Randomized Controlled Trial Examining the Efficacy of Adding Financial Incentives to Best Practices for Smoking Cessation Among Pregnant and Newly Postpartum Women

Abstract

We report results from a single-blinded randomized controlled trial examining financial incentives for smoking cessation among 249 pregnant and newly postpartum women. Participants included 169 women assigned to best practices (BP) or BP plus financial incentives (BP+FI) for smoking cessation available through 12-weeks postpartum. A third condition included 80 never-smokers (NS) sociodemographically-matched to women who smoked. Trial setting was Burlington, Vermont, USA, January, 2014 through January, 2020. Outcomes included 7-day point-prevalence abstinence antepartum and postpartum, and birth and other infant outcomes during 1st year of life. Reliability and external validity of results were assessed using pooled results from the current and four prior controlled trials coupled with data on maternal-smoking status and birth outcomes for all 2019 singleton live births in Vermont. Compared to BP, BP+FI significantly increased abstinence early- (AOR=9.97; 95%CI, 3.32–29.93) and late-pregnancy (primary outcome, AOR=5.61; 95%CI, 2.37–13.28) and through 12-weeks postpartum (AOR=2.46; CI,1.05–5.75) although not 24- (AOR=1.31; CI,0.54–3.17) or 48-weeks postpartum (AOR=1.33; CI,0.55–3.25). There was a significant effect of trial condition on small-for-gestational-age (SGA) deliveries (χ² [2]=9.01, P=.01), with percent SGA deliveries (±SEM) greatest in BP, intermediate in BP+FI, and lowest in NS (17.65±4.13, 10.81±3.61, and 2.53±1.77, respectively). Reliability analyses supported the efficacy of financial incentives for increasing abstinence antepartum and postpartum and decreasing SGA deliveries; external-validity analyses supported relationships between antepartum cessation and SGA risk. Adding financial incentives to Best Practice increases smoking cessation antepartum and postpartum women and improves other maternal-infant outcomes.

TRIAL REGISTRATION:

ClinicalTrials.gov Identifier: NCT02210832.

Introduction

Smoking is the leading preventable cause of poor pregnancy outcomes in developed countries, increasing risk for catastrophic pregnancy complications, preterm birth, small for gestational age (SGA) birth, sudden unexpected infant death (SUID), and latter-in-life metabolic disease (Anderson et al., 2019; Cnattingius et al., 2004; Gould et al., 2020; Higgins et al., 2020; USDHHS, 2020). Socioeconomically disadvantaged women are at especially high risk for smoking during pregnancy, which exacerbates health disparities (DeWolff et al., 2019; Higgins et al., 2009; Kurti et al., 2017).

Efficacious cessation treatments for pregnant women are available but quit rates are typically low (<15%) (Higgins & Solomon, 2016). Usual care for smoking cessation in the U.S. typically entails referral to a tobacco quitline; best practice entails coupling such referrals with follow-up and further referrals for women who continue smoking (CDC, 2014). Here we report results from a randomized controlled trial (RCT) examining the efficacy of best practices alone (BP) compared to best practices combined with financial incentives (BP+FI) on antepartum and postpartum smoking abstinence, breastfeeding, and infant birth and other health outcomes in the 1st year of life. A third condition included never-smokers (NS) sociodemographically matched to women who smoked. Systematic reviews and meta-analysis on smoking cessation with pregnant women support the efficacy of financial incentives for increasing antepartum cessation, but comparison conditions have varied and there is more to be learned regarding effects on birth outcomes, other health outcomes, and cost-effectiveness (Higgins & Solomon, 2016; Notley et al., 2019).

To enhance potential utility to policymakers, we also examined the reliability and external validity of trial results by examining consistency with those in a pooled data set comprising data from the current and four prior controlled trials conducted at the University of Vermont (UVM) examining similar financial incentive models and by examining associations between maternal smoking status and birth outcomes among all 2019 singleton live births in Vermont, the most recent calendar year that overlapped with the current trial. Note that what we refer to here as financial-incentive models are commonly referred to as contingency-management models in research on substance use disorders, especially illicit drug use disorders (e.g., Bolivar et al., 2021).

Methods

Trial Participants

Our goal was to enroll 345 women, including 230 still smoking at initiation of prenatal care and 115 never-smokers (biochemically-confirmed self-report of no smoking in past 7 days and < 100 cigarettes lifetime). Other inclusion criteria were gestational age ≤25 weeks, plans to remain in the area for ≥12 months following delivery, and English speaking. Exclusion criteria were incarceration, previous participation in a study on incentives for smoking cessation, residing with a current trial participant, and regular use of opioid, stimulant, or antipsychotic medications.

All women receiving prenatal care at referring clinics completed a questionnaire on smoking status. Women interested in study participation signed and returned the questionnaire to the clinic receptionist who forwarded it to study staff. 584 women who smoked began screening;126 failed to complete screening; 282 were ineligible; 176 were enrolled and randomly assigned to one of the two treatment conditions stratifying on the clinic where they received prenatal care (Figure 1). Most common reason for ineligibility was chronic opioid use (128/282, 45%), an unexpected obstacle to study enrollment. 759 women who reported never smoking initiated screening; 21 failed to complete screening; 657 were ineligible; 81 were enrolled (Figure 1). Never smoking was defined as biochemically verified self-report of no current smoking in the past 30 days and fewer than 100 cigarettes smoked lifetime (Centers for Disease Control and Prevention, 2017). Most common reason for ineligibility was failure to meet criteria for matching to women in the intervention conditions (574/657, 87%).

Figure 1.

CONSORT diagram on participant enrollment and randomization.

