Despite substantial research, maternal smoking during pregnancy (MSDP) continues to be a leading modifiable contributor to low-birth weight, a robust risk factor for adverse neonatal, neurodevelopmental, metabolic, and cardiovascular outcomes across the lifespan (1). The year 2017 marks the 40th anniversary since publication of the first prenatal smoking cessation intervention trial (2). Since this time, 96 randomized controlled intervention trials (RCTs) including over 31,000 pregnant smokers, testing 37 individual behavior change techniques (3, 4), and an additional 26 qualitative studies (5, 6) have led to important but modest gains. In the United States, the prevalence of MSDP fell most dramatically in the 1990s from 17.3 to 13.3%. However, in the following decade, the prevalence fell by just 1 to 12.3% in 2010 (7). This means that one in nine infants in the United States is born prenatally exposed to MSDP. This figure is as high as one in four for infants born to socioeconomically disadvantaged women and women from marginalized populations (7, 8). On average, intervention increases the chances a pregnant smoker will stop smoking before delivery by only 6% (4). Even for interventions involving financial incentives for biologically verified abstinence, well under half of pregnant smokers achieve abstinence by delivery (9). Intervention—especially financial-incentive-based interventions—are associated with improved birth outcomes (10). However, to achieve the goal of 1% prevalence in MSDP set by Healthy People 2020, an infusion of novel approaches is imperative (11).

With this goal in mind, we reflect on MSDP within the broader challenge of promoting healthy lifestyles in the general population that spans at least a half a century (12). Following the development of 83 largely overlapping theories of behavior change comprising more than 1,000 constructs and 23 different intervention targets (13), the leading causes of morbidity and mortality in the United States continue to be directly attributable to lifestyle behavior, namely, smoking, poor diet, and inactivity (14, 15). For example, when we consider that obesity has become more rather than less prevalent during this time, and is presenting earlier in life in the United States (16, 17), the progress made on reducing MSDP seems outstanding in comparison. And this is hardly surprising.

Pregnancy itself elicits abrupt temporary change in smoking behavior without formal intervention and does so more effectively than any intervention strategy tested to date (18). The presence of an other-oriented mechanism of change—concern about the unborn child—is well-documented in qualitative (5) and quantitative studies (3, 19, 20), yet remains poorly understood. This unique interpersonal aspect of MSDP could represent a powerful volitional target for intervention.

We describe a novel conceptual framework for mechanistic research on health behavior change motivated by concern for others, as is observed in pregnant smokers. Foundational to the proposed framework are cognitive, affective, motivational, and regulatory processes described in emerging neurobiological models of human caregiving behavior and how these processes interact dynamically with addictive processes. We illustrate the potential utility of this proposed mechanism, if validated, for intervention development and personalization of treatment. We conclude by outlining a roadmap for empirical validation of the proposed mechanism and its application to intervention development. A systematic review of the literature is not the aim of this article—we refer readers to recent high-quality systematic reviews of prenatal smoking cessation intervention studies (3); qualitative studies on MSDP (5, 6); and other relevant topics throughout this article.

Traditionally, health behavior change research has involved efficacy trials aimed at modifying one or more targets identified from behavior change theories [for a full taxonomy of these theories see Ref (13).]. In recent years, recognition of the limitations and inefficiency of this approach has fueled interest in an Experimental Medicine (EM) approach that prioritizes a mechanistic understanding of putative targets prior to conducting full efficacy trials (21). Clearly elucidated mechanisms of change enable scientists to conclude why a particular intervention worked or failed to work in a given study and affords the opportunity to translate knowledge gained from one particular trial to other contexts, behaviors, or populations (13). Moreover, behavior change theories are limited by (a) a lack of specification of population and context (i.e., in what populations or contexts are theories and targets identified by theories most predictive of behavior change); (b) a lack of are specification of volitional factors (why people would want to change in the first place); (c) the inability to infer how change in targets over time influences behavior since theories are derived primarily from correlational evidence; and (d) the rare examination of affective processes as putative targets for interventions (22).

