In a chart survey of pregnancy in Behçet's disease (BD), Noel and colleagues found that pregnancy favorably affected the clinical course of BD and that outcome of pregnancy was good (). Several issues require further discussion.
Disease duration was found to be significantly shorter (by a median of 4 years) in patients who experienced a greater number of flares during gestation. It is well recognized that BD becomes less severe with the passage of time (). Thus, the inherent disease course in these patients is another factor that might have influenced the smaller number of flares during pregnancies. Consistent with this, those patients who received colchicine had fewer flares. It is probable that patients who did not experience the more severe manifestations of BD did not require therapy with agents other than colchicine, which is usually reserved for the treatment of arthritis and some forms of skin/mucosa involvement in BD, while more severe BD requires therapy with immunosuppressive drugs and/or biologic agents (). More simply, from this survey, we cannot be sure whether it is the pregnancy that affected the disease course or the other way around.
It is worth noting that the number of ocular flares during the postpartum period significantly increased; this can be taken as evidence against the notion of the potential bias we have mentioned. This, however, brings us to the second important methodologic problem in this study. All outcome information, including incidence of flares, was reported per pregnancy rather than per patient, while the baseline information (assessed at the beginning of the pregnancy) was reported for the whole group. Thus, we do not know the concordance of the flares or the pregnancy outcomes in patients with >1 pregnancy. Further, as particularly related to the bias we note, we do not know what type of ocular disease patients had at baseline before experiencing an ocular flare during the postpartum period. Without correction for these important methodologic shortcomings, a fair interpretation of the data at hand, including statistical comparisons based on the data as displayed in Tables 2 and 3 of the article, is very difficult.