LCDR John C. Chin, MD, MC, USN

Cigarette smoking is an independent risk factor for lung cancer and atherosclerotic cardiovascular disease (ASCVD).^1-3 The National Lung Screening Trial (NLST) demonstrated both lung cancer mortality reduction with the use of surveillance low-dose computed tomography (LDCT) and ASCVD as the most common cause of death among smokers.^4,5 ASCVD remains the leading cause of death in the lung cancer screening (LCS) population.^2,3 After publication of the NLST results, the US Preventive Services Task Force (USPSTF) established LCS eligibility among smokers and the Center for Medicare and Medicaid Services approved payment for annual LDCT in this group.^1,6,7

Recently LDCT has been proposed as an adjunct diagnostic tool for detecting coronary artery calcium (CAC), which is independently associated with ASCVD and mortality.^8-13 CAC scores have been recommended by the 2019 American College of Cardiology/American Heart Association cholesterol treatment guidelines and shown to be cost-effective in guiding statin therapy for patients with borderline to intermediate ASCVD risk.^14-16 While CAC is conventionally quantified using electrocardiogram (ECG)-gated CT, these scans are not routinely performed in clinical practice because preventive CAC screening is neither recommended by the USPSTF nor covered by most insurance providers.^17,18 LDCT, conversely, is reimbursable and a well-validated ASCVD risk predictor.^18,19

In this study, we aimed to determine the validity of LDCT in identifying CAC among the military LCS population and whether it would impact statin recommendations based on 10-year ASCVD risk.

Methods

Participants were recruited from a retrospective cohort of 563 Military Health System (MHS) beneficiaries who received LCS with LDCT at Naval Medical Center Portsmouth (NMCP) in Virginia between January 1, 2019, and December 31, 2020. The 2013 USPSTF LCS guidelines were followed as the 2021 guidelines had not been published before the start of the study; thus, eligible participants included adults aged 55 to 80 years with at least a 30-pack-year smoking history and currently smoked or had quit within 15 years from the date of study consent.^6,7

Between November 2020 and May 2021, study investigators screened 287 patient records and recruited 190 participants by telephone, starting with individuals who had the most recent LDCT and working backward until reaching the predetermined 170 subjects who had undergone in-office consents before ECG-gated CT scans. Since LDCT was not obtained simultaneously with the ECG-gated CT, participants were required to complete their gated CT within 24 months of their last LDCT. Of the 190 subjects initially recruited, those who were ineligible for LCS (n = 4), had a history of angioplasty, stent, or bypass revascularization procedure (n = 4), did not complete their ECG-gated CT within the specified time frame (n = 8), or withdrew from the study (n = 4) were excluded. While gated CT scans were scored for CAC in the present time, LDCT (previously only read for general lung pathology) was not scored until after participant consent. Patients were peripherally followed, via health record reviews, for 3 months after their gated CT to document any additional imaging ordered by their primary care practitioners. The study was approved by the NMCP Institutional Review Board.

Coronary Artery Calcification Scoring

We performed CT scans using Siemens SOMATOM Flash, a second-generation dual-source scanner; and GE LightSpeed VCT, a single-source, 64-slice scanner. A step-and-shoot prospective trigger technique was used, and contiguous axial images were reconstructed at 2.5-mm or 3-mm intervals for CAC quantification using the Agatston method.²⁰ ECG-gated CT scans were electrocardiographically triggered at mid-diastole (70% of the R-R interval). Radiation dose reduction techniques involved adjustments of the mA according to body mass index and iterative reconstruction. LDCT scans were performed without ECG gating. We reconstructed contiguous axial images at 1-mm intervals for evaluation of the lung parenchyma. Similar dose-reduction techniques were used, to limit radiation exposure for each LDCT scan to < 1.5 mSv, per established guidelines.²¹ CAC on LDCT was also scored using the Agatston method. CAC was scored on the 2 scan types by different blinded reviewers.

Covariates

We reviewed outpatient health records to obtain participants’ age, sex, medical history, statin use, smoking status (current or former), and pack-years. International Classification of Diseases, Tenth Revision codes within medical encounters were used to document prevalent hypertension, hyperlipidemia, and diabetes mellitus. Participants’ most recent low-density lipoprotein value (within 24 months of ECG-gated CT) was recorded and 10-year ASCVD risk scores were calculated using the pooled cohorts equation.

Statistical Analysis

A power analysis performed before study initiation determined that a prospective sample size of 170 would be sufficient to provide strength of correlation between CAC scores calculated from ECG-gated CT and LDCT and achieve a statistical power of at least 80%. The Wilcoxon rank sum and Fisher exact tests were used to evaluate differences in continuous and categorical CAC scores, respectively. Given skewed distributions, Spearman rank correlations and Kendall W coefficient of concordance were respectively used to evaluate correlation and concordance of CAC scores between the 2 scan types. κ statistics were used to rate agreement between categorical CAC scores. Bland-Altman analysis was performed to determine the bias and limits of agreement between ECG-gated CT and LDCT.²² For categorical CAC score analysis, participants were categorized into 5 groups according to standard Agatston score cut-off points. We defined the 5 categories of CAC for both scan types based on previous analysis from Rumberger and colleagues: CAC = 0 (absent), CAC = 1-10 (minimal), CAC = 11-100 (mild), CAC = 101-400 (moderate), CAC > 400 (severe).²³ Of note, LDCT reports at NMCP include a visual CAC score using these qualitative descriptors that were available to LDCT reviewers. Analyses were conducted using SAS version 9.4 and Microsoft Excel; P values < .05 were considered statistically significant.

 

Results

The 170 participants had a mean (SD) age of 62.1 (4.6) years and were 70.6% male (Table 1). Hyperlipidemia was the most prevalent cardiac risk factor with almost 70% of participants on a statin. There was no incidence of ischemic ASCVD during follow-up, although 1 participant was later diagnosed with lung cancer after evaluation of suspicious pulmonary findings on ECG-gated CT. CAC was identified on both scan types in 126 participants; however, LDCT was discordant with gated CT in identifying CAC in 24 subjects (P < .001).

The correlation between CAC scores on ECG-gated CT and LDCT was 0.945 (P < .001) and the concordance was 0.643, indicating moderate agreement between CAC scores on the 2 different scans (Figure 1). Median CAC scores were significantly higher on ECG-gated CT when compared with LDCT (107.5 vs 48.1 Agatston units, respectively; P < .05). Table 2 shows the CAC score characteristics for both scan types. The κ statistic for agreement between categorical CAC scores on ECG-gated CT compared with LDCT was 0.49 (SEκ= 0.05; 95% CI, -0.73-1.71), and the weighted κ statistic was 0.71, indicating moderate to substantial agreement between the 2 scans using the specified cutoff points. The Bland-Altman analysis presented a mean bias of 111.45 Agatston units, with limits of agreement between -268.64 and 491.54, as shown in Figure 2, suggesting that CAC scores on ECG-gated CT were, on average, about 111 units higher than those on LDCT. Finally, there were 24 participants with CAC seen on ECG-gated CT but none identified on LDCT (P < .001); of this cohort 20 were already on a statin, and of the remaining 4 individuals, 1 met statin criteria based on a > 20% ASCVD risk score alone (regardless of CAC score), 1 with an intermediate risk score met statin criteria based on CAC score reporting, 1 did not meet criteria due to a low-risk score, and the last had no reportable ASCVD risk score.

In the study, there were 80 participants with reportable borderline to intermediate 10-year ASCVD risk scores (5% ≤ 10-year ASCVD risk < 20%), 49 of which were taking a statin. Of the remaining 31 participants not on a statin, 19 met statin criteria after CAC was identified on ECG-gated CT (of these 18 also had CAC identified on LDCT). Subsequently, the number of participants who met statin criteria after additional CAC reporting (on ECG-gated CT and LDCT) was statistically significant (P < .001 and P < .05, respectively). Of the 49 participants on a statin, only 1 individual no longer met statin criteria due to a CAC score < 1 on gated CT.

Discussion

In this study population of recruited MHS beneficiaries, there was a strong correlation and moderate to substantial agreement between CAC scores calculated from LDCT and conventional ECG-gated CT. The number of nonstatin participants who met statin criteria and would have benefited from additional CAC score reporting was statistically significant as compared to their statin counterparts who no longer met the criteria.