Lack of specification is reflected in the U.S. Preventive Services Task Force’s (USPSTF) recommendation for “Tobacco Smoking Cessation in Adults, Including Pregnant Women” using the 5 A’s intervention (Ask, Assess, Advise, Assist, and Arrange). The extent of specification for pregnant women involves (a) behavioral rather than pharmacologic interventions and (b) ensuring educational materials about risk are pregnancy-specific (23, 24). There are clearly overlaps in barriers to smoking cessation among non-pregnant and pregnant women. These include stress and its psychopathological manifestations (25, 26), limited internal and external resources (27, 28), and smoking-related factors (29, 30). However, smoking cessation among pregnant women is also associated with unique volitional factors. It is externally- rather than internally-motivated and appears to be temporary (19, 31). Yet, these unique factors have not been adequately explicated as change mechanisms.

The term “spontaneous quitters” has been used to describe a substantial subset of pregnant smokers—some 40–65% of privately insured smokers and 11–28% of smokers in publicly funded obstetric clinics—who stop smoking without intervention in early pregnancy and prior to the onset of obstetric care in the United States. The prevalence of spontaneous quitting is remarkably similar across the United Kingdom and other European countries (32, 33). Spontaneous quitters have lower smoking-related risks (i.e., nicotine dependence, partner smoking, age of smoking initiation, impulsivity) and comparatively favorable demographic profiles (being married, more highly educated, and nulliparous) (34). However, there is converging evidence that unmeasured factors or processes beyond the mere absence of risks facilitate sustained abstinence among spontaneous quitters.

First, spontaneous quitters achieve abstinence rates two to three times higher than non-pregnant women who quit “cold turkey” over a similar period of time (65–81% across gestation when biochemically verified) (34–36). Spontaneous quitters’ abstinence rates also compare favorably to other highly motivated populations including men and women who recently suffered a heart attack (37, 38). Moreover, spontaneous quitters report greater confidence about their ability to remain abstinent for the duration of pregnancy relative to non-pregnant quitters over the same period of time, yet do not use the behavioral strategies (i.e., coping skills, distractions, oral-substitutes) used by successful recent ex-smokers (39).

Even women who make adoption placements at birth report suspending tobacco, alcohol, and illicit drug use at rates comparable to rates reported in epidemiologic samples (28, 40, 41). This suggests that mechanisms that facilitate spontaneous quitters’ success are at least partially attributable to being pregnant, distinct from factors associated with the transition to parenthood (42). Moreover, these mechanisms may apply to other addictive behaviors besides cigarette smoking (43). Consistent with this are findings by Curry and colleagues (31) that pregnant smokers endorse different motives for abstinence from smoking during pregnancy, i.e., “I want to be healthy for my baby,” versus abstinence during the postpartum period “I do not want to be known as a smoker by my child” (31, 44).

Finally, the mechanisms that facilitate spontaneous quitting could reflect a broader pattern of harm reduction. Substantial inter- and intra-individual fluctuation in smoking across gestation has been documented in quantitative studies of pregnant women (45). In qualitative studies, many women who do not abstain from smoking during pregnancy reduce their smoking early in pregnancy with the stated intent of reducing harm to the fetus (since quitting altogether is not perceived as possible) (3, 46, 47). Additionally, the power of the imagined harm to the fetus caused by continued smoking during gestation is evidenced by intense guilt reported by women who cut down but do not abstain (5). In one of the largest studies specifically focused on spontaneous quitting which included 118 spontaneous quitters and 231 continuing smokers, the belief that “smoking will greatly harm my baby” was associated with a 14-fold increase in the likelihood of being a spontaneous quitter (27). In summary, in the absence of intervention, “being pregnant” seems to positively influence smoking trajectories via temporary suspension and reduction of smoking. Elucidating this change mechanism and how it can be amplified can yield new avenues for improving the immediate and long-term health of mothers and their children.

