Are empathic processes mechanisms of pregnancy’s protective effect on smoking? Identification of a novel target for preventive intervention

Abstract

Background:

Spontaneous cessation and reduction in smoking by pregnant women suggest that concern about others, or empathy, could be a malleable target for intervention. We examined various empathy-related processes in relations to reported and biochemically assessed smoking during pregnancy.

Methods:

Participants were 154 pregnant women (M = 12.4 weeks gestation, SD = 4.6) who were smoking cigarettes immediately prior to pregnancy recognition (85 had quit and 69 were still smoking at enrollment). Empathy-related processes were measured with performance-based paradigms (affect sharing, empathic concern, and theory of mind) and a speech sample (expressed emotion). Smoking was assessed with timeline follow back interviews and urine cotinine assays. Using zero-inflated Poisson regression models, we tested direct and interactive effects of empathy-related processes with respect to biologically verified smoking cessation (zero portion); and mean cigarettes/day smoked after pregnancy recognition among persistent smokers (count portion).

Results:

Affect sharing was inversely related to post-recognition cigarettes/day (B(SE) = −0.17(0.07), 95%C.I. −0.30,−0.04, p = .011) and moderated the relationship between pre-recognition smoking and post-recognition smoking consistent with a buffering effect (B(SE) = −.17(0.05); 95%C.I. − 0.28,−0.06; p = .002). Other empathy related processes showed neither direct nor interactive effects on smoking outcomes.

Conclusions:

Further research is recommended to clarify the role of empathy in pregnancy smoking.

1. Introduction

Spontaneous suspension or reduction in cigarette smoking by pregnant women is a well-documented phenomenon (Solomon and Quinn, 2004) widely viewed as evidence that women’s concern about fetal well-being can temporarily override the drive to smoke (Massey et al., 2012, 2015a; Stotts et al., 1996). Documentation of this ‘protective effect’ of pregnancy on women’s alcohol use and illicit drug use (Edwards et al., 2019; Kendler et al., 2017) support the view of pregnancy as a natural experimental paradigm for the specification of mechanisms that interrupt addictive processes (Massey et al., 2017; Massey and Wisner, 2018). Qualitative and quantitative evidence support the conceptualization of reduction in smoking by pregnant women as caregiving behaviors motivated by concerns about the unborn baby rather than typical health-related behaviors motivated by concerns about personal health (Flemming et al., 2013; Graham et al., 2014; Massey et al., 2017; Stotts et al., 1996). It has previously been proposed that processes related to concern for the well-being of others, sometimes described as empathy (Decety et al., 2016; Decety and Jackson, 2004), and related antecedents of sensitive caregiving behavior (Massey et al., 2018a) may underlie the vast within- and between-individual variation in smoking behavior observed across the perinatal period (Pickett et al., 2003; Wakschlag et al., 2003). Empathy is conceptualized as having distinct facets, including affective, motivational and cognitive components (Decety et al., 2015), each of which could differentially influence smoking behavior (Massey et al., 2017). Yet these facets, referred to for simplicity as empathy, have not been systematically examined in relation to smoking behavior during pregnancy (Massey et al., 2015a, 2018b; Massey and Wisner, 2018).

Finally, when viewed as a prosocial (helping) behavior that emerges in the context of the developing maternal-fetal relationship (Massey et al., 2012, 2015a; Zahn-Waxler and Radke-Yarrow, 1990), positive changes in pregnancy smoking might be expected to reflect the specific antecedent processes thought to motivate other types of prosocial behavior (Massey et al., 2017)—namely, perceiving the target as being in need (Decety et al., 2015), adopting the target’s perspective (i.e. perspective-taking) (Batson, 2014; Batson et al., 1987), and the ability to regulate internal emotions that might be elicited by the former two processes (Lockwood et al., 2014). While this well-studied neuroscientific framework formed the foundation for our selection of measures of empathy-related processes (Decety et al., 2016), this framework is not typically applied to understanding behavior of pregnant women (Massey et al., 2018e). Furthermore, the degree of overlap between empathy directed towards one’s child versus towards a general conspecific (i.e., another adult) is unclear.

