Recall bias in pain scores evaluating abdominal wall and groin pain surgery.

PURPOSE: To determine whether levels of pre-operative pain as recalled by a patient in the post-operative phase are possibly overestimated or underestimated compared to prospectively scored pain levels. If so, a subsequent misclassification may induce recall bias that may lead to an erroneous effect outcome. METHODS: Data of seven retrospective cohort studies on surgery for chronic abdominal wall and groin pain using three different pain scores were systematically analyzed. First, it was assessed whether retrospectively acquired pre-operative pain levels, as scored by the patient in the post-operative phase, differed from prospectively obtained pre-operative pain scores. Second, it was determined if errors associated with retrospectively obtained pain scores potentially lead to a misclassification of treatment outcome. Third, a meta-analysis established whether recall misclassifications, if present, affected overall study conclusions. RESULTS: Pain data of 313 patients undergoing remedial surgery were evaluated. The overall prevalence of misclassification due to a recall error was 13.7%. Patients not benefitting from surgery ('failures') judged their pre-operative pain level as more severe than it actually was. In contrast, patients who were pain free after remedial surgery ('successes') underestimated pre-operative pain scores. Recall misclassifications were significantly more present in failures than in successful patients (odds ratio 2.4 [95% CI 1.2-4.8]). CONCLUSION: One in seven patients undergoing remedial groin surgery is misclassified on the basis of retrospectively obtained pre-operative pain scores (success instead of failure, or vice versa). Misclassifications are relatively more present in failures after surgery. Therefore, the effect size of a therapy erroneously depends on its success rate.

Immunological alterations inducible by mercury compounds. III. H-2A acts as an immune response and H-2E as an immune "suppression" locus for HgCl2-induced antinucleolar autoantibodies.

In responder mouse strains repeated injections of subtoxic doses of HgCl2 induce formation of antinuclear autoantibodies (ANA) and antinucleolar autoantibodies (ANolA). Others have shown that responsiveness to HgCl2-induced formation of ANA and ANolA is linked to H-2. Here, we extend these studies to a variety of mouse strains not tested previously. After confirming that strain B10.S (H-2s) is a high responder we have shown that strains B10.D2 (H-2d) and B10.BR (H-2k) are nonresponders. By comparing a panel of strains carrying appropriate intra-H-2 recombinant haplotypes derived from d, k and s, we were able to map responsiveness to As. Interestingly, among four strains all of which were As, and thus responsive, only the two H-2E- ones, B10.S and B10.RSD2, were high responders whereas the two H-2E+ ones, B10.HTT and B10.S(9R), were significantly less responsive. Thus, the genetics of HgCl2-induced autoantibody formation follow the rules established for immune responses to a variety of different antigens in that expression of H-2E "suppressed" the response.