Only criterion for withdrawing someone from the trial was pregnancy termination/fetal demise; a total of 8 women (3 BP, 4 BP+FI, and 1 NS) were withdrawn on that criterion, leaving 169 women who smoked and 80 NSs (Table 1). Only women who smoked were included in analyses on smoking status and craving/withdrawal (n=169) while smokers and NSs were included in all other analyses (n=249). All participants provided written informed consent (see Supplement 1).

Table 1.

Participant characteristics.

Characteristics	BP1 (n=88)	BP+FI2 (n=81)	NS3 (n=80)	BP vs BP+FI p-valuea	All groups p-valueb
Demographics:
Age (years)	26.61 ± 5.47	25.40 ± 4.96	25.56 ± 4.96	.13	.25
Education				.08	.001
% <12 years	16	28	10c
% 12 years	64	48	80d
% >12 years	20	23	10c
% Non-Latino White	93	91	90	.66	.76
% Married	20	16	42d	.46	<.001
% Private insurance	28	27	40	.86	.15
% Employed outside of home	50	59	66	.23	.10
% 1st pregnancy	42	53	49	.15	.35
Weeks pregnant at intake	11.14 ± 4.07	12.37 ± 4.22	15.21 ± 6.44d	.05	<.001
Pre-pregnancy BMI	29.60 ± 8.45	28.20 ± 7.54	28.36 ± 8.30	.26	.47
Smoking Characteristics:
Cigs/day pre-pregnancy	18.27 ± 9.42	19.25 ± 9.87	NA	.51
Cigs/day at intake	9.92 ± 6.18	8.99 ± 5.21	NA	.29
Age started smoking (years)	15.47 ± 2.96	15.10 ± 2.92	NA	.41
% Living with other smoker(s)	77	79	22d	.75	<.001
% With no smoking allowed in home	70	68	91d	.72	<.001
% With none or few friends/family who smoke	26	25	78d	.83	<.001
% Attempted to quit pre-pregnancy	73	70	NA	.73
Number of quit attempts during pregnancy	0.73 ± 2.35	0.57 ± 0.97	NA	.56
Nicotine withdrawal questionnaire total scores	1.60 ± 0.75	1.37 ± 0.76	NA	.05
Fagerström total scores	4.23 ± 2.30	3.98 ± 2.07	NA	.46
Attitude Item:
% Endorsing that smoking will greatly harm baby	81	85	99d	.44	.001
Psychiatric Symptoms
Stress rating	6.26 ± 2.22	4.98 ± 2.47	3.34 ± 2.12d	<.001	<.001
Beck Depression Inventory	12.84 ±8.56	9.98 ± 7.36	5.78 ± 5.25d	.02	<.001
% Ever having depressive symptoms for 2+ weeks	52	42	20d	.18	<.001

Note: Reported values are M ± SD or %.

¹Best Practices

²Best Practices + Financial Incentives

³Never-Smokers

^aReported p-values are from t-tests for continuous measures and chi-square tests for categorical measures comparing the two intervention conditions.

^bReported p-values are from ANOVA or chi-square tests comparing the three trial conditions.

^cDenotes that Never-Smokers differ significantly from Best Practices + Financial Incentives condition; there were no instances where Never Smokers only differed from Best Practices condition.

^dDenotes that Never-Smokers differ significantly from both Best Practices and Best Practices + Financial Incentives conditions.

Trial Assessments

Participants completed questionnaires (e.g., Beck Depression Inventory (BDI, Beck et al., 1996), Minnesota Nicotine Withdrawal Scale (MNWS, Hughes & Hatsukami, 1986) examining sociodemographic, smoking, and psychiatric characteristics, and provided breath and urine specimens at intake. Breath specimens were analyzed using carbon monoxide (CO) monitors (Bedfont Scientific Ltd., Kent, UK) and urine specimens using an enzyme-immunoassay cotinine test (Microgenics Corp., Fremont, CA; Siemens Viva-E analyzer, Siemens Medical Solutions, Malvern, PA). Modified versions of this battery were completed one month after intake (early-pregnancy assessment), at ≥ 28-weeks gestation (late-pregnancy assessment), and 2-, 4-, 8-, 12-, 24-, and 48-weeks postpartum. To be considered abstinent, women had to report no smoking for the past 7 days and meet urine-cotinine abstinence criteria (≤80 ng/ml). Staff conducting follow-up assessments were not blinded to treatment condition.

Two serial ultrasound examinations were performed at approximately 30- and 34-weeks gestation to estimate fetal growth (Bernstein et al., 1997). Birth outcomes were obtained from maternal medical records. Infant-growth measures were obtained by study staff at 6- and 12-months; weight was obtained using an electronic scale; length by a measuring board. The Ages & Stages Questionnaire (ASQ) (Ages & Stages Questionnaires^®, 2021) was completed at these assessments to assess developmental milestones.

Trial Conditions

Following study enrollment, all participants assigned to BP chose a quit date within the subsequent two weeks. Once a quit date was selected, staff faxed a signed referral to the Vermont quitline which offered perinatal-specific individualized counseling (National Jewish Health, 2021); a maximum of five brief phone calls with a quit coach antepartum and four postpartum. To encourage call completion, the quitline offered up to $65 in incentives for completing calls. Women were eligible for free nicotine replacement if their medical provider agreed. In addition to quitline services, women received brief-cessation counseling from research staff at all assessments based on the brochure “Need Help Putting Out That Cigarette?” (National Partnership for Smoke Free Families, 2021).

