Clostridium difficile Infection in Liver Transplant Recipients: A Retrospective Study of Rates, Risk Factors and Outcomes

Authors


Abstract

Clostridium difficile infection (CDI) occurs in 3–7% of liver transplant recipients (LTR). However, few data exist on the recent epidemiology, predictors and outcomes of CDI in LTR. A cohort study was performed including LTR from 2000 to 2010 at a tertiary care hospital in Detroit. CDI was defined as diarrhea with a stool C. difficile positive test. Data analyzed included demographics, comorbidities, length of stay (LOS), severity of CDI, rates of recurrence (<12 weeks), relapse (<4 weeks) and overall mortality. Predictors of CDI were calculated using Cox proportional hazard model; 970 LTR were followed for years. Overall prevalence of CDI was 18.9%. Incidence of CDI within 1 year of transplant was 12.4%. Severe CDI occurred in 29.1%. CDI recurrence and relapse rates were 16.9% and 9.7%, respectively. Independent predictors of CDI were year of transplant (hazard ratio [HR] 1.137, 95% confidence interval [CI] 1.06–1.22; p < 0.001), white race (105/162 whites, HR 1.47, 95% CI 1.03–2.1; p = 0.035), Model for End-Stage Liver Disease score (HR 1.03, 95% CI 1.01–1.045, p = 0.003) and LOS (HR 1.01, 95% CI 1.005–1.02, p < 0.001). Significant mortality was observed among LTR with CDI compared to those without CDI (p = 0.003). We concluded that CDI is common among LTR and is associated with higher mortality.

Abbreviations
CDC

Centers for Disease Control and Prevention

CDI

Clostridium difficile infection

CI

confidence interval

EIA

enzyme immunoassay

HCF

health-care facility

HR

hazard ratio

LOS

length of stay

LT

liver transplantation

LTR

liver transplant recipients

MELD

Model for End-Stage Liver Disease

NAP-1

North American Pulsed-Field Type 1 strain

Introduction

Diarrhea is a major cause of morbidity in liver transplant recipients (LTR). Although the exact incidence of diarrhea in LTR is unknown, incidence rates of 10–43% have been reported in other solid organ transplant populations [1]. Diarrhea attributable to Clostridium difficile infection (CDI) is reported to occur in 3.5–9% of LTR [2-4]. A significant increase in the incidence and severity of CDI has been reported in recent years, in part believed to be due to the emergence of a new hypervirulent and easily transmissible strain called the North American Pulsed-Field Type 1 strain (NAP-1) of C. difficile [5]. Furthermore, asymptomatic carriage of C. difficile is believed to occur in about 7% of healthy adults [6] and in about 11–25% of hospitalized patients [7, 8] and may contribute to nosocomial transmission. Despite therapy, approximately 15–20% of patients will develop recurrence of CDI.

With high recurrence rates and few data on CDI in LTR reported in the context of recent changes in the epidemiology of CDI, this study aims to: (1) report rates of CDI in hospitalized LTR over the last 10 years, (2) evaluate relevant demographic, clinical and laboratory predictors for CDI, (3) examine changing trends in clinical characteristics of CDI and (4) report CDI-associated mortality.

Patients and Methods

Study population

All patients who underwent liver transplantation (LT) at Henry Ford Hospital, Detroit, MI, between January 2000 and December 2010, were included in this study. The Henry Ford Hospital is a 900-bed acute care tertiary care medical center that performs approximately 90–100 LT annually. Patients were identified from the LT registry database. Patients who died immediately after LT or were lost to follow-up were excluded. The study protocol was approved by the institutional research review board at Henry Ford Hospital. All LTR with a positive enzyme immunoassay (EIA) for C. difficile toxins A/B performed on a diarrheal stool were considered cases. In 2008, the EIA was switched to a more sensitive and specific combination two-step testing, using glutamate dehydrogenase screening test for C. difficile followed by EIA and molecular testing to confirm toxigenic C. difficile [9]. Date of infection was considered as the date of positive toxin assay as test results are available within few hours at our institution. Time to onset was the number of days between the date of LT and the date of positive toxin assay. We also evaluated patients that developed CDI up to 30 days pre-LT.

Study design

A retrospective cohort study was performed. All LTR with diarrhea and a positive C. difficile toxin assay performed on a diarrheal stool were considered cases and all those who did not have CDI were the controls.