1.1. Aims and hypotheses (Fig. 1)

Fig. 1.

Interactions tested (top) using the zero-inflated Poisson model (bottom) for examining likelihood of quitting (zero portion) and level of smoking in persistent smokers (count portion).

A substantial proportion of women who begin a pregnancy as smokers abruptly quit and abstain through delivery once they suspect or confirm a pregnancy (Solomon and Quinn, 2004). Those who do not quit often cut down the number of cigarettes smoked per day with the goal of minimizing harm to the fetus (Graham et al., 2014). While smoking cessation is the ideal outcome, understanding what facilitates reduction in cigarettes/day smoked among women who do not quit (Flemming et al., 2013; Graham et al., 2014) is also meaningful since the dose of prenatal exposure (mean CPD) is inversely related to birth weight (Massey et al., 2018d).

Thus, as illustrated in Fig. 1, we examined the role of empathy in quitting smoking and in spontaneous change in smoking in response to recognition of the pregnancy. We hypothesized that empathy would attenuate the predictive relationship between the level of smoking prior to pregnancy recognition (left) and (a) the likelihood of biologically verified smoking cessation; and (b) for women who did not quit, level of smoking after pregnancy recognition.

To inform future work aimed at validating empathy-related processes as a mechanism of pregnancy-associated reduction in smoking (Massey et al., 2017; Sheeran et al., 2017), an exploratory goal of this study was to elucidate interrelationships among different components of empathy (affective versus cognitive), and to whom empathic processes are directed (general others versus one’s own child). We hypothesized that affective and cognitive empathic abilities would be distinct, as in seen in non-pregnant adults (Gleichgerrcht and Decety, 2014; Kogler et al., 2020). Furthermore, considering the putative overlap in neuro-circuitry subserving parental and non-parental social behaviors (Swain et al., 2012), we hypothesized that different facets of empathic ability, and empathy with regard to general others versus towards the imagined future child would be intercorrelated.

2. Methods

2.1. Sample and recruitment

All procedures described were approved by the Northwestern University Institutional Review Board prior to conduct. Participants were N=154 English-speaking pregnant women at least 18 years of age in their second trimester or earlier, identified from either online advertisements or the electronic medical record of a large academic medical center located in a large city in the Midwestern United States (Fig. 2). Potential participants were invited to participate in “a study about smokers and how thoughts and feelings during pregnancy relate to smoking behavior.” From a total of 281 potential participants identified, 269 were screened for eligibility (Fig. 2). Because we were interested in the role of empathy related processes in the changes in smoking made specifically because of the pregnancy, women had to have been smoking cigarettes regularly immediately prior to recognizing the current pregnancy. For example, former smokers who were no longer smoking immediately prior to the current pregnancy were excluded (n = 72) because they may have quit for reasons other than pregnancy. Smokers who had quit smoking since realizing they were pregnant (and thus were not smoking at enrollment), and those who were still smoking at enrollment were intentionally included to ensure our sample reflected the full range of known variability in early pregnancy smoking behavior (Pickett et al., 2003; Solomon and Quinn, 2004).

Fig. 2.

Recruitment and derivation of analytic samples for primary analyses.

Exclusion criteria were inability to provide informed consent due to psychosis, delirium, or intellectual disability (n =3); having a condition characterized by severe deficits in social cognition (autism, schizophrenia) (none encountered); and women receiving treatment for smoking cessation (none encountered). One hundred ninety-four women screened met eligibility criteria; of these, n =154 women enrolled, completed the study visit, and constituted the analytic sample. Women who enrolled (n = 154) and those who declined to enroll (n = 40) did not differ with respect to characteristics collected from the screening questionnaire (i.e. age, race, ethnicity).

