The association between hypervirulent ribotype and severe Clostridium difficile infection was nonsignificant after adjustment for microbiologic, epidemiologic, and laboratory variables. This nonsignificant association was further validated using an independent data set.
Abstract
Background. Studies of Clostridium difficile outbreaks suggested that certain ribotypes (eg, 027 and 078) cause more severe disease than other ribotypes. A growing number of studies challenge the validity of this hypothesis.
Methods. We conducted a cross-sectional study of C. difficile infection (CDI) to test whether ribotype predicted clinical severity when adjusted for the influence of other predictors. Toxigenic C. difficile isolates were cultured from stool samples, screened for genes encoding virulence factors by polymerase chain reaction (PCR) and ribotyped using high-throughput, fluorescent PCR ribotyping. We collected data for 15 covariates (microbiologic, epidemiologic, and laboratory variables) and determined their individual and cumulative influence on the association between C. difficile ribotype and severe disease. We then validated this influence using an independent data set.
Results. A total of 34 severe CDI cases were identified among 310 independent cases of disease (11.0%). Eleven covariates, including C. difficile ribotype, were significant predictors of severe CDI in unadjusted analysis. However, the association between ribotypes 027 and 078 and severe CDI was not significant after adjustment for any of the other covariates. After full adjustment, severe cases were significantly predicted only by patients' white blood cell count and albumin level. This result was supported by analysis of a validation data set containing 433 independent CDI cases (45 severe cases; 10.4%).
Conclusions. Ribotype is not a significant predictor of severe CDI when adjusted for the influence of any other variables separately or in combination. White blood cell count and albumin level are the most clinically relevant predictors of severe CDI cases.
(See the Editorial Commentary by Barbut and Rupnik, on pages 1669–72.)
An increase in the incidence and severity of Clostridium difficile infection (CDI) throughout the United States, Canada, and Europe coincided with the emergence of a previously rare genotype [1]. This genotype, known as polymerase chain reaction (PCR) ribotype 027, North American Pulsed-field type 1 (NAP1), or restriction endonuclease analysis (REA) type BI, was reported to harbor an intrinsic ability to cause more severe disease compared to other pathogenic isolates [1, 2]. At least one other C. difficile lineage, ribotype 078, has been referred to as “hypervirulent” [3]. Laboratory studies identified numerous microbiologic properties to explain the increased virulence of 027 and/or 078 isolates, including antibiotic resistance [4], increased toxin production [5], enhanced ability for toxin B isoforms to bind target cells [6], and increased sporulation ability [7].
More recent data do not support the hypervirulent hypothesis [8–12], although the clinical definition of severe CDI or the methods used for data analysis are not consistent across all studies. Because institutions are often limited to retrospective review of patient records, it is important to define severe CDI using commonly recorded information. Such a definition has been recommended [13] but not universally used.
We sought to quantify the prevalence of C. difficile ribotypes at a single institution and to determine whether specific ribotypes were associated with severe disease. In particular, we tested the hypothesis that C. difficile ribotype predicts severe CDI cases even after adjustment for other clinical and laboratory variables. To do so, we developed models using an initial (derivation) data set and then validated our results with the same model fitted to a validation data set.
MATERIALS AND METHOD
Setting
The University of Michigan Health System (UMHS) includes a 930-bed, tertiary care inpatient facility and 5 off-site ambulatory care facilities. This study was approved by the University of Michigan Institutional Review Board.
Clinical Epidemiology
Severe CDI was defined as recommended by McDonald et al (intensive care unit admission, interventional surgery, or death within 30 days of diagnosis) [13]. We tested whether microbiologic, epidemiologic, and laboratory factors were predictive of severe CDI. Microbiologic factors included CDI cases caused by isolates of ribotypes 027 and 078–126 (hereafter referred to as 078) and/or those carrying previously recognized virulence factors (binary toxin and insertions/deletions, premature stop codons, or deletions of the tcdC gene). Epidemiologic factors included patient age, sex, and the setting of CDI onset or CDI surveillance definition [13] (healthcare facility-onset, healthcare facility-associated [HO-HCFA]; community-onset, healthcare facility associated [CO-HCFA]; community-associated [CA]; or indeterminate [IND]). Other epidemiologic factors included comorbid conditions as defined by the Charlson comorbidity index (CCI) [14]. Renal disease and a history of cancer were defined using the CCI and combined with age to create a modified age, renal function, and history of cancer (ARC) comorbidity score [15]. Laboratory values were recorded for white blood cell count, albumin level, creatinine, hematocrit, platelet count, total bilirubin, and blood urea nitrogen. Leukocytosis and leukopenia were defined as >12 000 and <4000 cells/mL, respectively. All laboratory values were collected within 72 hours of CDI diagnosis.
Two independent data sets were generated for this study. An initial data set of 310 cases of CDI (derivation set) was used for preliminary unadjusted and adjusted analyses and to derive a final model (see data analysis section below). A second data set (validation set) of 433 cases of CDI was used to validate the model fitted to the derivation data set. Data for each set were collected in the same manner.
