The Impact of Regular Vaginal pH Screening on the Diagnosis of Bacterial Vaginosis in Pregnancy

Preterm delivery is one of the most important—and most costly—problems in obstetric medicine. Despite recent advances, nearly 10% of all infants in this country are born preterm.¹ Advances in the care of premature newborns, such as the administration of antenatal corticosteroids, the use of surfactant, and other developments in respiratory technology have improved outcomes, yet a greater emphasis on identifying and managing the causes of premature delivery is needed.

Recent research has indicated that maternal infection probably plays a much greater role in the occurrence of preterm labor than was previously understood. Bacterial byproducts have been shown to stimulate endogenous phospholipase A2, which induces the formation of prostaglandin, an important stimulant of uterine contractions.² Further, neutrophils and their constituent enzymes, together with bacteria and their proteases, weaken the amniochorion, predisposing a woman to premature rupture of membranes.³ A study⁴ comparing placentas from 21 women who delivered before 36 weeks’ gestation with those from 66 women who delivered at term found the rate of infected placentas to be 62% and 12%, respectively.

Specific genital infections have been linked to preterm labor, including sexually transmitted infections secondary to Chlamydia trachomatis, Neisseria gonorrhoeae, and Trichomonas vaginalis.⁵ Bacterial vaginosis, which is more common and not necessarily sexually transmitted, is associated with a 1.4- to 6.9-fold increased risk of preterm delivery.^6-9 The relatively high prevalence of bacterial vaginosis in the obstetric population (10% to 32%),¹⁰ coupled with evidence that screening and treating the condition reduces the rate of preterm deliveries,^11-13 makes detection important for reducing perinatal morbidity and mortality.

There is scant evidence that can be used to determine the optimal screening regimen for bacterial vaginosis in pregnancy. It is unclear, for example, whether all women should be routinely screened, how often the screening should occur, and which tests should be used. The vaginal pH test may be a valuable screening tool. It is a quick, inexpensive test where values >4.5 indicate the presence of either bacterial vaginosis or trichomoniasis. The sensitivity for detecting bacterial vaginosis with pH testing ranges from 84% to 97%.¹⁴ An elevated vaginal pH level has been shown to be 100% sensitive and 92% specific in screening nonpregnant premenopausal women for aerobic bacterial pathogens (b-hemolytic streptococci, Gardnerella vaginalis, or mixed aerobic organisms).¹⁵ However, little is known about the utility of the vaginal pH level as a screening tool in pregnancy.

The purpose of our study was to determine whether regular prenatal vaginal pH level screening resulted in more frequent diagnoses and treatment of bacterial vaginosis, trichomoniasis, and other genital infections and thus fewer preterm deliveries. In addition, we sought to determine the effectiveness of using recent vaginal symptoms and vaginal pH levels to diagnose bacterial vaginosis and trichomoniasis.

Methods

We enrolled pregnant women of less than 34 weeks’ gestation attending 1 of 3 participating clinics located in the Minneapolis/St. Paul, Minnesota, metropolitan area between July 1, 1996, and December 31, 1997. We calculated that a sample size of 208 women would detect a 50% increase in the diagnosis of bacterial vaginosis with a power of .80 and an a of .05. Women who participated in the study were randomized into experimental and control groups at the initial prenatal visit, using computer-generated random number tables and sealed envelopes containing the patient’s group assignment and study protocol. A study protocol was placed in the patient’s chart to facilitate group-specific clinical procedures and data collection Table 1; therefore, health care providers and subjects were not blinded to the intervention (vaginal pH testing).

During the first prenatal visit, patients from both groups were managed the same: symptoms of preterm labor (regular uterine contractions) and vaginal infection (discharge, burning, itching, and odor) were investigated and recorded as present or absent. The following laboratory tests were performed: vaginal pH, wet mount, dieoxyribonucleic acid (DNA) probe for C trachomatis and N gonorrhoeae, and urinalysis.

At each subsequent prenatal visit, women from both groups were asked about symptoms of preterm labor, vaginal infection, and urinary tract infection. Women in either group who reported symptoms of preterm labor had a wet mount examination, DNA probe, and urinalysis performed; women with vaginal symptoms had a wet mount and DNA probe performed. Experimental and control group subjects differed in that experimental group members had a vaginal pH level tested by their physician at each prenatal visit (controls did not have pH testing after the initial visit), and women with a pH level >4.5 also had a wet mount examination performed. The vaginal pH level was tested using a strip of specially formulated pH paper with an enhanced range in the spectrum of interest (pH = 4.0 to 5.5; paper manufactured by Micro Essential Laboratory, Brooklyn, NY). Participating study physicians were instructed in the technique for determining vaginal pH levels. The patient’s physician recorded the pH value in the medical record.

