Duke 2008 | ASHE | Equity and Efficiency in Health and Healthcare

Background: A few studies have evaluated the effects of maternal smoking on birth outcomes (primarily birth weight) considering its potential endogenous selection using instrumental variable (IV) analysis and taxes of cigarettes and future smoking status as instruments. Comparable effects of smoking were found overall between exogenous and endogenous smoking estimates suggesting overall minimal omitted variable bias. Given that IV results are dependent on the used instruments and the strong theoretical support for self-selection into prenatal behaviors including smoking based on typically unobserved risk expectations and preferences, evaluating the effects of smoking on birth outcomes using alternative instruments is important. Genetic variants that affect smoking have been identified and these could be evaluated as instruments for smoking.

Objectives: This study aimed at evaluating the utility of a set of genetic variants as potential instruments for smoking when evaluating its effects on birth outcomes including occurrence of oral clefts (OC), a common and burdensome birth defect, and birth weight.

Methods: 15 variants in five genes (DBH, DDC, GABAB2, CHRNA4 and TPH) previously shown to be related to smoking were assessed in a sample of 382 infants with and without OC from Iowa. The effects of the genetic variants on smoking were evaluated and instrumental variable (IV) analyses of the effects of smoking on the studied birth outcomes were estimated using Two-Stage Least Squares regression and the genetic variants that were significantly related to smoking as instruments. Smoking was measured by any smoking immediately preceding or during the first trimester of pregnancy.

Results: Five genetic variants significantly increased smoking probability (11-17% per variant). Smoking had no effect on OC ignoring self-selection (relative risk RR=1.2, insignificant and comparable to previous estimates), but had larger effects under the IV model (RR= 1.5-2.3 using various instrument combinations; largest RR statistically significant). Smoking reduced birth weight by 200 grams ignoring self-selection and by 545-860 grams under the IV model (all effects statistically significant). The over-identification restrictions of the various genetic instrument combinations could not be rejected (p=0.5-0.7) and the partial F statistics of the instrument significance ranged from 4.8 to 7.9.

Conclusions: These results suggest that the selected genetic variants might be appropriate instruments for smoking in health production studies. The effects of smoking on OC and birth weight might have been underestimated in previous studies due to women at higher risks for OC and low birth weight outcomes being less likely to smoke. The study has important implications for future research in terms of replicating these analyses in larger samples and comparing, using the same sample, IV estimates based on genetic versus non-genetic instruments such as cigarette taxes to assess the sensitivity of IV estimates to different types of instruments.