all appropriate testing relevant for their particular condition. For persons with diabetes, this included having a hemoglobin A1c level checked within the past 6 months, a blood pressure check and foot and eye exams within the year, and cholesterol testing within 5 years. For those with heart disease or hypertension, it included having had blood pressure checked within the prior year and cholesterol testing within the prior 5 years. This approach is consistent with that of McBean et al, who have shown that a combination of appropriate tests is more predictive of glycemic control for diabetes.¹³

Because we were interested in different aspects of patient-initiated care seeking, we also evaluated delay in seeking care and adherence to physician recommendations as further measures of outcome. Specific questions were:

• “During the last 12 months, was there any time when you had a medical problem but put off, postponed, or did not seek medical care when you needed to?”
• “Has there been a time in the last two years when you didn’t follow the doctor’s advice or treatment plan, get a recommended test, or see a referred doctor?” (asked of respondents who had visited a doctor or clinic or had been admitted to the hospital in the last 2 years.)

Analysis

To test our first hypothesis—that persons of racial/ethnic minorities perceive negative experiences with the health care encounter more often than whites or English speakers—we examined

associations between demographic characteristics, utilization variables, and negative perceptions of the health care encounter using chisquared tests and multivariate logistic regression. In these analyses, we dichotomized education into high school graduate or less, and some technical school/college and more. We dichotomized the primary language spoken at home into non-English and English, and we used federal poverty level groupings (<100%, 100%–200% and >200%) to categorize household income. Almost 19% percent of respondents did not report incomes, so we created a dummy variable to account for those with unreported incomes.

We classified insurance status as none or any (either public or private); race/ethnicity as white, black, Hispanic, Asian, and other (Native American, mixed race, or other). We examined the effect of these variables alone and in concert. For example, we calculated predicted percentages to evaluate the combined effects of race and gender, as well as race and education, in relationship to our outcome variables.

Finally, we used multivariate logistic regression to test the relationship between negative perceptions of the patient-provider relationship and our utilization variables. In these analyses, perceptions were the covariate of interest; we controlled for patient characteristics that could also influence utilization, including education, income, insurance status, presence of a primary physician and existence of a comorbid condition (in this case, hypertension, diabetes, heart disease, asthma, and cancer.) This last variable was, by necessity, excluded from the analysis involving optimal chronic disease testing.

Results

Table 1 describes demographic characteristics and utilization measures for our sample. Consistent with prior literature, blacks and Hispanics had lower incomes and higher rates of non-insurance than both whites and Asians. Hispanics responded most frequently that English was not their primary language.

Hispanics and Asians were less likely than whites to have received optimal chronic disease care, while blacks and Hispanics were more likely than whites to have received optimal cancer screening. There were no differences between racial/ethnic groups in not following the doctor’s advice or in putting off care.

TABLE 1
Demographics/characteristics and health care utilization of study participants

Negative perceptions of the patient-provider relationship

Race. Over 14% of blacks, 19% of Hispanics, and 20% of Asians reported they had been treated with disrespect by their doctor. Members of these groups were also more likely than whites to report that they were treated unfairly because of their race or their language, and that they would have received better care had they belonged to a different race (Table 2).

Language. Persons for whom English was not the primary language were also more likely to say they had been treated with disrespect, and to report they would have received better care had they been of a different race. For each racial/ethnic group, bivariate relationships persisted after controlling for other respondent characteristics, including education and income (Table 2).

Sex. Men were significantly more likely than women to perceive being treated with disrespect by the doctor (15.9% vs 11.6%), and the percentage varied by race/ethnicity. Using our model to predict the combined effects of race and gender, we found that Asian and Hispanic men (24% and 23%, respectively) were more likely than black men (17%) or white men (11%) to perceive being treated with disrespect.

Education. Education was similarly associated with perceptions of disrespect. Almost 18% of persons without a college education believed they had been treated with disrespect, versus only 10% of those with a college education. Minorities with lower education were more likely to have this perception. Twenty-nine percent of Asians, 22% of Hispanics, and 19% of blacks without a college education reported being treated with disrespect or being looked down upon, versus 13% of whites.

TABLE 2
Relationship of demographic variables to measures of negative perceptions

Impact on care

Respondents who reported being treated with disrespect were significantly less likely to have had a physical exam within the prior year; those with diabetes, hypertension, or heart disease were less likely to have received optimal care. These respondents were also more likely to report not following the doctor’s advice and putting off needed care (Table 3). This relationship was not seen for optimal cancer screening.

Persons who believed they had been treated unfairly due to their race and who thought they would have received better care had they been of a different race were more likely to ignore the doctor’s advice and put off care when medically needed. Those who believed they would have received better care had they been of a different race were also less likely to receive optimal chronic disease care. In analyses not shown, we examined the independent effects of income and education, as well as interactions between these variables and insurance, and found the results basically unchanged.

TABLE 3
Relationship of negative perceptions to health care outcomes

Discussion

We hypothesized that patients who have negative perceptions of the patient-provider relationship would be less likely to seek needed care, and that reports of such feelings would be more prevalent among minority patients. As anticipated, large proportions of blacks, Hispanics, and Asians reported that they were treated with disrespect, were treated unfairly, or would have received better care if their race had been different. Male gender and lower educational attainment were also associated with perceived disrespect, particularly among minorities.

Negative experiences lead to suboptimal care

The finding of greater likelihood of perceived disrespect among minority groups, men, and those with lower levels of education is particularly important in light of the strong relationship between such reports and the quality of care that patients receive. Those who reported that they were treated unfairly because of race were less likely to get a routine physical exam, follow a doctor’s advice, or receive appropriate secondary preventive care for diabetes, heart disease, and hypertension. In other words, negative experiences within the health care environment may jeopardize care for medically needy patients. Receipt of suboptimal care, particularly in the context of chronic disease, is likely to be associated with worse health outcomes, and may contribute to disparities.