2.2. Study procedures

Following informed consent procedures, assessments described below were conducted, in the order described at an average of 12.5 weeks of gestation, (SD = 4.6). The entire study visits lasted between two and two-and-a half hours. All study assessments were conducted in a private office separate from any individuals who may have accompanied the participant to the lab. This was done to protect participant privacy and to minimize social desirability bias associated with reported smoking. Participants were compensated with a gift card for $100 at the end of the visit.

2.3. Measures

2.3.1. Retrospective reports of smoking before and after recognition of the pregnancy

We used timeline follow-back interviews to assess patterns of smoking over time (Sobell and Sobell, 1996). Using a paper calendar, participants were asked to identify salient recent events, including the date they recognized the current pregnancy. Next, they were instructed to record the number of cigarettes they smoked on each day beginning 4 weeks prior to the date the pregnancy was recognized through the present day. The primary predictor variable, pre-recognition smoking, a direct reflection of smoking-related processes occurring immediately prior to the putative onset of pregnancy-related protective processes (i. e., concerns about the pregnancy and the fetus), was derived from the mean cigarettes per day smoked during this four-week period. To capture change in smoking behavior specifically in response to the recognition of the pregnancy, known to be a salient event that motivates spontaneous changes in smoking behavior (Flemming et al., 2013; Graham et al., 2014), and not in response to other life stressors that may have occurred during gestation, we calculated the mean cigarettes per day smoked during the first week after the date the pregnancy was recognized to generate the continuous smoking outcome variable, post-recognition smoking-maternal report.

2.3.2. Assessment of smoking via cotinine bioassay of maternal urine

Smoking during pregnancy was also assessed from maternal urine specimens collected when convenient over the course of the study visit. Cotinine, a major nicotine metabolite of smoking (Benowitz et al., 2009), was quantified from urine samples with liquid chromatography-mass spectroscopy (LC/MS) performed at the United States Drug Testing Laboratories in nearby Des Plains, IL. Trimester-specific urine cotinine concentration (ng/ml) cut-offs reflective of increasing rates of gestational nicotine metabolism were used to recategorize women who reported smoking cessation on the TLFB calendar, but whose cotinine value suggested ongoing smoking (n = 4) (Stragierowicz et al., 2013).

2.3.3. Computerized neurocognitive assessments of empathic ability

Next, participants completed a 50-min battery of computerized neurocognitive tasks designed to elicit normal variations in adults’ capacity for affective empathy (affect sharing and empathic concern) and cognitive empathy (theory of mind). In the first section, we assessed two components of affective empathy - affect sharing, also known as emotional empathy, which refers to individuals’ perception of others’ pain, and empathic concern, which refers to individuals’ caring motivation to help others (Decety et al., 2015) using a well-validated paradigm linked to neuroanatomical circuits thought to subserve these processes in functional neuroimaging studies (Decety et al., 2009, 2015; Lamm et al., 2011).

Briefly, participants were presented with 40 randomly interspersed images consisting of 20 pain images (stimulus) and 20 non-pain images (control) of adults’ hands and feet in situations encountered in day-to-day life that would be expected to elicit somatic pain (i.e., hand directly under the cutting edge of a kitchen knife). Control images depicted similar non-pain situations (i.e., hand is next to, but not under the same knife). A neutral face was presented for 5 s prior to each image to enhance salience. Participants used a visual analog scale (1–100) to rate how much pain the depicted individual was experiencing (for affect sharing), and how sorry they felt for the depicted individual (empathic concern). Individuals were given as much time as they needed to respond to each question before advancing to the next image. Continuous scores were derived from net ratings (rating for stimulus image – rating for control image). The affective empathy paradigm took approximately 5 min to complete.

