Smoking Abstinence Symptoms across 67 Days Compared with Randomized Controls—Moderation by Nicotine Replacement Therapy, Bupropion, and Negative-Affect Traits

Abstract

Accurate knowledge of negative affect (NA)-related smoking abstinence symptoms (SAS) severity and duration and their moderation by pharmacotherapy and NA-related personality traits is critical for efficacious treatments given that elevated state and trait NA are predictors of relapse. However, SAS severity, duration, and moderation are not well characterized. To date, the longest randomized controlled trial (RCT) of NA-related SAS using randomized delayed-quit smoking controls only examined symptoms across 45 days, despite clinical evidence that SAS may last longer. The present RCT assessed SAS across 67 days in dependent smokers (N =95) who were randomized either to quit or to delay quitting for the course of the trial. The quit group was further randomized to receive either Nicotine Replacement Therapy (NRT), bupropion (BUP), or Placebo. Abstinence-related increases in anger-irritability, depressive, anxiety symptoms, and general NA did not resolve relative to the delayed quit group (DQG) levels across the 67 days in any of the three quit groups, though craving fell to below DQG and pre-quit levels. While NRT attenuated day-3 SAS relative to BUP and Placebo, BUP and NRT generally did not reduce SAS. High scores on trait measures of NA/neuroticism predicted greater increases in and duration of NA-related SAS, potentially indicating that smoking abstinence unmasks affective symptoms. Positive affect was not impacted by abstinence or treatment. The results support the views that: 1) pre-quit baseline values are not a valid index of NA SAS recovery, and 2) on average, NA-related SAS take longer than 67 days to resolve.

Mood-related smoking abstinence symptoms (SAS) are a primary predictor of smoking relapse. Yet the answers to two most basic questions about SAS are unknown: 1) how long does it take SAS (withdrawal symptoms, in particular) to resolve and 2) do affect-related individual differences or smoking cessation aids affect the duration or severity of SAS? Correlational studies suggest that negative-affect-related symptom improve after smoking cessation, but randomized controlled trials (RCTs) with control groups that continue to smoke are necessary to account for potential self-selection biases and testing effects (drift in self-reported negative affect [NA] associated with repeated testing). To date, only three such RCTs aiming to characterize SAS has lasted more than four weeks and none have examined SAS beyond 45 days. Therefore, the current study was designed to 1) extend the experimental investigation of SAS to 67 days, 2) assess the moderating influence of NA-related personality traits on SAS severity, duration, and type, and 3) assess the influence of two popular smoking cessation aids (nicotine replacement therapy [NRT] and bupropion SR [BUP]) on SAS.

Importance of SAS and differing prior estimates of duration.

Findings to date suggest that that elevated NA is a predictor of relapse (Shiffman & Paty, 2006) independent of smoking urge withdrawal symptoms (Piasecki, Niaura, Shadel, Abrams, Goldstein, Fiore, et al., 2000). Further, a major reason for smoking is to reduce NA and to increase positive affect (PA) (Gilbert, Sharpe, Ramanaiah, Detwiler, & Anderson 2000). Further, SAS duration may be moderated by NA-related personality traits and psychopathology (Gilbert, McClernon, Rabinovich, Plath, Jensen, & Meliska, 1998; Gilbert, McClernon, Rabinovich, Plath, Masson, Anderson, & Sly, 2002; Gilbert, Zuo, Rabinovich, Riise, Needham, & Huggenvik et al., 2009; Leventhal & Zvolensky, 2015) and by pre-quit baseline NA (Piasecki et al., 2000; Piasecki, Jorenby, Smith, Fiore, & Baker, 2003a).

Estimates of SAS resolution time have increased across the last 40 years as more rigorous and sophisticated analyses have been used. Early estimates suggested that SAS resolve within two weeks after quitting (Shiffman, 1979), which was later extended to 3–5 weeks (Cummings, Giovino, Jaen, & Emrich, 1985; Gross & Stitzer, 1989; Hughes, 1992; Hughes, Gust, Skoog, Keenan, & Fenwick, 1991; West, Hajek, & Belcher, 1987). However, more recent experimental evidence suggests that SAS resolution may take 8 weeks or even longer (Shiffman, Khayrallah, & Nowak, 2000; Piasecki, Jorenby, Smith, Fiore, & Baker, 2003b). Nonetheless, clinical and epidemiological studies suggest that symptom resolution may be far more rapid.

