Commentary on Mactier et al. (2014): Methadone-assisted treatment and the complexity of influences on fetal development

Mactier et al. [1] address an important issue regarding methadone-assisted treatment (MAT) for pregnant women: infant outcomes. However, a number of concerns detract from their conclusions.

First, there is an underlying assumption that reduced fetal growth, shown by infant birth size, among MAT pregnancies could be ‘fully explained by smoking or socio-economic deprivation’. Given the complexity of fetal development and the large variability inherent in the maternal population, it is highly unlikely that any two variables could fully explain any fetal or infant outcome. Thus, the paper proffers a ‘straw hypothesis’ that it then finds deficient. Covarying two fallible covariates runs the clear risk of reaching an incorrect conclusion [2].

Secondly, there is a concern regarding the clinical care provided to the women: ‘Women were prescribed sufficient methadone to eliminate physical withdrawals, with the aim of reducing towards the lowest acceptable dose of methadone in the weeks prior to delivery’. This practice contradicts United States clinical guidelines [3,4]; research has shown that pregnant MAT women will need, on average, three medication dose increases during pregnancy [5,6]. Unsurprisingly, 78% of the authors’ sample continued to use illicit opioids and/or benzodiazepines during pregnancy. This high degree of ongoing substance use may be due to inadequate opioid use disorder treatment, leading in turn to reductions in birth parameters, due probably to other exposures and/or associated maternal behaviors, here attributed mistakenly to methadone.

Thirdly, there is concern regarding the use of a binary variable to represent cigarette smoking (‘smoker/nonsmoker’). Such ‘crude classification of the observations’ costs power and leads to biased tests of significance [7,8]. Lack of quantification for a variable known to have significant and dose-dependent [9] effects on infant size presents a serious flaw in study design. Jones et al. [10] found that, in MAT samples with no concomitant illicit substance/alcohol use, a pregnant woman who smokes a pack of cigarettes daily is at risk for, among other outcomes, a more than 8% decrease in her neonate’s birth weight compared to a similar non-smoking MAT woman. The adjusted mean reduction in birth weight reported by Mactier et al. was 259 g—an amount approximately equal to 8% of the birth weight of their comparison sample.

Fourthly, omitted variables create a threat of overlooked etiologies. Although this threat overhangs all regression models, it is particularly worrisome in the present case, given no information about stimulant use and inadequate information about alcohol use (‘Four per cent of methadone-maintained mothers [self-] reported excess alcohol consumption in pregnancy’). Given the threat posed by any gestational alcohol use combined with its known teratogenic potential (including significant effects on infant size), and the likelihood of both under-reporting and greater prevalence of alcohol use in the opioid-dependent sample, this omission is probably substantial.

The single most important ‘take-home’ message is that the field needs to develop and evaluate conceptual models that comprehensively explain alterations in infant outcomes among pregnant women undergoing MAT. Opioid use disorder in pregnant women occurs in the context of numerous bio-psychosocial issues. As a result, fetal development/infant outcome must be understood in the context of many factors experienced by this population. These factors include other exposures to licit (e.g. medications, alcohol, nicotine) and/or illicit substances, environmental stressors [11] (e.g. poverty, crowding, air pollution) and familial factors (e.g. partner substance use, violence exposure, stress). There are maternal characteristics (e.g. psychiatric comorbidities and medications, social support, coping skills) that may further impact pregnancy/infant outcomes. Additionally, the plethora of medical issues (e.g. pre-existing medical conditions, poor prenatal care) facing this population cannot be ignored. To attribute reduced infant birth parameters in this highly vulnerable population to any one factor, particularly one that has shown to provide great benefit to mothers and infants, is inaccurate and potentially harmful.

Our conceptual models need to account for both the myriad factors involved in fetal development for pregnant women with opioid use disorder and population variability when examining infant outcomes in women undergoing MAT. Such models need to acknowledge three factors that have largely been ignored in this research: the uniformity myth, critical periods and synergy. The uniformity myth refers to the idea that by examining ‘pregnant women with opioid use disorder’ one has sufficiently isolated a population of interest. Therefore, it does not consider that distinct subpopulations exist for which there are other factors affecting fetal development/infant outcomes in the context of maternal opioid use. Critical periods are stages of fetal development that may be variably affected by timing of exposures to substances/life events/stress, etc. Synergy refers to the fact that there are probably interactions among some factors that adversely impact the developing fetus. The field needs to consider that there are putative causal factors that adversely impact or, conversely, protect, the MAT-exposed infant, such as genetic influences affecting neonatal abstinence syndrome severity [12]. Succinctly, failure to consider the complexity of influences on fetal development inherent in this population, and then attributing infant birth parameters to a single factor, does not enhance our understanding of this vulnerable and costly group, but rather detracts from both applications of optimal clinical care and the appreciation of how best to focus our research efforts.