Editors’ response

In 2021, concerns were raised by a reader about logistical and biological implausible findings in this single-centre, double-blind, placebo-controlled, RCT conducted at Kasr Al Ainy Medical School family planning clinic, Cairo University, Cairo, Egypt, between April 2017 and November 2017. Logistic concerns about this study were the unusually fast submission within 14 days after completion. Biologic concerns included that in both the Misoprostol group and the Placebo group, the difficulty scores during IUD insertion in women with previous 1 CS were significantly lower than those in women with previous ≥2 CS (3.2 vs. 5.94 and 4.6 vs. 6.7, respectively) which is difficult to understand.

According to the COPE guidelines, the editors asked Dr. Maged for the ethics document, protocol and original data, which he kindly provided. However, the editors ascertained many problematic findings in this dataset including that the dataset was not presented in the order in which it was randomised. On our request, Dr. Maged replied that the computerised form was available with the statistician who passed away the last years during the pandemic. For this reason, the files available in handwriting form were transformed by their team to the required form with the mentioned dates.

The provided dataset was not presented in the original randomised setting: it started with 15 participants randomised to group 1, followed by 60 participants to group 2, and then 45 participants randomised to group 1. Also, no participant ID numbers or study numbers were supplied. Other findings in the presented data set were: (a) from row 9 to 121, the value of ‘n of CS_3’ was completely conditioned on the value of ‘previous IUD’. The probability that this happens by chance is <0.00001. (b) Data on age and BMI did not follow normal distributions, with overrepresentation of some values and lacking values where some are expected.

On our request, a member of the Journal’s editorial team and an independent reviewer studied the provided dataset and they agreed that the dataset showed unusual patterns, but could not verify the validity of the data in the absence of an original randomisation list including the allocation sequence and dates of allocation for individual participants.

In order to obtain more clarity, the individual participant dataset of this RCT was sent out for confidential post-publication review by an independent statistician. This reviewer confirmed that the dataset purportedly contains the individual participant data from this RCT of vaginal misoprostol compared to placebo prior to IUD insertion in women with previous caesarean delivery. Analyses were conducted to examine whether the data were compatible with a genuine randomised controlled trial and whether they contained signs of fabrication. In short, the findings were as follows:

• Visual examination of numerical baseline variables age, BMI and previous CS shows (a) that the distribution in the treatment group substantially differs from the control group and (b) that variable values in the treatment group appear to have patterns that are suggestive of fabrication in blocks.

• Visual examination of numerical outcomes pain score and insertion difficulty score shows the same.

• The null hypothesis of no autocorrelation is strongly rejected for several variables in the treatment group (age, number of CS, difficulty inserting), as is the number of CS in the control group.

• These findings are confirmed by the Wald–Wolfowitz runs test of randomness.

• This all strongly suggests that the data have not arisen from a genuine RCT.

As a result of these findings of an independent statistician, the editors concluded that they no longer have confidence in the integrity of the data underlying the study as provided by Dr. Maged, and therefore have decided to retract the article.