Women assigned to BP+FI received the BP intervention combined with voucher-based financial-incentives available from quit date through 12-weeks postpartum. They were asked to select a Monday as their quit date and to attend clinic or be met by a staff member at an alternate site for the initial 5 days of the cessation effort; in Week 2, monitoring decreased to 2x weekly for next 7 weeks; weekly for 4 weeks; then once every other week until delivery. Following delivery, monitoring went back to weekly for 4 weeks, followed by every other week through 12-weeks postpartum when abstinence monitoring ended. Vouchers redeemable for retail items were earned contingent on submitting breath CO specimens ≤6 ppm during the initial five days of cessation. Beginning in Week 2, vouchers were delivered contingent on urine-cotinine levels ≤80 ng/ml determined using onsite enzyme immunoassay testing. These breath CO and urine-cotinine cut-points were previously validated with pregnant women (Higgins et al., 2007). Voucher delivery was based exclusively on meeting the biochemical-verification criterion. Unexcused failure to complete an assessment was treated as a positive-test result. For women smoking <10 CPD at intake, vouchers began at $6.25, and escalated by $1.25 per consecutive negative specimen to a maximum $45.00, where they remained barring positive-test results or missed abstinence-monitoring visits. Positive tests and missed visits reset voucher value back to the original low value, but two consecutive negative tests restored it to the pre-reset level. For women smoking ≥10 CPD, voucher values were offered on the same schedule but twice the values described above. Maximum total possible earnings for someone who participated in the voucher program for 32-weeks antepartum and 12-weeks postpartum and sustained abstinence throughout was $1,225 ($865 antepartum & $360 postpartum) for women smoking <10 CPD at intake and $2,450 ($1,730 antepartum & $720 postpartum) for those smoking ≥10 CPD. Because vouchers were only earned when women were abstinent, mean (+SEM) average earnings were $510.02+76.27 ($467.70+68.21 and $560.34+146.84 in lighter and heavier smokers, respectively). Women in the NS condition completed the same schedule of assessments minus counseling and quitline referrals. We hypothesized that women assigned to BP+FI would achieve greater antepartum smoking abstinence and their infants healthier birth outcomes than women assigned to BP and their infants.

Pooled Sample

Data from the 169 participants assigned to smoking-cessation arms in the current trial were combined with 284 participants from four prior controlled trials examining financial incentives (FI) for smoking cessation conducted in the same research clinic (see Heil et al., 2008; Higgins et al., 2004; Higgins et al., 2012; Higgins et al., 2014). Among the 453 women who contributed data, 245 were assigned to FI and 208 to control conditions. In the control conditions, the same voucher incentive schedule was used except vouchers were provided independent of smoking (non-contingently). A total of 313 women were enrolled in these four trials. Twenty-nine were excluded from the present study due to pregnancy termination/fetal demise, leaving 284 women who contributed data. All participants provided written informed consent.

Vermont State Sample

The Vermont state-wide sample consisted of all live singleton deliveries in 2019 (n=5,083 mother-infant pairs) (Vermont Vital Statistics Annual Report, 2019). Maternal insurance at delivery was 40.65% Medicaid, 55.12% private insurance, and 4.23% other. Maternal self-reported smoking status during the 3-months preceding and during pregnancy was obtained from birth records. SGA outcomes by maternal smoking status was determined from birth records to assess the external validity of trial outcomes detailed below. Data were de-identified and deemed exempt from informed consent.

Statistical Methods

The two intervention conditions were compared on baseline characteristics using t-test for continuous variables and chi-square tests for categorical variables. Additionally, the NS condition was then compared to the intervention conditions on baseline characteristics. If a characteristic differed significantly across study conditions and was predictive of outcome (P<.05), it was used as a covariate; smoking characteristics were not considered as covariates in analyses including the NS condition. Analyses of treatment effects on smoking status were limited to women randomized to the two intervention conditions using an intent-to-treat approach (Armitage et al., 1983) wherein all women randomized were included in the analyses except for exclusion for abortion/fetal demise. Late antepartum smoking abstinence and birth outcomes were the primary outcomes. Analysis of all other outcomes except craving/withdrawal included NSs. Comparisons of dichotomous outcomes with repeated measurements were analyzed using generalized-estimating equations utilizing a logistic-link function with treatment condition, assessment time, and the condition-by-assessment interaction as fixed effects. If the interaction was non-significant it was removed from the model (SAS: PROC GENMOD, 2021). Comparisons of treatment conditions on continuous outcomes with repeated measurements used linear-mixed models (SAS: PROC MIXED, 2021) with treatment condition, assessment, and the condition-by-assessment interaction as fixed effects. For outcomes measured at only one timepoint, comparisons were done using analysis of variance and logistic regression for continuous and dichotomous outcomes, respectively. For birth outcomes, additional potential covariates included maternal pre-pregnancy BMI, first pregnancy, and fetal gender. If they were associated with birth outcome at P<.25 they were included in a backward elimination. Only covariates at P<.20 were retained in the final model (Vittinghoff et al., 2012). Analysis of birthweight included gestational age at delivery in the model. Analyses for the combined effects for all five trials were performed in a similar manner including trial as a fixed effect. Analyses of sonographic estimated fetal growth were based on methods of Hadlock et al. (1985). To protect against preterm delivery for iatrogenic reasons potentially obscuring treatment effects on smoking-related fetal growth restriction (Benjamin et al., 2013; England et al., 2001; USDHHS, 2020) analyses on mean birth weight and percent SGA deliveries (<10^th percentile, INTERGROWTH-21^st [2021]) were analyzed including all study participants and limiting inclusion to only term deliveries. Significance level of alpha <.05 was used or all analyses. All analyses were performed using SAS 9.4 (SAS Institute, Cary, NC).

