C. difficile in HIV+ Patients: Risk Factors and Outcomes
C. difficile in HIV+ Patients: Risk Factors and Outcomes
We performed a retrospective cohort analysis of HIV-infected patients receiving longitudinal HIV primary care in the Johns Hopkins Hospital outpatient HIV clinic between 1 July 2003 and 31 December 2010. All patients who initiate care in the clinic for HIV primary care are offered enrolment in the Johns Hopkins HIV Clinical Cohort (98% acceptance rate).
Patients with CDI were identified from hospital electronic records. The medical records of each of these patients were then reviewed manually to confirm the laboratory result and to collect data on clinical presentation and disease course. Annual CDI incidence was computed using the number of initial cases per calendar year and person-years of follow-up in the cohort. A nested case–control study was performed with four HIV-infected controls with no known CDI matched to each case of CDI. Controls were individually matched on cohort enrolment date within 6 months, duration of follow-up within 6 months and CD4 cell count at cohort enrolment within 100 cells/μl.
To identify both inpatient and outpatient CDI cases, test results were extracted from a hospital electronic database. From 1 July 2003 through 9 May 2004, the stool C. difficile cell culture cytotoxicity neutralization assay was used for testing. From 10 May 2004 through 14 June 2009, stool samples were screened with the C. difficile glutamate dehydrogenase (GDH) antigen enzyme immunoassay (EIA) and positive samples were tested with the C. difficile cytotoxicity assay as previously described. Starting on 15 June 2009, stools screening positive with the C. difficile GDH antigen EIA were tested with the BD GeneOhm (BD Diagnostics, Sparks, Maryland, USA) PCR for C. difficile toxin B gene real-time PCR assay. From 1 January 2010 through the end of the study, stools were tested exclusively with the C. difficile toxin B PCR. The laboratory only accepted unformed stool for testing with the C. difficile toxin B PCR, but this requirement was not completely enforced for other CDI tests. No C. difficile strain identification data are collected by the laboratory.
An incident CDI case was defined as the first positive C. difficile test result during the study period in individuals with no known prior CDI. CDI testing results prior to the study period were obtained starting 1 January 1999 in order to exclude individuals with evidence of CDI prior to the study period from the incident case analyses.
Multiple aspects of clinical presentation were examined in the review of case records. Diarrhoea was defined as documented subjective patient or physician report of diarrhoea. Stools per day were recorded, where available. Bloody stools were defined as documented subjective patient or physician report of bloody stools. The diagnosis of colitis was based on colitis or signs of inflammation consistent with colitis on computerized tomography (CT) scan reports. Pseudomembranous colitis was defined as pseudomembranes reported on colonoscopy or pathology. White blood cell count (WBC) data were extracted from laboratory records as the laboratory value nearest to and at a maximum of 3 days from the date of the positive stool test result.
Healthcare exposure was defined according to the clinical practice guidelines for C. difficile infection in adults. Event date was defined as the date of the positive test result for cases. The same date was used as the event date for matched controls with approximated equivalent follow-up time as the corresponding case. Hospital onset-healthcare facility associated (HO-HCFA) exposure was defined as an event date in cases and matched control occurring from hospital day 3 through discharge. Community onset-healthcare facility associated (CO-HCFA) was defined as any inpatient, outpatient or skilled nursing facility exposure in the 12 weeks preceding the event date in cases and controls, including the first 3 days of hospitalization. Outpatient exposure was defined as any provider visits, physical therapy or documented nursing visit where treatment was provided. Healthcare exposure at outside facilities was included when documented in the available records. Community-associated CDI (CA) was defined as absence of healthcare facility exposure in the 12 weeks preceding the event date. Recurrent disease was defined as a positive CDI test or documented concern for recurrent CDI by the treating clinician prompting retreatment within 28 days after completion of the initial CDI treatment. Follow-up was defined as documented inpatient or outpatient visit or communication with clinical staff within 28 days after completion of CDI treatment.
We reviewed the outpatient clinic records and inpatient pharmacy order database to determine medications received in the 30 days prior to the date of CDI in cases and corresponding controls. Antibiotics were categorized into six groups on the basis of mechanism and antimicrobial spectrum: clindamycin, fluoroquinolones, penicillin derivatives with gram-negative activity, cephalosporins, trimethoprim-sulfamethoxazole (TMP-SMX) and macrolides. Fluoroquinolones included ciprofloxacin, moxifloxacin, gatifloxacin, norfloxacin and levofloxacin. Penicillin derivatives with gram-negative activity category (GN-PCN) included aminopenicillins and anti-pseudomonal penicillins. The cephalosporins category included any cephalosporin use. Macrolide use included azithromycin and erythromycin. Gastric acid suppressants included all histamine H2 blockers and proton pump inhibitors. Immunosuppressive agents included steroids, methotrexate, mycophenolate, calcineurin inhibitors, mammalian target of rapamycin (MTOR) inhibitors, mAbs and chemotherapeutic agents.