For the purposes of our study, a wet mount evaluation by a medical technologist was used to diagnose bacterial vaginosis, trichomoniasis, and vulvovaginal candidiasis. Participating medical technologists were formally trained in laboratory procedures and were required to have proficiency testing on wet mount examinations 3 times a year. The medical technologists were blinded to participants’ group assignments. Our selection of the wet mount examination as the confirmatory diagnostic test for vaginal infections was made on the basis of its applicability to each of the 3 types of infection, its utility for private practice (relatively easy to perform and inexpensive), and on previous data reporting its relatively high sensitivity for the diagnosis of G vaginalis, the most prevalent bacteria in bacterial vaginosis. The sensitivity of the wet mount for detecting G vaginalis and T vaginalis is 90% and 100%, respectively, when synthetic oligonucleotide probes are used as a gold standard.¹⁶ Other researchers have also used medical technologists’ readings of wet mounts as a standard for the diagnosis of bacterial vaginosis.¹⁷

Women from both experimental and control groups who were diagnosed with a specific vaginal or cervical infection were given standard antimicrobial treatments specified in the study protocol. Although most women diagnosed with bacterial vaginosis or trichomoniasis received oral antibiotics (metronidazole 500 mg twice daily for 7 days or clindamycin 300 mg twice daily for 7 days), a minority were treated with vaginal antimicrobials.

Hospital charts were also reviewed for gestational age at the time of delivery. Gestational age was determined using clinical parameters with ultrasound confirmation for cases where clinical dates were uncertain. Women who delivered at 20 weeks’ to less than 37 weeks’ gestation were categorized as having had a preterm delivery, consistent with accepted definitions.¹⁸

We used the chi-square test to look for significant differences between the 2 groups in infection, antibiotic treatment, and preterm delivery rates. The sensitivity and specificity of vaginal pH values and symptom reports for detecting bacterial vaginosis and T vaginalis were determined using data from the initial prenatal visits, where both vaginal pH and wet mount examinations were routinely performed.

Results

We invited a total of 243 women to participate in our study. Not all eligible patients who were pregnant at the time of the study were asked to participate because of staff changes and work demands. One hundred one declined to participate because of a reluctance to undergo routine vaginal examinations. Twenty-one additional women initially expressed interest in the study, but subsequently transferred care or had a spontaneous or induced abortion before their first formal prenatal visit. The 121 women randomized into our study (63 women in the experimental group; 58 in the control group) represented an overall response rate of 50% (number of participants divided by number of women invited to participate). Two women were pregnant twice during the course of the study, resulting in 123 total pregnancies. There were 910 prenatal visits during the course of the study, with an average of 7.4 prenatal visits per pregnancy. When our study population was compared with nonparticipating obstetrical patients at the 3 clinic sites, no differences were seen in age and proportion of women insured by public assistance. However, compared with nonparticipants, the study sample included a higher proportion of whites.

The mean age of the participants was 24.8 years (standard deviation = 5.4), and 60.2% of the women were on public assistance. The population included several ethnic groups: 47.2% were African American; 36.6% white; 4.1% Native American; 3.3% Asian; 1.6% Hispanic; and 7.2% unknown or other. There were no differences between experimental and control groups in age, ethnicity, and proportion of women on public assistance.

Women who underwent regular vaginal pH screening in addition to standard prenatal care showed significantly higher rates for the diagnosis of bacterial vaginosis, more frequent diagnosis of trichomoniasis (not statistically significant), higher antibiotic treatment rates for bacterial vaginosis and trichomoniasis (46.9% vs 27.1%, P = .024), and a lower rate of preterm births (not statistically significant; Table 2.

The sensitivity of a high vaginal pH level or vaginal symptoms in identifying women with bacterial vaginosis or trichomoniasis was 84.4% Table 3. Although the specificity of either of these tests for identifying bacterial vaginosis or trichomoniasis was modest (57.1%), it rose to 78.5% when both a high pH value and vaginal symptoms were present. Of the 27 women were given a diagnosis of and treatment for bacterial vaginosis and had follow-up wet mount evaluations,¹⁵ (55.6%) had recurrent or persistent bacterial vaginosis.

Discussion

In our study, regular testing for vaginal pH levels at prenatal examinations resulted in significantly more diagnoses of bacterial vaginosis and more antibiotic prescriptions. Increased recognition and treatment of bacterial vaginosis may have contributed to the reduced number of preterm deliveries seen in the experimental group; however, we did not calculate our sample size using preterm delivery as a primary outcome, and the number of women we recruited fell short of the number required to detect a significant difference in the preterm delivery rate.

