This is an ancillary study of the WHO CHAMPION trial which collected data on pre and post-delivery hemoglobin levels during the period of 12 months. The primary outcome was post-delivery 48-72 hours hemoglobin. The secondary outcome was the proportion of women with a drop in post-delivery hemoglobin ≥ 2 g/dl after 48-72 hours of delivery [11].
The WHO CHAMPION trial was a randomized, double-blind, non-inferiority trial comparing the effectiveness of HS carbetocin with oxytocin for the prevention of PPH after vaginal birth. The trial methods and results are described in detail elsewhere [11]. This ancillary study was conducted in one of the six participating sites in India, and it consisted of including the additional measurement of pre- and post-delivery hemoglobin to the WHO CHAMPION protocol. Design and procedures are the ones described in the main study [11].
After initial clinical assessment of women presenting for delivery to the KLES Dr Prabhakar Kore Charitable Hospital & Medical Research Center, Belgaum, a written informed consent was taken separately for the ancillary study along with the consent for the main trial.
Pre-delivery hemoglobin of each participant was recorded before she was randomized to the main trial. Estimation of hemoglobin was done using the HemoCue Hb 301 with the technology of absorbance measurement of whole blood at Hb/HbO2 isobestic point with a cuvette. Under aseptic conditions, 0.5 ml of the capillary blood was collected after finger prick with a sterile lancet and placed on the cuvette. The time required by the hemoglobinometer for the hemoglobin estimation was approximately 10 seconds. Hemoglobin values were recorded on a form specifically designed for this study.
During the second stage of labour when vaginal delivery was imminent, eligible women were randomized to receive either oxytocin 10 IU IM or HS carbetocin 100 µg IM based on the random allocation sequence generated centrally at WHO Headquarters (the CHAMPION trial coordination centre) using computer-generated random numbers. During the third stage of labour women received the randomly assigned treatment, and blood loss was measured for one hour or two hours if bleeding continued after the first hour [17].
Post-delivery hemoglobin was measured using the same standard procedure used for estimation of pre-delivery hemoglobin between 48-72 hours postpartum.
The present study was approved by JNMC Institutional Ethics Committee for Human Subjects. The trial was registered with Clinical Trial Registry of India CTRI/2016/06/006996.
Statistical analyses
The main analysis was conducted with the modified intention-to-treat population (as defined in Widmer et al 2018).
Descriptive statistics were calculated for characteristics of women and those of babies at birth. The number of participants, number of events and percentages in each category were calculated for the categorical variables. Number of participants, median and interquartile range (IQR) were calculated for the numeric variables.
Univariate analyses were conducted with the variables: pre-delivery hemoglobin, post-delivery haemoglobin, the drop from pre-delivery to post-delivery hemoglobin, and blood loss, to compare the treatments, using analysis of variance techniques. The statistical distribution of the variables was examined to decide whether a transformation was needed. The difference of means was used if the distribution was close to normal. In the case of blood loss, the distribution was asymmetric and the logarithmic transformation was used.
The primary outcome, post-delivery hemoglobin, was analysed using analysis of covariance to compare treatments, with pre-delivery hemoglobin as covariate to adjust the post-delivery hemoglobin for possible differences in baseline values (Model 1). Analysis of covariance is the preferred technique compared to the analysis of post-pre-difference [18]. Adjusted means of post-delivery hemoglobin were calculated, with 95% confidence intervals.
Regression analysis was used to assess the association of post-delivery hemoglobin and blood loss volume between the treatments (Model 2). This association was explored by including an interaction term of blood loss by arm in the model in order to allow the treatment regression lines to have different slopes (Model 3). The association between post-delivery hemoglobin and blood loss was quantified by the treatment-specific regression coefficients
The proportion of women with a drop in post-delivery hemoglobin of 2 g/dl or more from pre-delivery hemoglobin was calculated for each treatment and compared using relative risk (RR) with 95% confidence intervals and risk difference with 95% confidence intervals. Both unadjusted and adjusted analysis for pre-delivery hemoglobin were conducted.