Stool Samples and C. difficile Isolates
Suspected CDI-positive stool samples, defined as clinician-ordered specimens that were submitted for testing for the presence of toxigenic C. difficile, from hospitalized inpatients and ambulatory outpatients presenting to the main hospital or UMHS off-site facilities, were obtained from the Clinical Microbiology Laboratory between 14 January 2010 and 2 March 2011 (derivation set) and between 2 March 2011 and 5 March 2012 (validation set). No asymptomatic colonization cases were included.
Samples were cultured anaerobically on taurocholate-cycloserine-cefoxitin-fructose agar at 37°C [16]. A single colony was then subcultured in brain-heart infusion broth [16]. Aliquots were diluted in sterile water (UltraPure Distilled Water, Invitrogen) and used for PCR and ribotyping. The presence of other microbial causes of colitis or gastroenteritis was not evaluated.
Taxonomic and Toxigenic Verification
C. difficile–specific 16S ribosomal RNA (rRNA)-encoding gene PCR was used to verify isolate taxonomy [17]. A 5-plex PCR assay was also used to screen for the presence of an independent C. difficile–specific 16S locus along with C. difficile toxin (tcdA and tcdB) and binary toxin (cdtA and cdtB) genes [18]. Only tcdA- and/or tcdB-positive C. difficile isolates were analyzed. Toxin gene negative isolates were confirmed to be nontoxigenic using a Vero-cell cytotoxicity assay [19]. Published primers were used to amplify and sequence the tcdC gene [20]. Sequences were obtained in both directions and aligned/edited using SeqMan Pro (DNASTAR Lasergene 8.1.5, DNASTAR, Inc, Madison, Wisconsin). The MEGA5 program [21] was then used to identify tcdC mutations according to published alleles [22].
Fluorescent PCR Ribotyping
PCR ribotyping primers [23] were synthesized with a fluorescent label (Integrated DNA Technologies, Inc) and adjusted to 10 pmol/µL. A 25 µL PCR was performed using AmpliTaq Gold DNA Polymerase (Applied Biosystems) and the following conditions: 95°C (10 minutes); 35 cycles of 95°C (30 seconds), 55°C (30 seconds), and 72°C (1 minutes 30 seconds); final extension of 72°C (10 minutes). Amplicons were analyzed using an ABI3730xl DNA Analyzer and MapMaker 1000 ROX DNA sizing standard (BioVentures, Inc).
Reference strains included REA types BI-1, J-4, K-19, G-6, Y-6, CF-3, and BK-2 (PCR ribotypes 027, 001, 053–163, 002, 014–020, 017, and 078, respectively). Isolates of hyphenated ribotypes cluster into the same REA groups [24], so it is unclear whether they represent distinct genotypes. A similar ribotyping procedure was developed elsewhere [25]. Important differences are highlighted as Supplemental Data.
Data Analysis
Unconditional logistic regression was used to test our main hypothesis and to assess the predictive strength of factors discussed in the Clinical Epidemiology section above on severe CDI. Unadjusted and adjusted analyses were conducted with the glm() function using R software [26]. To generate models, only significant covariates on unadjusted analysis were considered. In addition, we performed likelihood ratio tests to evaluate significance for the effects of covariates and to develop a final model. Two-tailed level of significance was set at .05. Per our a priori hypothesis, ribotype was included in all models, and odds ratios were calculated for both the derivation and validation data sets. The rationale for deriving and validating a model was to assess the reproducibility of the results. In secondary analyses, the data were combined to minimize type 2 error, and a final model was fitted in the adjusted analysis. All covariates, regardless of their significance in the unadjusted analysis, were included in the model of the combined data set.
RESULTS
Inclusion of C. difficile Isolates and CDI Cases
We identified 331 C. difficile isolates obtained from symptomatic patients during the initial study period. Nontoxigenic isolates were excluded (n = 12 [4%]) along with samples representing repeat testing of the same patient (n = 9 [3%]). In total, 310 isolates from 310 different patients were considered for the derivation set. For the validation set we identified 460 additional C. difficile isolates in the second study period, of which 27 were excluded because they were nontoxigenic (n = 12 [3%]) or represented repeat sampling of the same patient (n = 15 [3%]). In total, 433 C. difficile isolates from 433 different patients were considered in the validation set. Patient demographics were similar in each data set (Table 1).
Table 1.
