We agree with Dr. Nardone that clinicians use patterns of clinical and laboratory variables to solve problems. We have analyzed the manner in which the sequential addition of our identified predictors affected the diagnosis of C. difficile disease (Table 2). Although this information may be useful, the results are not statistically significant. For practical purposes, physicians will suspect C. difficile-associated enteric complications in any patient with otherwise unexplained diarrhea that occurs with exposure to antibiotics. Decisions about use of laboratory testing are optimally dictated by the effect of the test results on management and by the probability of a positive result. Cephalosporins are the major agents implicated in C. difficile-associated diarrhea in recent years [1, 2]. Usage rates indicate that clindamycin is probably the most commonly used agent, but not enough patients in our series were receiving clindamycin to allow us to verify this well-known association.

Fang and colleagues summarize the findings of some of their previously published reports on the sensitivity of tests for C. difficile toxin. The major point of our paper was not that two to three stool specimens are needed for adequate diagnostic sensitivity. Rather, we prospectively evaluated the combination of clinical and laboratory data to establish a better clinical sense of how much testing needs to be done. Our previous studies show that the rate of false-negative test results is not greatly influenced by the selection of reagents that are commercially available for enzyme immunoassays [3]. Most important is the stool dilution used in the assay [4]. We believe that the toxigenic culture test Fang and colleagues propose is unrealistic for routine laboratory use in the United States. The attractive feature of this test is its sensitivity. Its disadvantages are that 1) most clinical laboratories do not offer C. difficile cultures, 2) the anticipated cost would be high [culture plus toxin assay], 3) results would be delayed 3 to 5 days, and 4) most laboratories do not do strain typing. An additional concern is specificity. MacFarland and colleagues [5] showed that about 30% of hospitalized patients have C. difficile in the absence of diarrhea; most of these strains are toxigenic.

We appreciate the remarks of Drs. Minochi and Richards and agree that the C. difficile toxin assay is associated with false-positive results. The literature they cite for this conclusion is our own, but it is limited to studies of patients receiving a few antibiotics. We have even greater concern about specificity with culture because of high carriage rates of C. difficile, as noted above. However, to keep the observations in perspective, it is important to emphasize that almost all enteric bacterial pathogens are associated with relatively high rates of false-positive results on standard tests. This applies to salmonella, shigella, Vibrio cholerae, Campylobacter jejuni, and other organisms. The requirement for the clinician is clinical correlation.

Dr. Holland correctly points out that both lactoferrin assay and Gram stain provide different ways of measuring the presence of fecal leukocytes. We wish to correct the impression that they were additively predictive of C. difficile disease; they were not. We separately analyzed lactoferrin and Gram stain for fecal leukocytes only to validate the utility of lactoferrin as an adjunct in diagnosing C. difficile disease.