Digestive Diseases and Sciences

, Volume 57, Issue 10, pp 2592–2599

Superiority of the DNA Amplification Assay for the Diagnosis of C. difficile Infection: A Clinical Comparison of Fecal Tests

Authors

    • Division of Gastroenterology, Department of MedicineUniversity of Miami Miller School of Medicine
  • Neilanjan Nandi
    • Division of Gastroenterology, Department of MedicineUniversity of Miami Miller School of Medicine
  • Nancimae Miller
    • Department of PathologyMount Sinai Medical Center
  • Alexandra Grace
    • Department of MedicineMount Sinai Medical Center
  • Jamie S. Barkin
    • Division of Gastroenterology, Department of MedicineUniversity of Miami Miller School of Medicine
    • Department of MedicineMount Sinai Medical Center
  • Daniel A. Sussman
    • Division of Gastroenterology, Department of MedicineUniversity of Miami Miller School of Medicine
Original Article

DOI: 10.1007/s10620-012-2200-x

Cite this article as:
Barkin, J.A., Nandi, N., Miller, N. et al. Dig Dis Sci (2012) 57: 2592. doi:10.1007/s10620-012-2200-x

Abstract

Background

Clostridium difficile infection (CDI) is a major infectious concern, accounting for substantial morbidity and resource utilization. Advances in microbiological and molecular techniques have resulted in an increasing number of testing options for CDI. A glutamate dehydrogenase (GDH) enzyme immunoassay (EIA) and a DNA amplification (DNA-A) test for the diagnosis of CDI have recently become commercially available.

Aims

The aim of this prospective study was to compare the test performance characteristics of the traditional diagnostic modality for CDI diagnosis, the toxin A/B (TOX) EIA, with those of the GDH EIA and DNA-A test, utilizing enriched toxigenic culture (TGC) as the gold standard. Clinical variables predictive of CDI were also studied.

Methods

Participants fulfilled one or more criteria placing them at increased risk for CDI. Each stool sample was tested by each of the methods mentioned above. Clinical data parameters were collected via a 12-month review of the electronic medical record prior to the index date of the first stool test.

Results

A total of 272 stool samples from 144 admissions of 139 patients were evaluated for CDI. The sensitivity and positive predictive value (PPV) of the TOX EIA were 86.1 and 58.4 %, respectively, whereas the sensitivity and PPV of the GDH EIA and DNA-A test were 100 %. 1.8 % of the GDH tests yielded inconclusive results. Using TGC as the gold standard, nosocomial exposure with emphasis on nursing home residence, history of previous CDI, and female gender were predictive of CDI.

Conclusions

Test performance characteristics of the DNA-A test and GDH EIA were superior to those of the traditional TOX EIA. The GDH test is limited by inconclusive test results and requires a multi-step diagnostic algorithm. Therefore, the DNA-A test should be implemented as the diagnostic method of choice for CDI. CDI clinical predictors are important for diagnostic decision-making.

Keywords

Clostridium difficileDiagnosisImmunoenzyme assaysDNA amplification

Introduction

Clostridium difficile infection (CDI) is a major infectious concern, accounting for substantial morbidity and resource utilization. From 2000 to 2005, there was a 23 % increase in nosocomial Clostridium difficile-associated diarrhea (CDAD) in the USA, with an increase in CDAD-related fatality rates from 1.2 to 2.2 % [1]. On average, patient length of stay was extended by 3.6 days and required an additional $3,669 expenditure per patient as compared to those without CDAD [2]. Data from the Healthcare Cost and Utilization Project in 2008 estimated that CDAD accounted for 348,950 hospitalizations annually, and a 2009 estimate placed the approximate costs of CDAD at $3.4 billion annually, more than a tripling of the estimates from only 7 years earlier [35].