Observational studies.

A growing number of researchers rely on observational clinical studies to argue that NA-related SAS resolve within several weeks following smoking cessation and that smoking may increase NA and decrease mental health (reviewed by Taylor, McNeill, Girling, Farley, Lindson-Hawley, Aveyard et al., 2014). For example, a recent meta-analysis of studies that evaluated mental health symptoms across time among quitters and non-quitters concluded:

“Whether or not smoking cessation directly causes the observed improvement in mental health, there are direct clinical implications. Smokers can be reassured that stopping smoking is associated with mental health benefits. This could also overcome barriers that clinicians have toward intervening with smokers with mental health problems” (Taylor et al., 2014, pg. 7).

Similarly, other clinical researchers have interpreted their findings as indicating that, “Whatever immediate effects of smoking may have on perceived stress, overall it may generate or aggravate negative emotional states. The results provide reassurance to smokers worried that stopping smoking may deprive them of a valuable resource,” (Hajek, Taylor & McRobbie, 2010, p. 1466). Further, others have stated that “a burgeoning body of research demonstrating that significant improvements in psychological functioning can be observed among those who successfully quit smoking even in the most severe psychiatric group [clinical depression]” (Mathew, Robinson, Norton, Cinciripini, Brown, & Blalock, 2013, p. 1807).

Other epidemiological studies have similarly found that ex-smokers experience better subjective wellbeing, less stress, and less NA than individuals who continue to smoke (Mulder, Tijhuis, Smit, & Kromhout, 2001). This finding has been interpreted by many as indicating that quitting smoking causes a reduction in NA and an enhancement of positive affect and wellbeing. For example, the authors of a study of 9,660 smokers and ex-smokers stated that, “Knowledge of the impact of smoking cessation on health-related quality of life may be important in encouraging smokers to quit” (Mulder et al., 2001, p. 653). If the conclusions of these researchers and the meta-analysis are valid, this would be good news and simplify tasks for clinicians and researchers who could inform smoker attempting to quit that they will feel better soon.

Evidence that NA-related SAS resolution may be protracted: Randomized controlled clinical trials.

However, a small number of controlled experimental studies have partially addressed the limitations of observational studies, finding that SAS resolution may take longer than 7 weeks (Gilbert & Pergadia, 2017; Gilbert, et al., 2002; Piasecki & Baker, 2000). This work has generally randomly assigned quitting smokers to a cessation aid or placebo (e.g., Piasecki et al., 2000; Shiffman Khayrallah, & Nowak, 2000). For example, Shiffman and colleagues (2000) found that the nicotine patch (NRT) reduced NA-related SAS compared to placebo even at 7–10 weeks after quitting, suggesting that abstinence symptoms may not resolve even after two months or more.

Importantly, one large (n = 893) well-designed RCT demonstrated large individual differences in SAS trajectories that, on average, did not return to baseline levels and actually increased in severity across time in some individuals across eight weeks (Piasecki, Jorenby, Smith, Fiore, & Baker, 2003b). The study demonstrated that while SAS largely returned to prequit baseline levels of NA after three to six weeks of quitting if one used the pre-quit day with the highest pre-quit symptom levels, NA SAS did not resolve within eight weeks if one of the more numerous lower prequit NA levels was considered the baseline. As expected, NA SAS were more severe in individuals who lapsed. It is also important to note that Piasecki et al. findings: 1) were based on mean values across treatments (NRT, BUP, or placebo, with about 75 percent on one or both of the drugs, which were found to significantly reduce NA-related SAS), 2) had only 22% of quitters maintain abstinence across the 8-week study, and 3) like almost all studies in this area, did not include a randomized no-quit control group. Thus, the Piasecki et al., 2003b study supports the hypothesis that the natural course of unmedicated SAS may be longer than eight weeks on average, especially in some individuals.

Problems with relying on correlational studies: Biased relapse, downward NA drift, and the need for RCTs with no-quit controls.

Biased dropout.

It is important to note that most RCT clinical studies are inherently correlational in nature given that individuals are not randomly assigned to maintain abstinence versus relapse or continue to smoke. Specifically, almost all RCTs have been designed to evaluate the relative efficacy of different treatment interventions, all of which have high relapse rates (Shiffman, West, & Gilbert, 2004). As such, they have not included no-quit controls because such controls are irrelevant to the primary study goal of characterizing treatment efficacy.