Given the randomized design of this study, the 1.8-fold increase in the rate of diagnosis of bacterial vaginosis among experimental subjects compared with controls (48.4% vs 27.1%) was likely a result of the increased screening with vaginal pH testing. It is noteworthy that the prevalence of bacterial vaginosis in the control group (27%) is near the upper end of the 10% to 32% range reported for other previously studied obstetric populations.¹⁰ This high-normal rate of infection may be related to the relatively high proportion of women of lower socioeconomic status, a group known to be at risk for preterm labor.¹⁹ In addition, our procedure of asking both experimental and control subjects about symptoms of vaginal infection as a routine part of each prenatal visit may have increased the rate of detecting infections for both groups. Given the high-risk nature of our population, it is notable that our controls had a preterm birth rate (10.2%) comparable with that of the general population,¹ and the experimental group’s preterm birth rate (4.7%) was less than half that of the controls.

A high vaginal pH value, vaginal symptoms, or both had 84% sensitivity to detect bacterial vaginosis or trichomoniasis. Inquiring about a history of vaginal symptoms and testing for vaginal pH levels are inexpensive procedures that could easily be implemented as part of routine prenatal care. If larger studies confirm its impact on the detection of infection and the preterm delivery rate, vaginal pH testing will merit wide use as a screening tool, with its results confirmed by more specific diagnostic procedures, such as the wet mount or Gram stain.

Nearly half of our sample received antimicrobial prescriptions for vaginal infections. Yet 56% of those treated who also had follow-up wet mount examinations had either persistent or recurrent bacterial vaginosis. This rate is 3 times higher than the 18% recurrence rate previously reported²⁰ and may indicate either noncompliance, resistance to the medication, or a tendency toward recurrence. Women who continue to have evidence of infection may remain at risk for preterm birth. A more effective strategy for eliminating bacterial vaginosis resulting in a lower rate of persistence or recurrence in the experimental group may have improved the preterm labor outcomes for women undergoing regular screening.

Limitations

An important limitation of this study is the relatively small sample size. Although we demonstrated a significant difference in our primary outcome (infection rates), the more compelling clinical outcome is preterm delivery rate. McGregor and colleagues¹¹ demonstrated a 50% reduction in bacterial-vaginosis–linked idiopathic preterm births with a program of screening for and treating bacterial vaginosis.¹¹ To have 80% power to show a reduction in preterm birth from 10% to 5%, we would have needed a sample size of 336 mothers in each group—a number beyond our capacity to fund and recruit.

Our difficulty in recruiting women resulted from our failure to invite all eligible women to participate and from the refusals to participate because we required frequent vaginal examinations. The observation that so many women disliked the idea of regular prenatal vaginal examinations suggests that we should investigate ways of making the vaginal pH test more acceptable, perhaps through self-testing. It is also possible that other populations (eg, women with previous preterm births or women of higher socioeconomic status) would be more open to regular pH testing.

Additional limitations of our study include potential selection bias, which may reduce the generalizability of our findings, and the fact that patients and providers were not blinded to the intervention. It would be helpful to minimize these potential sources of bias in a large-scale blinded randomized trial. Future studies should also differentiate the causes of positive results in follow-up tests and should examine the benefit of treating asymptomatic infections and doing follow-up tests for cure.

Conclusions

Our frequent screening procedures significantly increased the recognition of bacterial vaginosis and the treatment of infected women. Since vaginal symptoms and elevated pH levels appear to be useful in screening for bacterial vaginosis and trichomoniasis, frequent pH testing should be evaluated in larger studies.

Acknowledgments

This study was supported by a grant from the UCare Foundation.

Preterm delivery is one of the most important—and most costly—problems in obstetric medicine. Despite recent advances, nearly 10% of all infants in this country are born preterm.¹ Advances in the care of premature newborns, such as the administration of antenatal corticosteroids, the use of surfactant, and other developments in respiratory technology have improved outcomes, yet a greater emphasis on identifying and managing the causes of premature delivery is needed.

Recent research has indicated that maternal infection probably plays a much greater role in the occurrence of preterm labor than was previously understood. Bacterial byproducts have been shown to stimulate endogenous phospholipase A2, which induces the formation of prostaglandin, an important stimulant of uterine contractions.² Further, neutrophils and their constituent enzymes, together with bacteria and their proteases, weaken the amniochorion, predisposing a woman to premature rupture of membranes.³ A study⁴ comparing placentas from 21 women who delivered before 36 weeks’ gestation with those from 66 women who delivered at term found the rate of infected placentas to be 62% and 12%, respectively.

Specific genital infections have been linked to preterm labor, including sexually transmitted infections secondary to Chlamydia trachomatis, Neisseria gonorrhoeae, and Trichomonas vaginalis.⁵ Bacterial vaginosis, which is more common and not necessarily sexually transmitted, is associated with a 1.4- to 6.9-fold increased risk of preterm delivery.^6-9 The relatively high prevalence of bacterial vaginosis in the obstetric population (10% to 32%),¹⁰ coupled with evidence that screening and treating the condition reduces the rate of preterm deliveries,^11-13 makes detection important for reducing perinatal morbidity and mortality.