Cancer screening the exception

While the relationships between negative perceptions and receipt of care for chronic disease and receipt of a routine physical examination were strong, the correlation did not persist for cancer screening. Black and Hispanic respondents were more likely than whites to receive optimal cancer screening, a finding that has been reported elsewhere.^{14, 15}

We hypothesize that this is in part because a wide array of community programs make special outreach efforts allowing patients to “bypass” the traditional office environment.¹⁶ These settings may be more likely to use culturally sensitive approaches or may be so transient that negative perceptions based on race or income may be less likely to form.

However, based on our finding that care requiring follow-up (eg, diabetes management) is less likely to occur with individuals who report negative perceptions of the patient-provider relationship, we would hypothesize that individuals who receive initial cancer screening might be less likely to follow up on abnormal results once screened. It may be that in situations requiring long term relationships, such as chronic disease care, perceptions of discrimination and disrespect may take the greatest toll. This hypothesis is supported by previous literature consistently reporting excess mortality despite higher cancer screening rates among blacks.^17-20

Limitations

Our study has several limitations. Because we are relying on self-report, we could not assess which specific aspects of the patient-provider relationship may have influenced the reports of disrespect. Responses may have been affected by experiences completely outside of this relation ship, or outside of the health care system, that independently could have impacted health care utilization. We cannot disentangle how general life experience influences perceptions of the health care encounter or care-seeking; for example respondents who perceive racial bias in other environments such as the workplace may also be more likely to perceive it in the healthcare setting. In addition, self-reported utilization measures may not always be accurate, particularly regarding cancer screening.^21,22

Despite the deliberate oversampling of major racial/ethnic groups, we remain limited in our ability to examine important subgroups within them, whether related to ethnicity (eg, Cuban, Vietnamese) or chronic condition (eg, asthma, diabetes), even though some groups may differ dramatically from others.

We also had insufficient numbers of Native Americans to analyze separately.

We excluded respondents who did not have a regular doctor because they were unable to answer key questions about the health care encounter.

Finally, there is no agreement on the definition of age-appropriate breast and cervical cancer screening.^{23, 24} We conducted additional analyses varying the age criteria for testing, including starting the required age for screening at age 40 (for breast cancer) as well as setting the age cut off for required screening at 65 (for both breast and cervical cancer screening), and found that the results were essentially unchanged from those presented. Similarly, adding prostate cancer screening to our models for men over age 50 did not alter our results significantly.

Research should focus on improving perceptions of care

Although it is difficult to quantify or measure negative responses objectively, the strong relationship between patient perceptions of the encounter and utilization suggests an important area for further attention. These findings suggest there may still be a substantial core of individuals who will actively avoid care, perhaps based on previous negative interpersonal experiences in getting care. Interventions aimed at both doctors and potential patients will be required to address this. Research is needed to focus on what approaches can best improve perceptions of care within the patient-provider relationship and how such interventions can reduce racial disparities in health care.

Acknowledgments

This research was done in part by a grant from The Commonwealth Fund.

Corresponding author
Nicole Lurie, MD, MSPH, RAND Corporation, 1200 S. Hayes Street, Arlington, VA 22202. E-mail: [email protected]

Racial and ethnic disparities in health care have been catalogued across numerous diseases and care settings.¹ By clarifying the causes of these disparities, we can develop solutions. In a seminal study, Shulman found in patient simulations that identical presentations for heart disease received different recommendations for care based on the patient’s race and gender, thus pinpointing the patient-provider relationship as a potential source of disparities.² Other research suggests interactions with non-physician health care personnel might also be a source of negative experiences with care.³

Research is beginning to identify how the health care encounter might relate to disparities in use of services and quality of care. For example, race concordance between the physician and patient, at least for blacks, is associated with higher patient satisfaction and greater participatory decision-making. This in turn can impact compliance and possibly outcomes.^4-6 While black patients who have black physicians are more likely to report receipt of counseling about preventive care and cancer screening,⁷ race concordance does not appear to be independently associated with different patterns of utilization.⁸

Perceived discrimination has also been associated with lower levels of satisfaction with the health care system.⁹ In one survey, two thirds of respondents reported feeling discriminated against in their interactions with health care providers due to their race or socioeconomic status.¹⁰ How perceived discrimination influences quality and outcomes of care has not been fully explored.

We hypothesized that minority patients and those who do not speak English perceive negative experiences with the health care encounter more often than whites or English-speakers. We further hypothesized that patients who report such negative experiences are less likely to seek care initially or return for follow-up care. We tested these hypotheses using data from the Commonwealth Fund 2001 Health Care Quality Survey.

Methods

Sample

Respondents were from a nationally representative sample of 6722 adults, aged 18 years and older, living in the continental United States, and who speak English, Spanish, Mandarin, Cantonese, Vietnamese, or Korean.

The sampling frame was based on random-digit dialing; telephone exchanges with higher-than-average numbers of minority households were oversampled. In addition to the oversampling based on telephone exchanges, we interviewed members of 394 households identified from a nationwide demographic tracking survey as having an Asian/Asian American or African American family member. Interviews were conducted in English, Spanish, Mandarin, Cantonese, Vietnamese, or Korean, depending on the respondent preference. The response rate for the entire sample was 53.1%.

The final sample was weighted to correct for the disproportionate sample design and to ensure the sample was representative of all adults aged 18 years and older based on the March 2001 Current Population Survey (CPS). The final weighted sample is therefore reflective of the 193 million adults in the United States who have telephones. A more detailed description of the sampling and weighting methods can be found else-where.¹¹ Data were collected between April 30 and November 5, 2001.