Results

Participants

Participants assigned to BP and BP+FI were on average 25–26 yrs of age, most completed ≤12 years of education, and were Non-Latino White race/ethnicity (Table 1). Participants reported smoking on average 18–19 CPD pre-pregnancy and 9–10 CPD at intake. The only significant differences in baseline characteristics between women in BP versus BP+FI were greater stress ratings and depressive symptoms in BP. Women in NS were comparable to those in BP and BP+FI on age and race/ethnicity but completed more years of education and reported lower stress and depressive symptoms. These differences were included as covariates.

Smoking Abstinence

In the current trial, there was a significant interaction of trial condition and time on smoking abstinence (χ² (7)=21.93, P=.003) (Figure 2, upper panel). Compared to BP, women assigned to BP+FI had greater adjusted odds of early- (34.57% vs. 4.55%; AOR=9.97; 95%CI, 3.32–29.93) and late-pregnancy abstinence (primary outcome, 38.27% vs. 9.09%; AOR=5.61; 95%CI, 2.37–13.28). These effects were evident across lighter and heavier smokers, with late-pregnancy abstinence in BP+FI exceeding those in BP among women who smoked <10 (50.00% vs. 12.24%) and ≥10 CPD (24.32% vs. 5.13%) at intake. Adjusted odds of abstinence remained greater in BP+FI versus BP through the 12-week postpartum assessment (AOR=2.46; 95%CI, 1.05–5.75), although not at 24- or 48-week assessments conducted after incentives were discontinued (AOR=1.31; 95%CI, 0.54–3.17; AOR =1.33; 95%CI, 0.55–3.25, respectively) (Figure 2, upper panel).

Figure 2.

Seven-day point-prevalence abstinence outcomes for the Best Practices and Best Practices + Financial Incentives conditions in the current trial (top panel) and Financial Incentives and Control conditions in the pooled trials (bottom panel). X axes show consecutive assessments (early antepartum [Early AP], late antepartum [Late AP], weeks 2–48 postpartum (PP) (top panel) or 2–24 postpartum (PP) (bottom panel). Y axes show percent of participants abstinent. Asterisks represent assessments where abstinence was significantly greater in Best Practices + Financial Incentives compared to Best Practices (top panels) or Financial Incentives compared to Controls (bottom panels). Significance represents P < .05.

Abstinence levels in the pooled trials supported the reliability of results from the current trial with a significant interaction of trial condition and time (χ2(1)=26.67, P<.001) (Figure 2, lower panel). Compared to controls, women assigned to FI had greater adjusted odds of early- (38.37% vs. 6.25%; AOR=10.96; 95%CI, 5.80–20.71) and late-pregnancy abstinence (37.55% vs. 10.10%, AOR=6.20; 95%CI, 3.60–10.67). That effect was also evident among women who smoked <10 (49.29% vs. 14.81%) and ≥10 CPD (21.90% vs. 5.00%). Adjusted odds of abstinence remained greater among women in FI than controls through the 12-week postpartum assessment (AOR=3.02; 95%CI, 1.86–6.02) as well as 24-week assessment completed 12 weeks after incentives were discontinued (AOR=2.47; 95%CI, 1.32–4.63) (Figure 2, lower panel). Postpartum assessments were not conducted beyond 24 weeks in prior trials.

Craving and Withdrawal

In the current trial, there was a significant interaction of trial condition and time on mean MNWS craving (F[7,921]=4.16, P<.001) (Figure 3, upper left panel). Compared to BP, women in BP+FI reported less craving, with mean ratings significantly reduced from antepartum through 2-, 4- and 12-weeks postpartum (Ps≤.003 antepartum and <.05 postpartum). A comparable effect of trial condition was observed for MNWS total scores (F[7,922]=3.62, P<.001), with reductions in mean scores in BP+FI compared to BP through antepartum (Ps≤.001) although not postpartum (Figure 3, upper right panel).

Figure 3.

Mean Minnesota Nicotine Withdrawal Scale (MNWS) Craving (‘Desire to Smoke’ item) ratings and Total Scores from current trial are shown in the upper left and right panels, respectively, and those from the pooled trials are shown in the bottom left and right panels, respectively. Mean scores are shown on the Y axes and assessment times on the X axes (early antepartum (Early AP) and late antepartum (Late AP) assessments followed by postpartum (PP) weeks 2–48 (current trial) and 2–24 (pooled trials). Asterisks represent specific assessments where values in Best Practices +Financial Incentives were significantly lower than Best Practices in upper left and right panels; in the lower left panel asterisks represent specific assessments where values in Financial Incentives were significantly lower than Controls; the bracket in the lower right panel represents a significant main effect of trial condition where scores were significantly lower in Financial incentives than Controls and a significant main effect of time where scores in both trial conditions decreased over time, with no significant interaction of trial condition and time. Significance represents P < .05.

Results in the pooled trials supported the reliability of those results. There was a significant interaction of trial condition and time on mean craving (F[6,2226]=3.68, P=.001), with craving less in FI compared to controls throughout antepartum and initial four-weeks postpartum (Figure 3, lower left panel). There were also main effects of trial condition (F[1,419]=7.16, P=.008) and time (F[6,2235]=43.48, P<.001) on MNWS total scores, with scores lower in FI compared to controls (0.88±0.04 vs. 1.03±0.05) and decreasing over time in both conditions (Figure 3, lower right panel), with no significant interaction of trial condition and time.

Fetal Growth and Birth Outcomes

There were no significant effects of trial condition on sonographic fetal-growth outcomes (eTable 1, Supplement 2). Mean outcomes for estimated fetal weight and abdominal circumference were in the hypothesized direction (NS > BP+FI > BP) but non-significant and thus not examined in the pooled trials.