Advances in microbiological and molecular techniques have resulted in an increasing quantity of testing options for CDI. The advent of toxin (TOX) enzyme immunoassay (EIA) testing in the 1990s enhanced the diagnostic capability for CDI by enabling direct assessment of the presence of C. difficile toxins A and B. Though clinically useful, this test is prone to inaccuracies due to specimen collection and toxin degradation prior to processing. A glutamate dehydrogenase (GDH) EIA and a DNA amplification (DNA-A) method for the diagnosis of CDI have recently become commercially available, with the DNA-based test theoretically offering the advantage of more consistent diagnoses independent of specimen and environmental factors that affect the reliability of the toxin A/B EIA test. While not routinely used due to lengthy time for results reporting, toxigenic culture still remains the gold standard for CDI diagnosis [6].

In the current health care environment that emphasizes cost control measures and appropriate utilization of resources, identifying clinical predictors of disease can result in better clinical decisions and more efficient care. Using the TOX EIA to determine the presence of CDI, prior studies have identified a myriad of associated clinical predictors of CDI, including antibiotic use, proton pump inhibitor (PPI) use, and healthcare-associated exposures [3, 6, 7]. Although generally accepted as associated with CDI, these potential predictors have not been formally evaluated when CDI was determined with newer diagnostic modalities.

The aim of this prospective study was to compare the test performance characteristics of the traditional toxin A/B (TOX) EIA, glutamate dehydrogenase EIA, and DNA amplification (DNA-A) assay for diagnosis of CDI, utilizing toxigenic culture (TGC) as the gold standard. Clinical variables predictive of CDI were also identified.

Methods

Study Population

Subjects were derived from inpatients at Mount Sinai Medical Center (Miami Beach, FL), a large community teaching hospital, during the study period from September 1, 2010 to March 1, 2011. Daily orders placed and receipt of specimens by the laboratory for C. difficile toxin A/B (TOX) EIA were the basis by which physicians obtained consent for patient enrollment in the study. The electronic health record was reviewed to identify clinical factors meeting inclusion criteria. Patients eligible for inclusion in the study were 18 years of age or older, had diarrhea (defined as three or more bowel movements in 24 h), had stool samples submitted for CDI testing per clinician discretion, and were able to provide consent for study enrollment. Participants also fulfilled one or more of the following criteria placing them at increased risk for CDI: prior history of CDI, nosocomial exposure in the previous 6 months, antibiotic PPI use within the previous 3 months, age ≥65 year, or the presence of nasogastric or post-pyloric feeding tube. Subjects were excluded if they were currently being treated for documented CDI and then re-tested for CDI during the study period, or if stool samples were not tested by all four laboratory testing modalities under consideration herein. Institutional Review Board approval was obtained.

A total of 272 stool samples from 144 admissions of 139 patients were evaluated for CDI, with a maximum of three samples from any one patient admission being included in the study as per laboratory protocol for CDI testing. There were 36 samples from 31 patient admissions that were found to be positive for CDI by enriched TGC, with a prevalence of CDI in our study population of 13.2 % of stool samples (36/272 samples) and 21.5 % of patient admissions (31/144 admissions) (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs10620-012-2200-x/MediaObjects/10620_2012_2200_Fig1_HTML.gif
Fig. 1

Subject enrollment. CDIClostridium difficile infection

A total of 139 patients (80 men, 59 women, median age 70.5 years) were included in the study population. The median age of patients with CDI was 78 years, whereas that of patients included in the study without CDI was 69 years [p = 0.082, not significant (NS)].