We focus on the subset of questions addressing respondent’s experience with the health care encounter and their use of certain health care services. We used STATA Version 6.0¹² to conduct statistical analyses using the weighted sample.

Dependent variables

Negative perceptions of the patient-provider relationship. We identified factors that define the patient-provider relationship, and formulated questions based on these factors that would elicit meaningful responses. Specific questions included:

• “Did the doctor treat you with a great deal of respect and dignity, a fair amount, not too much, or none at all?” (4-point scale)
• Please tell me if you strongly agree, somewhat agree, somewhat disagree, or strongly disagree with the statement, “I often feel as if my doctor looks down on me and the way I live my life.” (4-point scale)
• “Thinking about all of the experiences you have had with health care visits in the last two years, have you ever felt that the doctor or medical staff you saw judged you unfairly or treated you with disrespect because of how well you speak English?” (yes/no)
• “Thinking about all of the experiences you have had with health care visits in the last two years, have you ever felt that the doctor or medical staff you saw judged you unfairly or treated you with disrespect because of your race or ethnic background?” (yes/no)
• “Do you think there was ever a time when you would have gotten better medical care if you had belonged to a different race or ethnic group?” (yes/no)

Those who said they did not have a doctor were unable to answer these questions and were excluded from our analysis. We combined the first 2 questions into a single dichotomous variable, characterized as “being treated with disrespect,” because both questions described negative perceptions of the health care encounter and because doing so preserved sample size for our analyses.

Utilization and optimal care

We examined self-reported use of specific services, including whether respondents had a physical exam within the past year. For cancer screening and chronic disease care, we created variables designed to represent optimal care. For example, optimal cancer screening was defined as being up to date on all tests for which the individual was eligible, based on age and gender.

These included:

• fecal occult blood testing for colon cancer screening within the prior year (both female and male respondents aged >50 years )
• cervical cancer screening within the prior 3 years (all women over the age of the 18)
• mammography within the prior year (women over the age of 50).

We excluded men younger than 50 years since colon cancer screening is not routinely recommended.

We considered respondents to have optimal chronic disease testing if they reported receiving all appropriate testing relevant for their particular condition. For persons with diabetes, this included having a hemoglobin A1c level checked within the past 6 months, a blood pressure check and foot and eye exams within the year, and cholesterol testing within 5 years. For those with heart disease or hypertension, it included having had blood pressure checked within the prior year and cholesterol testing within the prior 5 years. This approach is consistent with that of McBean et al, who have shown that a combination of appropriate tests is more predictive of glycemic control for diabetes.¹³

Because we were interested in different aspects of patient-initiated care seeking, we also evaluated delay in seeking care and adherence to physician recommendations as further measures of outcome. Specific questions were:

• “During the last 12 months, was there any time when you had a medical problem but put off, postponed, or did not seek medical care when you needed to?”
• “Has there been a time in the last two years when you didn’t follow the doctor’s advice or treatment plan, get a recommended test, or see a referred doctor?” (asked of respondents who had visited a doctor or clinic or had been admitted to the hospital in the last 2 years.)

Analysis

To test our first hypothesis—that persons of racial/ethnic minorities perceive negative experiences with the health care encounter more often than whites or English speakers—we examined

associations between demographic characteristics, utilization variables, and negative perceptions of the health care encounter using chisquared tests and multivariate logistic regression. In these analyses, we dichotomized education into high school graduate or less, and some technical school/college and more. We dichotomized the primary language spoken at home into non-English and English, and we used federal poverty level groupings (<100%, 100%–200% and >200%) to categorize household income. Almost 19% percent of respondents did not report incomes, so we created a dummy variable to account for those with unreported incomes.

We classified insurance status as none or any (either public or private); race/ethnicity as white, black, Hispanic, Asian, and other (Native American, mixed race, or other). We examined the effect of these variables alone and in concert. For example, we calculated predicted percentages to evaluate the combined effects of race and gender, as well as race and education, in relationship to our outcome variables.

Finally, we used multivariate logistic regression to test the relationship between negative perceptions of the patient-provider relationship and our utilization variables. In these analyses, perceptions were the covariate of interest; we controlled for patient characteristics that could also influence utilization, including education, income, insurance status, presence of a primary physician and existence of a comorbid condition (in this case, hypertension, diabetes, heart disease, asthma, and cancer.) This last variable was, by necessity, excluded from the analysis involving optimal chronic disease testing.

Results

Table 1 describes demographic characteristics and utilization measures for our sample. Consistent with prior literature, blacks and Hispanics had lower incomes and higher rates of non-insurance than both whites and Asians. Hispanics responded most frequently that English was not their primary language.

Hispanics and Asians were less likely than whites to have received optimal chronic disease care, while blacks and Hispanics were more likely than whites to have received optimal cancer screening. There were no differences between racial/ethnic groups in not following the doctor’s advice or in putting off care.

TABLE 1
Demographics/characteristics and health care utilization of study participants

Negative perceptions of the patient-provider relationship

Race. Over 14% of blacks, 19% of Hispanics, and 20% of Asians reported they had been treated with disrespect by their doctor. Members of these groups were also more likely than whites to report that they were treated unfairly because of their race or their language, and that they would have received better care had they belonged to a different race (Table 2).

Language. Persons for whom English was not the primary language were also more likely to say they had been treated with disrespect, and to report they would have received better care had they been of a different race. For each racial/ethnic group, bivariate relationships persisted after controlling for other respondent characteristics, including education and income (Table 2).

