Evaluation register study feasibility using maternal smoking as a risk factor in Finland

Abstract

Introduction Maternal exposures have been associated with birth outcomes and health effects later in life. Despite the well-known health effects of smoking, many knowledge gaps remain concerning prenatal exposure to maternal smoking. Our aims are (1) to summarize existing literature on maternal smoking as a risk factor for health effects in early life; (2) to assess the feasibility of a register based epidemiological study in Finland; and (3) to identify maternal smoking induced health effects in early life as a case study by using national heath registers. Methods A systematic literature review was conducted to identify relevant health endpoints. The needed study size was calculated for cohort and case-control design (α = 5%; study power = 90%). The smoking prevalence during pregnancy was estimated to be 15%. The study population comprises all 1.6 million births in Finland in 1987-2013. Results Childhood cancer, congenital malformations, sudden infant death syndrome (SIDS), stillbirth and other birth outcomes (preterm birth, low birth weight, small for gestational age) were identified as most relevant. Cohort design was shown to be more appropriate than case-control design for a Finnish register-based epidemiological study on the above endpoints. The cohort of all births in Finland in 1987 or later is big enough to investigate SIDS and the four birth outcomes with sufficient statistical power. The cohort is only sufficiently big for stillbirths with α = 10%. For the other health endpoint the available cohort is expected to be too small to reach a sufficient statistical power. Conclusions Finnish health registers are a promising source for health data for epidemiological studies. For a low prevalence of exposure in combination with rare outcomes the Finnish population register establishment is probably too small for sufficient statistical strength. International collaborations are needed to tackle this issue and increase the available study size and statistical reliability.