CAC screening using nongated CT has become an increasingly available and consistently reproducible means for stratifying ASCVD risk and guiding statin therapy in individuals with equivocal ASCVD risk scores.^24-26 As has been demonstrated in previous studies, our study additionally highlights the effective use of LDCT in not only identifying CAC, but also in beneficially impacting statin decisions in the high-risk smoking population.^24-26 Our results also showed LDCT missed CAC in participants, the majority of which were already on a statin, and only 1 nonstatin individual benefited from additional CAC reporting. CAC scoring on LDCT should be an adjunct, not a substitute, for ASCVD risk stratification to help guide statin management.^25,27

Our results may provide cost considerate implications for preventive CAC screening. While TRICARE covers the cost of ECG-gated CT for MHS beneficiaries, the same is not true of most nonmilitary insurance providers. Concerns about cancer risk from radiation exposure may also lead to hesitation about receiving additional CTs in the smoking population. Since the LCS population already receives annual LDCT, these scans can also be used for CAC scoring to help primary care professionals risk stratify their patients, as has been previously shown.^28-31 Clinicians should consider implementing CAC scoring with annual LDCT scans, which would curtail further risks and expenses from CAC-specified scans.

Although CAC is scored visually and routinely reported in the body of LDCT reports at our facility, this is not a universal practice and was performed in only 44% of subjects with known CAC by a previous study.³² In 2007, there were 600,000 CAC scoring scans and > 9 million routine chest CTs performed in the United States.³³ Based on our results and the growing consensus in the existing literature, CAC scoring on nongated CT is not only valid and reliable, but also can estimate ASCVD risk and subsequent mortality.^34-36 Routine chest CTs remain an available resource for providing additional ASCVD risk stratification.

As we demonstrated, median CAC scores on LDCT were on average significantly lower than those from gated CT. This could be due to slice thickness variability between the GE and Siemens scanners or CAC progression between the time of the retrospective LDCT and prospective ECG-gated CT. Aside from this potential limitation, LDCT has been shown to have a high level of agreement with gated CT in predicting CAC, both visually and by the Agatston technique.^37-39 Our results further support previous recommendations of utilizing CAC score categories when determining ASCVD risk from LDCT and that establishing scoring cutoff points warrants further development for potential standardization.^37-39 Readers should be mindful that LDCT may still be less sensitive and underestimate low CAC levels and that ECG-gated CT may occasionally be more optimal in determining ASCVD risk when considering the negative predictive value of CAC.⁴⁰

 

Limitations

Our study cohort was composed of MHS beneficiaries. Compared with the general population, these individuals may have greater access to care and be more likely to receive statins after preventive screenings. Additional studies may be required to assess CAC-associated statin eligibility among the general population. As discussed previously LDCT was not performed concomitantly with the ECG-gated CT. Although there was moderate to substantial CAC agreement between the 2 scan types, the timing difference could have led to absolute differences in CAC scores across both scan types and impacted the ability to detect low-level CAC on LDCT. CAC values should be interpreted based on the respective scan type.

Conclusions

LDCT is a reliable diagnostic alternative to ECG-gated CT in predicting CAC. CAC scores from LDCT are highly correlated and concordant with those from gated CT and can help guide statin management in individuals with intermediate ASCVD risk. The proposed duality of LDCT to assess ASCVD risk in addition to lung cancer can reduce the need for unnecessary scans while optimizing preventive clinical care. While coronary calcium and elevated CAC scores can facilitate clinical decision making to initiate statin therapy for intermediate-risk patients, physicians must still determine whether additional cardiac testing is warranted to avoid unnecessary procedures and health care costs. Smokers undergoing annual LDCT may benefit from standardized CAC scoring to help further stratify ASCVD risk while limiting the expense and radiation of additional scans.

Acknowledgments

The authors thank Ms. Lorie Gower for her contributions to the study.

Cigarette smoking is an independent risk factor for lung cancer and atherosclerotic cardiovascular disease (ASCVD).^1-3 The National Lung Screening Trial (NLST) demonstrated both lung cancer mortality reduction with the use of surveillance low-dose computed tomography (LDCT) and ASCVD as the most common cause of death among smokers.^4,5 ASCVD remains the leading cause of death in the lung cancer screening (LCS) population.^2,3 After publication of the NLST results, the US Preventive Services Task Force (USPSTF) established LCS eligibility among smokers and the Center for Medicare and Medicaid Services approved payment for annual LDCT in this group.^1,6,7

Recently LDCT has been proposed as an adjunct diagnostic tool for detecting coronary artery calcium (CAC), which is independently associated with ASCVD and mortality.^8-13 CAC scores have been recommended by the 2019 American College of Cardiology/American Heart Association cholesterol treatment guidelines and shown to be cost-effective in guiding statin therapy for patients with borderline to intermediate ASCVD risk.^14-16 While CAC is conventionally quantified using electrocardiogram (ECG)-gated CT, these scans are not routinely performed in clinical practice because preventive CAC screening is neither recommended by the USPSTF nor covered by most insurance providers.^17,18 LDCT, conversely, is reimbursable and a well-validated ASCVD risk predictor.^18,19

In this study, we aimed to determine the validity of LDCT in identifying CAC among the military LCS population and whether it would impact statin recommendations based on 10-year ASCVD risk.

Methods

Participants were recruited from a retrospective cohort of 563 Military Health System (MHS) beneficiaries who received LCS with LDCT at Naval Medical Center Portsmouth (NMCP) in Virginia between January 1, 2019, and December 31, 2020. The 2013 USPSTF LCS guidelines were followed as the 2021 guidelines had not been published before the start of the study; thus, eligible participants included adults aged 55 to 80 years with at least a 30-pack-year smoking history and currently smoked or had quit within 15 years from the date of study consent.^6,7

Between November 2020 and May 2021, study investigators screened 287 patient records and recruited 190 participants by telephone, starting with individuals who had the most recent LDCT and working backward until reaching the predetermined 170 subjects who had undergone in-office consents before ECG-gated CT scans. Since LDCT was not obtained simultaneously with the ECG-gated CT, participants were required to complete their gated CT within 24 months of their last LDCT. Of the 190 subjects initially recruited, those who were ineligible for LCS (n = 4), had a history of angioplasty, stent, or bypass revascularization procedure (n = 4), did not complete their ECG-gated CT within the specified time frame (n = 8), or withdrew from the study (n = 4) were excluded. While gated CT scans were scored for CAC in the present time, LDCT (previously only read for general lung pathology) was not scored until after participant consent. Patients were peripherally followed, via health record reviews, for 3 months after their gated CT to document any additional imaging ordered by their primary care practitioners. The study was approved by the NMCP Institutional Review Board.

Coronary Artery Calcification Scoring

We performed CT scans using Siemens SOMATOM Flash, a second-generation dual-source scanner; and GE LightSpeed VCT, a single-source, 64-slice scanner. A step-and-shoot prospective trigger technique was used, and contiguous axial images were reconstructed at 2.5-mm or 3-mm intervals for CAC quantification using the Agatston method.²⁰ ECG-gated CT scans were electrocardiographically triggered at mid-diastole (70% of the R-R interval). Radiation dose reduction techniques involved adjustments of the mA according to body mass index and iterative reconstruction. LDCT scans were performed without ECG gating. We reconstructed contiguous axial images at 1-mm intervals for evaluation of the lung parenchyma. Similar dose-reduction techniques were used, to limit radiation exposure for each LDCT scan to < 1.5 mSv, per established guidelines.²¹ CAC on LDCT was also scored using the Agatston method. CAC was scored on the 2 scan types by different blinded reviewers.

Covariates

We reviewed outpatient health records to obtain participants’ age, sex, medical history, statin use, smoking status (current or former), and pack-years. International Classification of Diseases, Tenth Revision codes within medical encounters were used to document prevalent hypertension, hyperlipidemia, and diabetes mellitus. Participants’ most recent low-density lipoprotein value (within 24 months of ECG-gated CT) was recorded and 10-year ASCVD risk scores were calculated using the pooled cohorts equation.

Statistical Analysis

A power analysis performed before study initiation determined that a prospective sample size of 170 would be sufficient to provide strength of correlation between CAC scores calculated from ECG-gated CT and LDCT and achieve a statistical power of at least 80%. The Wilcoxon rank sum and Fisher exact tests were used to evaluate differences in continuous and categorical CAC scores, respectively. Given skewed distributions, Spearman rank correlations and Kendall W coefficient of concordance were respectively used to evaluate correlation and concordance of CAC scores between the 2 scan types. κ statistics were used to rate agreement between categorical CAC scores. Bland-Altman analysis was performed to determine the bias and limits of agreement between ECG-gated CT and LDCT.²² For categorical CAC score analysis, participants were categorized into 5 groups according to standard Agatston score cut-off points. We defined the 5 categories of CAC for both scan types based on previous analysis from Rumberger and colleagues: CAC = 0 (absent), CAC = 1-10 (minimal), CAC = 11-100 (mild), CAC = 101-400 (moderate), CAC > 400 (severe).²³ Of note, LDCT reports at NMCP include a visual CAC score using these qualitative descriptors that were available to LDCT reviewers. Analyses were conducted using SAS version 9.4 and Microsoft Excel; P values < .05 were considered statistically significant.