Laboratory Analysis

Specimens were refrigerated upon arrival to the laboratory, and C. difficile TOX EIA testing was performed daily. In accordance with laboratory policy, specimens received more than 2 h after collection were rejected for testing, and only specimens of unformed stools were accepted for testing. Laboratory policy allows submission of a maximum of three stool specimens collected on different days for testing from any one patient admission. Samples from each patient were included in the study until there was one C. difficile TOX EIA positive result, or a maximum of three negative C. difficile TOX EIA results. Each stool sample underwent four tests: (1) TOX EIA (Meridian Premier Toxins A and B Microwell EIA; Meridian Bioscience, Cincinnati, OH); (2) illumigene C. difficile DNA Amplification Assay (Meridian Bioscience); (3) GDH EIA testing (ImmunoCard C. difficile; Meridian Bioscience); (4) enriched TGC. Positive results in the TOX EIA were indicated by measuring optical densities at 450 and 630 nm (OD450/630) of ≥0.100. The illumigene C. difficile assay utilizes loop-mediated isothermal DNA amplification (LAMP) to detect the pathogenicity locus (PaLoc) of toxigenic C. difficile. The PaLoc is a gene segment present in all known toxigenic C. difficile strains; it codes for both the Toxin A and Toxin B genes, has conserved border regions, and is found at the same site on the C. difficile genome in all toxigenic strains [8]. The assay detects the PaLoc by targeting a partial DNA fragment on the Toxin A gene selected as an intact conserved region remaining in all known A+B+ and A−B− toxinotypes. Each reaction device contains an internal control to monitor for amplification inhibition. The ImmunoCard C. difficile EIA detects the presence of GDH, which is a common antigen found in both toxigenic and non-toxigenic isolates of C. difficile. Positive results are detected visually by presence of a color change in the test reaction port. Specimens that lack detectable color change in the control reaction port are interpreted as invalid. Enriched TGC served as the gold standard for the diagnosis of CDI to which all other modalities were compared. Specimens were subjected to alcohol shock with 95 % alcohol followed by culture in chopped meat anaerobic broth for 24–48 h. The broths were filtered and tested for toxin production by the Premier Toxins A and B Microwell EIA after 24 h of incubation; if negative, they were tested again after 48 h of incubation. True positive and negative values were defined by the reference TGC results to calculate sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy. An external technical review team from Meridian Biosciences examined the laboratory procedures at the mid-point of the study to ensure technical fidelity for assays.

Clinical Analysis

Clinical data parameters were collected via a 12-month review of the electronic medical record prior to the index date of the first stool test (Table 1).
Table 1

Clinical predictors collected for analysis

Patient demographics

  Age

  Gender

Physical examination

  Fever, defined as temperature >38.0 °C

  Hemodynamic Instability (blood pressure <90/60 or heart rate >100)

  Abdominal distension

  Abdominal tenderness, rebound, or guarding

Laboratory markers

  WBC >11,500 cells/mm3

  Left shift of neutrophils (bands >20 %)

  Increase in creatinine >50 % of baseline

  Elevated lactate

Radiologic markers

  Abdominal X-ray or computed tomography performed

  Colonic distension

  Colonic wall thickening

  Peri-colonic fat stranding

  Ascites

  Ileus

Colonoscopic markers

  Presence of pseudomembranes

Exposures

  History of CDI

  Nosocomial exposure (hospital, nursing home, or both)

  Antibiotic use and antibiotic class

  PPI use, type and frequency

  ICU admission (current or previous)

  IBD (Crohn’s disease, UC, or indeterminate colitis)

  Human immunodeficiency virus

  Diabetes mellitus

  Immune-compromised state

  Systemic steroid use

  Chronic kidney disease

  Malignancy

  Nasogastric/gastrostomy/jejunostomy tube

WBC, White blood cells; CDI, Clostridium difficile infection; PPI, Proton pump inhibitors; ICU, intensive care unit; IBD inflammatory bowel disease; UC, ulcerative colitis

Statistical Analysis

Univariate logistic regression analysis using the chi-square test was performed to establish clinical variables significantly associated with CDI, and continuous variables were evaluated using the Student t test. PASW (SPSS) ver. 18 software was used for statistical analysis.

Results

Laboratory Testing for CDI

The sensitivity of the TOX test was 86.1 %, whereas the sensitivity of the GDH EIA and DNA-A tests were both 100 % (Table 2). The TOX test yielded 22 false positives and five false negatives. Neither the GDH EIA nor the DNA-A test had false positives or false negatives. There were no inconclusive TOX test results, five inconclusive GDH test results, and one inconclusive DNA-A test result. Within the group of five inconclusive GDH EIA results, one sample was also inconclusive with the DNA-A test, being the only inconclusive sample with this modality. Of the 35 GDH EIA positive results, 30 were also positive for the TOX EIA (85.7 %). One sample was positive for the TOX test, GDH test, and DNA-A test despite a negative result for the TGC; this sample was categorized as a false negative culture.
Table 2