2.3.4. Cognitive empathy (theory of mind)

Participants also completed a neurocognitive battery consisting of five validated theory of mind paradigms containing images of adult conspecifics of both sexes: the Reading the Mind in the Eyes Test (Baron-Cohen et al., 2001), the Emotion Recognition-40 Test (Benton et al., 1994), a morphed faces task for emotional intensity, a trustworthiness task (participants judged the trustworthiness of an individual whose face was shown), and an age discrimination task (participants rated which of two individuals was older). Details of this battery were described previously (Friesen and Ekman, 1976; Goldman et al., 2011). The order in which images within each task were presented was randomized and performance was not subject to time pressure. Performance (accuracy) from each of the five theory of mind tasks (Cronbach’s α = 0.82) was used to assess different facets of theory of mind. The entire battery of tests for affective and cognitive empathy took approximately 50 min to complete.

2.3.5. Empathy-related processes specific to the fetus (expressed emotion)

Expressed emotion is a measure of a parent’s criticism, over-protectiveness, or emotional over-involvement with their child, derived from qualitative analysis of parent’s verbalization of the relationship from the Five-Minute Speech Sample (Kazarian, 1992). There is recent interest in assessing maternal expressed emotion prenatally using an abbreviated (to 3 min) Five-Minute Speech Sample - Pregnancy (Lambregtse-van den Berg et al., 2013); maternal high expressed emotion during pregnancy has been associated with child psychopathology though mechanisms are unclear (Lucassen et al., 2015). Since other assessments of empathy-related processes in this study contained images of adult conspecifics, we sought to probe how emotional processes specific to the (imagined) child related to smoking behavior (Jussila et al., 2020; Magee et al., 2014; Massey et al., 2015a). Thus, we administered the Five-Minute Speech Sample - Pregnancy to provide a brief qualitative assessment of cognitive and affective processes relating to the future child. Participants were asked to speak openly about the relationship they wished to have with their future child, what they expected their child would be like, and how they imagined they would relate to their child. Speech was recorded for 3 min, then transcribed verbatim, noting pauses, sighs, and other non-verbal expressions discernible from audio-recordings. De-identified transcripts were scored a senior psychological research analyst at the University of California at Los Angeles (UCLA) family project (Weston et al., 2017) for criticism (negative initial statement or comments indicating disdain, annoyance) and emotional over-involvement (statements that indicated self-sacrifice, over-protective behavior or attitudes, excessive detail about, or excessive praise towards the child). The scorer was blinded to participants’ smoking data. Consistent with conventions for expressed emotion, we used a categorical variable of expressed emotion that reflected either high expressed emotion (evidence of either criticism or emotional over-involvement) or low expressed emotion (absence of these features).

2.3.6. Covariates

Smoking covariates included in regression models were selected a priori, based on demographic and smoking-related factors previously associated with inter-individual variability in patterns of pregnancy smoking. These covariates, annual household income (Cnattingius, 2004) and partner smoking (Homish et al., 2012) were included in all regression models.

2.4. Data analysis

All continuous variables were examined for distribution and outliers prior to entry into regression models. First, we examined psychosocial characteristics of the entire sample and as a function of quit status (Table 1) derived from timeline follow back and verified by urine cotinine concentration (Stragierowicz et al., 2013). Four women who reported quitting smoking were recategorized into the persistent smoking category due to urine cotinine values. As seen in our prior studies in different samples of pregnant smokers (Massey et al., 2020) variables for mean cigarettes per day both before and after pregnancy recognition were positively skewed [pre-recognition cigarettes/day (skew = 1.77; kurtosis = 3.48) and post-recognition smoking (skew = 2.96; kurtosis = 10.10)]. While we have previously employed square root transformation, most appropriate for data containing multiple zero values (Manikandan, 2010), we opted to utilize zero-inflated Poisson regression to examine hypothesized interactions shown in Fig. 1. This approach allowed us to simultaneously examine two important smoking outcomes: (a) smoking cessation after pregnancy recognition; and (b) mean cigarettes per day after pregnancy recognition via the zero and count portions of the model, respectively. Scores on the age discrimination and trustworthiness tasks showed bimodal distributions that were difficult to interpret and threatened assumptions of the Poisson model. Other empathy-related processes were tested in separate models to aid in interpretability and avoid potential multicollinearity. All continuous independent variables were standardized using z-score conversions prior to entry into models. We probed significant moderator effects using the procedure recommended by Preacher et al. (2006); we examined relevant effects 1.5 standard deviations above and 1.5 standard deviations below the mean (Poole and O’Farrell, 1971; Preacher et al., 2006). Descriptive statistics were performed in SPSS version 26. Zero-inflated Poisson regression modeling was conducted in R version 4.0.3, RStudio version 1.4.1103, and the R stats package version 4.0.3.