 

Results

The 170 participants had a mean (SD) age of 62.1 (4.6) years and were 70.6% male (Table 1). Hyperlipidemia was the most prevalent cardiac risk factor with almost 70% of participants on a statin. There was no incidence of ischemic ASCVD during follow-up, although 1 participant was later diagnosed with lung cancer after evaluation of suspicious pulmonary findings on ECG-gated CT. CAC was identified on both scan types in 126 participants; however, LDCT was discordant with gated CT in identifying CAC in 24 subjects (P < .001).

The correlation between CAC scores on ECG-gated CT and LDCT was 0.945 (P < .001) and the concordance was 0.643, indicating moderate agreement between CAC scores on the 2 different scans (Figure 1). Median CAC scores were significantly higher on ECG-gated CT when compared with LDCT (107.5 vs 48.1 Agatston units, respectively; P < .05). Table 2 shows the CAC score characteristics for both scan types. The κ statistic for agreement between categorical CAC scores on ECG-gated CT compared with LDCT was 0.49 (SEκ= 0.05; 95% CI, -0.73-1.71), and the weighted κ statistic was 0.71, indicating moderate to substantial agreement between the 2 scans using the specified cutoff points. The Bland-Altman analysis presented a mean bias of 111.45 Agatston units, with limits of agreement between -268.64 and 491.54, as shown in Figure 2, suggesting that CAC scores on ECG-gated CT were, on average, about 111 units higher than those on LDCT. Finally, there were 24 participants with CAC seen on ECG-gated CT but none identified on LDCT (P < .001); of this cohort 20 were already on a statin, and of the remaining 4 individuals, 1 met statin criteria based on a > 20% ASCVD risk score alone (regardless of CAC score), 1 with an intermediate risk score met statin criteria based on CAC score reporting, 1 did not meet criteria due to a low-risk score, and the last had no reportable ASCVD risk score.

In the study, there were 80 participants with reportable borderline to intermediate 10-year ASCVD risk scores (5% ≤ 10-year ASCVD risk < 20%), 49 of which were taking a statin. Of the remaining 31 participants not on a statin, 19 met statin criteria after CAC was identified on ECG-gated CT (of these 18 also had CAC identified on LDCT). Subsequently, the number of participants who met statin criteria after additional CAC reporting (on ECG-gated CT and LDCT) was statistically significant (P < .001 and P < .05, respectively). Of the 49 participants on a statin, only 1 individual no longer met statin criteria due to a CAC score < 1 on gated CT.

Discussion

In this study population of recruited MHS beneficiaries, there was a strong correlation and moderate to substantial agreement between CAC scores calculated from LDCT and conventional ECG-gated CT. The number of nonstatin participants who met statin criteria and would have benefited from additional CAC score reporting was statistically significant as compared to their statin counterparts who no longer met the criteria.

CAC screening using nongated CT has become an increasingly available and consistently reproducible means for stratifying ASCVD risk and guiding statin therapy in individuals with equivocal ASCVD risk scores.^24-26 As has been demonstrated in previous studies, our study additionally highlights the effective use of LDCT in not only identifying CAC, but also in beneficially impacting statin decisions in the high-risk smoking population.^24-26 Our results also showed LDCT missed CAC in participants, the majority of which were already on a statin, and only 1 nonstatin individual benefited from additional CAC reporting. CAC scoring on LDCT should be an adjunct, not a substitute, for ASCVD risk stratification to help guide statin management.^25,27

Our results may provide cost considerate implications for preventive CAC screening. While TRICARE covers the cost of ECG-gated CT for MHS beneficiaries, the same is not true of most nonmilitary insurance providers. Concerns about cancer risk from radiation exposure may also lead to hesitation about receiving additional CTs in the smoking population. Since the LCS population already receives annual LDCT, these scans can also be used for CAC scoring to help primary care professionals risk stratify their patients, as has been previously shown.^28-31 Clinicians should consider implementing CAC scoring with annual LDCT scans, which would curtail further risks and expenses from CAC-specified scans.

Although CAC is scored visually and routinely reported in the body of LDCT reports at our facility, this is not a universal practice and was performed in only 44% of subjects with known CAC by a previous study.³² In 2007, there were 600,000 CAC scoring scans and > 9 million routine chest CTs performed in the United States.³³ Based on our results and the growing consensus in the existing literature, CAC scoring on nongated CT is not only valid and reliable, but also can estimate ASCVD risk and subsequent mortality.^34-36 Routine chest CTs remain an available resource for providing additional ASCVD risk stratification.

As we demonstrated, median CAC scores on LDCT were on average significantly lower than those from gated CT. This could be due to slice thickness variability between the GE and Siemens scanners or CAC progression between the time of the retrospective LDCT and prospective ECG-gated CT. Aside from this potential limitation, LDCT has been shown to have a high level of agreement with gated CT in predicting CAC, both visually and by the Agatston technique.^37-39 Our results further support previous recommendations of utilizing CAC score categories when determining ASCVD risk from LDCT and that establishing scoring cutoff points warrants further development for potential standardization.^37-39 Readers should be mindful that LDCT may still be less sensitive and underestimate low CAC levels and that ECG-gated CT may occasionally be more optimal in determining ASCVD risk when considering the negative predictive value of CAC.⁴⁰

 

Limitations

Our study cohort was composed of MHS beneficiaries. Compared with the general population, these individuals may have greater access to care and be more likely to receive statins after preventive screenings. Additional studies may be required to assess CAC-associated statin eligibility among the general population. As discussed previously LDCT was not performed concomitantly with the ECG-gated CT. Although there was moderate to substantial CAC agreement between the 2 scan types, the timing difference could have led to absolute differences in CAC scores across both scan types and impacted the ability to detect low-level CAC on LDCT. CAC values should be interpreted based on the respective scan type.

Conclusions

LDCT is a reliable diagnostic alternative to ECG-gated CT in predicting CAC. CAC scores from LDCT are highly correlated and concordant with those from gated CT and can help guide statin management in individuals with intermediate ASCVD risk. The proposed duality of LDCT to assess ASCVD risk in addition to lung cancer can reduce the need for unnecessary scans while optimizing preventive clinical care. While coronary calcium and elevated CAC scores can facilitate clinical decision making to initiate statin therapy for intermediate-risk patients, physicians must still determine whether additional cardiac testing is warranted to avoid unnecessary procedures and health care costs. Smokers undergoing annual LDCT may benefit from standardized CAC scoring to help further stratify ASCVD risk while limiting the expense and radiation of additional scans.

Acknowledgments

The authors thank Ms. Lorie Gower for her contributions to the study.

Cigarette smoking is an independent risk factor for lung cancer and atherosclerotic cardiovascular disease (ASCVD).^1-3 The National Lung Screening Trial (NLST) demonstrated both lung cancer mortality reduction with the use of surveillance low-dose computed tomography (LDCT) and ASCVD as the most common cause of death among smokers.^4,5 ASCVD remains the leading cause of death in the lung cancer screening (LCS) population.^2,3 After publication of the NLST results, the US Preventive Services Task Force (USPSTF) established LCS eligibility among smokers and the Center for Medicare and Medicaid Services approved payment for annual LDCT in this group.^1,6,7

Recently LDCT has been proposed as an adjunct diagnostic tool for detecting coronary artery calcium (CAC), which is independently associated with ASCVD and mortality.^8-13 CAC scores have been recommended by the 2019 American College of Cardiology/American Heart Association cholesterol treatment guidelines and shown to be cost-effective in guiding statin therapy for patients with borderline to intermediate ASCVD risk.^14-16 While CAC is conventionally quantified using electrocardiogram (ECG)-gated CT, these scans are not routinely performed in clinical practice because preventive CAC screening is neither recommended by the USPSTF nor covered by most insurance providers.^17,18 LDCT, conversely, is reimbursable and a well-validated ASCVD risk predictor.^18,19

In this study, we aimed to determine the validity of LDCT in identifying CAC among the military LCS population and whether it would impact statin recommendations based on 10-year ASCVD risk.

Methods

Participants were recruited from a retrospective cohort of 563 Military Health System (MHS) beneficiaries who received LCS with LDCT at Naval Medical Center Portsmouth (NMCP) in Virginia between January 1, 2019, and December 31, 2020. The 2013 USPSTF LCS guidelines were followed as the 2021 guidelines had not been published before the start of the study; thus, eligible participants included adults aged 55 to 80 years with at least a 30-pack-year smoking history and currently smoked or had quit within 15 years from the date of study consent.^6,7

Between November 2020 and May 2021, study investigators screened 287 patient records and recruited 190 participants by telephone, starting with individuals who had the most recent LDCT and working backward until reaching the predetermined 170 subjects who had undergone in-office consents before ECG-gated CT scans. Since LDCT was not obtained simultaneously with the ECG-gated CT, participants were required to complete their gated CT within 24 months of their last LDCT. Of the 190 subjects initially recruited, those who were ineligible for LCS (n = 4), had a history of angioplasty, stent, or bypass revascularization procedure (n = 4), did not complete their ECG-gated CT within the specified time frame (n = 8), or withdrew from the study (n = 4) were excluded. While gated CT scans were scored for CAC in the present time, LDCT (previously only read for general lung pathology) was not scored until after participant consent. Patients were peripherally followed, via health record reviews, for 3 months after their gated CT to document any additional imaging ordered by their primary care practitioners. The study was approved by the NMCP Institutional Review Board.