Regarding birth outcomes, there was a significant main effect of trial condition on percent SGA deliveries with levels greatest in BP, intermediate in BP+FI, and lowest in NS (χ² [2]=9.01, P=.01) (Figure 4, upper-left panel; Table 2, upper panel). Compared to NS, women in BP had approximately nine-times greater adjusted odds of an SGA delivery (AOR=9.09; 95%CI, 1.94–41.67). Adjusted odds of SGA in BP+FI were intermediate between BP and NS and not significantly different from either; adjusted odds of SGA in BP+FI were approximately four-times greater compared to NS (AOR=3.86; 95%CI, 0.76–19.61) and approximately two-times greater in BP compared to BP+FI (AOR=2.33; 95%CI, 0.88–6.25), but 95%CIs included one in both comparisons.

Figure 4.

Small for gestational age (SGA) deliveries by trial condition in the current trial (top panel) and pooled trials (bottom panel). Left-most upper and lower panels show analyses including all infants while right-most upper and lower panels show results for term infants. BP = Best Practices (n=88 all infants; n=79 term infants), BP+FI = Best Practices + Financial Incentives (n=81 all infants; n=65 term infants), NS= Never-smokers (n=80 all infants; n=71 term infants), C = Control Conditions (n=202 all infants; n=179 term infants), and FI = Financial Incentives (n=234 all infants; n=213 term infants). Conditions not sharing a superscript letter differ significantly at P < .05.

Table 2.

Birth outcomes by trial condition.

	Current Trial
Measure	BP 1 (All infants, n=88) (Term infants, n=79)	BP + FI 2 (All infants, n=76) (Term infants, n=65)	NS 3 (All infants, n=79) (Term infants, n=71)	p-value for group effect
SGA (%) (all)	17.65 ± 4.13a	10.81 ± 3.61a,b	2.53 ± 1.77b	.01
SGA (%) (term)	17.72 ± 4.30a	7.69 ± 3.30b	1.41± 1.40b	.01
Birth Weight (g) (all)	3176.39 ± 46.74a	3214.84 ± 50.11a	3391.14 ± 49.64b	.005
Birth Weight (g) (term)	3241.39 ± 47.66a	3290.24 ± 52.94a	3471.15 ± 51.33 b	.004
Gestational Age (wks) (all)	39.18 ± 0.22a	38.69 ± 0.23a	39.35 ± 0.22a	.10
Preterm (%) (all)	7.06 ± 2.78a	13.16 ± 3.88a	7.59 ± 2.98a	.41
NICU (%) (all)	12.35 ± 3.66a	13.24 ± 4.11a	13.92 ± 3.89a	.77
	Pooled Trials
Measure	Controls 4 (All, n=202) (Term, n=179)	FI 5 (All, n=234) (Term, n=213)
SGA (%) (all)	15.92 ± 2.58a	8.62 + 1.84b	.02
SGA (%) (term)	15.64 ± 2.71a	7.08 + 1.76b	.01
Birth Weight (g) (all)	3198.23 ± 31.78a	3248.10 ± 29.79a	.23
Birth Weight (g) (term)	3286.77 ± 33.78a	3337.95 30.84a	.24

Note: Values represent percentages ± SEs for SGA (small for gestational age deliveries), Preterm (preterm deliveries), and NICU (neonatal intensive care unit admissions) and least square means ± SEMs for Birth Weight (mean birth weight adjusted for gestational age) and Gestational Age (gestational age in weeks at delivery). Data points not sharing a superscript letter differ significantly at p < .05.

¹BP = Best Practices

²BP+FI = Best Practices + Financial Incentives

³NS = Never-Smokers

⁴Controls = No-Incentive Control Condition

⁵FI = Financial Incentives

A similar pattern was seen when SGA analyses were limited to term infants, but differences between study conditions were larger. Rather than a nine-fold difference when all participants were included, adjusted odds of SGA in BP compared to NS were approximately 15-times greater and remained significant (AOR=15.15; 95%CI, 1.90–125.00); adjusted odds of SGA in BP+FI compared to NS increased from approximately four- to almost five-fold, but remained non-significant (AOR=4.72; 95%CI, 0.52–43.48); and adjusted odds of SGA among women assigned to BP compared to BP+FI increased from approximately two- to three-fold and were now significantly different (AOR=3.19; 95%CI, 1.02–10.00) (Table 2, upper panel).

The other birth outcome for which there was a significant main effect of trial condition was mean birth weight adjusted for gestational age (F[2,230]=5.33, P=.006). Mean birth weight was greatest in NS, intermediate in BP+FI, and lowest in BP (Table 2, upper panel). Mean birth weight in NS was significantly greater than in BP and BP+FI (Ps<.01), with no significant difference between the two smoking-intervention conditions (P=.58). Restricting this analysis to term infants was associated with no discernible changes beyond increases in mean birth weight across all conditions. There was a significant main effect of trial condition (F[2,206]=5.65, P=.004), with mean birth weight in the NS condition greater than BP (P=.001) and BP+FI (P=.02), with no significant difference between the two smoking-intervention conditions (P=.49).

There were no significant effects of trial condition on gestational age at delivery (F[2,235]=2.28, P=.10), preterm deliveries (χ²=1.79, P=.41), or neonatal intensive care unit (NICU) admissions (χ²=0.52, P=.77) (Table 2, upper panel). Hence, those effects were not examined in the pooled trials. Reliability of the two birth outcomes for which there were significant treatment effects was supported in the pooled trials (Table 2, lower panel). There was a significant main effect of trial condition on percent SGA deliveries (χ² [1]=5.59, P=.02; Figure 4, lower-left panel). Adjusted odds were approximately two-times greater in controls compared to FI (AOR=2.09; 95%CI, 1.14.–3.85). When analyses were restricted to term deliveries the main effect of trial condition remained significant (χ² [1]=6.63, P=.01) and adjusted odds of SGA increased from approximately 2.1- to 2.5-times greater in controls compared to FI (AOR=2.45; 95%CI, 1.24.–4.85) (Figure 4, lower-right panel).