There is scant evidence that can be used to determine the optimal screening regimen for bacterial vaginosis in pregnancy. It is unclear, for example, whether all women should be routinely screened, how often the screening should occur, and which tests should be used. The vaginal pH test may be a valuable screening tool. It is a quick, inexpensive test where values >4.5 indicate the presence of either bacterial vaginosis or trichomoniasis. The sensitivity for detecting bacterial vaginosis with pH testing ranges from 84% to 97%.¹⁴ An elevated vaginal pH level has been shown to be 100% sensitive and 92% specific in screening nonpregnant premenopausal women for aerobic bacterial pathogens (b-hemolytic streptococci, Gardnerella vaginalis, or mixed aerobic organisms).¹⁵ However, little is known about the utility of the vaginal pH level as a screening tool in pregnancy.

The purpose of our study was to determine whether regular prenatal vaginal pH level screening resulted in more frequent diagnoses and treatment of bacterial vaginosis, trichomoniasis, and other genital infections and thus fewer preterm deliveries. In addition, we sought to determine the effectiveness of using recent vaginal symptoms and vaginal pH levels to diagnose bacterial vaginosis and trichomoniasis.

Methods

We enrolled pregnant women of less than 34 weeks’ gestation attending 1 of 3 participating clinics located in the Minneapolis/St. Paul, Minnesota, metropolitan area between July 1, 1996, and December 31, 1997. We calculated that a sample size of 208 women would detect a 50% increase in the diagnosis of bacterial vaginosis with a power of .80 and an a of .05. Women who participated in the study were randomized into experimental and control groups at the initial prenatal visit, using computer-generated random number tables and sealed envelopes containing the patient’s group assignment and study protocol. A study protocol was placed in the patient’s chart to facilitate group-specific clinical procedures and data collection Table 1; therefore, health care providers and subjects were not blinded to the intervention (vaginal pH testing).

During the first prenatal visit, patients from both groups were managed the same: symptoms of preterm labor (regular uterine contractions) and vaginal infection (discharge, burning, itching, and odor) were investigated and recorded as present or absent. The following laboratory tests were performed: vaginal pH, wet mount, dieoxyribonucleic acid (DNA) probe for C trachomatis and N gonorrhoeae, and urinalysis.

At each subsequent prenatal visit, women from both groups were asked about symptoms of preterm labor, vaginal infection, and urinary tract infection. Women in either group who reported symptoms of preterm labor had a wet mount examination, DNA probe, and urinalysis performed; women with vaginal symptoms had a wet mount and DNA probe performed. Experimental and control group subjects differed in that experimental group members had a vaginal pH level tested by their physician at each prenatal visit (controls did not have pH testing after the initial visit), and women with a pH level >4.5 also had a wet mount examination performed. The vaginal pH level was tested using a strip of specially formulated pH paper with an enhanced range in the spectrum of interest (pH = 4.0 to 5.5; paper manufactured by Micro Essential Laboratory, Brooklyn, NY). Participating study physicians were instructed in the technique for determining vaginal pH levels. The patient’s physician recorded the pH value in the medical record.

For the purposes of our study, a wet mount evaluation by a medical technologist was used to diagnose bacterial vaginosis, trichomoniasis, and vulvovaginal candidiasis. Participating medical technologists were formally trained in laboratory procedures and were required to have proficiency testing on wet mount examinations 3 times a year. The medical technologists were blinded to participants’ group assignments. Our selection of the wet mount examination as the confirmatory diagnostic test for vaginal infections was made on the basis of its applicability to each of the 3 types of infection, its utility for private practice (relatively easy to perform and inexpensive), and on previous data reporting its relatively high sensitivity for the diagnosis of G vaginalis, the most prevalent bacteria in bacterial vaginosis. The sensitivity of the wet mount for detecting G vaginalis and T vaginalis is 90% and 100%, respectively, when synthetic oligonucleotide probes are used as a gold standard.¹⁶ Other researchers have also used medical technologists’ readings of wet mounts as a standard for the diagnosis of bacterial vaginosis.¹⁷

Women from both experimental and control groups who were diagnosed with a specific vaginal or cervical infection were given standard antimicrobial treatments specified in the study protocol. Although most women diagnosed with bacterial vaginosis or trichomoniasis received oral antibiotics (metronidazole 500 mg twice daily for 7 days or clindamycin 300 mg twice daily for 7 days), a minority were treated with vaginal antimicrobials.

Hospital charts were also reviewed for gestational age at the time of delivery. Gestational age was determined using clinical parameters with ultrasound confirmation for cases where clinical dates were uncertain. Women who delivered at 20 weeks’ to less than 37 weeks’ gestation were categorized as having had a preterm delivery, consistent with accepted definitions.¹⁸

We used the chi-square test to look for significant differences between the 2 groups in infection, antibiotic treatment, and preterm delivery rates. The sensitivity and specificity of vaginal pH values and symptom reports for detecting bacterial vaginosis and T vaginalis were determined using data from the initial prenatal visits, where both vaginal pH and wet mount examinations were routinely performed.

