Association between maternal SARS-CoV-2 infection & clinical outcomes in infants: A multicentric retrospective cohort study in India

Materials & Methods

This multicentric retrospective study was undertaken by the department of Health Systems and Implementation Research, International Institute of Health Management Research. Approval was obtained from Ethical Review Boards of all participating institutions (Agartala Government Medical College, SCB Medical College, Gandhi Medical College and SMS Medical College). Participants were explained the details and purpose of the study before initiating the telephonic interviews. Women who wished to participate provided their informed consent verbally, following which the interviews were conducted. After required information was obtained from the participant, her case file was identified from the hospital for recording required data. All data collected were anonymised and kept confidential with the principal investigators only. If any woman reported illness in her infant, she was advised to visit the hospital at the earliest to seek treatment. Under ideal circumstances, all examinations or assessments would have been done physically by a doctor. However, due to prevailing conditions owing to the pandemic, concerns were raised by the research team and the hospitals. On ethical grounds, contact with the child and mothers was kept to a minimum.

Results

A total of 1696 women were identified, of whom 1668 were included in the study after they provided consent. Of these, 789 women were recruited to the exposure cohort and 879 women were recruited to the comparison cohort. Exposure cohort reported COVID positive results through RT-PCR (n=271, 34.5%) or RAT (n=518, 65.6%), while the comparison group demonstrated negative results through RT-PCR (n=346, 39.4%) or RAT (n=533, 60.6%). Most women in both groups were tested for COVID-19 as part of routine testing in the third trimester of pregnancy and few underwent testing for the first time when they were admitted for delivery (those women availing antenatal checkups in other facilities but reported to the study sites for delivery).

Significantly greater proportions of women in the exposure group had anaemia, GDM and eclampsia. Proportion of stillbirths was higher among the exposure cohort [relative risk (RR) 1.94; 95% CI 1.17, 3.21; P=0.01]. Among livebirths, admission to SNCUs (RR 1.98; 95% CI 1.60, 2.45; P<0.001) and illness of the baby in the six months preceding the interview (RR 4.23; 95% CI 3.14, 5.69; P<0.001) were higher among babies born to COVID-positive mothers (Table I). Reasons for admission to SNCU/NICU among babies born to COVID-positive mothers included birth asphyxia/respiratory distress (n=48, 6.1%), preterm or very low birth weight (n=56, 7.1%), for observation since they were born to COVID-positive mother (n=76, 9.6%), and other causes (n=16, 2.0%). The illnesses reported among the exposure cohort babies included fever (n=132, 16.7%), upper respiratory infection (n=66, 8.4%), diarrhoeal disease (n=40, 5.1%), pneumonia (n=16, 2.0%) and jaundice (n=2, 0.3%). Among the comparison cohort babies, mothers reported febrile illness (n=29, 3.3%), upper respiratory infection (n=21, 2.4%), diarrhoeal disease (n=5, 0.6%) and jaundice (n=1, 0.1%).

Exposure to COVID was significantly associated with a higher risk of stillbirths [Adjusted OR (aOR) 2.63; 95% CI 1.39, 4.96] and SNCU admission (aOR 2.57; 95% CI 1.85, 3.58), when adjusted for region, cohort allocation, mother’s age, pregnancy order, preterm birth, previous h/o abortion, anaemia in pregnancy, PIH, GDM, eclampsia, preterm labour, type of delivery, infection in pregnancy and handwashing practices. No association was observed with preterm labour or preterm births.

Table II presents the results of PSM analysis. Average treatment effects on treated cases are computed from differences in outcome means within matched samples. Raw estimates, prior to matching, are obtained from the unmatched sample, while average treatment effect on the treated (ATT), average treatment effect on the untreated (ATU) and average treatment effect (ATE) represent estimates post-matching. It was observed that before matching, exposed mothers had a 2.8 per cent higher stillbirth risk compared to those with a negative result. The ATT values for the exposed and control groups are 0.047 and 0.027. This suggests that if mothers who tested positive for COVID-19 had not tested positive, the occurrence of stillbirths would have been only 2.7 per cent. Likewise, the differences in ATT values for birth weight among mothers testing positive for COVID and those testing negative indicate a 704 g lower birth weight for COVID-positive cases. The ATT values for SNCU admission for exposure and control groups were 0.257 and 0.153, respectively, indicating that if mothers testing positive for COVID-19 had not tested positive, SNCU admissions would have occurred only in 15.3 per cent compared to 25.7 per cent as observed. ATU value shows that had the COVID-negative mothers tested positive, their chances of delivering a stillborn baby would have been 1.3 per cent higher, and the chances of their babies being admitted to SNCU would have only been 13.6 per cent. The ATE for SNCU admission was 13.6 per cent which shows a positive impact of COVID-19 on SNCU admission.

Age of attainment of developmental milestones were similar in both the exposure and comparison cohort babies, indicating no clinically meaningful difference (Table III).