Sex. Men were significantly more likely than women to perceive being treated with disrespect by the doctor (15.9% vs 11.6%), and the percentage varied by race/ethnicity. Using our model to predict the combined effects of race and gender, we found that Asian and Hispanic men (24% and 23%, respectively) were more likely than black men (17%) or white men (11%) to perceive being treated with disrespect.

Education. Education was similarly associated with perceptions of disrespect. Almost 18% of persons without a college education believed they had been treated with disrespect, versus only 10% of those with a college education. Minorities with lower education were more likely to have this perception. Twenty-nine percent of Asians, 22% of Hispanics, and 19% of blacks without a college education reported being treated with disrespect or being looked down upon, versus 13% of whites.

TABLE 2
Relationship of demographic variables to measures of negative perceptions

Impact on care

Respondents who reported being treated with disrespect were significantly less likely to have had a physical exam within the prior year; those with diabetes, hypertension, or heart disease were less likely to have received optimal care. These respondents were also more likely to report not following the doctor’s advice and putting off needed care (Table 3). This relationship was not seen for optimal cancer screening.

Persons who believed they had been treated unfairly due to their race and who thought they would have received better care had they been of a different race were more likely to ignore the doctor’s advice and put off care when medically needed. Those who believed they would have received better care had they been of a different race were also less likely to receive optimal chronic disease care. In analyses not shown, we examined the independent effects of income and education, as well as interactions between these variables and insurance, and found the results basically unchanged.

TABLE 3
Relationship of negative perceptions to health care outcomes

Discussion

We hypothesized that patients who have negative perceptions of the patient-provider relationship would be less likely to seek needed care, and that reports of such feelings would be more prevalent among minority patients. As anticipated, large proportions of blacks, Hispanics, and Asians reported that they were treated with disrespect, were treated unfairly, or would have received better care if their race had been different. Male gender and lower educational attainment were also associated with perceived disrespect, particularly among minorities.

Negative experiences lead to suboptimal care

The finding of greater likelihood of perceived disrespect among minority groups, men, and those with lower levels of education is particularly important in light of the strong relationship between such reports and the quality of care that patients receive. Those who reported that they were treated unfairly because of race were less likely to get a routine physical exam, follow a doctor’s advice, or receive appropriate secondary preventive care for diabetes, heart disease, and hypertension. In other words, negative experiences within the health care environment may jeopardize care for medically needy patients. Receipt of suboptimal care, particularly in the context of chronic disease, is likely to be associated with worse health outcomes, and may contribute to disparities.

Cancer screening the exception

While the relationships between negative perceptions and receipt of care for chronic disease and receipt of a routine physical examination were strong, the correlation did not persist for cancer screening. Black and Hispanic respondents were more likely than whites to receive optimal cancer screening, a finding that has been reported elsewhere.^{14, 15}

We hypothesize that this is in part because a wide array of community programs make special outreach efforts allowing patients to “bypass” the traditional office environment.¹⁶ These settings may be more likely to use culturally sensitive approaches or may be so transient that negative perceptions based on race or income may be less likely to form.

However, based on our finding that care requiring follow-up (eg, diabetes management) is less likely to occur with individuals who report negative perceptions of the patient-provider relationship, we would hypothesize that individuals who receive initial cancer screening might be less likely to follow up on abnormal results once screened. It may be that in situations requiring long term relationships, such as chronic disease care, perceptions of discrimination and disrespect may take the greatest toll. This hypothesis is supported by previous literature consistently reporting excess mortality despite higher cancer screening rates among blacks.^17-20

Limitations

Our study has several limitations. Because we are relying on self-report, we could not assess which specific aspects of the patient-provider relationship may have influenced the reports of disrespect. Responses may have been affected by experiences completely outside of this relation ship, or outside of the health care system, that independently could have impacted health care utilization. We cannot disentangle how general life experience influences perceptions of the health care encounter or care-seeking; for example respondents who perceive racial bias in other environments such as the workplace may also be more likely to perceive it in the healthcare setting. In addition, self-reported utilization measures may not always be accurate, particularly regarding cancer screening.^21,22

Despite the deliberate oversampling of major racial/ethnic groups, we remain limited in our ability to examine important subgroups within them, whether related to ethnicity (eg, Cuban, Vietnamese) or chronic condition (eg, asthma, diabetes), even though some groups may differ dramatically from others.

We also had insufficient numbers of Native Americans to analyze separately.

We excluded respondents who did not have a regular doctor because they were unable to answer key questions about the health care encounter.

Finally, there is no agreement on the definition of age-appropriate breast and cervical cancer screening.^{23, 24} We conducted additional analyses varying the age criteria for testing, including starting the required age for screening at age 40 (for breast cancer) as well as setting the age cut off for required screening at 65 (for both breast and cervical cancer screening), and found that the results were essentially unchanged from those presented. Similarly, adding prostate cancer screening to our models for men over age 50 did not alter our results significantly.

Research should focus on improving perceptions of care

Although it is difficult to quantify or measure negative responses objectively, the strong relationship between patient perceptions of the encounter and utilization suggests an important area for further attention. These findings suggest there may still be a substantial core of individuals who will actively avoid care, perhaps based on previous negative interpersonal experiences in getting care. Interventions aimed at both doctors and potential patients will be required to address this. Research is needed to focus on what approaches can best improve perceptions of care within the patient-provider relationship and how such interventions can reduce racial disparities in health care.

Acknowledgments

This research was done in part by a grant from The Commonwealth Fund.