Results

We invited a total of 243 women to participate in our study. Not all eligible patients who were pregnant at the time of the study were asked to participate because of staff changes and work demands. One hundred one declined to participate because of a reluctance to undergo routine vaginal examinations. Twenty-one additional women initially expressed interest in the study, but subsequently transferred care or had a spontaneous or induced abortion before their first formal prenatal visit. The 121 women randomized into our study (63 women in the experimental group; 58 in the control group) represented an overall response rate of 50% (number of participants divided by number of women invited to participate). Two women were pregnant twice during the course of the study, resulting in 123 total pregnancies. There were 910 prenatal visits during the course of the study, with an average of 7.4 prenatal visits per pregnancy. When our study population was compared with nonparticipating obstetrical patients at the 3 clinic sites, no differences were seen in age and proportion of women insured by public assistance. However, compared with nonparticipants, the study sample included a higher proportion of whites.

The mean age of the participants was 24.8 years (standard deviation = 5.4), and 60.2% of the women were on public assistance. The population included several ethnic groups: 47.2% were African American; 36.6% white; 4.1% Native American; 3.3% Asian; 1.6% Hispanic; and 7.2% unknown or other. There were no differences between experimental and control groups in age, ethnicity, and proportion of women on public assistance.

Women who underwent regular vaginal pH screening in addition to standard prenatal care showed significantly higher rates for the diagnosis of bacterial vaginosis, more frequent diagnosis of trichomoniasis (not statistically significant), higher antibiotic treatment rates for bacterial vaginosis and trichomoniasis (46.9% vs 27.1%, P = .024), and a lower rate of preterm births (not statistically significant; Table 2.

The sensitivity of a high vaginal pH level or vaginal symptoms in identifying women with bacterial vaginosis or trichomoniasis was 84.4% Table 3. Although the specificity of either of these tests for identifying bacterial vaginosis or trichomoniasis was modest (57.1%), it rose to 78.5% when both a high pH value and vaginal symptoms were present. Of the 27 women were given a diagnosis of and treatment for bacterial vaginosis and had follow-up wet mount evaluations,¹⁵ (55.6%) had recurrent or persistent bacterial vaginosis.

Discussion

In our study, regular testing for vaginal pH levels at prenatal examinations resulted in significantly more diagnoses of bacterial vaginosis and more antibiotic prescriptions. Increased recognition and treatment of bacterial vaginosis may have contributed to the reduced number of preterm deliveries seen in the experimental group; however, we did not calculate our sample size using preterm delivery as a primary outcome, and the number of women we recruited fell short of the number required to detect a significant difference in the preterm delivery rate.

Given the randomized design of this study, the 1.8-fold increase in the rate of diagnosis of bacterial vaginosis among experimental subjects compared with controls (48.4% vs 27.1%) was likely a result of the increased screening with vaginal pH testing. It is noteworthy that the prevalence of bacterial vaginosis in the control group (27%) is near the upper end of the 10% to 32% range reported for other previously studied obstetric populations.¹⁰ This high-normal rate of infection may be related to the relatively high proportion of women of lower socioeconomic status, a group known to be at risk for preterm labor.¹⁹ In addition, our procedure of asking both experimental and control subjects about symptoms of vaginal infection as a routine part of each prenatal visit may have increased the rate of detecting infections for both groups. Given the high-risk nature of our population, it is notable that our controls had a preterm birth rate (10.2%) comparable with that of the general population,¹ and the experimental group’s preterm birth rate (4.7%) was less than half that of the controls.

A high vaginal pH value, vaginal symptoms, or both had 84% sensitivity to detect bacterial vaginosis or trichomoniasis. Inquiring about a history of vaginal symptoms and testing for vaginal pH levels are inexpensive procedures that could easily be implemented as part of routine prenatal care. If larger studies confirm its impact on the detection of infection and the preterm delivery rate, vaginal pH testing will merit wide use as a screening tool, with its results confirmed by more specific diagnostic procedures, such as the wet mount or Gram stain.

Nearly half of our sample received antimicrobial prescriptions for vaginal infections. Yet 56% of those treated who also had follow-up wet mount examinations had either persistent or recurrent bacterial vaginosis. This rate is 3 times higher than the 18% recurrence rate previously reported²⁰ and may indicate either noncompliance, resistance to the medication, or a tendency toward recurrence. Women who continue to have evidence of infection may remain at risk for preterm birth. A more effective strategy for eliminating bacterial vaginosis resulting in a lower rate of persistence or recurrence in the experimental group may have improved the preterm labor outcomes for women undergoing regular screening.