Comparison of diagnostic tests for CDI using toxigenic culture as the gold standard

Parameters

Toxin A/B (TOX) EIA

[95 % CI]

GDH EIA

[95 % CI]

DNA-A

[95 % CI]

Sensitivity

86.1 % (31/36)

[0.697–0.948]

100 % (35/35)

[0.877–1.000]

100 % (36/36)

[0.880–1.000]

Specificity

90.7 % (214/236)

[0.860–0.939]

100 % (232/232)

[0.980–1.000]

100 % (235/235)

[0.980–1.000]

PPV

58.4 % (31/53)

[0.442–0.716]

100 % (35/35)

[0.877–1.000]

100 % (36/36)

[0.880–1.000]

NPV

97.7 % (214/219)

[0.944–0.992]

100 % (232/232)

[0.980–1.000]

100 % (235/235)

[0.980–1.000]

Inconclusive tests

0

1.8 % (5/272)

0.4 % (1/272)

Accuracy

86.3 % (235/272)

100 % (267/267)

100 % (271/271)

EIA, Enzyme immunoassay; GDH, glutamate dehydrogenase; DNA-A, DNA amplification test; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value

Clinical Predictors of CDI

Forty variables were reviewed as potential clinical predictors of CDI (Table 1). Three variables were associated with CDI by univariate analysis. Female subjects comprised a greater frequency of the CDI-positive group (61.3 vs. 37.2 %; p = 0.016). A history of previous CDI was also more common in the CDI-positive group (12.9 vs. 2.7 %; p = 0.019). CDI was more frequent in those subjects with a nosocomial exposure (89.7 vs. 77.5 %, respectively; p = 0.004). This disparity was particularly evident with nursing home exposure prior to hospitalization and both nursing home and inpatients stays in combination (Table 3). All other variables analyzed, including antibiotic use, PPI use, admission to the intensive care unit, comorbid diseases, and numerous diagnostic and laboratory data points, were not found to be associated with CDI (Table 4).
Table 3

Clinical predictors significantly associated with CDI

Clinical predictor

Negative CDI

Positive CDI

p value

Nosocomial exposure

  

0.004

  None

25 (22.5 %)

3 (10.3 %)

 

  Nursing home only

3 (2.7 %)

5 (17.2 %)

 

  Hospital only

67 (60.4 %)

13 (44.8 %)

 

  Nursing home and hospital

16 (14.4 %)

8 (27.6 %)

 

History of CDI

  

0.019

  Negative history of CDI

110 (97.3 %)

27 (87.1 %)

 

  Positive history of CDI

3 (2.7 %)

4 (12.9 %)

 

Gender

  

0.016

  Male

71 (62.8 %)

12 (38.7 %)

 

  Female

42 (37.2 %)

19 (61.3 %)

 
Table 4

Clinical predictors not associated with CDI

Clinical predictor

Negative CDI

Positive CDI

p value

Physical examination

  Fever, defined as temperature >38.0 C

21.6 % (24/111)

20.0 % (6/30)

0.847

  Hemodynamic instability (blood pressure <90/60 or heart rate >100)

15.5 % (17/110)

16.1 % (5/31)

0.927

  Abdominal distension

34.5 % (38/110)

41.9 % (13/31)

0.449

  Abdominal tenderness, rebound, or guarding

45.5 % (50/110)

38.7 % (12/31)

0.504

Laboratory markers

  WBC >11,500 cells/mm3

40.5 % (45/111)

58.1 % (18/31)

0.083

  Left shift of neutrophils (bands >20 %)

41.1 % (39/95)

42.3 % (11/26)

0.908

  Increase in creatinine >50 % of baseline

17.9 % (20/112)

25.8 % (8/31)

0.324

  Elevated lactate

22.6 % (7/31)

28.6 % (4/14)

0.665

Radiologic markers

  Colonic distension

16.7 % (12/72)

15.0 % (3/20)

0.858

  Colonic wall thickening

36.1 % (26/72)