Corresponding author
Nicole Lurie, MD, MSPH, RAND Corporation, 1200 S. Hayes Street, Arlington, VA 22202. E-mail: [email protected]

Racial and ethnic disparities in health care have been catalogued across numerous diseases and care settings.¹ By clarifying the causes of these disparities, we can develop solutions. In a seminal study, Shulman found in patient simulations that identical presentations for heart disease received different recommendations for care based on the patient’s race and gender, thus pinpointing the patient-provider relationship as a potential source of disparities.² Other research suggests interactions with non-physician health care personnel might also be a source of negative experiences with care.³

Research is beginning to identify how the health care encounter might relate to disparities in use of services and quality of care. For example, race concordance between the physician and patient, at least for blacks, is associated with higher patient satisfaction and greater participatory decision-making. This in turn can impact compliance and possibly outcomes.^4-6 While black patients who have black physicians are more likely to report receipt of counseling about preventive care and cancer screening,⁷ race concordance does not appear to be independently associated with different patterns of utilization.⁸

Perceived discrimination has also been associated with lower levels of satisfaction with the health care system.⁹ In one survey, two thirds of respondents reported feeling discriminated against in their interactions with health care providers due to their race or socioeconomic status.¹⁰ How perceived discrimination influences quality and outcomes of care has not been fully explored.

We hypothesized that minority patients and those who do not speak English perceive negative experiences with the health care encounter more often than whites or English-speakers. We further hypothesized that patients who report such negative experiences are less likely to seek care initially or return for follow-up care. We tested these hypotheses using data from the Commonwealth Fund 2001 Health Care Quality Survey.

Methods

Sample

Respondents were from a nationally representative sample of 6722 adults, aged 18 years and older, living in the continental United States, and who speak English, Spanish, Mandarin, Cantonese, Vietnamese, or Korean.

The sampling frame was based on random-digit dialing; telephone exchanges with higher-than-average numbers of minority households were oversampled. In addition to the oversampling based on telephone exchanges, we interviewed members of 394 households identified from a nationwide demographic tracking survey as having an Asian/Asian American or African American family member. Interviews were conducted in English, Spanish, Mandarin, Cantonese, Vietnamese, or Korean, depending on the respondent preference. The response rate for the entire sample was 53.1%.

The final sample was weighted to correct for the disproportionate sample design and to ensure the sample was representative of all adults aged 18 years and older based on the March 2001 Current Population Survey (CPS). The final weighted sample is therefore reflective of the 193 million adults in the United States who have telephones. A more detailed description of the sampling and weighting methods can be found else-where.¹¹ Data were collected between April 30 and November 5, 2001.

We focus on the subset of questions addressing respondent’s experience with the health care encounter and their use of certain health care services. We used STATA Version 6.0¹² to conduct statistical analyses using the weighted sample.

Dependent variables

Negative perceptions of the patient-provider relationship. We identified factors that define the patient-provider relationship, and formulated questions based on these factors that would elicit meaningful responses. Specific questions included:

• “Did the doctor treat you with a great deal of respect and dignity, a fair amount, not too much, or none at all?” (4-point scale)
• Please tell me if you strongly agree, somewhat agree, somewhat disagree, or strongly disagree with the statement, “I often feel as if my doctor looks down on me and the way I live my life.” (4-point scale)
• “Thinking about all of the experiences you have had with health care visits in the last two years, have you ever felt that the doctor or medical staff you saw judged you unfairly or treated you with disrespect because of how well you speak English?” (yes/no)
• “Thinking about all of the experiences you have had with health care visits in the last two years, have you ever felt that the doctor or medical staff you saw judged you unfairly or treated you with disrespect because of your race or ethnic background?” (yes/no)
• “Do you think there was ever a time when you would have gotten better medical care if you had belonged to a different race or ethnic group?” (yes/no)

Those who said they did not have a doctor were unable to answer these questions and were excluded from our analysis. We combined the first 2 questions into a single dichotomous variable, characterized as “being treated with disrespect,” because both questions described negative perceptions of the health care encounter and because doing so preserved sample size for our analyses.

Utilization and optimal care

We examined self-reported use of specific services, including whether respondents had a physical exam within the past year. For cancer screening and chronic disease care, we created variables designed to represent optimal care. For example, optimal cancer screening was defined as being up to date on all tests for which the individual was eligible, based on age and gender.

These included:

• fecal occult blood testing for colon cancer screening within the prior year (both female and male respondents aged >50 years )
• cervical cancer screening within the prior 3 years (all women over the age of the 18)
• mammography within the prior year (women over the age of 50).

We excluded men younger than 50 years since colon cancer screening is not routinely recommended.

We considered respondents to have optimal chronic disease testing if they reported receiving all appropriate testing relevant for their particular condition. For persons with diabetes, this included having a hemoglobin A1c level checked within the past 6 months, a blood pressure check and foot and eye exams within the year, and cholesterol testing within 5 years. For those with heart disease or hypertension, it included having had blood pressure checked within the prior year and cholesterol testing within the prior 5 years. This approach is consistent with that of McBean et al, who have shown that a combination of appropriate tests is more predictive of glycemic control for diabetes.¹³

Because we were interested in different aspects of patient-initiated care seeking, we also evaluated delay in seeking care and adherence to physician recommendations as further measures of outcome. Specific questions were:

• “During the last 12 months, was there any time when you had a medical problem but put off, postponed, or did not seek medical care when you needed to?”
• “Has there been a time in the last two years when you didn’t follow the doctor’s advice or treatment plan, get a recommended test, or see a referred doctor?” (asked of respondents who had visited a doctor or clinic or had been admitted to the hospital in the last 2 years.)

