Is the association of replete status in vitamin D with better results in IVF demonstrated? Free

Sir,

We read with attention the interesting article ‘Vitamin D and assisted reproductive treatment outcomes: a systematic review and meta-analysis’ (Chu et al., 2018), since we are performing a similar work (Iliuta, 2018). After a thorough review of their data, we found some apparent mistakes in the data retrieval from the references and in the data classification process. Reanalysis of corrected data leads, in our view, to different results and inferences not in agreement with those proposed in the paper.

The authors reviewed 11 publications, including 2700 participants and compared vitamin D replete women to those with deficient or insufficient vitamin D status. It was concluded that replete vitamin D women were more likely to achieve pregnancy (OR 1.34 [1.04–1.73]), clinical pregnancy (OR 1.46 [1.05–2.02]) and live birth (OR 1.33 [1.08–1.65]) than vitamin D deficient and insufficient women.

One of the six included studies for the live birth calculation (Polyzos et al., 2014) was reported as having 61 events out of 129 replete women vs 139 in 368 ‘deficient and insufficient women’. However, (Polyzos et al., 2014) original manuscript presents a live birth rate of 61/129 for the ‘replete and insufficient’ group and 78/239 for the deficient group. Thus, if these data were excluded, statistical significance would not be reached, since the resulting OR would be 1.05 [0.83–1.33].

When biochemical pregnancy rate was considered, a similar problem seemed to occur in reporting the data of the same author (Polyzos et al., 2014). The meta-analysis (Chu et al., 2018) considered 86 events in 129 vitamin D replete women and 210 in 368 ‘insufficient and deficient’ women. However, the original work provided different data: 86/129 in vitamin D ‘insufficient and replete’ women and 124/239 in deficient women. We repeated the analysis, excluding these participants for not meeting the study criteria. The obtained OR was 1.02 [0.77–1.37], again being lower than reported by the authors, and without statistical significance.

When clinical pregnancy rates were analyzed, we observed some other discrepancies. The data from Polyzos were reported in the meta-analysis as follows: 70/129 in vitamin D replete women and 168/368 in ‘insufficient and deficient women’. However, the original data are: 21/37 in replete women, 49/92 in insufficient and 98/239 in deficient participants, which makes 147/331 in ‘insufficient and deficient’ women instead of 168/368. In the same outcome, the data extracted from another author (Garbedian et al., 2013) do not match with the original work. In the meta-analysis, it was reported that there were 66 events in 190 ‘insufficient and deficient’ women instead of 33/95 women as given in the original work (Garbedian et al., 2013). The obtained OR with the revised numbers is: OR 1.29 [1.06–1.55] which is somewhat lower than reported by the authors, but still statistically significant.

Finally, in the meta-analysis of the clinical pregnancy rates in autologous oocytes, Chu et al. (2018) provided the same confusion in the data for Polyzos et al. (2014), as shown in the previous paragraph.

According to our interpretation, in the aforementioned meta-analysis using the revised data, the replete vitamin D status only demonstrated to be significantly associated with the clinical pregnancy rates (OR 1.29 [1.06–1.55]), a significant but weak association. Regarding live birth rates (OR 1.05 [0.83–1.33]) and biochemical pregnancy rates (OR 1.02 [0.77–1.37]), a trend to better results was observed, but without reaching statistical significance.

Larger meta-analysis should be published in order to solve those questions.

Conflict of interest

None declared.

References