Coronary Artery Calcification Scoring

We performed CT scans using Siemens SOMATOM Flash, a second-generation dual-source scanner; and GE LightSpeed VCT, a single-source, 64-slice scanner. A step-and-shoot prospective trigger technique was used, and contiguous axial images were reconstructed at 2.5-mm or 3-mm intervals for CAC quantification using the Agatston method.²⁰ ECG-gated CT scans were electrocardiographically triggered at mid-diastole (70% of the R-R interval). Radiation dose reduction techniques involved adjustments of the mA according to body mass index and iterative reconstruction. LDCT scans were performed without ECG gating. We reconstructed contiguous axial images at 1-mm intervals for evaluation of the lung parenchyma. Similar dose-reduction techniques were used, to limit radiation exposure for each LDCT scan to < 1.5 mSv, per established guidelines.²¹ CAC on LDCT was also scored using the Agatston method. CAC was scored on the 2 scan types by different blinded reviewers.

Covariates

We reviewed outpatient health records to obtain participants’ age, sex, medical history, statin use, smoking status (current or former), and pack-years. International Classification of Diseases, Tenth Revision codes within medical encounters were used to document prevalent hypertension, hyperlipidemia, and diabetes mellitus. Participants’ most recent low-density lipoprotein value (within 24 months of ECG-gated CT) was recorded and 10-year ASCVD risk scores were calculated using the pooled cohorts equation.

Statistical Analysis

A power analysis performed before study initiation determined that a prospective sample size of 170 would be sufficient to provide strength of correlation between CAC scores calculated from ECG-gated CT and LDCT and achieve a statistical power of at least 80%. The Wilcoxon rank sum and Fisher exact tests were used to evaluate differences in continuous and categorical CAC scores, respectively. Given skewed distributions, Spearman rank correlations and Kendall W coefficient of concordance were respectively used to evaluate correlation and concordance of CAC scores between the 2 scan types. κ statistics were used to rate agreement between categorical CAC scores. Bland-Altman analysis was performed to determine the bias and limits of agreement between ECG-gated CT and LDCT.²² For categorical CAC score analysis, participants were categorized into 5 groups according to standard Agatston score cut-off points. We defined the 5 categories of CAC for both scan types based on previous analysis from Rumberger and colleagues: CAC = 0 (absent), CAC = 1-10 (minimal), CAC = 11-100 (mild), CAC = 101-400 (moderate), CAC > 400 (severe).²³ Of note, LDCT reports at NMCP include a visual CAC score using these qualitative descriptors that were available to LDCT reviewers. Analyses were conducted using SAS version 9.4 and Microsoft Excel; P values < .05 were considered statistically significant.

 

Results

The 170 participants had a mean (SD) age of 62.1 (4.6) years and were 70.6% male (Table 1). Hyperlipidemia was the most prevalent cardiac risk factor with almost 70% of participants on a statin. There was no incidence of ischemic ASCVD during follow-up, although 1 participant was later diagnosed with lung cancer after evaluation of suspicious pulmonary findings on ECG-gated CT. CAC was identified on both scan types in 126 participants; however, LDCT was discordant with gated CT in identifying CAC in 24 subjects (P < .001).

The correlation between CAC scores on ECG-gated CT and LDCT was 0.945 (P < .001) and the concordance was 0.643, indicating moderate agreement between CAC scores on the 2 different scans (Figure 1). Median CAC scores were significantly higher on ECG-gated CT when compared with LDCT (107.5 vs 48.1 Agatston units, respectively; P < .05). Table 2 shows the CAC score characteristics for both scan types. The κ statistic for agreement between categorical CAC scores on ECG-gated CT compared with LDCT was 0.49 (SEκ= 0.05; 95% CI, -0.73-1.71), and the weighted κ statistic was 0.71, indicating moderate to substantial agreement between the 2 scans using the specified cutoff points. The Bland-Altman analysis presented a mean bias of 111.45 Agatston units, with limits of agreement between -268.64 and 491.54, as shown in Figure 2, suggesting that CAC scores on ECG-gated CT were, on average, about 111 units higher than those on LDCT. Finally, there were 24 participants with CAC seen on ECG-gated CT but none identified on LDCT (P < .001); of this cohort 20 were already on a statin, and of the remaining 4 individuals, 1 met statin criteria based on a > 20% ASCVD risk score alone (regardless of CAC score), 1 with an intermediate risk score met statin criteria based on CAC score reporting, 1 did not meet criteria due to a low-risk score, and the last had no reportable ASCVD risk score.

In the study, there were 80 participants with reportable borderline to intermediate 10-year ASCVD risk scores (5% ≤ 10-year ASCVD risk < 20%), 49 of which were taking a statin. Of the remaining 31 participants not on a statin, 19 met statin criteria after CAC was identified on ECG-gated CT (of these 18 also had CAC identified on LDCT). Subsequently, the number of participants who met statin criteria after additional CAC reporting (on ECG-gated CT and LDCT) was statistically significant (P < .001 and P < .05, respectively). Of the 49 participants on a statin, only 1 individual no longer met statin criteria due to a CAC score < 1 on gated CT.

Discussion

In this study population of recruited MHS beneficiaries, there was a strong correlation and moderate to substantial agreement between CAC scores calculated from LDCT and conventional ECG-gated CT. The number of nonstatin participants who met statin criteria and would have benefited from additional CAC score reporting was statistically significant as compared to their statin counterparts who no longer met the criteria.

CAC screening using nongated CT has become an increasingly available and consistently reproducible means for stratifying ASCVD risk and guiding statin therapy in individuals with equivocal ASCVD risk scores.^24-26 As has been demonstrated in previous studies, our study additionally highlights the effective use of LDCT in not only identifying CAC, but also in beneficially impacting statin decisions in the high-risk smoking population.^24-26 Our results also showed LDCT missed CAC in participants, the majority of which were already on a statin, and only 1 nonstatin individual benefited from additional CAC reporting. CAC scoring on LDCT should be an adjunct, not a substitute, for ASCVD risk stratification to help guide statin management.^25,27

Our results may provide cost considerate implications for preventive CAC screening. While TRICARE covers the cost of ECG-gated CT for MHS beneficiaries, the same is not true of most nonmilitary insurance providers. Concerns about cancer risk from radiation exposure may also lead to hesitation about receiving additional CTs in the smoking population. Since the LCS population already receives annual LDCT, these scans can also be used for CAC scoring to help primary care professionals risk stratify their patients, as has been previously shown.^28-31 Clinicians should consider implementing CAC scoring with annual LDCT scans, which would curtail further risks and expenses from CAC-specified scans.

Although CAC is scored visually and routinely reported in the body of LDCT reports at our facility, this is not a universal practice and was performed in only 44% of subjects with known CAC by a previous study.³² In 2007, there were 600,000 CAC scoring scans and > 9 million routine chest CTs performed in the United States.³³ Based on our results and the growing consensus in the existing literature, CAC scoring on nongated CT is not only valid and reliable, but also can estimate ASCVD risk and subsequent mortality.^34-36 Routine chest CTs remain an available resource for providing additional ASCVD risk stratification.

As we demonstrated, median CAC scores on LDCT were on average significantly lower than those from gated CT. This could be due to slice thickness variability between the GE and Siemens scanners or CAC progression between the time of the retrospective LDCT and prospective ECG-gated CT. Aside from this potential limitation, LDCT has been shown to have a high level of agreement with gated CT in predicting CAC, both visually and by the Agatston technique.^37-39 Our results further support previous recommendations of utilizing CAC score categories when determining ASCVD risk from LDCT and that establishing scoring cutoff points warrants further development for potential standardization.^37-39 Readers should be mindful that LDCT may still be less sensitive and underestimate low CAC levels and that ECG-gated CT may occasionally be more optimal in determining ASCVD risk when considering the negative predictive value of CAC.⁴⁰

 

Limitations

Our study cohort was composed of MHS beneficiaries. Compared with the general population, these individuals may have greater access to care and be more likely to receive statins after preventive screenings. Additional studies may be required to assess CAC-associated statin eligibility among the general population. As discussed previously LDCT was not performed concomitantly with the ECG-gated CT. Although there was moderate to substantial CAC agreement between the 2 scan types, the timing difference could have led to absolute differences in CAC scores across both scan types and impacted the ability to detect low-level CAC on LDCT. CAC values should be interpreted based on the respective scan type.