45.0 % (9/20)

0.469

  Peri-colonic fat stranding

19.4 % (14/72)

30.0 % (6/20)

0.311

  Ascites

25.0 % (18/72)

25.0 % (5/20)

1.000

  Ileus

13.9 % (10/72)

15.0 % (3/20)

0.900

Colonoscopic markers

  Presence of pseudomembranes

14.8 % (4/27)

50.0 % (1/2)

0.204

Exposures

  Antibiotic use

68.1 % (77/113)

83.9 % (26/31)

0.086

  PPI use

60.2 % (68/113)

54.8 % (17/31)

0.592

  ICU admission (current or previous)

35.8 % (39/109)

48.4 % (15/31)

0.388

  IBD (Crohn’s disease, UC, or indeterminate colitis)

12.4 % (14/113)

6.4 % (2/31)

0.351

  Human immunodeficiency virus

3.5 % (4/113)

9.7 % (3/31)

0.159

  Diabetes mellitus

20.4 % (23/113)

25.8 % (8/31)

0.513

  Immune-compromised state

15.0 % (17/113)

3.2 % (1/31)

0.078

  Systemic steroid use

11.5 % (13/113)

6.4 % (2/31)

0.415

  Chronic kidney disease

20.4 % (23/113)

19.4 % (6/31)

0.902

  Malignancy

23.9 % (27/113)

9.7 % (3/31)

0.084

  Nasogastric/gastrostomy/jejunostomy tube

16.5 % (18/109)

29.0 % (9/31)

0.119

Discussion

Diagnostic Test Performance

The optimal diagnosis of CDI has become somewhat challenging in recent years, not due to a lack of available tests, but rather due to new testing modalities and diagnostic algorithms. With the increasing incidence of CDI and financial pressures on the American healthcare system with respect to resource utilization, focus has turned to improving diagnostic modalities for early and accurate detection of CDI. Three TOX EIA tests comprise the traditional method for diagnosing CDI [6]. However, opinions diverge with regards to what the most efficient strategy for the diagnosis of CDI is, especially with the recent availability of new diagnostic modalities, such as the GDH EIA and DNA-A test, as alternatives for diagnosing CDI. Information on the independent verification and comparison of these new modalities to the traditional TOX EIA is limited. Although microbiological studies comparing the TOX EIA tests with variable combinations of the GDH EIA, commercially available DNA-A assays, and in-house developed DNA-A assays have been published [918], a clear comparison of these modalities in the clinical arena did not exist until now. Our study differs from previous evaluations in that it examined stool specimens selected from a population of patients whose demographic and clinical histories placed them at increased risk for CDI. The importance of this distinction in the study population was to reduce potential false positive results, in stark contrast to previous studies that examined every stool sample submitted for evaluation.

The performance of the TOX test, GDH test, and DNA-A assay, respectively, in our study differed markedly from that reported in previous studies [918]. The midpoint review by our internal data safety monitoring board identified a large number of false positive TOX test results, prompting us to invite an independent technical review team from Meridian Biosciences to review laboratory procedures. This team verified that proper laboratory procedures had indeed been followed, confirming that false positive TOX results were due to the inherent test performance and not to laboratory error. The sensitivity of the TOX test in our series was 86.1 %, which is consistent with the best reported sensitivities of TOX tests (range 48.0–86.4 % [918]). Its NPV in our study, 97.7 %, was also in the high end of the reported range in previous studies (87.0–98.5 % [918]). However, our results deviated from those reported in the literature for the PPV, which was 58.4 % in our study and thereby markedly lower than the reported range of 68.9–100.0 % [918]. This low PPV was due to a large number of false positive TOX test results. This finding was particularly surprising given the higher prevalence population described herein. The correct performance of laboratory testing leads us to conclude that there is great potential for over-diagnosis associated with TOX EIA testing as the diagnostic modality for CDI, which is compounded by the conventional practice of performing TOX testing multiple times.