Clinical Characteristics of Patients With Clostridium difficile Infection in Derivation and Validation Data Sets
| Characteristic | Derivation Data Set (n = 310) | Validation Data Set (n = 433) |
|---|---|---|
| Age | ||
| Total No. of cases | 310 | 433 |
| Median | 57 y | 57 y |
| Range | 6 mo to 93 y | 6 mo to 95 y |
| Sex | ||
| Total No. of cases | 310 | 433 |
| Male | ||
| No. (%) | 153 (49.4) | 191 (44.1) |
| Female | ||
| No. (%) | 157 (50.6) | 242 (55.9) |
| CDI surveillance definition | ||
| Total No. of cases | 309 | 433 |
| HO-HCFA | ||
| No. (%) | 120 (38.7) | 183 (42.3) |
| CO-HCFA | ||
| No. (%) | 58 (18.7) | 84 (19.4) |
| IND | ||
| No. (%) | 42 (13.5) | 38 (8.8) |
| CA | ||
| No. (%) | 89 (28.7) | 128 (29.6) |
| Unknown | ||
| No. (%) | 1 (0.3) | 0 (0.0) |
| Lab values | ||
| White blood cell count (×1000 cells/mL) | ||
| Total No. of cases | 234 | 334 |
| Median (range) | 8.5 (0.1–58.5) | 8.0 (0.1–54.1) |
| Albumin level (g/dL) | ||
| Total No. of cases | 180 | 261 |
| Median (range) | 3.3 (1.8–4.9) | 3.3 (1.4–5.0) |
| Creatinine level (mg/dL) | ||
| Total No. of cases | 233 | 334 |
| Median (range) | 0.8 (0.1–9.0) | 0.8 (0.1–10.7) |
| Hematocrit level (%) | ||
| Total No. of cases | 234 | 334 |
| Median (range) | 30.8 (2.3–47.5) | 30.3 (18.8–51.6) |
| Platelet count (×1000 cells/mL) | ||
| Total No. of cases | 233 | 332 |
| Median (range) | 200.0 (10.0–1037.0) | 220 (7.0–951.0) |
| Total bilirubin level (mg/dL) | ||
| Total No. of cases | 172 | 250 |
| Median (range) | 0.5 (0.1–30.7) | 0.4 (0.1–27.1) |
| Urea nitrogen level (mg/dL) | ||
| Total No. of cases | 233 | 334 |
| Median (range) | 15.0 (0.5–158.0) | 16.0 (0.5–83.0) |
| Comorbidity scores | ||
| CCI | ||
| Total No. of cases | 280 | ND |
| Median (range) | 3.0 (0.0–10.0) | ND |
| ARC | ||
| Total No. of cases | 278 | ND |
| Median (range) | 2.0 (0.0–8.0) | ND |
Abbreviations: ARC, age, renal function, and history of cancer comorbidity score; CA, community acquired; CCI, Charlson comorbidity index; CDI, Clostridium difficile infection; CO-HCFA, community onset-health care facility associated; HO-HCFA, hospital onset-health care facility associated; IND, indeterminate; ND, not determined.
Ribotype Abundance
The 310 C. difficile isolates of the derivation set belonged to 75 distinct ribotypes (Figure 1A). The most common ribotypes observed were 014–020 (17%) and 027 (14%). Other ribotypes accounted for <20 cases (7%) each. Similarly, 91 distinct ribotypes were observed in the validation set (Figure 1B) and 014–020 and 027 isolates were again the most common (34%). These results illustrate that the C. difficile population at our institution was relatively stable over the sampling period, diverse, and not dominated by a single ribotype.
Figure 1.
Ribotype abundance and severe Clostridium difficile infection in the (A) derivation and (B) validation data sets. This bar chart shows the abundance of common (>3 cases) ribotypes identified in this study. Each bar represents the No. of severe (gray) and nonsevere (black) cases.
Predictors of Severe CDI
Unadjusted logistic regression implied that 2 microbiologic, 3 epidemiologic, and 6 laboratory covariates were significant predictors of severe CDI (Table 2). We then adjusted for covariates separately to determine their influence on the effect of C. difficile ribotype (Table 3). Each of the models resulted in a decreased adjusted odds ratio for C. difficile ribotype and a nonsignificant association with severe CDI.
Table 2.
Odds Ratios for Predictors of Severe Clostridium difficile Infection Based on Unadjusted Analysis of the Derivation Data Set (n = 310)
| Predictorsa | OR (95% CI) | P Value |
|---|---|---|
| Microbiologic | ||
| Hypervirulent ribotype | ||
| 027/078 vs other (reference) | 2.33 (1.03–5.02) | .035 |
| non-027/078 (reference) | ||
| cdtAB (binary toxin) vs | 1.66 (.75–3.53) | .195 |
| tcdC allele with any deletion vs full length (reference) | 2.64 (1.24–5.60) | .011 |
| Epidemiologic | ||
| Age (years) | 1.02 (1.00–1.04) | .047 |
| Sex | ||
| Male vs female (reference) | 0.90 (.44–1.84) | .778 |
| CDI surveillance definition | ||
| HO-HCFA | 23.16 (4.75–417.99) | .003 |
| CO-HCFA | 8.30 (1.29–161.36) | .056 |
| IND | 6.77 (.84–139.21) | .102 |
| CA (reference) | ||
| Charleson comorbidity index | 1.19 (1.03–1.38) | .022 |
| ARC score | 1.19 (.99–1.42) | .060 |
| Laboratory | ||
| White blood cell count | ||
| Abnormal (>12 000 or <4000) vs | 2.89 (1.36–6.46) | .007 |
| normal (reference) | ||
| Albumin levels (g/dL) | 0.22 (.10–0.45) | <.001 |
| Creatinine levels (mg/dL) | 1.22 (.97–1.52) | .072 |
| Hematocrit levels (%) | 0.88 (.81–.94) | <.001 |
| Platelet count (×1000/µL) | 0.995 (.992–.999) | .014 |
| Total bilirubin (mg/dL) | 1.16 (1.06–1.33) | .005 |
| Urea nitrogen levels (mmol/L) | 1.018 (1.005–1.032) | .007 |
Abbreviations: ARC, age, renal function, and history of cancer comorbidity score; CA, community acquired; CDI, Clostridium difficile infection; CI, confidence interval; CO-HCFA, community-onset, healthcare facility associated; HO-HCFA, healthcare facility-onset, healthcare facility-associated; IND, indeterminate; OR, odds ratio.