Conclusions

LDCT is a reliable diagnostic alternative to ECG-gated CT in predicting CAC. CAC scores from LDCT are highly correlated and concordant with those from gated CT and can help guide statin management in individuals with intermediate ASCVD risk. The proposed duality of LDCT to assess ASCVD risk in addition to lung cancer can reduce the need for unnecessary scans while optimizing preventive clinical care. While coronary calcium and elevated CAC scores can facilitate clinical decision making to initiate statin therapy for intermediate-risk patients, physicians must still determine whether additional cardiac testing is warranted to avoid unnecessary procedures and health care costs. Smokers undergoing annual LDCT may benefit from standardized CAC scoring to help further stratify ASCVD risk while limiting the expense and radiation of additional scans.

Acknowledgments

The authors thank Ms. Lorie Gower for her contributions to the study.

Acute pyogenic vertebral osteomyelitis is often due to hematogenous spread of aerobic bacteria.^1-4 Conversely, only 0.5% of anaerobic bacteremias lead to osteomyelitis.⁵ Anaerobic osteomyelitis typically results from the contiguous spread of polymicrobial infections through breaks in the gut mucosal barrier and involves the vertebral bodies in only 2% to 5% of cases.^5,6 Although Bacteroides fragilis (B fragilis) is the most common anaerobic pathogen cultivated from blood, accounting for about half of all anaerobic blood isolates, it seldom leads to osteomyelitis.^1,2,7-11 We report an uncommon case of B fragilis bacteremia and vertebral osteomyelitis confounded by uncertainties in anaerobic identification and susceptibilities.

Case Presentation

A healthy-appearing male aged 55 years presented to the Naval Medical Center Portsmouth (NMCP) with subacute low back pain and fevers of 103 °F for > 3 weeks. While traveling 4 weeks prior, he completed a course of oseltamivir for influenza B infection; afterward, he was diagnosed with community-acquired pneumonia and treated with a dose of ceftriaxone and a 7-day course of doxycycline. The patient presented to the same facility a week later for low back pain and nonresolving respiratory symptoms, and his therapy was changed to azithromycin, cefuroxime, prednisone, and inhalers. Additionally, after being treated for influenza, he developed constipation and hematochezia for which he did not seek care. The hematochezia was similar to a previous episode from an anal fissure 1 year prior that resolved with stool softeners. When he was finally seen at NMCP after 3 weeks of worsening back pain and fevers, lumbosacral magnetic resonance imaging (MRI) demonstrated vertebral osteomyelitis and discitis at L4-L5 and admitted to the hospital (Figure 1).

After a fluoroscopy-guided biopsy of the L4 vertebral body on hospital day 1, the patient was started on cefepime and vancomycin. The biopsy sample was inoculated onto solid media (blood agar, chocolate agar, and MacConkey agar) and incubated at 36 °C for 24 hours in a 5% CO₂ atmosphere, as well as onto Shaedler agar with vitamin K and chopped meat glucose broth and incubated at 36 °C for 48 hours under anaerobic conditions. Metronidazole was added and vancomycin discontinued after 2 anaerobic blood culture vials obtained on hospital day 1, incubated in a Becton Dickinson BACTEC FX automated system, which demonstrated Gram-negative bacilli after 48 hours. The blood culture isolates demonstrated a > 99% probability of being identified as ß-lactamase positive Prevotella loescheii using Thermo Fischer Scientific RapID ANA II biochemical testing. Nitrocefinase discs were used to detect ß-lactamase activity.

The biopsy demonstrated nongranulomatous focal areas of necrotic bone and neutrophilia in a hematopoietic background consistent with acute osteomyelitis (Figure 2); on hospital day 4, ß-lactamase positive B fragilis grew from the bone culture. Additionally, 1 anaerobic vial from a surveillance blood culture set that was obtained on hospital day 3 grew ß-lactamasepositive B fragilis using the same identification methods. With these results he was thought to have a polymicrobial infection (B fragilis and Prevotella loescheii [P loescheii]) from a suspected bowel source based on his hematochezia and history of anal fissure. No aerobic, Gram-negative enterobacteriaceae were isolated, but he had previously been on cefuroxime, which has potential activity against these organisms, for ≥ 2 weeks prior to hospitalization and cultures. He was discharged on moxifloxacin and metronidazole pending final culture results, including requested anaerobic susceptibility testing.

At 1-week follow-up, both aerobic and anaerobic vials from surveillance blood cultures remained negative for any microbes, so antibiotics were deescalated to moxifloxacin monotherapy. However, after 3 days the patient was readmitted for increasing C-reactive protein (CRP) levels and intractable back pain with worsening bilateral radiculopathy. A repeat MRI demonstrated interval disease progression with near obliteration of the L4-L5 disc space and hyperenhancement of the prevertebral soft tissues and adjacent psoas musculature without focal rim-enhancing fluid collection (Figure 3). After repeat L4 biopsy, metronidazole was restarted and ertapenem added for enterobacteriaceae coverage, given the known B fragilis and potential suppression from previous cephalosporin therapy; moxifloxacin was discontinued. L4 biopsy cultures showed no growth, and CRP levels trended down from 154.2 mg/L (start of first admission) to 42.4 mg/L (start of second admission) to 14.9 mg/L (day of discharge) (reference range, 5-9.9 mg/L). He was discharged on ertapenem and metronidazole. He completed a 6-week course without further complication.

During antibiotic therapy he had an unremarkable colonoscopy, CRP normalized to 2.6 mg/L (reference range, 0-4.9 mg/L), and he underwent successful L4-L5 transforaminal lumbar interbody fusion 2 weeks after finishing antibiotics.

We retroactively sent both P loescheii isolates and the 1 B fragilis isolate that grew from the surveillance blood culture to the Multidrug-resistant Organism Repository and Surveillance Network (MRSN) at the Walter Reed Army Institute of Research for identification confirmation and susceptibility analysis. Whole genome sequencing with single nucleotide polymorphism (SNP)-based analysis revealed all isolates were 100% identical and consistent with B fragilis and not P loescheii, based on clustering around other B fragilis sequences found in the National Center for Biotechnology Information (NCBI) Genbank database (Figure 4). All isolates carried the antibiotic resistance genes— cepA, sul(2), tetQ— encoding for possible resistance to cephalosporins, sulphonamides, and tetracyclines, respectively; as well as a point mutation in the gyrA gene (Ser82Phe). None of the isolates carried the nim gene, and screening for the 3 subtypes of B fragilis enterotoxin gene (bft-1, bft-2, bft-3) was negative. Eventual susceptibility testing at the Mayo Clinic several months after the conclusion of the case indicated that the B fragilis isolate was sensitive to piperacillin-tazobactam, ertapenem, clindamycin, and metronidazole; however, testing was not performed against moxifloxacin.

 

Discussion

In the era of growing antibiotic resistance patterns, antimicrobial stewardship programs recommend interventions to improve antimicrobial use through targeted narrow- spectrum antibiotics.¹² The Clinical and Laboratory Standards Institute (CLSI) maintains guidelines on the major indications for anaerobic antimicrobial susceptibility testing (AST) to help direct narrow-targeted antimicrobial therapy. However, in a 2008 practice survey Goldstein and colleagues reported that less than half of US hospitals performed anaerobic AST, and only 21% of these facilities did it in-house, while the remainder sent out their isolates for testing.^11-14 The CLSI major indications for AST include situations in which the selection of agents is important because of the (1) known resistance of a particular species; (2) confirmation of appropriate therapy for severe infections or for those that may require long-term therapy; (3) persistence of infection despite adequate treatment with an appropriate therapeutic regimen; and (4) difficulty in making empirical decisions based on precedent.¹⁴ Additionally, isolates from brain abscess, endocarditis, osteomyelitis, joint infection, infection of prosthetic devices or vascular grafts, bacteremia, and normally sterile body sites (unless contamination suspected) should be tested.¹⁴

Because of the lack of anaerobic AST, health care providers must base empiric treatment on reported sensitivities from the medical literature. Empiric selection of antimicrobials for anaerobic infections is made even more challenging by the increased rates of resistance reported in the literature, leading to recommendations to increase susceptibility testing to guide therapy.^13,15,16 Empiric therapy of deep-seated anaerobic infections may lead to use of inactive agents or overly broad-spectrum antibiotics. Current antimicrobial stewardship initiatives recognize the importance of narrow-spectrum antibiotics to minimize risk of adverse events and selective pressure for antimicrobial resistance.