The false positive rate of the TOX test may be due to an inherent test threshold that is overly sensitive in order to reduce false negatives, which underscores the need for a different diagnostic modality for CDI. The potential implications of false positive TOX test results are numerous. First and foremost, multiple patients would be unnecessarily subjected to prolonged and inappropriate antibiotic treatment. This may not only result in complications, but also contributes to the misdirection of significant healthcare resources from incorrect medications and from inpatient isolation practices. Incorrect diagnosis with CDI likely also contributes to poor clinical care outcomes with the resultant ignorance of the patient’s true etiology for diarrhea.

The GDH EIA and DNA-A test for diagnosis of CDI have superior test characteristics when compared to the traditional TOX test among populations at increased risk for CDI. The value of these newer assays lies in their low false positive rates and resultant improvement in specificity and PPV. Our study results for DNA-A testing are consistent with those of studies that examined similar DNA-A assays. We found the sensitivity of the DNA-A assay in our study to be 100.0 %, in comparison with the range proposed previously of 77.3–99.0 % [918]. Peterson et al. [12] examined 618 stool samples in their comparison of a different EIA test than that used in our study and an in-house designed DNA-A assay, with anaerobic culture as the gold standard, similar to our study. Compared to the commercially available DNA-A assay we evaluated in our study, the in-house DNA-A assay of these authors showed decreased performance statistics, with a sensitivity of 93.3 %, specificity of 97.4 %, PPV of 75.7, and NPV of 99.4 % [12]. Stamper et al. [13] examined 404 stool samples using a different commercially available EIA and a different commercially available DNA-A assay than that used in our study, with TGC as the gold standard. The test performance of the DNA-A assay used by these authors was superior to that of the TOX EIA, with a sensitivity of 83.6 %, specificity of 98.2 %, PPV of 89.5 %, and NPV of 97.1 % [13]. While the specific performance of each test may vary, the superiority of the DNA-A assay when compared to the TOX EIA is clear in their studies and was confirmed in our study.

This benefit of the DNA-A assay is likely to be related to the lack of sample degradation by fecal byproducts, in contrast to the toxin examined in the EIA, thus decreasing potential false negative results and increasing the sensitivity of the test. Specimen storage is also advantageous with assays involving DNA amplification. Before undergoing DNA-A testing, specimens may be stored up to 5 days under refrigeration (2–8 °C) or frozen (less than −20 °C) and may be thawed once without affecting test accuracy [19]. The primary drawbacks to DNA-A testing are the moderately increased cost of this modality and the additional training required for laboratory personnel performing the assay. These costs are decreasing with improvements in DNA-A technology and the widespread commercial use and availability of this modality.

Some investigators suggest the use of GDH testing as an initial screening tool, with positive test results to be confirmed by DNA-A testing [911]. In our study population, the test performance of the GDH EIA was similar to that of the DNA-A assay with respect to numbers of false positives and false negatives; however, the GDH test in our study had a number of inconclusive test results due to an invalid internal control. Further, GDH testing cannot distinguish between toxigenic and non-toxigenic strains of C. difficile, thereby requiring a second confirmatory assay, while DNA-A assays are capable of this differentiation in one step [20]. In our population, 85.7 % of the stools which tested positive for GDH also tested positive for TOX. Therefore, 14.3 % of the GDH-positive results had either a false negative TOX test result, or the patient may have been colonized with a non-toxigenic strain of C. difficile; each of these scenarios requires additional testing. Although both GDH testing and DNA-A modalities are superior to traditional TOX EIA testing for accurately identifying the presence of C. difficile, the increased cost of a two-step or three-step algorithm, time of test performance for a second, confirmatory assay, and high rate of inconclusive results of GDH testing make it unlikely that the GDH test will be useful as the primary test in the diagnostic algorithm for CDI in the patient population.

As with any bacterial or viral pathogen, it is possible that genomic variance among strains or the presence of differing ribotypes may affect the performance characteristics of diagnostic modalities. For C. difficile, this could account for minor variation among the GDH and DNA-A test results based on differences in the specific region that these tests identify. TOX EIA results would be less likely to be affected as the TOX test examines the presence of preformed toxin in the stool irrespective of differences in the genomic region that codes for the toxin. As with any microbiological test, the emergence of varied bacterial strains necessitates the re-evaluation and modification of diagnostics.