However, there is reason to believe that NA-related SAS severity is generally underestimated because individuals with the most severe NA-related SAS—those who experience random exogenous stressors, and those high in NA-related traits and contexts—are more likely to relapse than others. Thus, it is to be expected that epidemiological studies show that ex-smokers have better subjective wellbeing and less stress and NA compared to individuals who continue to smoke (Audrain-McGovern, Rodriguez, & Kassel, 2009; Vangeli, Stapleton, Smit, Borland, & West, 2011). Those with higher socioeconomic status (SES) and lower pre-quit stress levels and psychopathology are more successful in their quit attempts than individuals who fail to quit (Audrain-McGovern et al., 2009; Gilbert, 1995; Khantzian, 1997). Further, individuals with high NA and poor mental health are less likely to be become ex-smokers. Taken together, these results make it erroneously appear that relapse to smoking causes increased NA (Gilbert & Pergadia, 2017).

In this regard, it may be instructive to recall the hormone-replacement therapy (HRT) debacle of late 20th century medicine. For several decades, most physicians and patients were convinced of significant health benefits of HRT in post-menopausal women for the heart and health. Subsequently, the NIH conducted a large randomized clinical trial, the results of which compellingly demonstrated that this treatment increased health risk for cardiovascular disease and breast cancer rather than preventing them (Writing Group for the Women’s Health Initiative Investigators, 2002). Non-randomized, biased samples played a key role in the generation of the previous earnest but erroneous belief in the health benefits of HRT (Stampfer & Colditz, 1991). It is important to ensure that a similar mistake is not made in attempts to characterize SAS duration because assuming the SAS resolve sooner than they do would lead to inadequate treatments.

Downward drift.

An additional concern with correlational studies of symptomology is that they fail to account for downward drift—the tendency for participants to report progressively fewer negative affect-related symptoms when tested repeatedly. There is a large, though often-ignored, literature demonstrating that downward drift of 30 to 50% occurs in self-reported NA scores across repeated assessments in both smoking (Gilbert et al., 1998; Gilbert et al., 2002; Gilbert & Pergadia, 2017) and nonsmoking populations (Ahava, Iannone, Grebstein, & Shirling, 1998; Arrindell, 2001; Atkeson, Calhoun, Resick, & Ellis, 1982; Choquette & Hesselbrock, 1987; Hatzenbuehler, Parpal, & Matthews, 1983; Sharpe & Gilbert, 1998) regardless of assignment to treatment or condition. Thus, in smoking cessation studies without randomly assigned non-quit controls the downward drift in NA-related symptoms can easily be confused with resolution of NA SAS in smoking cessation studies without randomly assigned no-quit controls. Relatedly, NA symptoms tend to increase during the week before quitting (Gilbert & Pergadia, 2017), as was the case of Piasecki et al. (2003b). Without control groups, the upward drifts in NA frequently observed in the few days just before quitting can further bias symptom trajectory interpretations given that these elevations make it appear that resolution of symptoms has occurred because of elevated pre-quit baseline estimates of NA SAS (Gilbert & Pergadia, 2017). These upward shifts are situationally induced and occur across a larger and continuing overall downward drift that is independent of environmental factors.

Need for RCTs with randomized no-quit controls.

Despite these concerns, randomized no-quit controls have been used in only three RCTs that lasted more than four weeks and were designed with the primary aim of characterizing mood-related SAS. These RCTs demonstrated two important points. First, the abstainers did not return to no-quit control levels across the 31 or 45 days assessed. Second, these studies illustrated the importance of including a randomized control group that continues to smoke because of the large downward drift in NA (Gilbert et al., 1998; Gilbert et al., 2002; Gilbert & Pergadia, 2017). Although helpful for clarifying the change in NA over time, these RCTs followed participants for fewer than seven weeks (Gilbert et al., 1998 [31 days]; Gilbert et al., 2002 [31 days]; Gilbert et al., 2009 [45 days]). Thus, there is a critical need to better characterize longer-term SAS trajectories using well-controlled RCTs that include smokers randomized to delay their quitting. Importantly, it is not clear whether longer durations of abstinence result in full mood recovery and possibly even enhanced mood in most smokers.

The Current Study.

