Data

In Malaysia, Islam is the federal religion and religious practices are an integral part of daily life.^{(Reference Kamarzaman, Rajoli and Kamaruzaman26)} In Segamat, Malaysia, the Southeast Asia Community Observatory health and demographic surveillance system (SEACO HDSS) conducted a household-level health survey for the first time in 2013. The SEACO data collection team conducted annual visits to enumerate and update information on all individuals living in five of 11 sub-districts in Segamat, namely Bekok, Chaah, Gemereh, Jabi, and Sungai Segamat. Each household was visited up to three times if no one was present during the first two attempts. Since all individuals in the SEACO sub-districts had an equal opportunity to participate, selection bias was minimized. The collected data provide information on the general health of half of the estimated total population, including medical history, psychological well-being, risk factors for non-communicable disease and use of health services.^{(Reference Partap, Young and Allotey27)}

The data comprised all observations from the two combined SEACO HDSS survey rounds in 2013 and 2018. Twenty-five percent of the individuals participated in both surveys; thus, only their second observation was included in the analysis. Further, only participants aged 35 years or above with plausible reported age (age between the date of interview and the date of birth) were included as blood pressure and blood glucose measures were not available for younger individuals. The final dataset included information on sex, date of birth, date of interview, age at interview, religion, and measures of blood pressure and RBG.

Outcome measures

The SEACO HDSS data collection teams employed a digital automatic blood pressure device to measure the blood pressure of each participant on their left arm on three occasions, with a three-minute interval between each measurement. Prior to the initial blood pressure measurement, the participants were required to remain seated for a minimum of 15 minutes.⁽²⁸⁾ MAP was calculated using the second and third measurements, with one-third of the value attributed to the systolic and two-thirds attributed to the diastolic blood pressure.^{(Reference Kandil, Soliman and Alghamdi29)} RBG was determined from blood samples collected without prior fasting.^{(Reference Partap, Young and Allotey27)}

Ramadan exposure categories

Following the literature, we estimated in utero Ramadan exposure based on date of birth. Specifically, we calculated the overlap between date of birth and the historical Ramadan period in Malaysia, assuming an average pregnancy duration of 266 days after the conception.^{(Reference van Ewijk2,Reference Pradella and van Ewijk30)} The exposure was categorized to the following seven groups: (i) ‘certainly not exposed’, (ii) ‘probably not exposed’, (iii) ‘conceived during Ramadan’, (iv) ‘Ramadan during the first trimester’, (v) ‘Ramadan during the second trimester’, (vi) ‘Ramadan during the third trimester’, (vii) ‘born during Ramadan’ (see Table 1).

Table 1. Categorization of estimated in utero Ramadan exposure for different analysis

Statistical analyses

First, we checked for birth heaping using a bar graph and tested for pregnancy planning avoiding Ramadan by comparing the observed number of conceptions during Ramadan with the expected number assuming an even distribution using a two-sided binomial probability test.

Difference-in-differences (DiD) analyses were conducted using ‘being Muslim’, ‘being in utero during Ramadan (in different phases of gestation)’, and the interaction term ‘being Muslim $ \times $ being in utero during Ramadan’ utilizing different exposure categories as independent variables. Since non-Muslims do not observe Ramadan fasting, they act as a control group for residual confounding in the comparison between being in utero during Ramadan and not being in utero during Ramadan.

DiD Model 1 looked at a binary exposure by combining the exposure categories mentioned above into exposed (group (iii) to group (vii)) and not exposed (group (i) and group (ii)). Model 2 utilized again group (iii) to group (vii) as the exposure category but the unexposed are considered separately using group (i) and group (ii). Finally, Model 3 considered the timing of Ramadan exposure during pregnancy by using all exposure groups as independent variables. All models were separated by sex and adjusted for age, age squared, and month of birth to control for birth’s seasonal effect on lifetime disease risks.^{(Reference Doblhammer and Vaupel31–Reference Boland, Shahn and Madigan33)}

In a final step, we performed different robustness checks. First, we fitted logistic regression models to evaluate the association between Ramadan exposure and the likelihood of receiving treatment for hypertension or hyperglycaemia as well as a full set of interaction terms of Ramadan exposure with Muslim and the original controls. Second, to reflect the greater burden of hypertension and hyperglycaemia in later life, we re-estimated our primary models in the subsample aged ≥50 years, where high blood pressure and high blood glucose are more prevalent.^{(34,Reference Chandran, Abdullah and Abdul35)}

All statistical analyses were done using Stata version 17.0 (StataCorp LLC, 4905 Lakeway Drive, College Station, TX 77845, USA).