Analysis

To test our first hypothesis—that persons of racial/ethnic minorities perceive negative experiences with the health care encounter more often than whites or English speakers—we examined

associations between demographic characteristics, utilization variables, and negative perceptions of the health care encounter using chisquared tests and multivariate logistic regression. In these analyses, we dichotomized education into high school graduate or less, and some technical school/college and more. We dichotomized the primary language spoken at home into non-English and English, and we used federal poverty level groupings (<100%, 100%–200% and >200%) to categorize household income. Almost 19% percent of respondents did not report incomes, so we created a dummy variable to account for those with unreported incomes.

We classified insurance status as none or any (either public or private); race/ethnicity as white, black, Hispanic, Asian, and other (Native American, mixed race, or other). We examined the effect of these variables alone and in concert. For example, we calculated predicted percentages to evaluate the combined effects of race and gender, as well as race and education, in relationship to our outcome variables.

Finally, we used multivariate logistic regression to test the relationship between negative perceptions of the patient-provider relationship and our utilization variables. In these analyses, perceptions were the covariate of interest; we controlled for patient characteristics that could also influence utilization, including education, income, insurance status, presence of a primary physician and existence of a comorbid condition (in this case, hypertension, diabetes, heart disease, asthma, and cancer.) This last variable was, by necessity, excluded from the analysis involving optimal chronic disease testing.

Results

Table 1 describes demographic characteristics and utilization measures for our sample. Consistent with prior literature, blacks and Hispanics had lower incomes and higher rates of non-insurance than both whites and Asians. Hispanics responded most frequently that English was not their primary language.

Hispanics and Asians were less likely than whites to have received optimal chronic disease care, while blacks and Hispanics were more likely than whites to have received optimal cancer screening. There were no differences between racial/ethnic groups in not following the doctor’s advice or in putting off care.

TABLE 1
Demographics/characteristics and health care utilization of study participants

Negative perceptions of the patient-provider relationship

Race. Over 14% of blacks, 19% of Hispanics, and 20% of Asians reported they had been treated with disrespect by their doctor. Members of these groups were also more likely than whites to report that they were treated unfairly because of their race or their language, and that they would have received better care had they belonged to a different race (Table 2).

Language. Persons for whom English was not the primary language were also more likely to say they had been treated with disrespect, and to report they would have received better care had they been of a different race. For each racial/ethnic group, bivariate relationships persisted after controlling for other respondent characteristics, including education and income (Table 2).

Sex. Men were significantly more likely than women to perceive being treated with disrespect by the doctor (15.9% vs 11.6%), and the percentage varied by race/ethnicity. Using our model to predict the combined effects of race and gender, we found that Asian and Hispanic men (24% and 23%, respectively) were more likely than black men (17%) or white men (11%) to perceive being treated with disrespect.

Education. Education was similarly associated with perceptions of disrespect. Almost 18% of persons without a college education believed they had been treated with disrespect, versus only 10% of those with a college education. Minorities with lower education were more likely to have this perception. Twenty-nine percent of Asians, 22% of Hispanics, and 19% of blacks without a college education reported being treated with disrespect or being looked down upon, versus 13% of whites.

TABLE 2
Relationship of demographic variables to measures of negative perceptions

Impact on care

Respondents who reported being treated with disrespect were significantly less likely to have had a physical exam within the prior year; those with diabetes, hypertension, or heart disease were less likely to have received optimal care. These respondents were also more likely to report not following the doctor’s advice and putting off needed care (Table 3). This relationship was not seen for optimal cancer screening.

Persons who believed they had been treated unfairly due to their race and who thought they would have received better care had they been of a different race were more likely to ignore the doctor’s advice and put off care when medically needed. Those who believed they would have received better care had they been of a different race were also less likely to receive optimal chronic disease care. In analyses not shown, we examined the independent effects of income and education, as well as interactions between these variables and insurance, and found the results basically unchanged.

TABLE 3
Relationship of negative perceptions to health care outcomes

Discussion

We hypothesized that patients who have negative perceptions of the patient-provider relationship would be less likely to seek needed care, and that reports of such feelings would be more prevalent among minority patients. As anticipated, large proportions of blacks, Hispanics, and Asians reported that they were treated with disrespect, were treated unfairly, or would have received better care if their race had been different. Male gender and lower educational attainment were also associated with perceived disrespect, particularly among minorities.

Negative experiences lead to suboptimal care

The finding of greater likelihood of perceived disrespect among minority groups, men, and those with lower levels of education is particularly important in light of the strong relationship between such reports and the quality of care that patients receive. Those who reported that they were treated unfairly because of race were less likely to get a routine physical exam, follow a doctor’s advice, or receive appropriate secondary preventive care for diabetes, heart disease, and hypertension. In other words, negative experiences within the health care environment may jeopardize care for medically needy patients. Receipt of suboptimal care, particularly in the context of chronic disease, is likely to be associated with worse health outcomes, and may contribute to disparities.

Cancer screening the exception

While the relationships between negative perceptions and receipt of care for chronic disease and receipt of a routine physical examination were strong, the correlation did not persist for cancer screening. Black and Hispanic respondents were more likely than whites to receive optimal cancer screening, a finding that has been reported elsewhere.^{14, 15}

We hypothesize that this is in part because a wide array of community programs make special outreach efforts allowing patients to “bypass” the traditional office environment.¹⁶ These settings may be more likely to use culturally sensitive approaches or may be so transient that negative perceptions based on race or income may be less likely to form.