The current study aims to more clearly characterize abstinence trajectories by controlling for testing effects over time and biased dropout. Specifically, dependent smokers were randomized to an immediate Quit Group (while on placebo, nicotine replacement therapy [NRT], or bupropion SR [BUP]) or to a Delayed Quit Smoke (DQS) Group. We then assessed affect-related SAS across a 67-day (9.5-week) period while using monetary incentives to minimize dropout due to relapse. We predicted that most measures of NA would not resolve within the 67-day abstinence period.

Further, based on evidence that NRT and BUP may reduce NA SAS (Piasecki et al., 2003a), we hypothesized that NRT and BUP would significantly reduce NA and increase PA relative to placebo controls during the 45 days on active treatment. While NRT has been found to reduce NA-related abstinence symptoms 7–8 weeks after quitting relative to placebo (Shiffman, Khayrallah, & Nowak, 2000), the effects of BUP on these symptoms is not clear given their inconsistent effects in randomized, placebo-controlled trials (Lerman, Roth, Kaufmann, Audrain, Hawk, Liu, et al., 2002; Jorenby, Leischow, Nides, Rennard, Johnston, Hughes, et al., 1999; McCarthy, Piasecki, Lawrence, Jorenby, Shiffman & Baker, 2008; Hurt, Sachs, Glover, Offord, Johnston, Dale, et al., 1997; Piper, Federman, McCarthy, Bolt, Smith, Fiore, et al., 2008) and experimental studies (Shiffman, Johnston, Khayrallah, Elash, Gwaltney, Paty, et al., 2000; Teneggi, Tiffany, Squassante, Milleri, Ziviani, & Bye, 2005). In contrast, there is some evidence that both NRT and BUP may increase PA affect relative to placebo (McCarthy, Piasecki, Lawrence, Jorenby, Fiore, & Baker, 2008). Thus, we also predicted that this would be the case in the present investigation.

Finally, we tested the hypothesis that individuals high in trait measures of depressive affect, anger, and anxiety would experience greater treatment benefits from BUP and NRT on NA SAS than those low in these traits. Additionally, we predicted that individuals high in a specific affective trait (anxiety, anger, or depression) would exhibit relatively greater increases in the corresponding negative affective state during abstinence—that is, trait-dependent SAS severity. This latter finding may be due to the unmasking of underlying psychopathology following abstinence from smoking (Audrain-McGovern et al., 2009; Gilbert, 1995; Khantzian, 1997).

Method

Participants

Individuals who reported smoking 7 or more cigarettes/day (cpd) for at least one year and who expressed a desire to permanently quit smoking were recruited from 9/01/06 to 9/01/11 by newspaper ads and community and university postings. As few as 7 cpd were included because of the growing number of light smokers in the general smoking population (Shiffman, 2009). Exclusionary criteria (in large part due to requirements of the parent study) included current use of psychoactive drugs (n = 158) or medications (n = 217) (other than nicotine, alcohol and caffeine), alcohol use > 28 alcoholic drinks per week (n = 7), age < 18 years, poor English ability, education < 12 years without high-school equivalency exam, atypical sleep cycles, pregnancy or currently breast feeding, uncorrected poor vision, history of seizure disorders, bulimia or anorexia, current diagnoses of a mood, bipolar, or psychotic disorder, hypertension, serious health problems (n=69), and acute stressors (n=31). Demographic characteristics of enrollees are presented in Table 1. Participation required medical clearance by an in-house physician. All procedures were approved by the Springfield Committee for Research Involving Human Subjects Institutional Review Board (IRB) for Southern Illinois University School of Medicine, Study Title: “NRT and Bupropion Mechanisms in Smokers”, 12–639, reference number: 011265. Study follow-up was completed in March of 2014.

Table 1.

Prequit baseline demographics by group.

