Table 4 Variables to adjust for or to use in subgroup analyses and exposure variables

Confounders/covariates to adjust for in analyses:

Wealth score constructed out of household assets

Education status of pregnant woman (none, primary, secondary, higher secondary +)

Parity (n)

Age of pregnant woman in years

Gestational age at measurement in weeks

Baseline (study enrolment) measure of Hb (for primary outcome analysis)

Adjustments for study design

Religion (Muslim versus Hindu)

Ethnicity (hills versus plains)

Travel time to health facility

Cluster size (number of eligible women identified in the cluster during the census as used during constrained randomisation)

Cluster (as random effect)

Variables to use in subgroup analyses

Baseline (study enrolment) anaemia status

Categories of BMI at enrolment (kg/m²)

Wealth categories constructed out of household assets

Main exposure (independent) variable for analyses of outcomes above:

Study arm: Home visiting +PLA versus Control (which is the reference group)

Within the intervention arm the levels of exposure to PLA women’s groups and home visits are coded in three (independent) variables for analyses of a dose-response effect upon the outcomes above (or a subset of them):

Exposure to women’s group—score of exposure constructed out of number of meetings attended by pregnant woman and number attended by family members

Exposure to home visiting with tailored counselling and dose of iron-folic acid supplements—score of exposure constructed out of number of meetings attended by pregnant woman and number attended by family members

IFA consumption during pregnancy (count or ordinal score constructed from the count)