We read with interest the recently published article by Reibaldi and associates, “Enhanced Depth Imaging Optical Coherence Tomography of the Choroid in Idiopathic Macular Hole: A Cross-sectional Prospective Study.” The authors state that the study followed a cross-sectional, prospective design. Cross-sectional studies involve sampling from a reference population at a single point in time, evaluating the exposure and the disease simultaneously. As described, the authors did not, in fact, conduct a cross-sectional study; rather, it was a matched case-control study. Patients were selected if they had a full-thickness idiopathic macular hole in a single eye (ie, cases) at presentation and an unaffected eye, which was used as a control. Additionally, a second group of controls were selected from the general outpatient population without macular hole and were matched to cases based on age and sex. Unfortunately, this is more than a semantic issue because, as a result of failing to appreciate their actual study design, the analysis of the study data was inappropriate.
The statistical tests used by the authors in their analysis, namely the analysis of variance and Student t test, require the assumption of independence between comparison groups. However, because controls were either an unaffected eye from the same person or an age- and sex-matched control, this is a violation of independence. Statistical methods that account for this dependency are required, such as the paired t test and repeated-measures analysis of variance.
Additionally, although the correlations between subfoveal choroidal thickness and axial length and age were conducted appropriately, the manner in which the results were presented can be misleading. Based on the data presented in Table 3, subfoveal choroidal thickness was correlated negatively with age in the macular hole group and fellow eye group; however, these correlations were not statistically significant. Given the limited sample size and resulting limited power, the lack of statistical significance could be the result of a type II error. Therefore, it is important to consider the magnitude of the association, rather than relying on a P value.
We hope the authors will consider these points and correct their analysis.