However, based on our finding that care requiring follow-up (eg, diabetes management) is less likely to occur with individuals who report negative perceptions of the patient-provider relationship, we would hypothesize that individuals who receive initial cancer screening might be less likely to follow up on abnormal results once screened. It may be that in situations requiring long term relationships, such as chronic disease care, perceptions of discrimination and disrespect may take the greatest toll. This hypothesis is supported by previous literature consistently reporting excess mortality despite higher cancer screening rates among blacks.^17-20

Limitations

Our study has several limitations. Because we are relying on self-report, we could not assess which specific aspects of the patient-provider relationship may have influenced the reports of disrespect. Responses may have been affected by experiences completely outside of this relation ship, or outside of the health care system, that independently could have impacted health care utilization. We cannot disentangle how general life experience influences perceptions of the health care encounter or care-seeking; for example respondents who perceive racial bias in other environments such as the workplace may also be more likely to perceive it in the healthcare setting. In addition, self-reported utilization measures may not always be accurate, particularly regarding cancer screening.^21,22

Despite the deliberate oversampling of major racial/ethnic groups, we remain limited in our ability to examine important subgroups within them, whether related to ethnicity (eg, Cuban, Vietnamese) or chronic condition (eg, asthma, diabetes), even though some groups may differ dramatically from others.

We also had insufficient numbers of Native Americans to analyze separately.

We excluded respondents who did not have a regular doctor because they were unable to answer key questions about the health care encounter.

Finally, there is no agreement on the definition of age-appropriate breast and cervical cancer screening.^{23, 24} We conducted additional analyses varying the age criteria for testing, including starting the required age for screening at age 40 (for breast cancer) as well as setting the age cut off for required screening at 65 (for both breast and cervical cancer screening), and found that the results were essentially unchanged from those presented. Similarly, adding prostate cancer screening to our models for men over age 50 did not alter our results significantly.

Research should focus on improving perceptions of care

Although it is difficult to quantify or measure negative responses objectively, the strong relationship between patient perceptions of the encounter and utilization suggests an important area for further attention. These findings suggest there may still be a substantial core of individuals who will actively avoid care, perhaps based on previous negative interpersonal experiences in getting care. Interventions aimed at both doctors and potential patients will be required to address this. Research is needed to focus on what approaches can best improve perceptions of care within the patient-provider relationship and how such interventions can reduce racial disparities in health care.

Acknowledgments

This research was done in part by a grant from The Commonwealth Fund.

Corresponding author
Nicole Lurie, MD, MSPH, RAND Corporation, 1200 S. Hayes Street, Arlington, VA 22202. E-mail: [email protected]

Survival curves illustrate prognosis. The percentage of patients reaching an endpoint (eg, death, recurrence of disease, or cure) is plotted on the y(vertical) axis against time on the x(horizontal) axis.

Plotting a survival curve

Two common plotting methods are used. With the actuarial method, the x axis is divided into regular intervals (eg, by month) and percent survival is calculated for each interval. With the Kaplan-Meier method, percent survival is recalculated each time a patient dies (or reaches a different endpoint). Consider the example here (Figure).¹

Time zerois when each patient entered the trial. Survival is the percentage of patients still alive thereafter. Median survivalis found by extending a horizontal line from the 50% survival point until it intersects the curve (24 months in this case).

FIGURE
Sample survival curve

Survival curves have limitations. Consider a study that enrolls patients between 1996 and 2002 and ends in 2005. All that is known about a patient enrolled in 2002 who survived until 2005 is that he or she survived 3 years. Some patients also drop out of the study early or are lost to follow-up. Some patients die from causes other than the one under study.

Censoringis the process of excluding data from survival curves when information about survival is unknown. For a patient who drops out early, for example, only data obtained when the patient was followed would be included. The result is a more accurate picture of survival for the patients under study.

Correspondence
Stephen A. Wilson, MD, UPMC Family Praxctice Residency, 200 Lothrop St., Pittsburgh, PA 15213-2582. E-mail: [email protected].

Survival curves illustrate prognosis. The percentage of patients reaching an endpoint (eg, death, recurrence of disease, or cure) is plotted on the y(vertical) axis against time on the x(horizontal) axis.

Plotting a survival curve

Two common plotting methods are used. With the actuarial method, the x axis is divided into regular intervals (eg, by month) and percent survival is calculated for each interval. With the Kaplan-Meier method, percent survival is recalculated each time a patient dies (or reaches a different endpoint). Consider the example here (Figure).¹

Time zerois when each patient entered the trial. Survival is the percentage of patients still alive thereafter. Median survivalis found by extending a horizontal line from the 50% survival point until it intersects the curve (24 months in this case).

FIGURE
Sample survival curve

Survival curves have limitations. Consider a study that enrolls patients between 1996 and 2002 and ends in 2005. All that is known about a patient enrolled in 2002 who survived until 2005 is that he or she survived 3 years. Some patients also drop out of the study early or are lost to follow-up. Some patients die from causes other than the one under study.

Censoringis the process of excluding data from survival curves when information about survival is unknown. For a patient who drops out early, for example, only data obtained when the patient was followed would be included. The result is a more accurate picture of survival for the patients under study.

Correspondence
Stephen A. Wilson, MD, UPMC Family Praxctice Residency, 200 Lothrop St., Pittsburgh, PA 15213-2582. E-mail: [email protected].

Survival curves illustrate prognosis. The percentage of patients reaching an endpoint (eg, death, recurrence of disease, or cure) is plotted on the y(vertical) axis against time on the x(horizontal) axis.

Plotting a survival curve

Two common plotting methods are used. With the actuarial method, the x axis is divided into regular intervals (eg, by month) and percent survival is calculated for each interval. With the Kaplan-Meier method, percent survival is recalculated each time a patient dies (or reaches a different endpoint). Consider the example here (Figure).¹

Time zerois when each patient entered the trial. Survival is the percentage of patients still alive thereafter. Median survivalis found by extending a horizontal line from the 50% survival point until it intersects the curve (24 months in this case).