a Significant (α < 0.05) predictors are indicated with bold values.
Table 3.
Odds Ratios for the Association Between Hypervirulent Ribotype (027/078) and Severe Clostridium difficile Infection Based on Adjusted Logistic Regression Models Fitted to Derivation Data Set (n = 310)
| Effect of Hypervirulent Ribotypea | OR (95% CI) | P Value |
|---|---|---|
| Unadjusted | 2.33 (1.03–5.02) | .035 |
| Adjusted for the following covariates: | ||
| tcdC deletion | 1.26 (.38–4.94) | .722 |
| Age | 2.11 (.92–4.59) | .065 |
| CDI surveillance definition | 1.94 (.83–4.33) | .111 |
| Charleson comorbidity index | 2.12 (.92–4.66) | .068 |
| White blood cell count | 1.90 (.82–4.24) | .123 |
| Albumin levels (g/dL) | 1.57 (.53–4.28) | .394 |
| Hematocrit levels (%) | 2.23 (.94–5.10) | .060 |
| Platelet count (×1000/µL) | 2.13 (.92–4.78) | .070 |
| Total bilirubin (mg/dL) | 1.62 (.52–4.48) | .376 |
| Urea nitrogen levels (mmol/L) | 1.88 (.81–4.19) | .129 |
Abbreviations: CDI, C. difficile infection; CI, confidence interval; OR, odds ratio.
a All odds ratios compare 027/078 versus non-027/078 (reference) ribotypes.
In the fully adjusted analysis (Table 4), we adhered to the following strategy. First, we developed a “full” model consisting of the 11 significant covariates in unadjusted analysis. Because of missing data, primarily for laboratory tests, this model was fit to complete data for 148 cases (48%) in the derivation data set. Covariates were evaluated for their significance relative to a model where they were excluded in step-wise fashion (see Methods). C. difficile ribotype was included as a covariate, regardless of significance, to assess whether its effect was influenced by adjusting for other covariates. This analysis implied that the effect of C. difficile ribotype is not significant after adjusting for the effects of white blood cell count (leukocytosis/leukopenia) and albumin level. Results for this reduced model are shown in Table 4. In a second step of validation, we fit the reduced model to complete data for 244 cases (56%) in the validation data set (Table 4). This analysis again supported the hypothesis that adjustment for other factors attenuates the association between hypervirulent ribotypes and severe CDI and again implied that white blood cell count (leukocytosis/leukopenia) and albumin level are significant predictors of a severe clinical outcome.
Table 4.
Odds Ratios for the Association Between Hypervirulent Ribotype (027/078) and Severe Clostridium difficile Infection in the Final Model (Adjusted Analysis) Fitted to the Derivation and Validation Data Sets
| Predictora | Derivation OR (95% CI) | P Value | Validation OR (95% CI) | P Value |
|---|---|---|---|---|
| Hypervirulent ribotype: | ||||
| 027/078 vs non-027/078 (reference) | 0.82 (.07–10.0) | .874 | 1.34 (.53–3.16) | .516 |
| White blood cell count: | ||||
| Leukocytosis (>12 000 cells/mL) or leukopenia | ||||
| (<4000 cells/mL) vs normal (reference) | 4.27 (1.14–19.46) | .041 | 2.32 (1.07–5.18) | .035 |
| Albumin level (g/dL) | 0.25 (.07–.77) | .025 | 0.47 (.25–.87) | .018 |
Abbreviations: CI, confidence interval; OR, odds ratio.
a Significant values are shown in bold.
Complete information was available regarding age, sex, and CDI surveillance definition (HO-HCFA, etc) for all but one patient in the derivation data set and all patients in the validation data set. We were unable to obtain complete laboratory data on all patients because the decision of which tests to order was made during the course of routine clinical care by the clinician at the time of diagnosis. To assess the association between hypervirulent ribotype and severe CDI with the most complete data (n = 309 and 433), we adjusted for only age, sex, and CDI surveillance definition in a secondary analysis. This yielded nonsignificant adjusted odds ratios for hypervirulent ribotype and severe CDI of 1.92 (95% confidence interval [CI]: .82–4.32) in the derivation set and 1.73 (95% CI: .83–3.48) in the validation set. This analysis also yielded a significant adjusted odds ratio for HO-HCFA (vs CA; CDI surveillance definition) and severe CDI of 19.96 (4.03–361.83) in the derivation data set, although it did not reach significance in the validation data set.