Table 1.

Characteristics of study participants (N = 154) by smoking status based on timeline follow back interview and urine cotinine concentration.

	Total N = 154	Quitters n = 85	Persistent smokers n = 69	t or χ2	P
	Mean (SD) or % (n)
Age in years	28.3 (4.1)	28.6 (3.9)	27.9 (4.4)	−1.05	.293
Gestational age in weeks	12.4 (4.6)	11.8 (4.8)	13.2 (4.3)	1.89	.060
Married (Y/N)	55.8% (86)	70.6% (60)	37.7% (26)	16.73	<.001
Multiparous (Y/N)	72.7% (112)	74.1% (63)	71.0% (49)	0.19	.667
Unplanned pregnancy (Y/N)	54.5% (84)	52.9% (45)	56.5% (39)	0.20	.657
Educational attainment				2.50	.778
Less than high school	7.8% (12)	5.2% (8)	2.6% (4)
High school/GED	18.8% (29)	9.7% (15)	9.1% (14)
Vocational/Assoc. degree	39.0% (60)	20.8% (32)	18.2% (28)
Any four-year college	17.5% (27)	9.7% (15)	7.8% (12)
Bachelor’s degree or more	16.9% (26)	9.7% (15)	7.2% (11)
Annual household income				11.76	.038
<$20,000	11.0% (17)	3.2% (5)	7.8% (12)
$20,000-$50,000	13.0% (20)	5.2% (8)	7.8% (12)
$50,000-$80,000	33.1% (51)	21.4% (33)	11.7% (18)
$80,000-$100,000	29.2% (45)	15.6% (24)	13.6% (21)
>$100,000	11.7% (18)	7.8% (12)	3.9% (6)
Works outside home (Y/N)	88.3% (136)	48.7% (75)	39.6% (61)	<.01	.974
Partner smokes (Y/N)	72.7% (112)	38.3% (59)	34.4% (53)	1.05	.305
Aversion to cigarettes (Y/N)	8.4% (13)	5.2% (8)	3.2% (5)	0.23	.631
Race/ethnicity				17.71	.003
African American/Black	23.4% (36)	9.1% (14)	14.3% (22)
American Indian/Alaska Native	3.2% (5)	1.3% (2)	1.9% (3)
Asian	1.3% (2)	0% (0)	1.3% (2)
Native Hawaiian/Pacific Islander	1.9% (3)	0.6% (1)	1.3% (2)
White	61.0% (94)	41.6% (64)	19.5% (30)
Hispanic/Latino	9.1% (14)	2.6% (4)	6.5% (10)
Fagerstrom score (current)	0.6 (1.3)	0.3 (0.8)	1.2 (1.6)	3.99	<.001
Urine cotinine concentration (ng/mL)	17.54 (89.84)	0.12 (0.82)	39.0 (131.6)	2.73	.007
Mean CPD pre-recognition	4.55 (4.02)	3.77 (3.08)	5.51 (4.81)	2.73	.007
Mean CPD post-recognition	1.13 (2.51)	0	2.46 (3.28)	6.69	<.001
Mean post-recognition Δ in CPD	3.41 (3.65)	3.71 (2.95)	3.05 (4.37)	−1.10	.272

3. Results

3.1. Characteristics of participants, total, and by quit status (Table 1)

Participants were primarily high school graduates with some vocational or other advanced training in their late twenties (M = 28.3 years, SD = 4.1), most of whom who worked outside the home (88.3%), had at least one other child (72.7% multiparous), and a partner who smoked (72.7%). Racial/ethnic composition of the sample was 61.0% White, 23.4% African American/Black, 9.1% Hispanic/Latino, 3.2% American Indian/Alaska Native, and 3.2% Asian/Native Hawaiian/Pacific Islander. About half of participants were unmarried (44.2%) and reported that the current pregnancy was unplanned (54.5%). A minority of women (8.4%) reported a current aversion to the taste and/or smell of cigarettes. Mean gestational age at the time of the study visit was 12.4 weeks (SD = 4.6 weeks).