The pooled-trial outcome for mean birth weight adjusted for gestational age in FI and controls was graded in the hypothesized direction but not significantly different (F[1,424]=1.46, P=.23) (Table 2, lower panel). Restricting this analysis to term infants resulted in expected increases in mean birth weights in FI and Controls but with no significant difference (F[1,380]=1.40, P=.24)

External validity of treatment effects on SGA seen in the current and pooled trials was supported by results in the state-wide sample. That sample included mothers who reported no smoking in the three months before or during pregnancy (NS, n=4264), smoked before or during pregnancy but quit before third trimester (S+Q, n=268), or smoked into the third trimester (S, n=551). Consistent with patterns seen in the trials, there was a significant graded decrease in odds of SGA deliveries when comparing S versus S+Q versus NS in the entire sample (χ² [2]=150.34, P<.001; Figure 5, upper-left panel). Overall, odds of SGA deliveries were approximately two-times greater among women in S compared to S+Q (OR=2.04; 95%CI, 1.36–3.06) and four-times greater among women in S compared to NS (OR=4.16; 95%CI, 3.30–5.24). That same pattern was observed when restricted to those with Medicaid insurance (χ²[2]=74.59, P<.001; Figure 5, upper-right panel) and private insurance (χ² [2]=17.21, P<.001; Figure 5, lower-left panel), although not other insurance where sample sizes were relatively small (χ² [2]=2.50, P=.29; Figure 5, lower-right panel). These analyses were not repeated restricting the sample to term deliveries.

Figure 5.

Small for gestational age (SGA) deliveries for all 2019 singleton births in Vermont overall and separated by maternal insurance status. NS= Never-smokers, S = Smokers throughout pregnancy, and S+Q = Smokers who quit during the pregnancy. Conditions not sharing a superscript letter differ significantly at P < .05.

Breastfeeding

There were no significant differences between trial conditions in overall breastfeeding duration across the 1^st year postpartum follow-up period (χ² [2]=4.34, P=.11); rates decreased over time in all conditions (χ² [5]=151.73, P<.001) with no significant interaction of trial condition and time (χ²[10]=9.52, P=.48; Table 3, left-most columns in upper panel).

Table 3.

Percent of women breastfeeding (left-most rows within each panel) and breastfeeding while abstinent from smoking (right-most rows within each panel) at all postpartum assessments from the current trial (top panel) and pooled trials (bottom panel).

	Current Trial
	% Breastfeeding *			% Breastfeeding and abstinent #
Assessments	BP1 (n=88)	BP + FI2 (n=81)	NS3 (n=80)	BP (n=88)	BP + FI (n=81)	NS (n=80)
	Percent (n/d)	Percent (n/d)	Percent (n/d) a	Percent (n/d)	Percent (n/d)	Percent (n/d) c
2 wks	68.06 (49/72)a	78.13 (50/64)a	78.67 (59/75)a	17.05 (15/88)b	35.80 (29/81)a	78.67 (59/75)c
4 wks	52.63 (40/76)a	60.61 (40/66)a	72.73 (56/77)a	12.50 (11/88)b	29.63 (24/81)a	72.73 (56/77)c
8 wks	44.74 (34/76)a	49.25 (33/67)a	63.16 (48/76)a	10.23 (9/88)b	23.46 (19/81)a	63.16 (48/76)c
12 wks	36.84 (28/76)a	47.06 (32/68)a	56.58 (43/76)a	7.95 (7/88)b	18.52 (15/81)a	56.58 (43/76)c
24 wks	28.00 (21/54)a	35.48 (22/62)a	50.00 (36/72)a	7.95 (7/88)b	11.11 (9/81)a	50.00 (36/72)c
48 wks	12.16 (9/74)a	15.00 (9/60)a	32.86 (23/70)a	3.41 (3/88)b	6.12 (5/81)a	32.86 (23/70)c
	Pooled Trials
	% Breastfeeding #		% Breastfeeding and abstinent *
Assessments	Controls4 (n=208)	FI5 (n=245)	Controls (n=208)	FI (n=245)
	Percent (n/d)	Percent (n/d)	Percent (n/d)	Percent (n/d)
2 wks	61.18 (104/170)a	67.31 (140/208)a	12.98 (27/208)a	27.76 (68/245)b
4 wks	47.73 (84/176)a	52.58 (112/213)a	9.62 (20/208)a	22.86 (56/245)b
8 wks	38.64 (68/176)a	43.98 (95/216)a	7.21 (15/208)a	18.37 (45/245)b
12 wks	29.05 (52/179)a	37.79 (82/217)a	4.81 (10/208)a	14.69 (36/245)b
24 wks	22.78 (41/180)a	24.76 (52/210)a	3.37 (7/208)a	7.35 (18/245)b
48 wks	-----------	-----------	-----------	-----------

Note: Data points represent percents (numerator/denominator).

^#No significant main effect of trial condition or interaction of trial condition and time.

^*Significant main effect of trial condition across assessments; data points not sharing a superscript letter differ significantly, P<.01

¹BP + FI = Best Practices

²BP + FI = Best Practices + Financial Incentives

³NS = Never-Smokers

⁴Controls = No-Incentives Control Condition.