Demographic, clinical and laboratory data and outcomes were collected from review of the electronic medical records. Patient data were reviewed for the 90 days preceding CDI to identify risk factors for CDI including antibiotic and antacid use. We also reviewed data related to LT, including indication for LT, pre-LT Model for End-Stage Liver Disease (MELD) score and length of stay (LOS) for LT. The Charlson's Comorbidity Index was used to grade the severity of comorbid illnesses. CDI-related characteristics were reviewed, including temporal relationship of CDI to transplant and prior hospitalization, severity of CDI, treatment used for CDI, relapse and recurrence rates and number of recurrences. Outcomes evaluated included the need for colectomy related to CDI, mortality attributable to CDI, all-cause 1-year mortality and all-cause graft loss.

The follow-up period began from date of LT to date of death or last day of follow-up until December 31, 2011. Date of death was obtained from the social security database and supplemented by our institution's transplant registry that ensured no loss to follow-up.

Definitions

There is no standardized definition of severe CDI; however, there are several parameters associated with severe disease including, white cell count >15 000 cells/mm3, elevation of serum creatinine >1.5 times the baseline, abdominal distension, low albumin, ICU treatment and CDI-related colectomy [10, 11]. Since several of these parameters may be abnormal in patients with end-stage liver disease and LTR, for the purposes of this study severe CDI was defined as the presence of the following—white cell count >15 000 cells/mm3 within 1 week of onset of CDI (in the absence of other cause), or CDI-related colectomy. Recurrent CDI was defined as new-onset of diarrhea and a positive stool toxin assay within 12 weeks of previous CDI. If recurrent CDI occurred within 4 weeks of prior episode of CDI, it was considered a relapse of CDI.

CDI was classified by the onset of symptoms according to the Centers for Disease Control and Prevention (CDC) National Healthcare Safety Network surveillance definition [9]. Community onset CDI—onset of symptoms within 72 h of hospital admission; hospital onset CDI—symptom onset after 72 h of admission; health-care facility (HCF) associated CDI—CDI in a patient discharged from an HCF within the last 12 weeks; Community-associated CDI—disease in patients with no prior HCF stay in the previous 12 weeks [12]. Cases considered as indeterminate in the CDC definition were considered HCF associated.

Statistical methodology

Data analysis of all LTR and outcomes were stratified according to presence of CDI posttransplant. Baseline characteristics were summarized using basic frequencies to obtain descriptive information such as age, sex, race and mean pretransplant MELD scores. Continuous variables were expressed in mean ± standard deviation since there was minimal skewing of data and categorical variables were mentioned in percentages. Precipitating factors of CDI were calculated using Cox proportional hazard model with adjusted analysis for age, gender, race and indication for liver transplant. Survival analysis was performed using the Kaplan–Meier method. Categorical variables were compared using a chi-squared test or Fisher's exact test if sample sizes were small. Continuous variables were compared using the two-sample t-test. Values of p < 0.05 were considered statistically significant. Data were analyzed with SAS software version 9.2 (SAS Institute, Inc., Cary, NC).

Results

A total of 970 LTR (mean age 53.2 ± 10.0 years, women 34.6% and white 64.5%) were identified between January 2000 and December 2010. Table 1 summarizes the baseline characteristics of the patients with post-LT CDI (cases) and those without CDI (controls).