During the month immediately preceding the date of pregnancy recognition, women smoked an average of five cigarettes per day (M = 4.7, SD = 4.1, range 0.2–19.8). Smoking dropped to about one cigarette per day (M =1.1, SD = 2.5, range 0–16.0) during the first week after the pregnancy was recognized. Mean cotinine concentration was 17.5 ng/mL (SD = 89.8, range 0–947.0). Eighty-nine women (58%) reported having quit smoking the week after recognizing the pregnancy while n = 65 (42%) reported continuing to smoke. Four women who reported smoking cessation were recategorized as persistent smokers after incorporating cotinine assay results.

3.2. Intercorrelations among empathy-related processes

Scores on facets of affective empathy (affect sharing, empathic concern) were intercorrelated (r = 0.52, p < .001), inversely related to (high) expressed emotion (r = −0.19, p < .05 r = −0.31, p < .001, respectively), and not correlated with any of the cognitive empathy tasks. Among cognitive empathy tasks, the Reading the Mind in the Eyes, Emotion Recognition-40, morphed faces, and age discrimination tasks were intercorrelated (r’s range from 0.22 to 0.80, p < .001 for all correlations). Meanwhile, performance on the trustworthiness task was inversely related to performance on the Reading the Mind in the Eyes (r = −0.19, p < .05) and morphed faces tasks (r = −0.21, p < .01).

3.3. Main and interactive effects on smoking cessation (Table 2, bottom)

Table 2.

Coefficients from count and zero portions of zero-inflated Poisson regression model testing moderation of pre-recognition smoking by affect sharing.

		Estimate	Std. Error	z value	Pr(>∣ z∣)
Count Predictors of CPD among persistent smokers	Intercept	2.733	0.078	35.135	<.001
	Mean cigs/day (CPD)	0.187	0.024	7.869	<.001
	Affect sharing (AS)	−0.002	0.039	−0.05	0.96
	Income	−0.226	0.034	−6.67	<.001
	Partner smoking	0.345	0.075	4.572	<.001
	Mean CPD x AS	−0.093	0.018	−5.1	<.001
Zero Predictors of smoking cessation	Intercept	0.874	0.434	2.013	0.044
	CPD	−0.977	0.272	−3.592	<.001
	AS	0.452	0.224	2.013	0.044
	Income	0.672	0.236	2.843	0.004
	Partner smoking	0.24	0.503	0.476	0.634
	Mean CPD x AS	0.383	0.292	1.315	0.189

CPD = Cigarettes per day.

The zero portion of the Poisson regression model indicated an inverse association between pre-recognition smoking and the likelihood of smoking cessation [B(SE) = −0.977(0.27), z =3.592, p < .001] and direct associations between income [B(SE) = 0.672 (0.236), z =2.843, p = .004] and affect sharing and smoking cessation [B(SE) = 0.452(0.22), z = 2.013, p = .044]. The interaction between pre-recognition smoking and affect sharing was not significantly associated with smoking cessation [B(SE) = 0.383(0.292), z = 1.315, p = .189]. Other empathy-related processes did not show direct nor interactive effects on smoking cessation.