	Delayed Quit		Placebo		Bupropion SR		NRT
	Comp. 18	Drops 02	Abstain 25	Relapse 10	Abstain 24	Relapse 10	Abstain 28	Relapse 10
Age	27.2 (10.9)	23.0 (4.2)	32.8 (11.5)	28.2 (6.7)	32.8 (12.4)	28.2 (10.1)	31.5 (11.3)	28.2 (10.1)
Gender	10M, 8F	2M, 0F	13M, 12F	4M, 6F	14M, 10F	6M, 4F	16M, 12F	4M, 6F
Ethnicity	AA-1 AI-0 Asn-1 Cau-14 Other-2	AA-0 AI-0 Asn-0 Cau-2 Other-0	AA-0 AI-0 Asn-1 Cau-24 Other-0	AA-0 AI-0 Asn-0 Cau-10 Other-0	AA-2 AI-0 Asn-0 Cau-19 Other-3	AA-1 AI-1 As-0 Cau-8 Other-0	AA-0 AI-0 Asn-0 Cau-27 Other-1	AA-1 AI-0 Asn-0 Cau-9 Other-0
Hispanic	1	0	2	0	1	0	1	1
Education	3.78 (1.35)	3.50 (0.71)	4.32 (1.46)	3.50 (0.85)	4.00 (1.77)	3.40 (1.75)	3.82 (1.52)	3.40 (1.43)
Age first smoked daily*	18.28 (4.23)	15.00 (4.24)	17.84 (2.59)	17.40 (2.17)	17.75 (3.47)	16.20 (1.75)	17.68 (3.33)	15.3 (2.75)
FTND	3.78 (2.34)	4.50 (0.71)	3.44 (1.94)	5.50 (1.72)	4.29 (1.14)	4.50 (2.42)	4.25 (2.19)	4.10 (1.79)
CPD	18.06 (6.38)	16.00 (4.25)	16.56 (6.04)	20.70 (1.90)	19.25 (9.23)	15.5 (4.97)	17.79 (7.18)	17.50 (5.48)
Cotinine	5.86 (.41)	6.0 (0.0)	5.92 (.28)	6.0 (.0)	6.0 (.0)	6.0 (.0)	5.95 (.21)	6.0 (.0)
Hx of Depress.	1	1	6	4	7	0	8	3
Hx of Anxiety Disorder	5	1	5	2	5	1	10	3
Hx of PTSD	5	0	4	3	3	0	8	2

Note. Means, standard deviations (in parentheses), and Counts: Education: 1= HS equivalency exam, 2=Completed HS, 3= Some College, 4=Associates degree; 5=4-yr degree; 6=Some Grad school; 7=Completed Grad School. Cotinine: 4= 200–500 ng/ml, 5 = 500–1000 ng/ml and 6 = over 1000 ng/ml urine concentrations. Ethnicity (self-reported): AA = African American, AI= American Indian, Asn=Asian, Cau = Caucasian, Oth = Other or mixed.

^*p <.05 completers = older age of daily smoking onset than relapsers. Hx = history of SCID-assessed major depressive disorder, anxiety disorder, or post-traumatic stress disorder.

Materials and Procedure

Smoking status and nicotine dependence.

Smoking was monitored using salivary cotinine, breath-CO concentrations, and self-report during each pre-quit and post-quit session. Cotinine concentration was assessed using Jant Accutest® NicAlert™ salivary test strips that provide a semi-quantitative estimate of cotinine concentrations across seven steps from 1–10 ng/ml to over 1000 ng/ml (Jant Pharmaceutical Corporation, 2005). The Fagerström Test of Nicotine Dependence (FTND, Heatherton, Kozlowski, Frecker, & Fagerström, 1991) assessed dependence at baseline.

Mood, craving, and personality questionnaires.

The mood state measures included the Profile of Mood States (POMS, McNair, Lorr, & Droppelman, 1971) Anger-Irritability, Sadness-Depression, and Tension-Anxiety subscales; the Positive and Negative Affect Scale (PANAS, Watson, Clark, & Tellegen, 1988); the Beck Depression Inventory (Beck & Steer, 1987); the Center for Epidemiology Study of Depression (CES-D, Radloff, 1977); the Shiffman-Jarvik (SJ) Craving and Irritability subscales (Shiffman & Jarvik, 1976); the state forms of the State-Trait Personality Inventory (STPI) Anger, Anxiety, and Depression subscales (Spielberger, 1995). Craving was measured using the two factors of the Questionnaire of Smoking Urges (QSU, Tiffany & Drobes, 1991) and the SJ Craving score. The trait measures obtained during the pre-quit baseline included the NEO-PI-R (Costa & McCrae, 1992) Anger, Anxiety, Depression, and overall Neuroticism scores.

Study Phases and Procedures.

Study Phase 1 was a two-week baseline during which participants smoked at their usual rate and attended biweekly in-lab mood and smoke monitoring sessions. Phase 1 ended with random assignment to treatment groups. Phase 2 included afternoon in-lab monitoring sessions on the first day of abstinence (or corresponding day in the DQS Group), at day 3 of abstinence, then every 72 hours through day 24, and then weekly until the final two days of the study when symptoms were assessed on both abstinence days 66 and 67 to assess possible variations in SAS on the final day of abstinence. All monitoring sessions were conducted in-lab and included the assessment of SAS, breath CO, saliva samples for cotinine, and timeline follow-back assessments of tobacco and substance use.