Table 1. Characteristics of liver transplant recipients with and without C. difficile infection (CDI)
VariablesCase patients with CDI, N (%) (n = 162)Control patients without CDI, N (%) (n = 808)Hazard ratio95% Confidence intervalp-Value
  • LT, liver transplantation; MELD, Model for End-Stage Liver Disease; SD, standard deviation.
  • *Statistically significant, p < 0.05.
Age in years (mean ± SD)53.2 ± 10.953.2 ± 9.81.0020.986–1.0170.825
Male gender104 (64.2)527 (65.2)0.9330.677–1.2860.672
White race105 (69.1)456 (62.2)1.4491.026–2.0460.035*
Black race27 (17.8)137 (18.7)0.9430.622–1.4300.784
Charlson score (mean ± SD)6.2 ± 3.05.8 ± 3.11.0380.990–1.0880.122
Charlson score ≥3146 (90.1)691 (85.5)1.6350.975–2.7400.062
No. of antibiotics <90 days of CDI (mean ± SD)4.7 ± 3.23.8 ± 3.11.1281.077–1.180<0.001*
Pretransplant CDI11 (6.8%)21 (2.6%)2.884  
Year of LT (mean ± SD)2006.2 ± 2.42005.5 ± 3.01.2021.128–1.281<0.001*
Length of hospital stay for LT (mean ± SD)25.5 ± 23.716.7 ± 316.91.0181.013–1.023<0.001*
MELD score (mean ± SD)21.0 ± 9.817.8 ± 9.11.0421.026–1.059<0.001*
Indication for LT
Hepatitis C liver disease69 (43.1%)375 (46.6%)0.8740.649–1.1950.398
Hepatitis B liver disease6 (3.8%)21 (2.6%)1.5180.671–3.4330.317
Alcoholic liver disease31 (19.4%)136 (16.9%)1.1890.803–1.7600.388
Cryptogenic liver disease6 (3.8%)37 (4.6%)0.8370.370–1.8920.669
Hepatocellular carcinoma6 (3.8%)49 (6.1%)0.6610.292–1.4940.320
Hepatic failure13 (8.1%)63 (7.8%)1.0050.570–1.7730.985
Other indication29 (18.1%)123 (15.3%)1.1520.770–1.7230.490
No. of LTs (mean ± SD)1.1 ± 0.41.0 ± 0.22.8031.743–4.507<0.001*
No. of endoscopies (mean ± SD)1.5 ± 2.01.6 ± 2.10.9650.891–1.0450.378

Rates

CDI occurred in 32 (3.3%) patients pretransplant (within 30 days before LT), and 162 (16.7%) patients developed CDI posttransplant. The overall prevalence of CDI in LTR was 18.9% (n = 183). The incidence of CDI in LTR within 1 year of transplant (early CDI) was 12.4% (120/970) and after 1 year of transplant (late CDI) was 6.6% (63/954). Overall the median time to CDI after LT was 51 days. Of the patients with CDI, 29.1% (n = 53) had severe CDI, 53 had white cell count >15 000 and 5 required a colectomy. Of the 53 patients with white cell count >15 000, 3 required a colectomy for CDI. A new rise in serum creatinine was noted in 45% (n = 72). The CDI recurrence rate was 16.9%, with an average time to recurrence (mean ± SD) of 62.5 ± 26.5 days. The CDI relapse rate was 9.7%. Among patients with severe CDI, 29% developed recurrence compared to 13% recurrence in patients without severe CDI.

Graft loss rate was 3.6% but was not attributable to CDI. Hospital onset of infection versus community onset was 62% versus 38%, respectively. Statistical significance was detected for hospital onset of CDI predicting severe infection and mortality versus community onset (Table 2). The HCF acquired CDI rate was 77% as compared to community acquired rate of 23%.

Table 2. Comparison results for community versus hospital onset of C. difficile infection (CDI)
VariablesHospital onset of CDI (N = 103)Community onset of CDI (N = 63)Comparison, p-value
  • Severe CDI defined as white cell count >15 000 within 1 week of CDI diagnosis or CDI-related colectomy c; graft loss and all-cause death were determined at 1 year postliver transplantation.
  • Data are given as frequency (percent of group). (C) Chi-squared test; (F) Fisher's exact test.
  • *Statistically significant, p < 0.05.
Severe CDI38 (37.6%)14 (22.2%)0.003 (C)*
Recurrence of CDI4 (4.0%)10 (16.1%)0.802 (C)
Graft loss4 (4.06%)2 (3.2%)1.000 (F)
Death46 (44.73%)14 (22.2%)0.003 (C)*

Single drug therapy was used in 32.7% (oral metronidazole), 1.2% (IV metronidazole) and 4.3% (oral vancomycin), respectively. Combination therapy was used in 15.4% (oral metronidazole and oral vancomycin), 9.3% (IV metronidazole and oral vancomycin) and 2% (oral metronidazole, IV metronidazole and oral vancomycin).

Predictors of posttransplant CDI

The statistically significant predictors of posttransplant infection were year of transplant, race, MELD score and transplant LOS (Table 3). All (100%) of LTR in the study received antibiotics before the diagnosis of CDI. The number of antibiotic received was significantly higher in the CDI group on univariate analysis but not in the regression model. Similarly, 95% of LTR were on proton-pump inhibitors with an equal distribution in both groups.