HIV transmission risk factors were IDU, MSM and heterosexual transmission. Individuals may have multiple HIV risk factors. Laboratory variables ascertained within 90 days before or 30 days after the CDI event included HIV-1 RNA, CD4 cell count and serum creatinine. Test results preceding the event were used when available. CD4 cell counts were divided into quartiles of 50 or less, 51–200, 201–350 and more than 350 cells/μl. Stages of chronic kidney disease (CKD) were defined according to the National Kidney Foundation/Kidney Disease Outcomes Quality Initiative (NKF/KDOQI) guidelines.
Statistical analyses were performed using STATA 12.1. Two-sided testing was used, with a P value of less than 0.05 considered significant. We used the Poisson distribution to calculate the standard error and 95% confidence interval (CI) for the incidence rate. Separate simple conditional logistic regression (SLR) analyses were performed to identify individual variables associated with the development of CDI. Conditional logistic regression models were used to avoid bias resulting from matching. Biologically plausible interactions between pairs of significant variables in the SLR analyses were tested. Variables included in the multiple conditional logistic regression (MLR) model were selected on the basis of clinical relevance and biologic plausibility. These variables were CD4 cell count category, log10 copies/ml HIV-1 RNA, HO-HCFA, CO-HCFA, clindamycin use, fluoroquinolone use, GN-PCN use, cephalosporin use, TMP-SMX use, macrolide use, use of gastric acid suppression, age categories, sex and CKD stage. We exponentiated the model coefficients to obtain the adjusted odds ratio (AOR) and corresponding 95% CI. Linear combination of estimates was used to make comparisons between nonreferent categories. Conditional logistic regression models estimate AOR, which approximates the relative risk given low CDI incidence and use of incidence density sampling. A log binomial model was used to estimate the relative risk of CDI recurrence by CD4 cell count category. Three sensitivity analyses of the MLR model were performed: including only individuals with diarrhoea documented in the medical record, including statistically significant and mechanistically plausible interaction terms, and including imputed missing data for CD4 cell count category and log10 HIV viral load. The multiple imputation utilized chained equations: ordinal logistic regression for CD4 cell count category and linear regression for log10 HIV viral load with MLR model covariates as predictors.
Materials and Methods
Study Design
We performed a retrospective cohort analysis of HIV-infected patients receiving longitudinal HIV primary care in the Johns Hopkins Hospital outpatient HIV clinic between 1 July 2003 and 31 December 2010. All patients who initiate care in the clinic for HIV primary care are offered enrolment in the Johns Hopkins HIV Clinical Cohort (98% acceptance rate).
Patients with CDI were identified from hospital electronic records. The medical records of each of these patients were then reviewed manually to confirm the laboratory result and to collect data on clinical presentation and disease course. Annual CDI incidence was computed using the number of initial cases per calendar year and person-years of follow-up in the cohort. A nested case–control study was performed with four HIV-infected controls with no known CDI matched to each case of CDI. Controls were individually matched on cohort enrolment date within 6 months, duration of follow-up within 6 months and CD4 cell count at cohort enrolment within 100 cells/μl.
Clostridium Difficile Tests
To identify both inpatient and outpatient CDI cases, test results were extracted from a hospital electronic database. From 1 July 2003 through 9 May 2004, the stool C. difficile cell culture cytotoxicity neutralization assay was used for testing. From 10 May 2004 through 14 June 2009, stool samples were screened with the C. difficile glutamate dehydrogenase (GDH) antigen enzyme immunoassay (EIA) and positive samples were tested with the C. difficile cytotoxicity assay as previously described. Starting on 15 June 2009, stools screening positive with the C. difficile GDH antigen EIA were tested with the BD GeneOhm (BD Diagnostics, Sparks, Maryland, USA) PCR for C. difficile toxin B gene real-time PCR assay. From 1 January 2010 through the end of the study, stools were tested exclusively with the C. difficile toxin B PCR. The laboratory only accepted unformed stool for testing with the C. difficile toxin B PCR, but this requirement was not completely enforced for other CDI tests. No C. difficile strain identification data are collected by the laboratory.
Definitions
An incident CDI case was defined as the first positive C. difficile test result during the study period in individuals with no known prior CDI. CDI testing results prior to the study period were obtained starting 1 January 1999 in order to exclude individuals with evidence of CDI prior to the study period from the incident case analyses.
