The study by Linna et al posited that “eating disorders appear to be associated with several adverse perinatal outcomes, particularly in offspring.” The adverse outcomes included anemia, slow fetal growth, premature contractions, and perinatal death. However, this conclusion cannot be supported by the data because the authors failed to correct the standard value of P = .05 to account for the large number of hypothesis tests. This leads to what is known as type 2 error and causes a hypothesis to be accepted that is actually false.

In this study there are a total of 78 hypothesis tests conducted, as shown in Tables 1 and 2 of the study cited previously. By chance alone, 1 in 20 of these, or about 5 tests, will be statistically significant if the data are random numbers. The critical P value can be corrected by many methods including the Bonferroni correction.

To obtain a true measurement of statistical significance, P = .05 must be divided by that number of tests in the study, which would be 0.05 of 78 = 0.000641. If the corrected value is used to avoid a type 2 error, only 1 of the 78 tests in the study is statistically significant: the association between a binge eating disorder and hypertension at P = .0000146 (adjusted) or P = .0000107 (adjusted for age, parity, smoking status, and marital status), both of which were below .000641.

Given the importance of correcting the P value for the number of comparisons conducted, we suggest that the conclusion in the study mentioned in previous text be limited to the link between a binge eating disorder and hypertension during pregnancy. Hypertension in binge eating disorder appears to be an important problem that requires early detection and intervention. On the other hand, the data do not support the conclusion that there is a link of anemia, slow fetal growth, premature contractions, or fetal death to anorexia nervosa, bulimia nervosa, or binge eating disorder.

This study has several important strengths, such as a large sample size, the inclusion of disordered eating other than restriction (bulimia and binge eating disorder), comprehensive statistical analysis, and examination of a large number of perinatal complications, among others. However, many of the conclusions cannot be supported because of the use of an uncorrected P value as the decision criteria for the scores of comparisons. We propose that obstetricians and midwives can monitor patients with moderate eating disorders and their newborns with regular protocols unless further evidence to the contrary develops.