Table 3. Multivariable Cox regression analysis results for predicting postliver transplantation C. difficile infection (CDI)
VariablesHazard ratio95% Hazard ratio confidence limitsp-Value
  • HCV, hepatitis C virus; LT, liver transplantation; MELD, Model for End-Stage Liver Disease; SD, standard deviation.
  • *Statistically significant, p < 0.05.
Age0.9990.9831.0160.930
Male gender0.9850.6831.4200.936
White race1.4701.0282.1030.035*
Charlson score1.0350.9751.0990.262
Pre-LT CDI1.4990.7692.9260.235
Year of transplant1.1371.0641.216<0.001*
Length of stay for LT1.0121.0051.019<0.001*
MELD score1.0271.0091.0450.003*
Number of transplants1.8650.9863.5280.055
Indication for LT
Hepatitis B liver disease (vs. HCV)0.6640.1612.7310.570
Alcoholic liver disease (vs. HCV)1.1100.7141.7240.644
Cryptogenic liver disease (vs. HCV)0.4970.1781.3880.182
Hepatocellular carcinoma (vs. HCV)0.4890.1741.3770.176
Hepatic failure (vs. HCV)0.7550.3891.4660.406
Other indication (vs. HCV)1.3920.8662.2400.172
Number of endoscopies0.9470.8691.0310.210

Figure 1 shows the trends in rates of CDI per year. The 1-year probability of freedom from infection decreased slightly within the more recent transplant years—2006 onward (Table 4), with a statistically significant log-rank test result for comparing the underlying yearly Kaplan–Meier curves of p < 0.001 (Table 3). Nosocomial CDI rates per 10 000 patient days available in the hospital Infection control database were 17, 22, 15 and 8 for the years 2007, 2008, 2009 and 2010, respectively.

Figure 1.

Number of C. difficile infection cases per year of transplant.

Table 4. One-year Kaplan–Meier estimates of freedom from postliver transplantation C. difficile infection
Year of liver transplantation1-Year freedom from C. difficile infection probability
20000.98
20011.00
20020.96
20030.89
20040.91
20050.93
20060.86
20070.76
20080.76
20090.88
20100.86

Mortality

The mortality rate in the entire population was 27.8% (n = 270). Mortality rates in the CDI population versus the non-CDI group were 35% (64/183) versus 26% (205/787), respectively. The Kaplan–Meier curve for freedom from death between the CDI-infected and noninfected patients yielded an overall significantly higher mortality rate in the CDI subset than the non-CDI patients (log-rank p-value = 0.003, Figure 2). In none of the patients was death directly attributable to CDI.

Figure 2.

Kaplan–Meier curves for freedom from death by infection status (log-rank p-value = 0.003).

Discussion

This large retrospective single-center study shows that CDI is a common problem in LTR with potential life-threatening complications. About one-sixth of LTR developed CDI in our series over 10 years which is similar to previously reported rates of 4% to up to 21% [13]. However, the reported rates vary depending on whether incidence or prevalence was calculated. We report both incidence rate within 1 year posttransplant and prevalence rates over a mean follow-up period of about 5 years. A recent nationwide analysis [4] showed that LTRs were the highest risk group among solid organ transplant recipients and the overall prevalence was 2.7%.

Other interesting finding in our study was that about 3.3% patients developed CDI in the 30-day pretransplant period, suggesting that the risk factors for CDI are similar to other hospitalized patients. Liver transplant candidates undergo multiple hospitalizations, antibiotic treatments, proton-pump inhibitor use and endoscopies before transplant, all of which are known risk factors for CDI [14, 15]. Although number of antibiotics used was statistically higher in our CDI population (Table 1), it was not a risk factor in multivariate analysis probably because all LTRs receive perioperative antibiotics. Also, almost all patients were on proton-pump inhibitors, so a definite conclusion regarding their effects on the rate of CDI could not be analyzed.

Our recurrence rate of 16.9% was almost double the previously reported rates. This may partly be explained by the longer follow-up period in our study. Also, we found that more than half of the re-infections occur within 4 weeks of the index infection. The relapse rates in our liver transplant population are higher than other hospitalized patients, likely due to immunocompromised state and comorbid conditions.