Clinical Predictor Analysis

Previous studies have identified a myriad of clinical variables associated with CDI [3, 6, 7]. These studies have primarily relied upon TOX EIA testing for CDI diagnosis. As we have shown, TOX EIA testing has marked inaccuracies. Accordingly, we sought to identify clinical variables associated with CDI diagnosis using enriched TGC as the gold standard for determining the presence of CDI.

Nosocomial exposure, defined as nursing home residence or hospitalization in the last 3 months, or current hospitalization with inpatient onset of diarrhea after 3 days, was strongly associated with CDI among our patient population (p = 0.004). Of the 29 patients for whom data were available in the group positive for CDI, 26 had been recently—or were currently— hospitalized or had nursing home exposure. Our study population had a high prevalence of nursing home exposure as our hospital system services an older population, and nursing home exposure was identified as the largest contributor to nosocomial exposure. The communal living environment of nursing homes may facilitate repeated exposure to C. difficile once a carrier enters the population [3]. Further examination of the specific interactions between hospital and nursing home exposure in increasing CDI acquisition is warranted.

Second, a history of previous CDI was positively associated with a diagnosis of CDI on the current admission (p = 0.019). Our study population only included patients with history of CDI who had received treatment and had cleared the infection. Patients with documented CDI prior to the study period who were on treatment and then re-tested for CDI during the study period were excluded. While the total number of patients in the study with prior history of CDI is small (n = 7), the distribution of these patients is strongly skewed toward the group positive for CDI.

Among our population, female gender was found to be associated with CDI (p = 0.016). This finding is unique to our study, and to the best of our knowledge such an association has not been previously published in the literature. We examined gender in relation to history of CDI and nosocomial exposure, particularly nursing home residence, and found no significant difference in the CDI-positive group based on gender. In a recently presented study examining fecal microbiota transplant as therapy for recurrent CDI by Mellow et al. [21], there was an increased predominance of women with recurrent CDI. The reason for this finding merits further investigation, but may be due to the effects of estrogen as a bacterial growth factor or increased exposure to antibiotics.

Many clinical characteristics, specifically medication use (antibiotics, PPIs, and immunosuppressants), ICU admission, malignancy, and diabetes, have been previously implicated as potential predispositions to CDI [3, 6, 7, 22]. We did not find these associations among our population. The major difference in methodology between our study and those previously reported is the assay used as the gold standard. Our study identified CDI via enriched TGC, while many other previous studies used the TOX EIA. Study limitations may explain differences in our findings compared to previous studies. In particular, our study was powered to demonstrate a 15 % difference in the sensitivity of the TOX EIA test and the DNA-A assay, but was not powered to show small differences in clinical predictors of CDI as this was a secondary study endpoint. Data on the clinical predictors were also collected by retrospective review of the electronic medical record, which may under-report clinical endpoints. Additionally, our strict inclusion criteria encompassing traditionally accepted high-risk features for CDI result in each of our patients being different from the average inpatient, making our study findings applicable to a higher risk group of individuals. Thus, the previously identified predisposing factors to CDI must still be considered.

Conclusions

The test performance characteristics of the DNA-A assay and GDH EIA are superior to that of the traditional TOX EIA. The GDH assay is limited by inconclusive test results and the inability to distinguish between toxigenic and nontoxigenic C. difficile strains, which necessitate additional confirmatory testing. These findings support the implementation of the DNA-A assay as the diagnostic method of choice for CDI and have resulted in a change in the diagnostic method used by our institutional laboratory. Using enriched TGC as the gold standard, we identified nosocomial exposure, with an emphasis on nursing home residence exposure, history of previous CDI, and female gender as predictive of CDI. Identifying patient characteristics associated with CDI is important to guide diagnostic decision-making to efficiently utilize healthcare resources.

Acknowledgments

Funding for laboratory testing provided by Mount Sinai Medical Center, Miami Beach, FL. No other funding sources identified.

Copyright information

© Springer Science+Business Media, LLC 2012