Unadjusted and adjusted analyses of a combined data set (derivation + validation) are shown in supplemental Table 1. These analyses yielded a nonsignificant odds ratio (OR, 1.06; 95% CI: .49–2.20) for C. difficile ribotype upon adjustment for other covariates. White blood cell count and albumin were significant predictors after adjustment, as was a marginal increase in blood urea nitrogen level (OR, 1.02; 95% CI: 1.01–1.04).
DISCUSSION
The global incidence of CDI has increased markedly [27, 28] and coincided with the emergence of a previously less common and sporadically encountered C. difficile genotype, known as ribotype 027 [29]. An increase in the severity of CDI cases was reported in Quebec, where the 30-day CDI-associated mortality rose from 4.7% in 1991–1992 to 13.8% in 2003 [1]. When patients were examined in 2004, 12 of 13 cases (92%) were found to be caused by 027 isolates [1]. Similarly, a CDI outbreak investigation of US healthcare facilities between 2001 and 2003 found ≥50% of isolates at 5 of 8 facilities belonged to ribotype 027 [2]. However, it was unclear from the clinical data whether these isolates necessarily caused more severe disease. For example, results from 2 facilities suggested that 027 isolates were hypervirulent, but results from 2 other facilities did not [2]. Evidence that ribotype 078 isolates are hypervirulent [3] is similarly inconclusive. For example, cases of severe diarrhea caused by 078 isolates were more common than cases caused by non-078 isolates [3]. However, this observation was marginally significant at an α level of 0.10.
A number of hypotheses have been proposed to explain the apparent increase in virulence of 027 and 078 isolates. It was suggested that elevated production of C. difficile toxins (TcdA and TcdB) due to a dysfunctional negative regulatory protein (TcdC) was important [2]. However, recent evidence suggests that 027 isolates, at least, do not necessarily produce more toxins in vitro than isolates of other lineages [7]. Similarly, the functional status of TcdC based on nucleotide sequencing does not necessarily correlate with disease severity [9, 12]. Our results offer additional support to these reports, as we found no association between mutations in tcdC and severe disease after adjusting for other covariates.
A 2-component binary toxin (CdtA/CdtB) has also been implicated in CDI severity [2, 30], although its role in disease pathogenesis is uncertain and binary toxin-carrying isolates that do not express TcdA/TcdB have been isolated from asymptomatic patients [31]. A recent study failed to identify an association between binary toxin and case severity [9]. Our results similarly suggest that binary toxin is not involved in the development of a severe clinical outcome within 30 days of diagnosis.
In general, our results support the findings of at least 5 other studies that failed to detect a ribotype association with CDI case severity [8, 10, 11, 32, 33]. Notably, all of these studies used different criteria to define cases of severe CDI; they considered the 027 ribotype only; and only 2 of them used a similar logistic regression approach where adjustments were made for the influence of other covariates [8, 10]. The most similar severity definition to ours [8] considered the same clinical measures (intensive care unit admission, interventional surgery, or death) but was based on 60-day postdiagnosis and not 30 days as recommended by McDonald et al [13]. We recognize that the McDonald et al definition is not without faults. For example, we only considered outcomes where C. difficile was listed in patients' charts as a contributing factor, and it is possible that CDI was not the attributable cause of all of these outcomes. Similarly, we cannot rule out the possibility that severe cases were missed due to omission of contributing factors in patient charts. In light of these limitations, it remains possible that a link between ribotype and severity will be found if alternative definitions are used.
Our results illustrate the importance of adjustment for the effects of covariates when considering the clinical importance of pathogenic C. difficile ribotypes. Isolate characterization at the ribotype level is time consuming and costly compared to evaluating laboratory values that are often routinely collected on patients at risk of severe CDI. In addition, other adjusted analyses of CDI cases support our hypothesis that an abnormal white blood cell count [8, 10] and hypoalbuminemia [34] are significant predictors of a poor clinical outcome even after adjustment for other covariates.
An advantage of our study design is a validated model, which suggested reliability of the results. Moreover, when the data were combined to minimize type 2 error, the main finding of no significance remained for hypervirulent ribotype. A limitation of the models presented here is that they were generated from CDI cases at a single institution. Studies that incorporate broader geographic, temporal, and socioeconomic heterogeneity are needed to assess the generalizability of the results.
The percentage of HO-HCFA cases in this study is somewhat lower than other reports. For example, the mean percentages of HO-HCFA and CA cases in a 5-institution study of 6906 cases were 65.5% and 7.6%, respectively [35]. The percentages observed in our study were quite different (HO-HCFA = 38.7% in the derivation set and 42.3% in the validation set; CA = 28.7% in the derivation set and 29.6% in the validation set). A number of differences between our institution and the others may account for this discrepancy, including differences in clinical practice (eg, more patients treated in the outpatient setting), the number of outpatient clinics that are serviced by our diagnostic laboratory (eg, more outpatient clinics), regional differences in patient populations (eg, northern vs central US), and the influence of time when the studies took place (ie, 2000–2006 vs 2010–2012). However, a significant proportion of CDI cases occurred in the community in a recent US population-based cohort study [36]. It is tempting to speculate that HO-HCFA cases are decreasing while CA cases are increasing, although more data are needed to verify this trend.