Hospital onset of infection was about twice as common as community onset suggesting that nosocomial transmission is common despite improved infection control measures. Similarly, HCF-acquired CDI was much more common as compared to community acquired implying that LTR have significant exposure to hospital environment and possible nosocomial acquisition. However, without strain typing, it is not possible to assess if these infections represent transmission of clonally related strains.

We found important risk factors for CDI including year of transplant, MELD score and LOS for transplant. White race was also found to be an important risk factor for CDI. The clinical implications of this finding are unknown but this finding is consistent with previously done nationwide analysis on CDI [4]. Further studies are required to confirm, analyze and find a biological plausibility for this observation. The almost universal use of the antibiotics and proton pump inhibitor in LTR population makes it difficult to determine the role of these “traditional” risk factors and predisposition for CDI.

The rates of infection increased steadily with increasing MELD score and year of transplant. However, the rates were highest in 2007 and 2008 and then declined. The reasons for this change are multifold. This observed trend corresponds to the changes in rates of nosocomial CDI in hospitalized patients at our institution. The increase in the years 2006–2008, also reflects the nationwide increase in the incidence and severity of CDI since 2000 [5]. This increase is believed to be partly due to the emergence of a more virulent and transmissible strain of C. difficile known as NAP-1 that is now endemic across the nation [5]. The subsequent decline of CDI in our LTR and the general hospital population probably relates to the introduction of a “CDI prevention bundle” at our institution. The bundle includes the placement of patients with suspected or confirmed CDI in contact isolation, hand washing with soap and water, and the use of bleach for environmental disinfection. This strategy along with antibiotic stewardship and real-time reporting of positive C. difficile test via a telephone call, leading to early treatment, is the likely explanation of reduced rates of CDI since 2008. This decrease was observed despite the introduction of a more sensitive two-step C. difficile testing in 2008, which would be expected to increase the rates of detection and hence incidence.

A notable feature of our study was categorization of CDI into hospital/community acquired infections and health-care associated/community-associated infections. These terms have been utilized for the epidemiological classification of CDI. Utilizing these definitions about a fifth of cases of CDI in LTR were community-associated. This is in keeping with recent estimates of 20–28% of community-associated CDI in the general population [16]. In a recent nationwide study of community-associated CDI in the United States, factors associated with increased risk were low-level outpatient health-care exposure, use of antibiotics or proton-pump inhibitors and exposure to infants [16]. The increased severity of CDI noted in LTR with hospital onset of CDI is likely due to CDI occurring in a more severely ill patient population. However, the possibility a more virulent, nosocomially acquired, strain of C. difficile causing severe CDI cannot be excluded [17].

The overall mortality was higher in CDI group than in non-CDI group, despite advancements in diagnosis and treatment of CDI, further reiterating the importance of identifying this infection in high-risk populations like LTR. Charlson's comorbidity score is an important measure of how sick the patients were. In our study population, mortality was higher in CDI group even though the Charlson's scores were comparable in both groups. Although neither graft loss or death was directly attributable to CDI in our study. However, the occurrence of CDI might identify a subset of patients at risk for a more complicated posttransplant course and poor outcome. This is suggested by the significantly higher MELD scores and longer LOS after LT noted in the multivariate analysis.

Also, newer medications like fidaxomicin, which has proven to reduce recurrence rates [18], may be considered for treatment of CDI in LTR given the high recurrence rates and increased mortality.

There are many limitations to our study. First, the retrospective cohort design limits data availability and analysis. Second, strain-typing data were not available. The observed trends in incidence and recurrence rates may be due, in part, to the emergence of a hypervirulent epidemic NAP-1 strain of C. difficile. The reported rates in our study may be higher due to the single tertiary center nature of study design. Although most LTR are followed up at tertiary care centers, cases diagnosed at other institutions may have been missed.

In conclusion, CDI remains a common problem in LTR with high rates of recurrence and increased mortality. Meticulous attention to infection control measures is required to reduce the rates of nosocomial transmission, which contributes to a significant number of cases.

Acknowledgments

No sources of funding were used in performing this study or writing this manuscript. The authors thank everyone that contributed to this study.

Disclosure

The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.

Ancillary