We appreciate Drs. Rhee and Baker's interest in our systematic review and would like to offer some comments on their letter. It is true that the findings of our meta-analysis should be interpreted with caution, and we specifically mentioned this in the Discussion section of the article. As was demonstrated in the sensitivity analysis, the results of the meta-analysis, e.g., the overall diagnostic OR of 27.26, were highly influenced by 2 of the studies included. When these 2 studies are excluded from the meta-analysis, the diagnostic OR decreases to half this value (13.46). In 1 of these studies (1), anti-p155 determination had perfect sensitivity and specificity (100%) for the diagnosis of cancer-associated myositis, and we cautioned that it is extremely improbable that such optimal performance would occur in standard clinical practice.
We also agree with Rhee and Baker that studies are more likely to be published if significant associations are found, giving rise to publication bias, which can affect the validity of the results of a meta-analysis. Nonetheless, the discovery of this antibody is quite recent and its characteristics have not yet been fully defined; hence, any study related to anti-p155 would be relevant and novel, even if the association with cancer-associated myositis was modest or nonexistent. Actually, we also included a caveat in the Discussion section stating that the power of the statistical test used did not suffice to detect this potential bias and that the “absence of statistical evidence of possible publication bias in our analysis should, however, be viewed with caution.” Thus, we are in complete agreement with the comment that the P values should be deemphasized.
Another point on which we agree is that evidence of publication bias would compromise the validity of our meta-analysis, but this evidence would have to come from an unpublished study that yielded negative results for anti-p155, not from a statistical technique. The “trim and fill” method repeats the meta-analysis, including data from fictitious studies to compensate for asymmetry in the funnel-plot (2). The authors who described this method caution that the results do not always reflect correction of publication bias since funnel-plot asymmetry can be due to causes other than this bias. For this reason, they recommend use of the trim and fill method as part of a sensitivity analysis and advise that the results obtained with this technique should not be used to form a final conclusion (2). After applying trim and fill, Rhee and Baker obtained a diagnostic OR of 13.6, which is very similar to the 13.46 obtained in our sensitivity analysis when the 2 studies mentioned above (1, 3) are excluded.
In conclusion, in accordance with the opinion of Rhee and Baker, we believe that the results of our systematic review should be viewed with caution, since they are influenced by studies with extreme results, as was seen in our sensitivity analysis. Nevertheless, in our opinion, these comments do not detract from the main conclusion of the study, that anti-p155 determination is useful in diagnosing cancer-associated myositis, at least from the clinical perspective. Only the use of anti-p155 autoantibody testing over time will verify its true diagnostic value in myositis patients.