We thank Dr. Koulaouzidis and colleagues for their interest in our study. Data presented at the European Association for the Study of the Liver meeting in 2005 reported 2123 ascitic samples.1 For each patient, a test was repeated during the 5-month period of the study, and at the time the abstract was submitted, 1069 patients were included. However, some of them were still duplicates and we thus corrected the number of patients, which was in fact 1041. We also recalculated sensitivity, which was slightly different and remained low, at a threshold of 2+. Two Multistix 8 SG strips were tested independently. The 2 dipstick readers examined and read the dipstick they had independently immersed in the ascitic sample. The reported concordance refers to a total of 4246 readings.
Dr. Lopes and colleagues and Dr. Castellote point out the poor sensitivity of the dipsticks and suggest lowering the cut-off for diagnosis of spontaneous bacterial peritonitis (SBP). When the cut-off was lowered to 1+ (corresponding to 70 leukocytes), sensitivity was 64.1%, specificity was 95.4%, positive predictive value was 44.9%, and negative predictive value was 97.9%. When the cut-off was lowered to “traces” (corresponding to 15 leukocytes), sensitivity was 78.6%, specificity was 92.2%, positive predictive value was 36.9%, and negative predictive value was 98.7%. Dr. Lopes and colleagues suggested using the receiver operating characteristic (ROC) curve to determine the best cut-off for diagnosis. Indeed, using the ROC curve, the best cut-off would be that of detection of traces. In that case, sensitivity was 78.6%; however, positive predictive value was quite low (36.9%). As Dr. Lopes and colleagues emphasize, a diagnostic test must be highly sensitive in this potentially lethal disease, and we feel that sensitivity should be higher than 95%. Taking into account these thresholds, sensitivity in our study was still too low for a potential lethal disease (less than 65% and 80%, respectively).
We thank Dr. Castellote for his comments. Sensitivity of the test depends on the defined cut-off, not on results observed by the physician or nurse. For this reason, diagnosis of SBP was based on the highest value when there was discordance between the 2 readers. In our Discussion section, we showed that differences between previous studies and ours were due to inclusion of a greater number of samples with fewer polymorphonuclear cells (<1000/μL), and this may account for the weak sensitivity of the test. We agree with Dr. Castellote that a clear positive test is indicative of SBP, but users of reagent strips should be cautious when a test is negative, particularly in a patient with clinical symptoms. Dr. Gülberg and colleagues report their study comparing Multistix 10 SG and Combur reagent strips for the diagnostic value of SBP, and they conclude that sensitivity of the 2 reagent tests is low, although that of the Combur reagent strips is slightly better, in accordance with previous studies.2 They raise an interesting hypothesis, suggesting that high ascitic protein content may alter results of the test. We also observed a higher mean ascitic protein content in patients with false-negative results than in patients with true-positive results (1.9 versus1.4 g/dL, respectively; P = 0.016). There may exist an interaction between ascitic protein content and the activity of leukocyte esterase, which should be tested in other studies.
Specific studies for ascites should be performed by dipstick manufacturers. In the meantime, however, our conclusion is that relying on a negative result of the dipstick to avoid diagnosis of SBP may be too risky; measurement of polymorphonuclear cell count in ascitic fluid should be retained for diagnosis of SBP.