Although we attempted to confirm the identification of the anaerobic isolates via commercially available methods, it was not until we performed genetic testing that we were able verify the isolates as B fragilis. Furthermore, earlier susceptibility testing would have allowed for more narrow-targeted antimicrobial therapy and could have potentially prevented our patient’s readmission and use of ertapenem, despite its > 98% susceptibility rates against B fragilis.^13,17

All of the B fragilis isolates carried the cepA gene, which is a cephalosporinase that encodes for resistance to cephalosporins and aminopenicillins but not to ß-lactam ß-lactamase inhibitor combinations.¹³ Although not a substitution for susceptibility analysis, genetic testing showed that all of the isolates carried a nonsynonymous mutation from serine to a phenylalanine at amino acid position 82 (S82F) in the gyrA gene. The S82F mutation has been implicated in fluoroquinolone resistance, via inhibition of substrate–target recognition and binding between fluoroquinolones and the target topoisomerase protein,¹⁸ and may potentially explain why our patient clinically worsened while on moxifloxacin monotherapy. Although moxifloxacin susceptibility was not performed, susceptibility rates remain highly variable, ranging from 50% to 70% for B fragilis.^13,15,16

It is important to note that the metronidazole the patient received during his first hospital admission could have sterilized the vertebral body without completely eradicating the microbe; thus could explain his clinical worsening while on moxifloxacin monotherapy despite no growth from the repeat biopsy culture. Our rationale for initially continuing moxifloxacin was based on its excellent bioavailability and bone penetration properties. Additionally, of the fluoroquinolones it has the most reliable anaerobic activity and is the only one recommended as monotherapy for complicated intraabdominal infections.¹⁹ However, guidelines recommend avoiding its use in patients who have received a fluoroquinolone in the past 90 days or at institutions with high rates of resistance. At our institution Escherichia coli has a > 90% susceptibility rate to fluoroquinolones. Given this rate and our concern that the patient had a polymicrobial infection, we felt that moxifloxacin would provide appropriate anaerobic and aerobic coverage, especially since he had no previous fluoroquinolone exposure.

 

Additionally, none of the isolates carried the nim or bft toxin genes. Although the nim gene is associated with metronidazole resistance,its presence does not invariably result in resistant strains of B fragilis; in fact, metronidazole resistance is relatively uncommon, with the majority of B fragilis showing < 1% resistance, based on CLSI breakpoints (≥ 32 mg/L).^13,20,21 However, one recent epidemiologic study on anaerobic wound isolates from Iraq and Afghanistan casualties found that 12% (2/17) of B fragilis isolates were resistant to metronidazole.¹⁵ Given the improvement of the patient’s symptoms while on metronidazole, it is likely that the B fragilis was susceptible. Nevertheless, susceptibility testing with minimum inhibitory concentrations is necessary to verify this result. Also, although enterotoxigenic strains of B fragilis have been associated with bloodstream infections, our patient’s isolates lacked the 3 subtypes of B fragilis enterotoxin gene.²²

 

Conclusions

We report a case of B fragilis bacteremia and vertebral osteomyelitis complicated by challenges in anaerobic identification and sensitivities that led to brief use of a possibly inactive antimicrobial and the subsequent use of carbapenem therapy, which may have been avoided if susceptibility testing were more readily available. This case led to changes in our hospital’s processing of anaerobic isolates to include susceptibility testing on request.

Acknowledgments

We thank Keith Thompson, MD (staff pathologist, Naval Medical Center Portsmouth Virginia), for providing the pathology images from the initial vertebral biopsy, and Dr. Kate Hinkle (director, Multidrug-Resistant Organism Repository and Surveillance Network, Silver Spring, Maryland ) for providing the whole genome sequencing results from the B fragilis isolates.

Acute pyogenic vertebral osteomyelitis is often due to hematogenous spread of aerobic bacteria.^1-4 Conversely, only 0.5% of anaerobic bacteremias lead to osteomyelitis.⁵ Anaerobic osteomyelitis typically results from the contiguous spread of polymicrobial infections through breaks in the gut mucosal barrier and involves the vertebral bodies in only 2% to 5% of cases.^5,6 Although Bacteroides fragilis (B fragilis) is the most common anaerobic pathogen cultivated from blood, accounting for about half of all anaerobic blood isolates, it seldom leads to osteomyelitis.^1,2,7-11 We report an uncommon case of B fragilis bacteremia and vertebral osteomyelitis confounded by uncertainties in anaerobic identification and susceptibilities.

Case Presentation

A healthy-appearing male aged 55 years presented to the Naval Medical Center Portsmouth (NMCP) with subacute low back pain and fevers of 103 °F for > 3 weeks. While traveling 4 weeks prior, he completed a course of oseltamivir for influenza B infection; afterward, he was diagnosed with community-acquired pneumonia and treated with a dose of ceftriaxone and a 7-day course of doxycycline. The patient presented to the same facility a week later for low back pain and nonresolving respiratory symptoms, and his therapy was changed to azithromycin, cefuroxime, prednisone, and inhalers. Additionally, after being treated for influenza, he developed constipation and hematochezia for which he did not seek care. The hematochezia was similar to a previous episode from an anal fissure 1 year prior that resolved with stool softeners. When he was finally seen at NMCP after 3 weeks of worsening back pain and fevers, lumbosacral magnetic resonance imaging (MRI) demonstrated vertebral osteomyelitis and discitis at L4-L5 and admitted to the hospital (Figure 1).

After a fluoroscopy-guided biopsy of the L4 vertebral body on hospital day 1, the patient was started on cefepime and vancomycin. The biopsy sample was inoculated onto solid media (blood agar, chocolate agar, and MacConkey agar) and incubated at 36 °C for 24 hours in a 5% CO₂ atmosphere, as well as onto Shaedler agar with vitamin K and chopped meat glucose broth and incubated at 36 °C for 48 hours under anaerobic conditions. Metronidazole was added and vancomycin discontinued after 2 anaerobic blood culture vials obtained on hospital day 1, incubated in a Becton Dickinson BACTEC FX automated system, which demonstrated Gram-negative bacilli after 48 hours. The blood culture isolates demonstrated a > 99% probability of being identified as ß-lactamase positive Prevotella loescheii using Thermo Fischer Scientific RapID ANA II biochemical testing. Nitrocefinase discs were used to detect ß-lactamase activity.

The biopsy demonstrated nongranulomatous focal areas of necrotic bone and neutrophilia in a hematopoietic background consistent with acute osteomyelitis (Figure 2); on hospital day 4, ß-lactamase positive B fragilis grew from the bone culture. Additionally, 1 anaerobic vial from a surveillance blood culture set that was obtained on hospital day 3 grew ß-lactamasepositive B fragilis using the same identification methods. With these results he was thought to have a polymicrobial infection (B fragilis and Prevotella loescheii [P loescheii]) from a suspected bowel source based on his hematochezia and history of anal fissure. No aerobic, Gram-negative enterobacteriaceae were isolated, but he had previously been on cefuroxime, which has potential activity against these organisms, for ≥ 2 weeks prior to hospitalization and cultures. He was discharged on moxifloxacin and metronidazole pending final culture results, including requested anaerobic susceptibility testing.

At 1-week follow-up, both aerobic and anaerobic vials from surveillance blood cultures remained negative for any microbes, so antibiotics were deescalated to moxifloxacin monotherapy. However, after 3 days the patient was readmitted for increasing C-reactive protein (CRP) levels and intractable back pain with worsening bilateral radiculopathy. A repeat MRI demonstrated interval disease progression with near obliteration of the L4-L5 disc space and hyperenhancement of the prevertebral soft tissues and adjacent psoas musculature without focal rim-enhancing fluid collection (Figure 3). After repeat L4 biopsy, metronidazole was restarted and ertapenem added for enterobacteriaceae coverage, given the known B fragilis and potential suppression from previous cephalosporin therapy; moxifloxacin was discontinued. L4 biopsy cultures showed no growth, and CRP levels trended down from 154.2 mg/L (start of first admission) to 42.4 mg/L (start of second admission) to 14.9 mg/L (day of discharge) (reference range, 5-9.9 mg/L). He was discharged on ertapenem and metronidazole. He completed a 6-week course without further complication.

During antibiotic therapy he had an unremarkable colonoscopy, CRP normalized to 2.6 mg/L (reference range, 0-4.9 mg/L), and he underwent successful L4-L5 transforaminal lumbar interbody fusion 2 weeks after finishing antibiotics.