An IRB-approved consent was signed by each participant. Participants in the Quit Groups received an abbreviated form of the American Lung Association smoking cessation program, while those in the DQS Group received this training after the completion of their 67-day assessment period. Among those randomized to one of the three quit groups, completion of all study requirements resulted in earning about $700 minus any penalties for smoking (The first cigarette penalty was $25, the second to the fourth cigarette penalties were $50 each for a maximum loss of $175). Lapse penalties were applied to nine individuals, each on one occasion: Placebo n =3; BUP n = 5; and NRT n = 1. Participants were excluded from the study without payment for smoking more than four cigarettes total across the abstinence period. Those in the DQS Group were paid about $800 for study completion, a greater amount than the quit groups to maintain motivation to quit after the 67-day delay. Both groups received task performance bonuses and $25 each for 3-, 6-, and 9-month follow-up sessions and $40 for the 12-month follow-up. After the 67^th day of reporting moods and providing biological samples confirming smoking status at the same intervals as the quit groups, individuals in the DQS Group were provided the same cessation training as the quit groups along with nicotine patches to aid their subsequent quit effort. The present report focuses solely on self-reported SAS, details on brain activity measures, behavioral tasks, and follow-up periods will be reported in a future publication. The 67-day abstinence period was chosen because it was three weeks longer than our previous longest study and allowed us to assess the effects of treatment across three weeks of post-treat, something that allowed us to assess potential rebound effects resulting from going off treatment.

Randomization and Treatment.

As seen in Figure 1, a total 127 participants (aged 18–62) entered the experimental sessions of the parent study, six of whom dropped out after completing the pretreatment sessions. Participants completed a pre-quit baseline phase that included mood, breath carbon monoxide (CO) concentration, and salivary cotinine on monitoring days. They then were randomly assigned by an independent pharmacy using an urn technique without replacement approach via a 28:28:28:16 ratio to one of four groups: NRT, BUP, Placebo, or a Delayed Quit Group (DQG) that continued to smoke through the end of the 67-day abstinence period of the other three groups. Researchers and participants in the quit groups were blind to pill and patch type. Biochemically verified abstinence across the 67-day quit period with no more than four lapse cigarettes is depicted in Figure 1. Abstinence rates were roughly comparable for the three quit groups: 1) Placebo n = 25 (13 m, 12 f) of 35 (17 m, 18 f) [71%]; 2) BUP n = 24 (14 m, 10 f) of 34 (20 m, 14 f) [71%], and 3) NRT n = 28 (16 m, 12 f) of 38 (20 m, 18 f) [74%]. Of the 20 DQS individuals (12 m, 8 f), 18 (10 m, 8 f) [90%] continued to smoke and completed all aspects of the study, including quitting after serving for 67 days as smoking controls. Participant demographics are presented in Table 1. All participants were native English speakers except for 4 who spoke fluent English.

Figure 1.

Study Flow Chart

The treatments regimens were as follows: 1) BUP, beginning two weeks prior to quitting took one encapsulated 150-mg pill per day for the first 3 days and two 150-mg pills for 56 days, followed by a 3-day stepdown period at one 150-mg pill per day; 2) NRT (NicoDerm CQ® patch) beginning 1st day cessation, 21 mg/24 days, 14 mg/14 days, 7 mg/7 days; and 3) Placebo (patch and capsule), placebo capsule and patch on same days as the active BUP and NRT groups. Self-reported adherence to the medication regimen was 100% of prescribed doses in both the placebo and active medication groups. BUP adherence was confirmed by blood assays for BUP major metabolites (morpholinol, bupropion-erytho-alcohol and bupropion-threo-alcohol). NRT adherence was confirmed by salivary samples for cotinine assays taken during each screening, orientation, and monitoring session.

Data Analysis

Verification of abstinence and smoking.

Abstinence was confirmed by CO concentrations of < 6 ppm (a range lower than the standard in the field of < 10 ppm, but in our experience more accurate than < 10 ppm as assessed by plasma cotinine concentration in similar studies— (Gilbert et al., 1998; Gilbert et al., 2002), self-report, and < 101 ng/ml urine cotinine concentration (except in the NRT group). Prior to quitting, groups had CO concentrations of about 18–20 ppm. After quitting the CO concentration of the three quit groups were maintained at about 2 ppm while the DQ Group (DQG) continued at baseline values (Figure 2). Of the 20 assigned to the DQG, 18 maintained study requirements through Day 67.