3.4. Main and interactive effects on post-recognition smoking in persistent smokers (Table 2, top)

From the count portion of the Poisson regression model, we observed direct effects of pre-recognition smoking [B(SE) = 0.383(0.024), z = 7.869, p < .001] and partner smoking [B(SE) = 0.345(0.075), z = 4.572, p < .001] on post-recognition smoking. Income was inversely associated with post-recognition cigarettes/day [B(SE) = −0.226(0.034), z = −6.67, p < .001]. Additionally, affect sharing interacted with pre-recognition smoking in the prediction of post-recognition smoking consistent with a buffering effect [B(SE) = −0.093(0.018), z = −5.1, p < .001]. As depicted graphically in Fig. 3, the relationship between pre-recognition smoking and post-recognition smoking was stronger when affect sharing was low and attenuated when affect sharing was high. Other empathy-related processes did exhibit direct nor interactive effects on post-recognition smoking.

Fig. 3.

Association between pre- and post-recognition smoking in persistent smokers with mean, high (+1.5 SD), and low (−1.5 SD) affect sharing (N = 154) from count portion of zero-inflated Poisson regression. Covariates: household income and partner smoking.

4. Discussion

Empathic processes antecedent to helping (Batson, 2014) or caregiving behavior (Boorman et al., 2019) are plausible mechanisms underlying widely-observed spontaneous changes in smoking among pregnant women (Massey et al., 2017), and thus potential targets for intervention development. Our prior work supports associations between empathic processes (Massey et al., 2015a; Massey et al., 2018b; Massey et al., 2012; Massey et al., 2018e) and their biological substrates (Massey et al., 2015b) in and patterns of pregnancy smoking. This is the first study to our knowledge to examine empathy-related processes in pregnant smokers using performance-based measures in relation to smoking cessation versus persistent smoking and to mean cigarettes/day smoked after the pregnancy was recognized among persistent smokers.

4.1. Empathy-related processes and smoking outcomes

Affect sharing, operationalized as the ability to perceive others’ somatic pain, was associated with likelihood of biochemically verified smoking cessation (Table 2, bottom). For the remainder of women who did not quit smoking, affect sharing buffered the expected relationship between mean cigarettes/day smoked before pregnancy recognition and mean cigarettes/day smoked after pregnancy recognition (Table 2, top, and Fig. 3). Other facets of empathy tested did not show direct nor interactive effects on smoking outcomes. We carefully considered whether validity of these paradigms for use in pregnant women played a role in null findings. However, as shown in Table 1, the intercorrelation between affect sharing and empathic concern, and distinctions between affective and cognitive components of empathy previously described in non-pregnant adults (Eres et al., 2015), were broadly mirrored in this sample of pregnant smokers. High expressed emotion was also inversely related to affective components of empathy, consistent with the view that empathic concern and prosocial behavior is dependent on emotional regulation (Brethel-Haurwitz et al., 2020; López-Pérez and Ambrona, 2015). These interrelationships provided some measure of reassurance about the validity of participant scores.

4.2. Other correlates of smoking cessation and post-recognition level of smoking

Mirroring results from other samples (Riaz et al., 2018), pre-recognition smoking level and income showed direct and inverse associations, respectively, with post-recognition smoking level and with smoking cessation. Partner smoking was associated with post-recognition smoking level, but not with the likelihood of smoking cessation. This raises the possibility that partner smoking is a marker of heavier pre-pregnancy smoking rather than a barrier to smoking cessation as is commonly believed. Clarifying mechanisms in this regard is recommended to inform prevention strategies.

4.3. Implications for prevention

Standard smoking cessation interventions designed for non-pregnant women and men have had limited efficacy and acceptability in pregnant smokers; we have previously proposed that specialized interventions for pregnant women necessitate consideration of pregnancy smoking in the context of a maternal-fetal dynamic (Massey et al., 2017). Results of the current study are the first to link smokers’ affect sharing, or the ability to perceive others’ pain, to the likelihood of smoking cessation. Affect sharing also appeared to buffer the effect of smoking-related processes among women who did not quit.