⁵FI = Financial Incentives

n/d = numerator/denominator

Where trial conditions differed significantly was in percent of women breastfeeding while sustaining smoking abstinence (χ² [2]=33.51, P<.0001) (Table 3, right-most column in upper panel), with rates highest in NS, intermediate in BP+FI, and lowest in BP. Adjusted odds in NS were approximately 12-times greater compared to BP (AOR=12.53; 95% CI, 5.24–29.96) but only three-times greater compared to BP+FI (AOR=3.75; 95% CI, 1.89–7.42), with adjusted odds in BP+FI approximately three-times greater than in BP (AOR=3.34; 95% CI, 1.51–7.40). Percent of women breastfeeding while sustaining abstinence also decreased over time (χ² [5]=77.39, P<.0001), but with no significant interaction of trial condition and time (χ² [10]=5.00, P=.89), meaning that the aforementioned differences between trial conditions were discernible throughout the one-year follow-up period.

That same pattern was observed in the pooled trials, with no significant effect of incentives on overall percent of women breastfeeding (χ²(1)=2.39, P=.12) (Table 3, left-most columns in lower panel), but approximately three-times greater adjusted odds of women in FI compared to controls breastfeeding while sustaining smoking abstinence throughout postpartum (χ²(1)=17.78, P<.001; AOR=3.07; 95%CI, 1.82–5.16) (Table 3, right-most columns in lower panel). Rates decreased over time in both conditions (χ²(4)=83.48, P<.001) but with no significant interaction of trial condition and time (χ²(4)=1.48, P=.83), meaning that the incentives effect remained discernible throughout follow-up. Breastfeeding outcomes were not examined in the state-wide sample.

1st-year Infant Growth and Developmental Milestones

There was a significant interaction of trial condition and time on infant growth (BMI percentile) during the 1st year of life (F[4,351]=3.43, P=.009) (eTable 2, Supplement 2). At birth, mean BMI percentile was greater in NS compared to BP (P=.04), with values in BP+FI intermediate between them and not significantly different from either. Changes in growth from birth to the 24-week postpartum assessment were greater in BP and BP+FI versus NS infants (Ps≤.003), with no significant difference between BP and BP+FI (P≥.05). Those growth patterns were such that BMI percentiles at the 24-week assessment were greater in BP and BP+FI versus NS (Ps≤.04) and again not different between BP and BP+FI (P≥.05). Those growth patterns changed again from the 24- to 48-week assessments, with infants born to mothers in NS on average growing at a significantly greater rate than in BP+FI (P=.04) and trending in that direction compared to BP (P=.09) such that there were no longer significant differences in BMI percentiles between trial conditions at 48-weeks postpartum (Ps≥.05) (i.e., infants in the two intervention conditions had “caught up” to those in NS).

Regarding ASQ developmental milestones (eTable 3, Supplement 2), there was a significant main effect of trial condition on maternal ratings of infant problem solving (χ2(2)=6.76, P=.03), with adjusted odds of possible delays greater in BP versus BP+FI (AOR=2.50; 95%CI, 1.08–5.88) and NS versus BP+FI (AOR=2.67; 95%CI; 1.21–5.91). There were no significant effects of trial condition on the other ASQ developmental outcomes (Ps≥.05).

Discussion

The current RCT trial is the fifth in a series examining this financial-incentives model, with each trial supporting its efficacy for increasing smoking abstinence (Heil et al., 2008; Higgins et al., 2004; Higgins et al., 2012; Higgins et al., 2014). The current trial is the first demonstration of its efficacy when compared against best practices. Lumley et al. noted in a 2009 report from the Cochrane meta-analysis series on smoking cessation during pregnancy that financial incentives produce the largest increases in late-pregnancy abstinence of any intervention tested in controlled studies, which at that time included 72 trials involving more than 25,000 women. That remains the case, with abstinence levels in the current and pooled trials representing what to our knowledge are the largest treatment effects reported for late-pregnancy abstinence in a literature that has grown to include 100 trials and greater than 30,000 women (Chamberlain et al., 2017; Coleman et al., 2015). The current results also fill a gap in the literature on smoking cessation during pregnancy noted by Chamberlain et al. (2017) regarding a paucity of trials comparing financial incentives against low-intensity interventions although there are excellent examples (Berlin et al., 2021; Tappin et al., 2015). The current trial demonstrates that combining financial incentives with best practices increases adjusted odds of late-pregnancy abstinence approximately six-times above best practices alone and that odds of abstinence remain several-fold higher through 12-weeks postpartum when incentives were discontinued. The only meaningful difference noted between results from the current and pooled trials is in magnitude of the treatment effect after incentives were discontinued; adjusted odds of abstinence were approximately 1.3-times greater than controls at 24- and 48-weeks postpartum in the current trial although these differences were not statistically significant. By contrast, in the pooled trials odds of abstinence were more than 2.4-times greater in the incentives compared to control condition and statistically significant at the 24-week assessment. The pooled trials had a substantially larger sample size suggesting that the failure to discern significance in the current trial may have been influenced by insufficient statistical power. The current trial is the only one in this series that went out to 48-weeks postpartum where odds of abstinence remained largely unchanged from 24-weeks. Whether the 1.3 or 2.4-fold estimated odds of abstaining after incentives are discontinued is the more reliable estimate will have to be determined in future trials, but both suggest benefit beyond discontinuation of the incentives. Continuing abstinence-contingent incentives further into the postpartum period to curtail relapse warrants investigation. Another important matter that has been largely ignored and warrants examination is whether combining incentives with other interventions (e.g., Community Reinforcement Approach therapy (Higgins et al., 2003), episodic future thinking (Stein et al., 2016), or substitution of non-combusted tobacco products for conventional cigarettes for women who cannot or do not wish to discontinue nicotine use (Benowitz et al., 2021) might help sustain longer-term smoking-abstinence.