It is interesting that patient CDI surveillance definition was not a significant predictor in the final model. This model was based on less than half of all cases (48%), so it is possible that these patients were not representative. Results from a secondary analysis of complete data were not consistent between data sets; therefore, no clear hypothesis postadjustment was supported for this covariate. These results suggest that similarly powered studies (n ∼ 300) reporting marginal associations should be interpreted with caution. Because the model presented here was validated, we feel it would be useful for others to consider as they begin assessing the importance of clinical factors in the prediction of severe CDI cases at other institutions.
Although previous studies found that CCI and ARC score were predictive of severe CDI [15, 37, 38], neither factor was significant in the present study upon adjustment. Differences in definitions used for severity (eg, diffuse diarrhea, shock index, and 90-day mortality) may explain differences in study results. However, it seems worthwhile to identify new CDI-specific comorbidity criteria that better reflect CDI outcomes. Further assessment and refinement of surveillance groups (HO-HCFA, CO-HCFA, and IND) is required to determine whether they are predictive of severe disease.
Another limitation of this study was the consideration of a single C. difficile isolate per patient stool sample as it is possible that an individual patient is infected with multiple ribotypes concurrently. We recently addressed this limitation directly, and although the data are part of a forthcoming manuscript (unpublished), the co-occurrence of multiple toxigenic ribotypes appears to be no more than that of a previous report (approximately 13% of patients with >1 genotype) [39].
Our results add support to the hypothesis that 027/078 ribotypes are not more virulent than other C. difficile ribotypes. These data should temper enthusiasm for using ribotype, or the presence of binary toxin or tcdC mutations, to influence patient care. Gaps in our knowledge of the determinants of C. difficile pathogenesis limit our capacity to predict clinical behavior based solely on microbial characteristics. It is important to recognize, however, that there may be other attributes of the C. difficile genome that can significantly influence virulence. More discriminant characterizations are needed to address this hypothesis. Until these data become available, patient rather than pathogen characteristics should be used to guide treatment strategies.
Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.
Notes
Acknowledgments. We thank the personnel of the UMHS Clinical Microbiology Laboratory, and especially Carol Young and William LeBar for their expertise and support with the execution of this project. We also thank Lynn Holevinski for database support as part of the University of Michigan Health System Medical Clinical Information Technology team. Reference isolates for ribotyping were provided by Dr Dale Gerding (Edward Hines Jr VA Hospital, Loyola University Medical Center, Maywood, Illinois).
Financial support. This work was supported by Award number UL1RR024986 from the National Center for Research Resources (S. T. W. and D. M. A.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Center for Research Resources or the National Institutes of Health. Additional support came from the National Institutes of Health grant 1U19AI090871–01 (S. T. W., L. W, V. B. Y., and D. M. A), 1K01AI09728101A1 (S. T. W.), and the Department of Veterans Affairs and Geriatric Research Education and Clinical Center (P. N. M). We also gratefully acknowledge support from the Claude Pepper Center grants AG08808 and AG024824 from the National Institute of Aging. S. E. received grant support from the American Society for Microbiology and institutional grant support from Enterics Research Investigational Network. R. J. received grant support from the National Institutes of Health (National Center for Research Resources, National Institute of Aging). M. R. received institutional grant support from the National Institutes of Health (National Center for Research Resources, National Institute of Aging).
