We read with interest the article by Cui et al. [1]. In the article, the efficacy and safety of conbercept and ranibizumab for neovascular age-related macular degeneration were compared in a retrospective case-controlled non-inferiority study. However, the authors made some mistakes in describing the study design.

First, a case-control study is often used to identify potential risk factors for a disease by comparing the frequencies of the risk factors of an illness group to one or more control groups [2]. The researchers first group the participants in a case-control study according to their outcome status and then looks backward to compare the levels of exposure between groups. In Cui’s study, however, 180 patients were divided by the status of intervention/exposure and then were followed longitudinally for the outcomes, which was essentially a forward-looking study instead of a backward-looking case-control study.

Second, the retrospective design mentioned in the article was also doubtful. By definition, a retrospective study is always an observational study, which is more subject to bias and confoundings [3]. Missing data is also the Achille’s heel of a retrospective study. But what we saw in the article was a well-controlled multicentre study. The interventions were chosen by the participants after only been informed with the names of the drugs. The baseline characteristics were balanced. And the attrition proportion was merely 6.7% with no documented treatment switching. However in a similar study in California, 14.4% of the 452 participants were either lost to follow-up or died, and another 17.3% had changed their treatments [4].

Finally, a non-inferiority design is almost always used in randomized control trials [5].

We believe quasi-experiment should be a more appropriate term for Cui’s study. A quasi-experiment prospectively enrolls participants and assigns them to different arms according to a pre-specified non-random allocation strategy [3, 6]. The study in question had a so-called weak quasi-experimental design as the participants chose to receive either conbercept or ranibizumab treatment all by themselves [6]. Thus, the groups would have been different in a number of ways.

To carefully design a study and transparently report it are important, not only for minimizing the risks of bias and controlling for potential confounders but for properly interpreting the results of studies and correctly ranking the evidence.