Statistics from Altmetric.com
Editor,—Ford and his colleagues in New Zealand found that consumption of ⩾400 mg of caffeine during the third trimester was associated with an increased risk of sudden infant death (SIDS) months after birth.1 Consumption of lower concentrations of caffeine were not associated with any increased risk of cot death. The lack of a dose-response relation between maternal caffeine consumption and risk of cot death should have alerted the authors to the possibility that the relation is not causal. Nevertheless, the authors proceeded to seek biological explanations for their findings. Let me offer an epidemiologist’s interpretation.
SIDS is a tobacco associated disorder
The risk of SIDS has repeatedly been increased among infants exposed to parents’ cigarette smoke, often in a dose-response pattern.2 3 Thus, parents’ smoking should be viewed as an important risk factor for SIDS.
Tobacco consumption measured inadequately
The authors classified maternal smoking during pregnancy as a yes/no variable. Maternal smoking during the months preceding the interview was not assessed, nor was paternal cigarette smoking.
As the greater the number of cigarettes smoked means the higher the risk of SIDS,2 3 the authors should have measured parental (and not just maternal) cigarette smoking as a continuous variable (that is, number of cigarettes smoked in the presence of the infant each day) or perhaps as a categorical variable providing a reasonable approximation of a continuous variable (for example, 1–10, 11–20, etc).
Caffeine measured better than smoking
Caffeine consumption was assessed as a categorical variable (small, light, moderate, heavy). Because those who smoke the most cigarettes also drink the most coffee/caffeine,4 caffeine will carry information about magnitude of cigarette smoking that was not carried by the “maternal tobacco: yes/no” variable, but should have been carried by a “parental tobacco: small, light, moderate, heavy” variable.4-10
The result of having better information about caffeine than about tobacco is that adjustment for tobacco will not eliminate tobacco confounding of the association between caffeine consumption and the risk of SIDS.4 11-13
Consider the possibility that the caffeine consumption variable tends to carry tobacco consumption information at the highest levels of caffeine consumption. During the third trimester, when caffeine degredation is slowed,14-16 those who are able to consume ⩾400 mg of caffeine/day tend to be those who smoke the most cigarettes. Thus, those women who consume the most caffeine and smoke the most cigarettes probably put their infants at risk of SIDS.
The authors have committed an epidemiological error that students of introductory epidemiology are taught to avoid. At a minimum, they should have identified in the discussion section of their paper that their failure to measure adequately one of the most important variables is a major, potential limitation.
Drs Ford and Schluter comment:
Based upon a New Zealand nationwide case control study we found that heavy caffeine consumption (⩾400 mg/day) in pregnancy was associated with an increased risk of sudden infant death syndrome (SIDS) after adjusting for likely confounding factors, including smoking.1-1 The adjusted odds ratio (aOR) for mothers who consumed heavy caffeine throughout their pregnancy was estimated at 1.65 (95% confidence interval (CI) 1.15, 2.36). However, Dr Leviton disputes this finding, claiming that our dichotomous yes/no response variable for maternal smoking over the last two weeks was inadequate and that residual confounding explained the reported association.2Instead, in Dr Leviton’s view, the analyses should have been conducted using parental (mother and father) daily cigarette consumption measured as a continuous variable. To assure your readers that the reported prenatal caffeine effect is real and different from the effect associated with smoking, we present a supplementary analysis of our data.
We examined the effect of heavy caffeine consumption in pregnancy on SIDS risk for non-smoking mothers only. Employing identical confounding variables and statistical analyses as before,1-1 the aOR for mothers who consumed heavy caffeine throughout their pregnancy was estimated at 1.71 (95% CI 0.85, 3.45). This estimate is slightly higher and more variable (due to the smaller sample size) than that previously reported when information for all mothers was analysed together.
We repeated the analyses previously reported,1-1 but left cigarette consumption as a continuous variable (that is, average number of cigarettes smoked each day over the last two weeks). Replacing the dichotomous maternal smoking variable with its continuous analogue had little effect on the resultant estimate of theaOR for mothers who consumed heavy caffeine throughout their pregnancy, aOR = 1.69 (95% CI 1.17, 2.43). Similarly, the inclusion of the paternal cigarette consumption variable had little effect on this estimated adjusted relative risk (aOR = 1.73; 95% CI 1.19, 2.50) as did the variable combining both maternal and paternal cigarette consumption (aOR = 1.73; 95% CI 1.20, 2.49).
Undoubtedly, parental smoking is an important risk factor associated with SIDS. However, our analyses disclose a link between heavy caffeine consumption in pregnancy and increased risk of SIDS. This association is separate from the effect of parental smoking on SIDS risk.