Potential conflicts of interest. All authors: No reported conflicts.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
References
- 1.Warny M, Pepin J, Fang A, et al. Toxin production by an emerging strain of Clostridium difficile associated with outbreaks of severe disease in North America and Europe. Lancet. 2005;366:1079–84. doi: 10.1016/S0140-6736(05)67420-X. [DOI] [PubMed] [Google Scholar]
- 2.McDonald LC, Killgore GE, Thompson A, et al. An epidemic, toxin gene-variant strain of Clostridium difficile. N Engl J Med. 2005;353:2433–41. doi: 10.1056/NEJMoa051590. [DOI] [PubMed] [Google Scholar]
- 3.Goorhuis A, Bakker D, Corver J, et al. Emergence of Clostridium difficile infection due to a new hypervirulent strain, polymerase chain reaction ribotype 078. Clin Infect Dis. 2008;47:1162–70. doi: 10.1086/592257. [DOI] [PubMed] [Google Scholar]
- 4.Blossom DB, McDonald LC. The challenges posed by reemerging Clostridium difficile infection. Clin Infect Dis. 2007;45:222–7. doi: 10.1086/518874. [DOI] [PubMed] [Google Scholar]
- 5.Akerlund T, Svenungsson B, Lagergren A, Burman LG. Correlation of disease severity with fecal toxin levels in patients with Clostridium difficile–associated diarrhea and distribution of PCR ribotypes and toxin yields in vitro of corresponding isolates. J Clin Microbiol. 2006;44:353–8. doi: 10.1128/JCM.44.2.353-358.2006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Stabler RA, Dawson LF, Phua LT, Wren BW. Comparative analysis of BI/NAP1/027 hypervirulent strains reveals novel toxin B-encoding gene (tcdB) sequences. J Med Microbiol. 2008;57:771–5. doi: 10.1099/jmm.0.47743-0. [DOI] [PubMed] [Google Scholar]
- 7.Merrigan M, Venugopal A, Mallozzi M, et al. Human hypervirulent Clostridium difficile strains exhibit increased sporulation as well as robust toxin production. J Bacteriol. 2010;192:4904–11. doi: 10.1128/JB.00445-10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Cloud J, Noddin L, Pressman A, Hu M, Kelly C. Clostridium difficile strain NAP-1 is not associated with severe disease in a nonepidemic setting. Clin Gastroenterol Hepatol. 2009;7:868–73. doi: 10.1016/j.cgh.2009.05.018. e2. [DOI] [PubMed] [Google Scholar]
- 9.Goldenberg SD, French GL. Lack of association of tcdC type and binary toxin status with disease severity and outcome in toxigenic Clostridium difficile. J Infect. 2011;62:355–62. doi: 10.1016/j.jinf.2011.03.001. [DOI] [PubMed] [Google Scholar]
- 10.Morgan OW, Rodrigues B, Elston T, et al. Clinical severity of Clostridium difficile PCR ribotype 027: a case-case study. PLoS One. 2008;3:e1812. doi: 10.1371/journal.pone.0001812. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Venugopal AA, Riederer K, Patel SM, et al. Lack of association of outcomes with treatment duration and microbiologic susceptibility data in Clostridium difficile infections in a non-NAP1/BI/027 setting. Scand J Infect Dis. 2011;44:243–9. doi: 10.3109/00365548.2011.631029. [DOI] [PubMed] [Google Scholar]
- 12.Verdoorn BP, Orenstein R, Rosenblatt JE, et al. High prevalence of tcdC deletion-carrying Clostridium difficile and lack of association with disease severity. Diagn Microbiol Infect Dis. 2010;66:24–8. doi: 10.1016/j.diagmicrobio.2009.08.015. [DOI] [PubMed] [Google Scholar]
- 13.McDonald LC, Coignard B, Dubberke E, Song X, Horan T, Kutty PK. Recommendations for surveillance of Clostridium difficile–associated disease. Infect Control Hosp Epidemiol. 2007;28:140–5. doi: 10.1086/511798. [DOI] [PubMed] [Google Scholar]
- 14.Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373–83. doi: 10.1016/0021-9681(87)90171-8. [DOI] [PubMed] [Google Scholar]
- 15.Welfare MR, Lalayiannis LC, Martin KE, Corbett S, Marshall B, Sarma JB. Co-morbidities as predictors of mortality in Clostridium difficile infection and derivation of the ARC predictive score. J Hosp Infect. 2011;79:359–63. doi: 10.1016/j.jhin.2011.08.015. [DOI] [PubMed] [Google Scholar]
- 16.Sorg JA, Dineen SS. Laboratory maintenance of Clostridium difficile. Curr Protoc Microbiol. 2009 doi: 10.1002/9780471729259.mc09a01s12. Chapter 9:Unit9A 1. [DOI] [PubMed] [Google Scholar]
- 17.Rinttila T, Kassinen A, Malinen E, Krogius L, Palva A. Development of an extensive set of 16S rDNA-targeted primers for quantification of pathogenic and indigenous bacteria in fecal samples by real-time PCR. J Appl Microbiol. 2004;97:1166–77. doi: 10.1111/j.1365-2672.2004.02409.x. [DOI] [PubMed] [Google Scholar]
- 18.Persson S, Torpdahl M, Olsen KE. New multiplex PCR method for the detection of Clostridium difficile toxin A (tcdA) and toxin B (tcdB) and the binary toxin (cdtA/cdtB) genes applied to a Danish strain collection. Clin Microbiol Infect. 2008;14:1057–64. doi: 10.1111/j.1469-0691.2008.02092.x. [DOI] [PubMed] [Google Scholar]
