Inconsistencies in a Study of Rifampicin‐Miconazole–Impregnated Catheters versus Standard Catheters

To the Editor—We read with interest the article by Lorente et al. [1] regarding the comparison of rifampicin-miconazole–impregnated catheters with standard catheters. We are puzzled by several factual errors and inaccuracies throughout the article. First, Lorente et al. [1] report using a test that compares continuous outcomes (the Kruskal-Wallis test) to analyze categorical data (namely, sex, diagnosis group, and reason for catheter removal) across catheter types. Second, figures 1 and 2, table 1, and the Results section in their article are internally inconsistent. The Kaplan-Meier curves in the figures suggest that 100% of patients in the standard catheter group experienced catheter-related bacteremia by the end of the follow-up period. In contrast, table 1 and the Results section clearly report that only a minority of patients experienced an event (8 events among 111 femoral standard catheter placements and 6 events among 127 jugular standard catheter placements). Most likely, both figures 1 and 2 are incorrect. Third, additional information is missing. Because the study was probably retrospective (described as involving a “historical cohort”), the assertion by Lorente et al. [1] that outcome assessment was blinded should be supported by a clear description of the blinding mechanism. The authors should clarify that either the type of catheter was never mentioned in patient notes in their hospital or that an investigator (who was not part of the outcome adjudication panel) was blinded to all patient notes with respect to catheter type before the notes were given to the panel. Furthermore, although we know the total number of catheters, it is unclear whether a single catheter or multiple catheters were used per patient. Table 1 suggests the latter (at least for some patients), because it reports the frequency of catheters inserted first, second, and third. If this is true, the authors (1) should have used different statistical analyses to account for the nesting of multiple catheters per patient and (2) should have clarified whether there were crossovers between catheter types (and how they were accounted for in the analyses). Use of the catheter as the unit of the analysis may render the comparison of clinical characteristics across groups misleading (e.g., patients who received 3 catheters would be represented 3 times in the counts in table 1). However, even if only 1 catheter per patient was studied, the authors should justify why they studied the first catheter in some patients and the second or third in others. Most importantly, the study by Lorente et al. [1] was nonrandomized, unadjusted for potential confounders. This is a major flaw, irrespective of the apparent balance between the compared groups in table 1. It is well appreciated that nonrandomized studies are subject to a plethora of biases [2]; even when they are analyzed properly (e.g., with use of propensity score matching [3] or instrumental variable regression analyses [4]), one can never be confident that the findings are unconfounded. Because of the aforementioned caveats, we are skeptical about the findings of the study. Potential conflicts of interest. N.A.T. and T.A.T: no conflicts.