To do so, we take an interdisciplinary approach that blends advances in social neuroscience (48), parenting neurobiology (49), and the interruption of parenting processes by addiction (50). Drawing from these literatures, we hypothesize that key cognitive, affective, motivational, and regulatory processes that facilitate responsive caregiving behavior, as characterized across separate theoretical foundations (51–53), may be highly relevant for identifying putative targets for intervention during pregnancy. Moreover, we illustrate how this developmental framework, if validated, can be used to personalize interventions. Finally, we describe a detailed research agenda aimed at validating the proposed mechanism.

Striking similarities in the mechanisms of affiliation and maternal caregiving behavior across mammalian species and their modulation by the neuropeptide hormone oxytocin support their highly conserved role in survival, adaptation, and evolution (54). In recent decades, functional magnetic neuroimaging (fMRI) studies in human parents presented with simulated cues salient to caregiving (i.e., infant faces, cries, and odors) have facilitated the identification of a network of neural regions implicated in parental caregiving commonly referred to as a parenting network, parenting circuitry, or the “parental brain” [for a recent review of these fMRI studies, see Ref (55)]. Optimal spatial resolution offered by fMRI is nicely complemented by temporally sensitive neuroimaging techniques such as electroencephalography (EEG) with event-related potentials (ERPs) and magneto-encephalography from which processing functions of specific neural structures may be inferred. For example, the temporal sequence of signals following an experimental stimulus is used to infer sensory perception, detection of salience, and attentional processes, followed by “top-down” regulation and reappraisal (56). Neural processes associated with parenting behavior, combined with their regulation by oxytocinergic, hypothalamic–pituitary–adrenal axis and autonomic functions are referred to, collectively, as the caregiving system (49, 51). This system is also thought to regulate altruistic (other-oriented) prosocial (helping) behaviors outside of the parenting context (51, 57–59).

Common to both sensitive parenting behavior and prosocial behavior is empathy (60), defined here, as the ability to perceive others’ need, coupled with a motivation to help (61, 62). Different definitions of empathy within and across scientific disciplines have created confusion, detracting from its highly conserved and critical role in social behavior (48, 58, 63). The dual function of empathy-related mechanisms in promoting other-oriented responding while regulating internal stress responses is, however, increasingly recognized (though not explicitly) by intense inquiry regarding the therapeutic potential of its biological substrate, oxytocin (48, 49, 64). For heuristic purposes, we will use the terms, caregiving processes and empathic processes interchangeably to refer collectively, to the perceptual, affective, motivational, and regulatory processes necessary for the expression of behavior aimed at helping others in parental and non-parental contexts, respectively. Specific cognitive and affective facets of empathy (i.e., affective sharing, empathic concern, perspective taking or cognitive empathy, and personal distress) will also be defined herein to illustrate how they map onto perceptual, affective, motivational, and regulatory processes described in parental caregiving models.

The component processes involved in parental caregiving behavior may be broadly characterized as: (a) the perception of cues signaling need or distress in another; (b) affective processing of these cues; (c) competing other-oriented versus self-oriented motivational states; and (d) subsequent behaviors (to help versus to withdraw) moderated by internal regulatory capacity (51, 65, 66). Below, we elaborate on these processes and describe how they may facilitate motivation to reduce or suspend smoking following the recognition of pregnancy.

Whether and to what extent the caregiving processes described in parents occur during pregnancy is not clear due to the relative paucity of clinical research studies conducted in pregnant women (129). However, tremendous plasticity observed during the early postpartum period (130) and the prediction of maternal–infant attachment by maternal–fetal attachment and plasma oxytocin during pregnancy support the emergence of caregiving processes prenatally (85). There is also early evidence that pregnant women show heightened vigilance to novelty and threat cues close to delivery when compared with non-pregnant women (131–133).

The most compelling evidence for prenatal neuroadaptions to caregiving comes from a recent longitudinal structural MRI study. Comparison of pre-pregnancy and postpartum scans showed reductions in gray matter volume in regions implicated in social information processing (i.e., the theory of mind network) among primiparous pregnant women. These structural changes predicted observed and reported attachment to infants postpartum and were neither observed in women’s male partners, multiparous pregnant women, nor in a comparison sample of non-pregnant women over the same period of time (134). These findings suggest that structural changes related to postpartum caregiving occur during a woman’s first pregnancy but not in successive pregnancies. Could this observed neuroplasticity during a woman’s first pregnancy facilitate spontaneous behavior change?

