Response to Mitter et al. re: “The Association of Assisted Reproductive Technology and Breastfeeding Rates: A Systematic Review and Meta-Analysis”

We appreciate your opinion ¹ and I would like to reply to the points that you mentioned. In the current meta-analysis, we assessed the association between assisted reproductive technology and breastfeeding rates. ² To achieve this goal, we selected studies in which the breastfeeding outcome was categorized based on the mode of conception (assisted reproductive technology and spontaneous conceptions). In the qualitative synthesis of studies, we have 18 studies and 6 articles were excluded.^3–20 Therefore, 12 articles with 4,929 subjects were retrieved in this systematic review.

The reasons for exclusion are as follows: In McDonald's study, ¹⁵ the breastfeeding outcome did not have categorized by the mode of conception and assisted reproductive technology conceptions were entered as a covariate in a multivariable logistic regression model. In Purtschert's study, ¹⁶ breastfeeding outcomes have been categorized by the in vitro fertilisation (IVF) and Swiss Infant Feeding Study population, which includes the general population without sufficient explanations about the type of conception in it. Also, IVF was entered as a covariate in a multivariable logistic regression model. The studies of Zegers-Hochschild et al, ¹⁷ Kermani et al, ¹⁸ Wiffen and Fetherston, ¹⁹ and Thevenet et al ²⁰ had insufficient data.

The odds ratio and relative risk for breastfeeding outcome were extracted from the studies. For articles that did not report these data, if the number of subjects in assisted reproductive technology and spontaneous conceptions categories was available, the odds ratio and 95% confidence interval were calculated using Stata software.

In Saleh's study, breastfeeding outcomes have been assessed in different follow-up periods (1st week—1-hour—24-hour—at discharge—<6 months—6–12 months, and >12 months after delivery), all of which had a frequency above zero, so they were included in the study. Whereas zero the frequency of breastfeeding that you have mentioned, since the follow-up period of this finding was not known, it was not included in the analysis. ⁵

In included studies, each follow-up period was considered as separate data. Since the follow-up period is an important factor, the analysis was performed based on follow-up periods. In this manner, the authors' names have been located independently in each follow-up period. It should be noted that different divisions of the follow-up period give different results. We performed the analysis based on the follow-up period of <6 months and >6 months in two separate forest plots because each of the studies included several follow-up periods and we did not want to lose data. Whereas in your forest plot only significant results of Saleh's study has been included. In this way, we will lose other data.

In this study, the “metan” command of Stata software was used, according to its instructions, there is no need to enter the log of the odds ratio or relative risk. In our study, the reference line is 1, which is visible as a dotted line in forest plot.

There are several strategies for finding the source of heterogeneity in a meta-analysis. ²¹ Depending on the type of covariates, subgroup meta-analysis or meta-regression may be used to explore the between-study heterogeneity. Subgroup meta-analysis is commonly used with categorical covariates, whereas meta-regression is used when at least one of the covariates is continuous. The goal of subgroup analysis is to compare these overall estimates across groups and determine whether the considered grouping helps explain some of the observed between-study heterogeneity known in the meta-analysis literature as moderators.

In our study heterogeneity was high, but subgroup analysis showed some source of heterogeneity. Researchers in future studies can design subsequent studies based on it.

All studies have strengths and weaknesses. Conducting a meta-analysis and its results do not necessarily mean using its results in practice. The results of our study give researchers a clear view of the need to conduct more studies with different follow-up periods to reach appropriate conclusions.