A reassuring observation in the current study that has received scant prior attention is that increases in antepartum and postpartum smoking abstinence achieved with financial incentives do not result in untoward craving or withdrawal. Instead, results show that incentives decrease craving and withdrawal antepartum and early postpartum. Berlin et al. (2021) also reported significant reductions in antepartum craving among pregnant women randomly assigned to an incentive intervention for abstinence compared to usual care although no difference in withdrawal total scores. These results are reassuring considering the vulnerability of peripartum women to labile mood (Biaggi et al., 2015; Stewart & Vigod, 2019).

This study also provides new evidence that adjusted odds of SGA among women who receive best practices alone or other control conditions are approximately two- to three-times greater than women who receive financial incentives. We previously reported that incentives reduced odds of low-birth-weight deliveries (Higgins et al., 2010a) but examined SGA in the present study to avoid potential confounding by differences in gestational age in the earlier report (USDHHS, 2020). When comparing BP and BP+FI in the current trial among all infants, adjusted odds were approximately two-times greater in the former but CIs included one. When that same comparison of BP to BP+FI was restricted to term infants, adjusted odds increased to approximately three-times greater in BP without CIs including one consistent with preterm deliveries potentially obscuring treatment effects on infant growth restriction (Benjamin-Garner et al, 2013; England et al., 2021; USDHHS, 2020). In the pooled trials, adjusted odds of SGA were approximately two-times greater in the control than FI conditions when including all participants and without overlap in CIs and that effect increased to approximately two-and-a-half-times greater when analyses were restricted to term infants also consistent with at least a limited influence of preterm deliveries in obscuring treatment effects on SGA. We saw no evidence that limiting analyses to term infants had a meaningful influence on treatment effects on mean birth weight adjusted for gestational age. While there have been prior trials that have restricted analyses to term deliveries when examining effects of smoking on SGA and mean birth weight (Benjamin-Garner et al, 2013; England et al., 2021), to our knowledge, the present study may be unique in reporting SGA outcomes with and without preterm infants included. Berlin et al. (2021) did not report SGA outcomes. They did report a trend towards a greater proportion of low birth weight (< 2500 g) deliveries among infants assigned to the control compared to the incentive intervention (unadjusted OR=1.95; 95%CI, 0.99–3.85, P=.055) that was significant when adjusted for infant sex (AOR=2.05; 95%CI, 1.03–4.10, P=.041) but not sex and prematurity (AOR=2.06; 95CI, 0.90–4.71), P=0.086).

We previously reported that financial incentives increased overall breastfeeding duration, with differences at 8- and 12-weeks postpartum in the incentive and control conditions in that study being 41% versus 26% and 35% versus 17%, respectively (Higgins et al., 2010b). That outcome was not replicated in the present study. What appears to have changed is not duration of breastfeeding in the incentives condition but duration in the control conditions, with rates at 8- and 12-weeks postpartum in the BP+FI versus BP conditions in the current trial, for example, being 49% versus 45% and 47% versus 37%, respectively. Breastfeeding recommendations have evolved over time such that the potential maternal-infant benefits of breastfeeding by women who smoke are thought to outweigh potential risks of infant exposure to toxins in breastmilk and second-hand smoke (ACOG, 2021). Importantly, though, that position does not ignore potential harms. Abstinence from smoking during lactation is still recommended as the safest course of action for protecting maternal and infant health (ACOG, 2021; Napierala et al., 2016).

We did not detect significant effects of incentivizing smoking abstinence on infant growth patterns during the 1st year of life and a significant effect on only one of five ASQ developmental categories. These were largely exploratory outcomes that we have not investigated previously. There was a prior pilot study using financial incentives to promote antepartum cessation wherein infants born to quitters appeared to exhibit less catch-up growth in the 1st year of life than infants born to mothers who continued smoking (Wen et al., 2019). Because impacting catch-up growth has potential to improve health throughout the life span (Katz et al., 2013; Ludvigsson et al., 2018; Kesavan & Devaskar, 2019), we wanted to examine whether the BP+FI intervention altered that outcome. While we observed significantly different growth patterns between infants born to mothers in the two smoking-cessation conditions compared to NS consistent with catch-up growth, we saw no evidence that greater smoking cessation in BP+FI compared to BP was sufficient to significantly alter catch-up growth patterns.

There are several study limitations that merit mention. First, we were unable to enroll the proposed sample size for the current trial in large part due to the unanticipated increase in opioid use disorder in the population of interest; opioid use was an exclusion criterion because opioids enhance the reinforcing effects of smoking which risks altering treatment response (Chait & Griffiths, 1984). Having the additional statistical power that a larger sample would afford could only have improved the precision of the estimates regarding the impact of incentives on outcomes. Second, participants in the current trial, pooled trials, and the state-wide analyses were all from a single rural U.S. state with a predominately non-Hispanic White population raising questions about generalizability to more diverse samples. We encourage extending investigation of this financial-incentives model to improving perinatal outcomes among racial/ethnic minority pregnant women (see Washio et al., 2017) and have developed a smartphone adaptation of this intervention to extend reach to women in geographically remote areas (Kurti et al., 2020).

These limitations notwithstanding, the present study provides important new knowledge on the efficacy, reliability, external validity, and cost-effectiveness (see Shepard et al., 2022, this issue) of financial incentives for decreasing smoking during pregnancy and early postpartum that has potential to improve maternal and infant health outcomes and reduce health disparities.

Highlights.

Financial incentives increase smoking abstinence during pregnancy.

Financial incentives decrease smoking craving and withdrawal.

Financial incentives decrease small for gestational age births.

Financial incentives increase women breastfeeding while abstinent.