- 19.Walk ST, Jain R, Trivedi I, et al. Non-toxigenic Clostridium sordellii: clinical and microbiological features of a case of cholangitis-associated bacteremia. Anaerobe. 2011;17:252–6. doi: 10.1016/j.anaerobe.2011.06.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Spigaglia P, Mastrantonio P. Molecular analysis of the pathogenicity locus and polymorphism in the putative negative regulator of toxin production (TcdC) among Clostridium difficile clinical isolates. J Clin Microbiol. 2002;40:3470–5. doi: 10.1128/JCM.40.9.3470-3475.2002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Tamura K, Peterson D, Peterson N, Stecher G, Nei M, Kumar S. MEGA5: molecular evolutionary genetics analysis using maximum likelihood, evolutionary distance, and maximum parsimony methods. Mol Biol Evol. 2011;28:2731–9. doi: 10.1093/molbev/msr121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Curry SR, Marsh JW, Muto CA, O'Leary MM, Pasculle AW, Harrison LH. tcdC genotypes associated with severe TcdC truncation in an epidemic clone and other strains of Clostridium difficile. J Clin Microbiol. 2007;45:215–21. doi: 10.1128/JCM.01599-06. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Bidet P, Barbut F, Lalande V, Burghoffer B, Petit JC. Development of a new PCR-ribotyping method for Clostridium difficile based on ribosomal RNA gene sequencing. FEMS Microbiol Lett. 1999;175:261–6. doi: 10.1111/j.1574-6968.1999.tb13629.x. [DOI] [PubMed] [Google Scholar]
- 24.Cheknis AK, Sambol SP, Davidson DM, et al. Distribution of Clostridium difficile strains from a North American, European and Australian trial of treatment for C. difficile infections: 2005–2007. Anaerobe. 2009;15:230–3. doi: 10.1016/j.anaerobe.2009.09.001. [DOI] [PubMed] [Google Scholar]
- 25.Indra A, Huhulescu S, Schneeweis M, et al. Characterization of Clostridium difficile isolates using capillary gel electrophoresis-based PCR ribotyping. J Med Microbiol. 2008;57:1377–82. doi: 10.1099/jmm.0.47714-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.R Project. R: A Language and environment for statistical computing. Available at: http://www.R-project.org/ [Google Scholar]
- 27.Archibald LK, Banerjee SN, Jarvis WR. Secular trends in hospital-acquired Clostridium difficile disease in the United States, 1987–2001. J Infect Dis. 2004;189:1585–9. doi: 10.1086/383045. [DOI] [PubMed] [Google Scholar]
- 28.McDonald LC, Owings M, Jernigan DB. Clostridium difficile infection in patients discharged from US short-stay hospitals, 1996–2003. Emerg Infect Dis. 2006;12:409–15. doi: 10.3201/eid1203.051064. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.O'Connor JR, Johnson S, Gerding DN. Clostridium difficile infection caused by the epidemic BI/NAP1/027 strain. Gastroenterology. 2009;136:1913–24. doi: 10.1053/j.gastro.2009.02.073. [DOI] [PubMed] [Google Scholar]
- 30.Bacci S, Molbak K, Kjeldsen MK, Olsen KE. Binary toxin and death after Clostridium difficile infection. Emerg Infect Dis. 2011;17:976–82. doi: 10.3201/eid1706.101483. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Geric B, Johnson S, Gerding DN, Grabnar M, Rupnik M. Frequency of binary toxin genes among Clostridium difficile strains that do not produce large clostridial toxins. J Clin Microbiol. 2003;41:5227–32. doi: 10.1128/JCM.41.11.5227-5232.2003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Sirard S, Valiquette L, Fortier LC. Lack of association between clinical outcome of Clostridium difficile infections, strain type, and virulence-associated phenotypes. J Clin Microbiol. 2011;49:4040–6. doi: 10.1128/JCM.05053-11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Wilson V, Cheek L, Satta G, et al. Predictors of death after Clostridium difficile infection: a report on 128 strain-typed cases from a teaching hospital in the United Kingdom. Clin Infect Dis. 2010;50:e77–81. doi: 10.1086/653012. [DOI] [PubMed] [Google Scholar]
- 34.Bishara J, Peled N, Pitlik S, Samra Z. Mortality of patients with antibiotic-associated diarrhea: the impact of Clostridium difficile. J Hosp Infect. 2008;68:308–14. doi: 10.1016/j.jhin.2008.01.033. [DOI] [PubMed] [Google Scholar]
- 35.Dubberke ER, Butler AM, Yokoe DS, et al. Multicenter study of Clostridium difficile infection rates from 2000 to 2006. Infect Control Hosp Epidemiol. 2010;31:1030–7. doi: 10.1086/656245. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Khanna S, Pardi DS, Aronson SL, et al. The epidemiology of community-acquired Clostridium difficile infection: a population-based study. Am J Gastroenterol. 2012;107:89–95. doi: 10.1038/ajg.2011.398. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Cadena J, Thompson GR, 3rd, Patterson JE, et al. Clinical predictors and risk factors for relapsing Clostridium difficile infection. Am J Med Sci. 2010;339:350–5. doi: 10.1097/MAJ.0b013e3181d3cdaa. [DOI] [PubMed] [Google Scholar]
- 38.Hardt C, Berns T, Treder W, Dumoulin FL. Univariate and multivariate analysis of risk factors for severe Clostridium difficile–associated diarrhea: importance of co-morbidity and serum C-reactive protein. World J Gastroenterol. 2008;14:4338–41. doi: 10.3748/wjg.14.4338. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Wroblewski D, Hannett GE, Bopp DJ, et al. Rapid molecular characterization of Clostridium difficile and assessment of populations of C. difficile in stool specimens . J Clin Microbiol. 2009;47:2142–8. doi: 10.1128/JCM.02498-08. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.