We retroactively sent both P loescheii isolates and the 1 B fragilis isolate that grew from the surveillance blood culture to the Multidrug-resistant Organism Repository and Surveillance Network (MRSN) at the Walter Reed Army Institute of Research for identification confirmation and susceptibility analysis. Whole genome sequencing with single nucleotide polymorphism (SNP)-based analysis revealed all isolates were 100% identical and consistent with B fragilis and not P loescheii, based on clustering around other B fragilis sequences found in the National Center for Biotechnology Information (NCBI) Genbank database (Figure 4). All isolates carried the antibiotic resistance genes— cepA, sul(2), tetQ— encoding for possible resistance to cephalosporins, sulphonamides, and tetracyclines, respectively; as well as a point mutation in the gyrA gene (Ser82Phe). None of the isolates carried the nim gene, and screening for the 3 subtypes of B fragilis enterotoxin gene (bft-1, bft-2, bft-3) was negative. Eventual susceptibility testing at the Mayo Clinic several months after the conclusion of the case indicated that the B fragilis isolate was sensitive to piperacillin-tazobactam, ertapenem, clindamycin, and metronidazole; however, testing was not performed against moxifloxacin.

 

Discussion

In the era of growing antibiotic resistance patterns, antimicrobial stewardship programs recommend interventions to improve antimicrobial use through targeted narrow- spectrum antibiotics.¹² The Clinical and Laboratory Standards Institute (CLSI) maintains guidelines on the major indications for anaerobic antimicrobial susceptibility testing (AST) to help direct narrow-targeted antimicrobial therapy. However, in a 2008 practice survey Goldstein and colleagues reported that less than half of US hospitals performed anaerobic AST, and only 21% of these facilities did it in-house, while the remainder sent out their isolates for testing.^11-14 The CLSI major indications for AST include situations in which the selection of agents is important because of the (1) known resistance of a particular species; (2) confirmation of appropriate therapy for severe infections or for those that may require long-term therapy; (3) persistence of infection despite adequate treatment with an appropriate therapeutic regimen; and (4) difficulty in making empirical decisions based on precedent.¹⁴ Additionally, isolates from brain abscess, endocarditis, osteomyelitis, joint infection, infection of prosthetic devices or vascular grafts, bacteremia, and normally sterile body sites (unless contamination suspected) should be tested.¹⁴

Because of the lack of anaerobic AST, health care providers must base empiric treatment on reported sensitivities from the medical literature. Empiric selection of antimicrobials for anaerobic infections is made even more challenging by the increased rates of resistance reported in the literature, leading to recommendations to increase susceptibility testing to guide therapy.^13,15,16 Empiric therapy of deep-seated anaerobic infections may lead to use of inactive agents or overly broad-spectrum antibiotics. Current antimicrobial stewardship initiatives recognize the importance of narrow-spectrum antibiotics to minimize risk of adverse events and selective pressure for antimicrobial resistance.

Although we attempted to confirm the identification of the anaerobic isolates via commercially available methods, it was not until we performed genetic testing that we were able verify the isolates as B fragilis. Furthermore, earlier susceptibility testing would have allowed for more narrow-targeted antimicrobial therapy and could have potentially prevented our patient’s readmission and use of ertapenem, despite its > 98% susceptibility rates against B fragilis.^13,17

All of the B fragilis isolates carried the cepA gene, which is a cephalosporinase that encodes for resistance to cephalosporins and aminopenicillins but not to ß-lactam ß-lactamase inhibitor combinations.¹³ Although not a substitution for susceptibility analysis, genetic testing showed that all of the isolates carried a nonsynonymous mutation from serine to a phenylalanine at amino acid position 82 (S82F) in the gyrA gene. The S82F mutation has been implicated in fluoroquinolone resistance, via inhibition of substrate–target recognition and binding between fluoroquinolones and the target topoisomerase protein,¹⁸ and may potentially explain why our patient clinically worsened while on moxifloxacin monotherapy. Although moxifloxacin susceptibility was not performed, susceptibility rates remain highly variable, ranging from 50% to 70% for B fragilis.^13,15,16

It is important to note that the metronidazole the patient received during his first hospital admission could have sterilized the vertebral body without completely eradicating the microbe; thus could explain his clinical worsening while on moxifloxacin monotherapy despite no growth from the repeat biopsy culture. Our rationale for initially continuing moxifloxacin was based on its excellent bioavailability and bone penetration properties. Additionally, of the fluoroquinolones it has the most reliable anaerobic activity and is the only one recommended as monotherapy for complicated intraabdominal infections.¹⁹ However, guidelines recommend avoiding its use in patients who have received a fluoroquinolone in the past 90 days or at institutions with high rates of resistance. At our institution Escherichia coli has a > 90% susceptibility rate to fluoroquinolones. Given this rate and our concern that the patient had a polymicrobial infection, we felt that moxifloxacin would provide appropriate anaerobic and aerobic coverage, especially since he had no previous fluoroquinolone exposure.

 

Additionally, none of the isolates carried the nim or bft toxin genes. Although the nim gene is associated with metronidazole resistance,its presence does not invariably result in resistant strains of B fragilis; in fact, metronidazole resistance is relatively uncommon, with the majority of B fragilis showing < 1% resistance, based on CLSI breakpoints (≥ 32 mg/L).^13,20,21 However, one recent epidemiologic study on anaerobic wound isolates from Iraq and Afghanistan casualties found that 12% (2/17) of B fragilis isolates were resistant to metronidazole.¹⁵ Given the improvement of the patient’s symptoms while on metronidazole, it is likely that the B fragilis was susceptible. Nevertheless, susceptibility testing with minimum inhibitory concentrations is necessary to verify this result. Also, although enterotoxigenic strains of B fragilis have been associated with bloodstream infections, our patient’s isolates lacked the 3 subtypes of B fragilis enterotoxin gene.²²

 

Conclusions

We report a case of B fragilis bacteremia and vertebral osteomyelitis complicated by challenges in anaerobic identification and sensitivities that led to brief use of a possibly inactive antimicrobial and the subsequent use of carbapenem therapy, which may have been avoided if susceptibility testing were more readily available. This case led to changes in our hospital’s processing of anaerobic isolates to include susceptibility testing on request.

Acknowledgments

We thank Keith Thompson, MD (staff pathologist, Naval Medical Center Portsmouth Virginia), for providing the pathology images from the initial vertebral biopsy, and Dr. Kate Hinkle (director, Multidrug-Resistant Organism Repository and Surveillance Network, Silver Spring, Maryland ) for providing the whole genome sequencing results from the B fragilis isolates.

Acute pyogenic vertebral osteomyelitis is often due to hematogenous spread of aerobic bacteria.^1-4 Conversely, only 0.5% of anaerobic bacteremias lead to osteomyelitis.⁵ Anaerobic osteomyelitis typically results from the contiguous spread of polymicrobial infections through breaks in the gut mucosal barrier and involves the vertebral bodies in only 2% to 5% of cases.^5,6 Although Bacteroides fragilis (B fragilis) is the most common anaerobic pathogen cultivated from blood, accounting for about half of all anaerobic blood isolates, it seldom leads to osteomyelitis.^1,2,7-11 We report an uncommon case of B fragilis bacteremia and vertebral osteomyelitis confounded by uncertainties in anaerobic identification and susceptibilities.

Case Presentation

A healthy-appearing male aged 55 years presented to the Naval Medical Center Portsmouth (NMCP) with subacute low back pain and fevers of 103 °F for > 3 weeks. While traveling 4 weeks prior, he completed a course of oseltamivir for influenza B infection; afterward, he was diagnosed with community-acquired pneumonia and treated with a dose of ceftriaxone and a 7-day course of doxycycline. The patient presented to the same facility a week later for low back pain and nonresolving respiratory symptoms, and his therapy was changed to azithromycin, cefuroxime, prednisone, and inhalers. Additionally, after being treated for influenza, he developed constipation and hematochezia for which he did not seek care. The hematochezia was similar to a previous episode from an anal fissure 1 year prior that resolved with stool softeners. When he was finally seen at NMCP after 3 weeks of worsening back pain and fevers, lumbosacral magnetic resonance imaging (MRI) demonstrated vertebral osteomyelitis and discitis at L4-L5 and admitted to the hospital (Figure 1).

After a fluoroscopy-guided biopsy of the L4 vertebral body on hospital day 1, the patient was started on cefepime and vancomycin. The biopsy sample was inoculated onto solid media (blood agar, chocolate agar, and MacConkey agar) and incubated at 36 °C for 24 hours in a 5% CO₂ atmosphere, as well as onto Shaedler agar with vitamin K and chopped meat glucose broth and incubated at 36 °C for 48 hours under anaerobic conditions. Metronidazole was added and vancomycin discontinued after 2 anaerobic blood culture vials obtained on hospital day 1, incubated in a Becton Dickinson BACTEC FX automated system, which demonstrated Gram-negative bacilli after 48 hours. The blood culture isolates demonstrated a > 99% probability of being identified as ß-lactamase positive Prevotella loescheii using Thermo Fischer Scientific RapID ANA II biochemical testing. Nitrocefinase discs were used to detect ß-lactamase activity.