Limitations

An important limitation of this study is the relatively small sample size. Although we demonstrated a significant difference in our primary outcome (infection rates), the more compelling clinical outcome is preterm delivery rate. McGregor and colleagues¹¹ demonstrated a 50% reduction in bacterial-vaginosis–linked idiopathic preterm births with a program of screening for and treating bacterial vaginosis.¹¹ To have 80% power to show a reduction in preterm birth from 10% to 5%, we would have needed a sample size of 336 mothers in each group—a number beyond our capacity to fund and recruit.

Our difficulty in recruiting women resulted from our failure to invite all eligible women to participate and from the refusals to participate because we required frequent vaginal examinations. The observation that so many women disliked the idea of regular prenatal vaginal examinations suggests that we should investigate ways of making the vaginal pH test more acceptable, perhaps through self-testing. It is also possible that other populations (eg, women with previous preterm births or women of higher socioeconomic status) would be more open to regular pH testing.

Additional limitations of our study include potential selection bias, which may reduce the generalizability of our findings, and the fact that patients and providers were not blinded to the intervention. It would be helpful to minimize these potential sources of bias in a large-scale blinded randomized trial. Future studies should also differentiate the causes of positive results in follow-up tests and should examine the benefit of treating asymptomatic infections and doing follow-up tests for cure.

Conclusions

Our frequent screening procedures significantly increased the recognition of bacterial vaginosis and the treatment of infected women. Since vaginal symptoms and elevated pH levels appear to be useful in screening for bacterial vaginosis and trichomoniasis, frequent pH testing should be evaluated in larger studies.

Acknowledgments

This study was supported by a grant from the UCare Foundation.

Preterm delivery is one of the most important—and most costly—problems in obstetric medicine. Despite recent advances, nearly 10% of all infants in this country are born preterm.¹ Advances in the care of premature newborns, such as the administration of antenatal corticosteroids, the use of surfactant, and other developments in respiratory technology have improved outcomes, yet a greater emphasis on identifying and managing the causes of premature delivery is needed.

Recent research has indicated that maternal infection probably plays a much greater role in the occurrence of preterm labor than was previously understood. Bacterial byproducts have been shown to stimulate endogenous phospholipase A2, which induces the formation of prostaglandin, an important stimulant of uterine contractions.² Further, neutrophils and their constituent enzymes, together with bacteria and their proteases, weaken the amniochorion, predisposing a woman to premature rupture of membranes.³ A study⁴ comparing placentas from 21 women who delivered before 36 weeks’ gestation with those from 66 women who delivered at term found the rate of infected placentas to be 62% and 12%, respectively.

Specific genital infections have been linked to preterm labor, including sexually transmitted infections secondary to Chlamydia trachomatis, Neisseria gonorrhoeae, and Trichomonas vaginalis.⁵ Bacterial vaginosis, which is more common and not necessarily sexually transmitted, is associated with a 1.4- to 6.9-fold increased risk of preterm delivery.^6-9 The relatively high prevalence of bacterial vaginosis in the obstetric population (10% to 32%),¹⁰ coupled with evidence that screening and treating the condition reduces the rate of preterm deliveries,^11-13 makes detection important for reducing perinatal morbidity and mortality.

There is scant evidence that can be used to determine the optimal screening regimen for bacterial vaginosis in pregnancy. It is unclear, for example, whether all women should be routinely screened, how often the screening should occur, and which tests should be used. The vaginal pH test may be a valuable screening tool. It is a quick, inexpensive test where values >4.5 indicate the presence of either bacterial vaginosis or trichomoniasis. The sensitivity for detecting bacterial vaginosis with pH testing ranges from 84% to 97%.¹⁴ An elevated vaginal pH level has been shown to be 100% sensitive and 92% specific in screening nonpregnant premenopausal women for aerobic bacterial pathogens (b-hemolytic streptococci, Gardnerella vaginalis, or mixed aerobic organisms).¹⁵ However, little is known about the utility of the vaginal pH level as a screening tool in pregnancy.

The purpose of our study was to determine whether regular prenatal vaginal pH level screening resulted in more frequent diagnoses and treatment of bacterial vaginosis, trichomoniasis, and other genital infections and thus fewer preterm deliveries. In addition, we sought to determine the effectiveness of using recent vaginal symptoms and vaginal pH levels to diagnose bacterial vaginosis and trichomoniasis.

Methods

We enrolled pregnant women of less than 34 weeks’ gestation attending 1 of 3 participating clinics located in the Minneapolis/St. Paul, Minnesota, metropolitan area between July 1, 1996, and December 31, 1997. We calculated that a sample size of 208 women would detect a 50% increase in the diagnosis of bacterial vaginosis with a power of .80 and an a of .05. Women who participated in the study were randomized into experimental and control groups at the initial prenatal visit, using computer-generated random number tables and sealed envelopes containing the patient’s group assignment and study protocol. A study protocol was placed in the patient’s chart to facilitate group-specific clinical procedures and data collection Table 1; therefore, health care providers and subjects were not blinded to the intervention (vaginal pH testing).