FIGURE
Sample survival curve

Survival curves have limitations. Consider a study that enrolls patients between 1996 and 2002 and ends in 2005. All that is known about a patient enrolled in 2002 who survived until 2005 is that he or she survived 3 years. Some patients also drop out of the study early or are lost to follow-up. Some patients die from causes other than the one under study.

Censoringis the process of excluding data from survival curves when information about survival is unknown. For a patient who drops out early, for example, only data obtained when the patient was followed would be included. The result is a more accurate picture of survival for the patients under study.

Correspondence
Stephen A. Wilson, MD, UPMC Family Praxctice Residency, 200 Lothrop St., Pittsburgh, PA 15213-2582. E-mail: [email protected].

Randomization is a crucial part of most clinical trials. The purpose of randomization in a trial comparing 2 groups is to ensure that the groups differ only with respect to the interventions being compared. Randomization determines not only which treatment subjects receive (eg, drug vs placebo), but also how the results are analyzed at the end of the trial.

Intention to treat prevents biased outcomes

The intention-to-treat principle states that all subjects must be analyzed with respect to the group to which they were randomized. Consider a recent study by Spector et al¹ in which patients with dementia were randomized to receive an intervention known as cognitive stimulation therapy (CST) or control (equivalent to placebo in a drug trial). The investigators used an intention-to-treat analysis. The intervention was a complex 14-session training program. Many patients did not complete all the sessions. Whether they had completed all the sessions or not, patients initially randomized to the intervention were still considered to have received the intervention when the results were analyzed.

At first, intention to treat doesn’t seem logical. If we are testing an intervention, doesn’t it make sense to evaluate its effect among patients who complied with it fully, and then compare them with patients who were not assigned to the intervention or failed to comply? The problem is, patients who fail to comply with an intervention for whatever reason (not attending all training sessions in the example above) may differ in an important way from those who do.

In the Spector study, for example, it is possible that patients who attended very few of the sessions were more likely to have some subtle cognitive deficits that limited their participation. If placed in the control group, the 2 groups would these patients were excluded from the analysis or differ not only with respect to the intervention, but also with respect to these subtle factors. The value of randomization, therefore, would be compromised. The treatment arm may even appear to be more effective than it really is.

Correspondence
Kiame J. Mahaniah, MD, Greater Lawrence Family Health Center, 34 Haverhill Street, Lawrence, MA 01841. E-mail: [email protected] .

Randomization is a crucial part of most clinical trials. The purpose of randomization in a trial comparing 2 groups is to ensure that the groups differ only with respect to the interventions being compared. Randomization determines not only which treatment subjects receive (eg, drug vs placebo), but also how the results are analyzed at the end of the trial.

Intention to treat prevents biased outcomes

The intention-to-treat principle states that all subjects must be analyzed with respect to the group to which they were randomized. Consider a recent study by Spector et al¹ in which patients with dementia were randomized to receive an intervention known as cognitive stimulation therapy (CST) or control (equivalent to placebo in a drug trial). The investigators used an intention-to-treat analysis. The intervention was a complex 14-session training program. Many patients did not complete all the sessions. Whether they had completed all the sessions or not, patients initially randomized to the intervention were still considered to have received the intervention when the results were analyzed.

At first, intention to treat doesn’t seem logical. If we are testing an intervention, doesn’t it make sense to evaluate its effect among patients who complied with it fully, and then compare them with patients who were not assigned to the intervention or failed to comply? The problem is, patients who fail to comply with an intervention for whatever reason (not attending all training sessions in the example above) may differ in an important way from those who do.

In the Spector study, for example, it is possible that patients who attended very few of the sessions were more likely to have some subtle cognitive deficits that limited their participation. If placed in the control group, the 2 groups would these patients were excluded from the analysis or differ not only with respect to the intervention, but also with respect to these subtle factors. The value of randomization, therefore, would be compromised. The treatment arm may even appear to be more effective than it really is.

Correspondence
Kiame J. Mahaniah, MD, Greater Lawrence Family Health Center, 34 Haverhill Street, Lawrence, MA 01841. E-mail: [email protected] .

Randomization is a crucial part of most clinical trials. The purpose of randomization in a trial comparing 2 groups is to ensure that the groups differ only with respect to the interventions being compared. Randomization determines not only which treatment subjects receive (eg, drug vs placebo), but also how the results are analyzed at the end of the trial.

Intention to treat prevents biased outcomes

The intention-to-treat principle states that all subjects must be analyzed with respect to the group to which they were randomized. Consider a recent study by Spector et al¹ in which patients with dementia were randomized to receive an intervention known as cognitive stimulation therapy (CST) or control (equivalent to placebo in a drug trial). The investigators used an intention-to-treat analysis. The intervention was a complex 14-session training program. Many patients did not complete all the sessions. Whether they had completed all the sessions or not, patients initially randomized to the intervention were still considered to have received the intervention when the results were analyzed.

At first, intention to treat doesn’t seem logical. If we are testing an intervention, doesn’t it make sense to evaluate its effect among patients who complied with it fully, and then compare them with patients who were not assigned to the intervention or failed to comply? The problem is, patients who fail to comply with an intervention for whatever reason (not attending all training sessions in the example above) may differ in an important way from those who do.

In the Spector study, for example, it is possible that patients who attended very few of the sessions were more likely to have some subtle cognitive deficits that limited their participation. If placed in the control group, the 2 groups would these patients were excluded from the analysis or differ not only with respect to the intervention, but also with respect to these subtle factors. The value of randomization, therefore, would be compromised. The treatment arm may even appear to be more effective than it really is.

Correspondence
Kiame J. Mahaniah, MD, Greater Lawrence Family Health Center, 34 Haverhill Street, Lawrence, MA 01841. E-mail: [email protected] .