The biopsy demonstrated nongranulomatous focal areas of necrotic bone and neutrophilia in a hematopoietic background consistent with acute osteomyelitis (Figure 2); on hospital day 4, ß-lactamase positive B fragilis grew from the bone culture. Additionally, 1 anaerobic vial from a surveillance blood culture set that was obtained on hospital day 3 grew ß-lactamasepositive B fragilis using the same identification methods. With these results he was thought to have a polymicrobial infection (B fragilis and Prevotella loescheii [P loescheii]) from a suspected bowel source based on his hematochezia and history of anal fissure. No aerobic, Gram-negative enterobacteriaceae were isolated, but he had previously been on cefuroxime, which has potential activity against these organisms, for ≥ 2 weeks prior to hospitalization and cultures. He was discharged on moxifloxacin and metronidazole pending final culture results, including requested anaerobic susceptibility testing.

At 1-week follow-up, both aerobic and anaerobic vials from surveillance blood cultures remained negative for any microbes, so antibiotics were deescalated to moxifloxacin monotherapy. However, after 3 days the patient was readmitted for increasing C-reactive protein (CRP) levels and intractable back pain with worsening bilateral radiculopathy. A repeat MRI demonstrated interval disease progression with near obliteration of the L4-L5 disc space and hyperenhancement of the prevertebral soft tissues and adjacent psoas musculature without focal rim-enhancing fluid collection (Figure 3). After repeat L4 biopsy, metronidazole was restarted and ertapenem added for enterobacteriaceae coverage, given the known B fragilis and potential suppression from previous cephalosporin therapy; moxifloxacin was discontinued. L4 biopsy cultures showed no growth, and CRP levels trended down from 154.2 mg/L (start of first admission) to 42.4 mg/L (start of second admission) to 14.9 mg/L (day of discharge) (reference range, 5-9.9 mg/L). He was discharged on ertapenem and metronidazole. He completed a 6-week course without further complication.

During antibiotic therapy he had an unremarkable colonoscopy, CRP normalized to 2.6 mg/L (reference range, 0-4.9 mg/L), and he underwent successful L4-L5 transforaminal lumbar interbody fusion 2 weeks after finishing antibiotics.

We retroactively sent both P loescheii isolates and the 1 B fragilis isolate that grew from the surveillance blood culture to the Multidrug-resistant Organism Repository and Surveillance Network (MRSN) at the Walter Reed Army Institute of Research for identification confirmation and susceptibility analysis. Whole genome sequencing with single nucleotide polymorphism (SNP)-based analysis revealed all isolates were 100% identical and consistent with B fragilis and not P loescheii, based on clustering around other B fragilis sequences found in the National Center for Biotechnology Information (NCBI) Genbank database (Figure 4). All isolates carried the antibiotic resistance genes— cepA, sul(2), tetQ— encoding for possible resistance to cephalosporins, sulphonamides, and tetracyclines, respectively; as well as a point mutation in the gyrA gene (Ser82Phe). None of the isolates carried the nim gene, and screening for the 3 subtypes of B fragilis enterotoxin gene (bft-1, bft-2, bft-3) was negative. Eventual susceptibility testing at the Mayo Clinic several months after the conclusion of the case indicated that the B fragilis isolate was sensitive to piperacillin-tazobactam, ertapenem, clindamycin, and metronidazole; however, testing was not performed against moxifloxacin.

 

Discussion

In the era of growing antibiotic resistance patterns, antimicrobial stewardship programs recommend interventions to improve antimicrobial use through targeted narrow- spectrum antibiotics.¹² The Clinical and Laboratory Standards Institute (CLSI) maintains guidelines on the major indications for anaerobic antimicrobial susceptibility testing (AST) to help direct narrow-targeted antimicrobial therapy. However, in a 2008 practice survey Goldstein and colleagues reported that less than half of US hospitals performed anaerobic AST, and only 21% of these facilities did it in-house, while the remainder sent out their isolates for testing.^11-14 The CLSI major indications for AST include situations in which the selection of agents is important because of the (1) known resistance of a particular species; (2) confirmation of appropriate therapy for severe infections or for those that may require long-term therapy; (3) persistence of infection despite adequate treatment with an appropriate therapeutic regimen; and (4) difficulty in making empirical decisions based on precedent.¹⁴ Additionally, isolates from brain abscess, endocarditis, osteomyelitis, joint infection, infection of prosthetic devices or vascular grafts, bacteremia, and normally sterile body sites (unless contamination suspected) should be tested.¹⁴

Because of the lack of anaerobic AST, health care providers must base empiric treatment on reported sensitivities from the medical literature. Empiric selection of antimicrobials for anaerobic infections is made even more challenging by the increased rates of resistance reported in the literature, leading to recommendations to increase susceptibility testing to guide therapy.^13,15,16 Empiric therapy of deep-seated anaerobic infections may lead to use of inactive agents or overly broad-spectrum antibiotics. Current antimicrobial stewardship initiatives recognize the importance of narrow-spectrum antibiotics to minimize risk of adverse events and selective pressure for antimicrobial resistance.

Although we attempted to confirm the identification of the anaerobic isolates via commercially available methods, it was not until we performed genetic testing that we were able verify the isolates as B fragilis. Furthermore, earlier susceptibility testing would have allowed for more narrow-targeted antimicrobial therapy and could have potentially prevented our patient’s readmission and use of ertapenem, despite its > 98% susceptibility rates against B fragilis.^13,17

All of the B fragilis isolates carried the cepA gene, which is a cephalosporinase that encodes for resistance to cephalosporins and aminopenicillins but not to ß-lactam ß-lactamase inhibitor combinations.¹³ Although not a substitution for susceptibility analysis, genetic testing showed that all of the isolates carried a nonsynonymous mutation from serine to a phenylalanine at amino acid position 82 (S82F) in the gyrA gene. The S82F mutation has been implicated in fluoroquinolone resistance, via inhibition of substrate–target recognition and binding between fluoroquinolones and the target topoisomerase protein,¹⁸ and may potentially explain why our patient clinically worsened while on moxifloxacin monotherapy. Although moxifloxacin susceptibility was not performed, susceptibility rates remain highly variable, ranging from 50% to 70% for B fragilis.^13,15,16

It is important to note that the metronidazole the patient received during his first hospital admission could have sterilized the vertebral body without completely eradicating the microbe; thus could explain his clinical worsening while on moxifloxacin monotherapy despite no growth from the repeat biopsy culture. Our rationale for initially continuing moxifloxacin was based on its excellent bioavailability and bone penetration properties. Additionally, of the fluoroquinolones it has the most reliable anaerobic activity and is the only one recommended as monotherapy for complicated intraabdominal infections.¹⁹ However, guidelines recommend avoiding its use in patients who have received a fluoroquinolone in the past 90 days or at institutions with high rates of resistance. At our institution Escherichia coli has a > 90% susceptibility rate to fluoroquinolones. Given this rate and our concern that the patient had a polymicrobial infection, we felt that moxifloxacin would provide appropriate anaerobic and aerobic coverage, especially since he had no previous fluoroquinolone exposure.

 

Additionally, none of the isolates carried the nim or bft toxin genes. Although the nim gene is associated with metronidazole resistance,its presence does not invariably result in resistant strains of B fragilis; in fact, metronidazole resistance is relatively uncommon, with the majority of B fragilis showing < 1% resistance, based on CLSI breakpoints (≥ 32 mg/L).^13,20,21 However, one recent epidemiologic study on anaerobic wound isolates from Iraq and Afghanistan casualties found that 12% (2/17) of B fragilis isolates were resistant to metronidazole.¹⁵ Given the improvement of the patient’s symptoms while on metronidazole, it is likely that the B fragilis was susceptible. Nevertheless, susceptibility testing with minimum inhibitory concentrations is necessary to verify this result. Also, although enterotoxigenic strains of B fragilis have been associated with bloodstream infections, our patient’s isolates lacked the 3 subtypes of B fragilis enterotoxin gene.²²

 

Conclusions

We report a case of B fragilis bacteremia and vertebral osteomyelitis complicated by challenges in anaerobic identification and sensitivities that led to brief use of a possibly inactive antimicrobial and the subsequent use of carbapenem therapy, which may have been avoided if susceptibility testing were more readily available. This case led to changes in our hospital’s processing of anaerobic isolates to include susceptibility testing on request.

Acknowledgments

We thank Keith Thompson, MD (staff pathologist, Naval Medical Center Portsmouth Virginia), for providing the pathology images from the initial vertebral biopsy, and Dr. Kate Hinkle (director, Multidrug-Resistant Organism Repository and Surveillance Network, Silver Spring, Maryland ) for providing the whole genome sequencing results from the B fragilis isolates.