Discussion

Our retrospective cohort study showed a significantly higher risk of stillbirths among COVID-positive mothers after adjusting for covariates. Infants of COVID-positive mothers were at greater risk of admission to SNCUs as compared to their COVID-negative counterparts.

Prenatal SARS-CoV-2 exposure has been consistently associated with poor neonatal outcomes. Though two previous studies from India assessing risk of admission to Intensive care units (ICU) among babies with prenatal SARS-CoV-2 exposure report conflicting findings^18,19, systematic reviews and meta-analyses report findings similar to our study^5,20. Women who were exposed to COVID during pregnancy were also twice more likely to have stillbirths. This is similar to the pooled results reported in other studies^21-24.

Unlike previous evidence^18,20,25, we did not find an increased risk of preterm birth or low birth weight among our exposure cohort. The risk of preterm deliveries is likely to be associated with the gestation period when any infection is contracted. In our case, it was not possible to ascertain this fact because every pregnant woman was subjected to a COVID test as a routine activity mandated during the latter part of the study. Though greater odds of low birth weight among COVID-exposed pregnant women are reported, studies from India did not show any significant association^19,24.

There is limited evidence on developmental outcomes among children born to COVID-positive mothers, especially from India. Recent studies suggest that prenatally exposed infants are at risk of developing neurological disorders later in life^26,27. Infection during the first and second trimesters and symptomatic infections posed a higher risk of developmental delay^28,29. However, some studies report no association between maternal COVID-19 and infant neurodevelopment^30-33.

We did not observe any clinically meaningful differences in the age of attainment of milestones between the exposure and comparison cohorts. The findings related to developmental milestones have to be considered with caution, as all information was elicited from the mothers over the telephone. Clinical examination was not possible as physical contact with participants was kept to a minimum due to the pandemic. Nevertheless, mothers are generally considered fairly reliable for reporting developmental history, especially for major milestones³⁴. Literature suggests a greater awareness in mothers regarding gross motor and social milestones of their infants than language and fine motor milestones³⁵. Another limitation was seeking information telephonically. Although technically not so robust, we continued with telephonic interviews (despite the waning off of the pandemic during later part of data collection), in order to maintain uniformity. Even if this affected accuracy, it would be for both the exposed and non-exposed groups, thereby diluting the measures of effect without impacting directionality.

The basis of analysis was the allocation of every woman to either the exposed or comparison group decided based on antigen or molecular test results. This itself is likely to introduce an error of misclassification due to differences in sensitivity and specificity of testing methods and symptom status of cases^36,37. In the initial phase of our study, tests were conducted on symptomatic cases, but later this became a usual practice on all women as part of an advisory, leading to possible selection bias. The PSM analysis addressed this problem to a large extent.

Our study addressed some shortcomings of previous studies on neurodevelopment. Firstly, though our study was not powered to test differences in developmental assessment between exposure and comparison groups, we had a reasonably good sample size, probably the largest of all published studies on development to date. Our multicentric study included a geographically diverse sample of women who delivered in public sector hospitals, therefore making our findings more generalisable. Although all infants could be followed up till one year of age, we were able to assess age-appropriate milestones for each infant as a proxy assessment for neurodevelopmental trajectory.

However, there were also certain limitations to our study. Clinical neurodevelopmental assessments by a paediatrician were not possible as visiting the hospital during the repeated surges of the pandemic was challenging for study participants. Telephonic ascertainment of ages of milestone attainment through history provided by the mother may have led to some degree of recall bias. Further, we cannot rule out the possibility of differential recall bias between the exposure and comparison cohorts. Future prospective studies with face-to-face follow ups using validated neurodevelopmental assessment scales and considering maternal disease severity and the probable protective effect of maternal antibodies on infant development are warranted. Since we collected data during the pandemic, it was challenging to completely avoid interviewer biases, as the interviews were telephonic, and blinding was not logistically feasible in the interest of minimising non-response. We were unable to meet our required sample size since recruitment and interview of women was telephonic, and many women could not be contacted, but we were able to achieve 100 per cent power to detect differences in SNCU admission. Since most mothers in our study tested positive at the time of delivery when routine testing was mandated, it was difficult to undertake subgroup analysis; also, the infections were unlikely to affect infant development. All pregnancy and birth-related information was obtained from hospital records, thus making it impossible to detect any routine measurement errors. However, even if they exist, they are not expected to be systematically different between exposure and comparison groups and are therefore unlikely to affect the direction of associations observed.

Despite certain limitations, our study found an association between COVID exposure during pregnancy and clinical outcomes such as stillbirth and SNCU admission. The results strengthen the evidence that in-utero COVID exposure is associated with immediate adverse pregnancy outcomes, such as stillbirths and illnesses warranting intensive care. There were no clinically meaningful differences in the attainment of developmental outcomes. Future studies conducting longer-term follow-ups of infant developmental outcomes using validated neurodevelopmental assessment methods, and adjusting for antepartum, perinatal and postpartum confounders, are recommended.