Multiple aspects of clinical presentation were examined in the review of case records. Diarrhoea was defined as documented subjective patient or physician report of diarrhoea. Stools per day were recorded, where available. Bloody stools were defined as documented subjective patient or physician report of bloody stools. The diagnosis of colitis was based on colitis or signs of inflammation consistent with colitis on computerized tomography (CT) scan reports. Pseudomembranous colitis was defined as pseudomembranes reported on colonoscopy or pathology. White blood cell count (WBC) data were extracted from laboratory records as the laboratory value nearest to and at a maximum of 3 days from the date of the positive stool test result.
Healthcare exposure was defined according to the clinical practice guidelines for C. difficile infection in adults. Event date was defined as the date of the positive test result for cases. The same date was used as the event date for matched controls with approximated equivalent follow-up time as the corresponding case. Hospital onset-healthcare facility associated (HO-HCFA) exposure was defined as an event date in cases and matched control occurring from hospital day 3 through discharge. Community onset-healthcare facility associated (CO-HCFA) was defined as any inpatient, outpatient or skilled nursing facility exposure in the 12 weeks preceding the event date in cases and controls, including the first 3 days of hospitalization. Outpatient exposure was defined as any provider visits, physical therapy or documented nursing visit where treatment was provided. Healthcare exposure at outside facilities was included when documented in the available records. Community-associated CDI (CA) was defined as absence of healthcare facility exposure in the 12 weeks preceding the event date. Recurrent disease was defined as a positive CDI test or documented concern for recurrent CDI by the treating clinician prompting retreatment within 28 days after completion of the initial CDI treatment. Follow-up was defined as documented inpatient or outpatient visit or communication with clinical staff within 28 days after completion of CDI treatment.
We reviewed the outpatient clinic records and inpatient pharmacy order database to determine medications received in the 30 days prior to the date of CDI in cases and corresponding controls. Antibiotics were categorized into six groups on the basis of mechanism and antimicrobial spectrum: clindamycin, fluoroquinolones, penicillin derivatives with gram-negative activity, cephalosporins, trimethoprim-sulfamethoxazole (TMP-SMX) and macrolides. Fluoroquinolones included ciprofloxacin, moxifloxacin, gatifloxacin, norfloxacin and levofloxacin. Penicillin derivatives with gram-negative activity category (GN-PCN) included aminopenicillins and anti-pseudomonal penicillins. The cephalosporins category included any cephalosporin use. Macrolide use included azithromycin and erythromycin. Gastric acid suppressants included all histamine H2 blockers and proton pump inhibitors. Immunosuppressive agents included steroids, methotrexate, mycophenolate, calcineurin inhibitors, mammalian target of rapamycin (MTOR) inhibitors, mAbs and chemotherapeutic agents.
Statistical Analysis
HIV transmission risk factors were IDU, MSM and heterosexual transmission. Individuals may have multiple HIV risk factors. Laboratory variables ascertained within 90 days before or 30 days after the CDI event included HIV-1 RNA, CD4 cell count and serum creatinine. Test results preceding the event were used when available. CD4 cell counts were divided into quartiles of 50 or less, 51–200, 201–350 and more than 350 cells/μl. Stages of chronic kidney disease (CKD) were defined according to the National Kidney Foundation/Kidney Disease Outcomes Quality Initiative (NKF/KDOQI) guidelines.
Statistical analyses were performed using STATA 12.1. Two-sided testing was used, with a P value of less than 0.05 considered significant. We used the Poisson distribution to calculate the standard error and 95% confidence interval (CI) for the incidence rate. Separate simple conditional logistic regression (SLR) analyses were performed to identify individual variables associated with the development of CDI. Conditional logistic regression models were used to avoid bias resulting from matching. Biologically plausible interactions between pairs of significant variables in the SLR analyses were tested. Variables included in the multiple conditional logistic regression (MLR) model were selected on the basis of clinical relevance and biologic plausibility. These variables were CD4 cell count category, log10 copies/ml HIV-1 RNA, HO-HCFA, CO-HCFA, clindamycin use, fluoroquinolone use, GN-PCN use, cephalosporin use, TMP-SMX use, macrolide use, use of gastric acid suppression, age categories, sex and CKD stage. We exponentiated the model coefficients to obtain the adjusted odds ratio (AOR) and corresponding 95% CI. Linear combination of estimates was used to make comparisons between nonreferent categories. Conditional logistic regression models estimate AOR, which approximates the relative risk given low CDI incidence and use of incidence density sampling. A log binomial model was used to estimate the relative risk of CDI recurrence by CD4 cell count category. Three sensitivity analyses of the MLR model were performed: including only individuals with diarrhoea documented in the medical record, including statistically significant and mechanistically plausible interaction terms, and including imputed missing data for CD4 cell count category and log10 HIV viral load. The multiple imputation utilized chained equations: ordinal logistic regression for CD4 cell count category and linear regression for log10 HIV viral load with MLR model covariates as predictors.
Source...