There is growing interest in the perception of pain and its modifiability via non-pharmacologic intervention, such as loving-kindness meditation a mindfulness-based approach that is beginning to be examined in pregnancy (Babbar et al., 2021). Using the same pain paradigm as used in the current study, mere exposure to loving kindness language, not the practice of loving kindness meditation per se, resulted in a reduction in individuals’ own pain, and surprisingly, an increase in individuals’ perception of others’ pain (Williams et al., 2018). Future research could extend our findings by testing whether loving kindness meditation can increase affect sharing in pregnant smokers, increase chances of quitting, and reduce subsequent smoking if cessation is not achieved. This approach would be the first step to validation of affect sharing as a mechanism of change in smoking during pregnancy (Sheeran et al., 2017).

Results of this study may also have implications beyond pregnancy. Addiction research has primarily focused on how social processes are interrupted by addictive substance use. A growing body of neuroimaging research has shown that smoking and substance-using mothers of infants exhibit impairments in empathic processing—specifically, in the processing of infant cues (Lowell et al., 2020). How social processes, including empathic processes, might interrupt addictive processes, however, is far less studied (Heilig et al., 2016). This is surprising since preclinical studies indicate that social interactions involving empathy are rewarding (Venniro and Golden, 2020) and that social rewards effectively compete with rewards from addictive drugs (Venniro et al., 2018). Our results suggest that this could also be true in pregnant women. Affect sharing may relate to social (maternal-fetal) processes that temporarily compete with smoking-related processes or addictive processes more broadly (Kendler et al., 2017; Massey and Wisner, 2018). Extending this line of research across clinical and translational domains is recommended as new and more effective treatment for addiction is sorely needed.

Findings also have implications for elucidating mechanisms linking prenatal tobacco exposure with behavioral and emotional dysregulation in children (Clark et al., 2016; Eiden et al., 2015; Massey et al., 2020; Wakschlag et al., 1997). We have previously shown that maternal responsiveness buffers the effect of pregnancy smoking on child behavior problems (Wakschlag and Hans, 2002). More recently, we found that women who quit smoking during pregnancy (as verified by prospective cotinine assays) exhibited more responsive parenting behavior when observed at home with their children five years later (Massey et al., 2018a). Results of the current study extend evidence that implicates maternal empathic processes as one mechanism through which prenatal smoking (and smoking cessation during pregnancy) influences neurodevelopment.

4.4. Limitations

Findings should be interpreted in the context of a modest sample size, recruitment of participants from a single geographic location, and relatively low mean levels of smoking before pregnancy (about five cigarettes/day). Next, it is important to emphasize that directionality in the relationship between affect sharing and smoking is unclear. While affect sharing assessed at baseline could buffer the effect of pre-recognition smoking levels on post-recognition smoking, we cannot rule out the possibility that smoking, or withdrawal from smoking, adversely influenced affect sharing. Rutherford and colleagues have demonstrated, for example, that substance-using mothers of infants exhibit impairments in recognizing infant cues and differences in underlying neural signaling, relative to non-substance-using mothers (Landi et al., 2011; Lowell et al., 2020). Future studies that incorporate validated functional neuroimaging assessments with the pain paradigm used in this study (Decety et al., 2009), and/or assessments involving infant cues (Rutherford et al., 2020) could confirm and extend support for empathy-related processes as a mechanism of prenatal smoking behavior change. Further, the development of empathy assessments specific to the maternal-fetal relationship could be ideal and well-justified considering the far-reaching ramifications of prenatal smoking (Massey et al., 2017, 2018c). Finally, others have demonstrated links between nicotine dependence, impaired identification of subjective emotional experiences (alexithymia) and differences in functional connectivity between the anterior insula and ventromedial prefrontal cortex, regions that also subserve empathic processing (Stoeckel et al., 2016; Sutherland et al., 2013, 2016). Thus, we cannot rule out the possibility that symptoms of nicotine withdrawal, expected to be greater among smokers who had quit, influenced affect sharing ability.

5. Conclusions

Findings from this study are the first to link affect sharing to patterns of reported and biochemically assessed smoking during pregnancy. Replication in a larger sample using more comprehensive multi-level assessments is recommended to determine whether affect sharing and other social processes are capable of overriding reward from addictive drugs (Massey and Wisner, 2018).