Increasingly sophisticated methods for assessing the maternal–fetal dynamic via fetal neurobehavioral assessments have led to a greater understanding of the complex mechanisms through which maternal mood and behavior during pregnancy influence fetal neurodevelopment and long-term developmental outcomes for children (135, 136). These approaches can also provide rich experimental and therapeutic opportunities for intervention on MSDP. Realizing the potential scientific gains from these diverse empirical frameworks, however, requires scientific collaboration that transcends disciplines and methodologic approaches (137). These disciplines include but are not limited to the science of behavior change, parenting and addiction neurobiology, attachment theory, social neuroscience, developmental psychology, and developmental neurobiology. With this spirit of interdisciplinary collaboration in mind, we present a research agenda for the application of this collective knowledge toward identifying novel targets for prenatal smoking cessation interventions.

The Science of Behavior Change working group of the National Institutes of Health recommends the following 4-step EM approach that focuses on the importance of understanding behavior change mechanisms prior to conducting efficacy trials (21). (1)

Identify a modifiable factor that relates to behavior as a putative target for intervention.

A mother’s behavior during pregnancy is arguably the most robust modifiable determinant of the intrauterine environment, increasingly recognized as a critical contributor to the lifelong health of her unborn child (150, 151). In this way, improving the impact of prenatal intervention constitutes a far-reaching preventive investment (152). The challenges of behavior change, even during the uniquely opportune window of pregnancy, however, are substantial and felt by pregnant women, their health-care providers, and intervention scientists alike. Pregnant women who smoke and use other addictive substances face enormous biopsychosocial risks that are often perpetuated inter-generationally. These risks compound the effects of smoking on mothers’ own health and the health of their families. A deeper mechanistic understanding of MSDP that transcends interdisciplinary divisions to leverage scientific advances generated from multiple perspectives and theoretical foundations is proposed to most effectively and efficiently address this major public health problem.

We underscore the importance of scientific progress on MSDP made in the past 40 years. Intervention is associated with increases in birth weight even when total abstinence is not achieved. Moreover, interventions that provide financial incentives for confirmed abstinence (the most effective approach to date) are clearly cost-saving when considering the neonatal health-care costs saved (3, 10). But this may not be a satisfactory solution. The long-term ethical, policy, and personal implications of widespread financial incentive-based interventions for MSDP are unknown. Perceptions regarding the use of public resources for these interventions could compound the stigma and marginalization already felt by populations disproportionately affected by MSDP leading to more harm than benefit (6). Moreover, reducing prenatal smoking exposure alone may not reduce children’s long-term mental health risks. Unlike animal models, in which prenatal nicotine exposure has demonstrated clear and causal effects on offspring brain development, causality in humans has not been established due to numerous confounders linking MSDP with child outcomes (153). Genetically informed studies designed to disentangle prenatal exposure from familial factors suggest that the independent effect of MSDP on neurodevelopmental risk is either very small (154, 155) or absent (156, 157).

What is clear is that families transmit mental health risk—through genes, parenting behavior, the rearing environment, and their complex interactions. This means that improving long-term mental health outcomes in children exposed to MSDP requires modification of maternal behaviors other than smoking. The proposed developmental framework which embeds MSDP within the greater landscape of the maternal–child relationship is an important step toward characterizing the insidious processes for which MSDP may simply be a marker. Even a minute possibility for successful preventive intervention justifies scientific investment and collaboration when the risks associated with the status quo are so high and so far-reaching.

SM drafted the initial manuscript. SM, JD, KW, and LW made substantial contributions to the conception of this manuscript, revising it critically for important intellectual content, and provided final approval for the current version to be published. All authors agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

SM, JD, and LW declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as potential conflicts of interest. The Department of Psychiatry at Northwestern University receives contractual fees for KW consultation to Quinn Emanuel Urquhart & Sullivan, LLP (New York City), who represent Pfizer Pharmaceutical Company.