Intergovernmental committee on Drugs working party on Fetal Alcohol Spectrum Disorders
Fetal Alcohol Spectrum Disorders in Australia: An Update
4.3 Methodological Limitations of the Evidence
Many studies of FAS, FASD, and alcohol and pregnancy more generally, are limited by small sample size, retrospective collection of potential confounders and use of different diagnostic criteria, all of which impede comparison between studies. The methodology varies between studies and findings need to be interpreted with caution due to issues of bias and confounding. Issues of selection bias (e.g. case control studies often rely on clinical samples to select cases), bias in relation to reporting of alcohol consumption need to be considered (Sayal 2007), and loss-to-follow-up in longitudinal studies.
The most recent examination of systematic reviews, meta-analyses and new articles suggest that the reported significant effects from low levels of prenatal alcohol exposure are likely to be due to methodological issues such as confounding or misclassification of the exposure or outcome (O'Leary and Bower 2011). Of the large body of research that has been conducted on the impact of low levels of alcohol during pregnancy there are fewer than thirty papers for any outcome and most outcomes have fewer than ten eligible studies that meet the stringent inclusion criteria used in a systematic review. This is due to a range of factors such as:
- inadequate reporting of raw data
- inadequate blinding to case/control status
- inadequate delineation of alcohol consumption
- lack of a control group that was not exposed to alcohol (although some systematic reviews have included studies with infrequent drinkers in the reference group as shown in Table 4.4)
- the timing of questions about maternal alcohol consumption (antenatal versus postnatal data collection)
- failure to define the trimester(s) in which drinking took place.
There is some evidence that the lack of consistency in research findings across countries may be due to differences in the patterns of drinking during pregnancy and the quantity of alcohol consumed at each occasion (Testa et al. 2003; Henderson et al. 2007). Henderson et al. also question whether the findings from the USA, where the majority of the studies have been conducted, can be generalised to other countries where there may be differences in the extent to which alcohol use in pregnancy is under-reported and in the ascertainment of outcomes (Henderson et al. 2007).
The issue of misclassification of the dose or pattern of exposure can be addressed by a ‘composite’ classification method which considers the dose, pattern and timing of maternal alcohol consumption. A recent study compared traditional methods of classifying prenatal alcohol exposure, such as averaging alcohol consumption, with the ‘composite’ method (O'Leary et al. 2010). This study found that the traditional methods lacked discrimination and resulted in some women who reported drinking at binge and heavy levels to be classified as low to moderate drinkers and vice versa. Using the composite method the lowest level of alcohol consumption to increase fetal effects was an increased odds of child behaviour problems following moderate alcohol exposure (30-40g per occasion and no more than 70g per week) (O'Leary et al. 2010). This finding was completely masked by the traditional methods of classification (O'Leary et al. 2010). It is therefore critical for future studies to consider the dose, pattern and timing of consumption.
Controlling for potential confounding factors is important when trying to elicit the independent effect of maternal alcohol consumption during pregnancy on child development. However, there is a lack of consistency in the adjustment for confounders across studies (Testa et al. 2003; Henderson et al. 2007). Some studies have not adjusted for factors well known to be confounders such as smoking, low socioeconomic status and ethnicity and this may have resulted in residual confounding (Henderson et al. 2007). The impact of residual confounding is well demonstrated by the findings of the systematic review by Testa et al., which showed that at the 12-13 month assessment there was a negative effect from prenatal alcohol exposure prior to adjustment for known confounders and a lessening of the effect following adjustment. Other studies have over adjusted by controlling for previous malformations or miscarriages which may have been associated with alcohol exposure leading Testa et al. to recommend an ‘a priori selection’ of confounding factors.