During the first prenatal visit, patients from both groups were managed the same: symptoms of preterm labor (regular uterine contractions) and vaginal infection (discharge, burning, itching, and odor) were investigated and recorded as present or absent. The following laboratory tests were performed: vaginal pH, wet mount, dieoxyribonucleic acid (DNA) probe for C trachomatis and N gonorrhoeae, and urinalysis.

At each subsequent prenatal visit, women from both groups were asked about symptoms of preterm labor, vaginal infection, and urinary tract infection. Women in either group who reported symptoms of preterm labor had a wet mount examination, DNA probe, and urinalysis performed; women with vaginal symptoms had a wet mount and DNA probe performed. Experimental and control group subjects differed in that experimental group members had a vaginal pH level tested by their physician at each prenatal visit (controls did not have pH testing after the initial visit), and women with a pH level >4.5 also had a wet mount examination performed. The vaginal pH level was tested using a strip of specially formulated pH paper with an enhanced range in the spectrum of interest (pH = 4.0 to 5.5; paper manufactured by Micro Essential Laboratory, Brooklyn, NY). Participating study physicians were instructed in the technique for determining vaginal pH levels. The patient’s physician recorded the pH value in the medical record.

For the purposes of our study, a wet mount evaluation by a medical technologist was used to diagnose bacterial vaginosis, trichomoniasis, and vulvovaginal candidiasis. Participating medical technologists were formally trained in laboratory procedures and were required to have proficiency testing on wet mount examinations 3 times a year. The medical technologists were blinded to participants’ group assignments. Our selection of the wet mount examination as the confirmatory diagnostic test for vaginal infections was made on the basis of its applicability to each of the 3 types of infection, its utility for private practice (relatively easy to perform and inexpensive), and on previous data reporting its relatively high sensitivity for the diagnosis of G vaginalis, the most prevalent bacteria in bacterial vaginosis. The sensitivity of the wet mount for detecting G vaginalis and T vaginalis is 90% and 100%, respectively, when synthetic oligonucleotide probes are used as a gold standard.¹⁶ Other researchers have also used medical technologists’ readings of wet mounts as a standard for the diagnosis of bacterial vaginosis.¹⁷

Women from both experimental and control groups who were diagnosed with a specific vaginal or cervical infection were given standard antimicrobial treatments specified in the study protocol. Although most women diagnosed with bacterial vaginosis or trichomoniasis received oral antibiotics (metronidazole 500 mg twice daily for 7 days or clindamycin 300 mg twice daily for 7 days), a minority were treated with vaginal antimicrobials.

Hospital charts were also reviewed for gestational age at the time of delivery. Gestational age was determined using clinical parameters with ultrasound confirmation for cases where clinical dates were uncertain. Women who delivered at 20 weeks’ to less than 37 weeks’ gestation were categorized as having had a preterm delivery, consistent with accepted definitions.¹⁸

We used the chi-square test to look for significant differences between the 2 groups in infection, antibiotic treatment, and preterm delivery rates. The sensitivity and specificity of vaginal pH values and symptom reports for detecting bacterial vaginosis and T vaginalis were determined using data from the initial prenatal visits, where both vaginal pH and wet mount examinations were routinely performed.

Results

We invited a total of 243 women to participate in our study. Not all eligible patients who were pregnant at the time of the study were asked to participate because of staff changes and work demands. One hundred one declined to participate because of a reluctance to undergo routine vaginal examinations. Twenty-one additional women initially expressed interest in the study, but subsequently transferred care or had a spontaneous or induced abortion before their first formal prenatal visit. The 121 women randomized into our study (63 women in the experimental group; 58 in the control group) represented an overall response rate of 50% (number of participants divided by number of women invited to participate). Two women were pregnant twice during the course of the study, resulting in 123 total pregnancies. There were 910 prenatal visits during the course of the study, with an average of 7.4 prenatal visits per pregnancy. When our study population was compared with nonparticipating obstetrical patients at the 3 clinic sites, no differences were seen in age and proportion of women insured by public assistance. However, compared with nonparticipants, the study sample included a higher proportion of whites.

The mean age of the participants was 24.8 years (standard deviation = 5.4), and 60.2% of the women were on public assistance. The population included several ethnic groups: 47.2% were African American; 36.6% white; 4.1% Native American; 3.3% Asian; 1.6% Hispanic; and 7.2% unknown or other. There were no differences between experimental and control groups in age, ethnicity, and proportion of women on public assistance.

