Intervening to Improve Infant Health in Ghana

• Overview of design and analysis [per CONSORT 2010] 2.1 Trial design [CONSORT 3a] The study is a three-arm cluster randomized trial, with two intervention arms and a control arm. The primary justification for intervening at the village level is that an individual level randomization, which would entail treatments and controls living side-by-side in a village, could lead to conflict within villages.

2.2 Study Participants [CONSORT 4a, 4b] The study will take place in the catchment area of Kintampo Health Research Centre. This comprises a population of 146,000. The study population is primarily rural and agricultural, and cooks primarily with biomass fuels. The study area is well described in Owusu-Agyei et al. 2012 (38).

A cluster was eligible to participate if it:

• Is located in Kintampo North or South Districts (this is the core study area for KHRC)
• Is primarily rural (in practice, this excludes Kintampo, which is a small city of approximately 40,000 people);
• Is operationally feasible (in practice, this excluded a handful very small, isolated clusters that would have presented extraordinary logistical challenges);
• Is home to women who primarily deliver at one of our four staffed birth facilities (in practice this excluded one village on the edge of the study area, in which women travel to another district for deliveries).

A woman will be eligible to participate in the study if she:

• Is in the first or second trimester of pregnancy (gestational age ≤ 24 weeks gestation; this is to ensure that the intervention is actually delivered prior to 27 weeks)
• Is carrying a live singleton fetus (twins will be excluded)
• Is the primary cook in her household or compound; and
• Is a non-smoker. 2.3 Interventions [CONSORT 5]

The study has three arms:

• LPG intervention arm, in which households receive a two burner LPG cookstove and monthly gas deliveries;
• An efficient biomass arm, in which households receive two BioLite cookstoves; and
• A control arm 2.4 Outcomes [CONSORT 6a]

The primary health outcome measures for the study are:

• Infant weight at birth, and
• Physician assessed pneumonia. Baseline assessments of outcomes No baseline data for ALRI prevalence from Ghana is available. Our study will thus provide very valuable baseline data regarding this important disease.

KHRC has, in the course of conducting other research projects, captured some information regarding birth weight; statistics calculated from these data were used as inputs into power calculations. These data are described in (39).

2.5 Sample size [CONSORT 7a, 7b] For the three-arm design, we calculate that we need to enroll households to achieve our target of 1225 births (assuming that 85% of the pregnancies will result in a recorded birth after, accounting for move out, drop out, and still birth rates). These will be spread across 35 or more clusters, as outlined in the following table.

Clusters Births Enrollment BioLite 13 455 525 Control 13 455 525 LPG 9 315 365 total 35 1225 1415

A cluster was eligible to participate if it:

• Is located in Kintampo North or South Districts (this is the core study area for KHRC)
• Is primarily rural (in practice, this excludes Kintampo, which is a small city of approximately 40,000 people);
• Is operationally feasible (in practice, this excluded a handful very small, isolated clusters that would have presented extraordinary logistical challenges);
• Is home to women who primarily deliver at one of our four staffed birth facilities (in practice this excluded one village on the edge of the study area, in which women travel to another district for deliveries).

A woman will be eligible to participate in the study if she:

• Is in the first or second trimester of pregnancy (gestational age ≤ 24 weeks gestation; this is to ensure that the intervention is actually delivered prior to 27 weeks)
• Is carrying a live singleton fetus (twins will be excluded)
• Is the primary cook in her household or compound; and
• Is a non-smoker.