Women who underwent regular vaginal pH screening in addition to standard prenatal care showed significantly higher rates for the diagnosis of bacterial vaginosis, more frequent diagnosis of trichomoniasis (not statistically significant), higher antibiotic treatment rates for bacterial vaginosis and trichomoniasis (46.9% vs 27.1%, P = .024), and a lower rate of preterm births (not statistically significant; Table 2.

The sensitivity of a high vaginal pH level or vaginal symptoms in identifying women with bacterial vaginosis or trichomoniasis was 84.4% Table 3. Although the specificity of either of these tests for identifying bacterial vaginosis or trichomoniasis was modest (57.1%), it rose to 78.5% when both a high pH value and vaginal symptoms were present. Of the 27 women were given a diagnosis of and treatment for bacterial vaginosis and had follow-up wet mount evaluations,¹⁵ (55.6%) had recurrent or persistent bacterial vaginosis.

Discussion

In our study, regular testing for vaginal pH levels at prenatal examinations resulted in significantly more diagnoses of bacterial vaginosis and more antibiotic prescriptions. Increased recognition and treatment of bacterial vaginosis may have contributed to the reduced number of preterm deliveries seen in the experimental group; however, we did not calculate our sample size using preterm delivery as a primary outcome, and the number of women we recruited fell short of the number required to detect a significant difference in the preterm delivery rate.

Given the randomized design of this study, the 1.8-fold increase in the rate of diagnosis of bacterial vaginosis among experimental subjects compared with controls (48.4% vs 27.1%) was likely a result of the increased screening with vaginal pH testing. It is noteworthy that the prevalence of bacterial vaginosis in the control group (27%) is near the upper end of the 10% to 32% range reported for other previously studied obstetric populations.¹⁰ This high-normal rate of infection may be related to the relatively high proportion of women of lower socioeconomic status, a group known to be at risk for preterm labor.¹⁹ In addition, our procedure of asking both experimental and control subjects about symptoms of vaginal infection as a routine part of each prenatal visit may have increased the rate of detecting infections for both groups. Given the high-risk nature of our population, it is notable that our controls had a preterm birth rate (10.2%) comparable with that of the general population,¹ and the experimental group’s preterm birth rate (4.7%) was less than half that of the controls.

A high vaginal pH value, vaginal symptoms, or both had 84% sensitivity to detect bacterial vaginosis or trichomoniasis. Inquiring about a history of vaginal symptoms and testing for vaginal pH levels are inexpensive procedures that could easily be implemented as part of routine prenatal care. If larger studies confirm its impact on the detection of infection and the preterm delivery rate, vaginal pH testing will merit wide use as a screening tool, with its results confirmed by more specific diagnostic procedures, such as the wet mount or Gram stain.

Nearly half of our sample received antimicrobial prescriptions for vaginal infections. Yet 56% of those treated who also had follow-up wet mount examinations had either persistent or recurrent bacterial vaginosis. This rate is 3 times higher than the 18% recurrence rate previously reported²⁰ and may indicate either noncompliance, resistance to the medication, or a tendency toward recurrence. Women who continue to have evidence of infection may remain at risk for preterm birth. A more effective strategy for eliminating bacterial vaginosis resulting in a lower rate of persistence or recurrence in the experimental group may have improved the preterm labor outcomes for women undergoing regular screening.

Limitations

An important limitation of this study is the relatively small sample size. Although we demonstrated a significant difference in our primary outcome (infection rates), the more compelling clinical outcome is preterm delivery rate. McGregor and colleagues¹¹ demonstrated a 50% reduction in bacterial-vaginosis–linked idiopathic preterm births with a program of screening for and treating bacterial vaginosis.¹¹ To have 80% power to show a reduction in preterm birth from 10% to 5%, we would have needed a sample size of 336 mothers in each group—a number beyond our capacity to fund and recruit.

Our difficulty in recruiting women resulted from our failure to invite all eligible women to participate and from the refusals to participate because we required frequent vaginal examinations. The observation that so many women disliked the idea of regular prenatal vaginal examinations suggests that we should investigate ways of making the vaginal pH test more acceptable, perhaps through self-testing. It is also possible that other populations (eg, women with previous preterm births or women of higher socioeconomic status) would be more open to regular pH testing.

Additional limitations of our study include potential selection bias, which may reduce the generalizability of our findings, and the fact that patients and providers were not blinded to the intervention. It would be helpful to minimize these potential sources of bias in a large-scale blinded randomized trial. Future studies should also differentiate the causes of positive results in follow-up tests and should examine the benefit of treating asymptomatic infections and doing follow-up tests for cure.

Conclusions

Our frequent screening procedures significantly increased the recognition of bacterial vaginosis and the treatment of infected women. Since vaginal symptoms and elevated pH levels appear to be useful in screening for bacterial vaginosis and trichomoniasis, frequent pH testing should be evaluated in larger studies.

Acknowledgments

This study was supported by a grant from the UCare Foundation.