Fig 2.

Expired breath carbon monoxide concentration in parts per million prior and subsequent to quitting in the three quit groups relative to the delayed quit group (smokers) that continued to smoke during the depicted times.

Change from baseline SAS z-score computations.

As in our previous work (Gilbert et al., 2009), composite z-score measures of change from baseline for each of the three NA SAS (anger-irritability, depression, anxiety), and of craving were computed as follows. Z-scores change scores were generated for each of the four pre-quit baseline sessions and the 14 post-quit monitoring sessions for each participant by subtracting the individual’s score on a given day on a given measure from that individual’s mean of the final two baseline scores (b3 + b4) for that measure divided by the mean SD (across all 4 groups) of the 3^rd and 4^th baseline sessions. For example, an individual’s (i) $\text{PANAS-NA z-score for post-quit day}{67}_{\text{i}}=$

$${\text{Day67 PANAS-NA}}_{\text{i}}-{\text{Mean (b3PANAS-NA}}_{\text{i}}+{\text{b4PANAS-NA}}_{\text{i}})/{\text{Mean (b3PANAS-NA SD}}_{\text{1–95}}+{\text{b4PANAS-NA SD}}_{\text{1–95}}\text{),}$$

where the SDs are for the entire group of 95 study completers, but the difference scores are for the individual participant (i). PANAS-NA and PANAS-PA scores were not included in composites scores because composite scores were designed to focus on specific affects, as opposed to generalized negative or positive affect.

The composite score measures for an individual for a given post-quit time were simply the mean of that individual’s z-score components for that day for the composite. For example, an individual’s abstinence day1 Depression composite =

$$\begin{array}{c}{\text{mean (Day1 BDI depression z-score}}_{\text{i}}+\text{Day}1{\text{CESD depression z-score}}_{\text{i}}+\\ {\text{Day 1 STPI depression z-score}}_{\text{i}}+{\text{POMS depression z-score}}_{\text{i}}\text{)}.\end{array}$$

The use of z-score changes from pre-quit baseline provides equal weight to each of the component SAS questionnaires and allows one to compare the SAS for different composite scores on a common metric. The use of these z-score provides a normalized score for the individual and group means such that a composite score of 1.0 would correspond to an increase in that score relative to the baseline value by an equivalent of 1.0 standard deviation of the group baseline value.

The Anger composite was formed from the mean of POMS anger, STPI anger, and SJ irritability z-scores. The Depression composite was the mean of BDI depression, CESD depression, STPI depression, and POMS depression z-scores. The anxiety composite was the mean of the POMS tension–anxiety and STPI anxiety z-scores. The Craving (urge to smoke) composite was the mean of QSU Factor1, QSU Factor2, and SJ craving z-scores.

Analytic plan.

Based on our goal of assessing SAS duration, our first planned analyses assessed differences between DQG controls and the three treatment groups across 67 days of abstinence. Our second planned analysis analyses used subjects’ own baseline scores to assess whether the apparent duration of withdrawal differ if quitters’ own baseline scores were used as the benchmark rather than the delayed quitter group. These comparisons were made by identifying the first post-quit assessment day at which the SAS value was no longer statistically significantly (two-tailed t test, p < .05) greater than the mean baseline of the final two baseline sessions without two subsequent days achieving significance. Our third planned analysis compared NRT, BUP, and placebo treatment group trajectories across the first 24 days, when both NRT and BUP were at full strength. These 24-day analyses were conducted using only the three quit groups to emulate typical treatment efficacy studies and to provide greater power to differentiate the two active treatments with each other and the placebo group. These 24-day, quit-groups-only analyses failed for the most part to demonstrate any significant benefit of either of the two active treatments (BUP or NRT) relative to the placebo group and thus their discussion is limited to the Supplement Material. The one exception to this lack of beneficial treatment effect, was that NRT attenuated SAS on the third day of abstinence relative to BUP and Placebo in analyses reported (Figures 4, 5, 7 & 8). Our final planned analysis was to assess the ability of individual differences in NEO-PI-assessed NA-related personality traits to predict SAS severity.

Fig 4.

Multivariate

Figure 5.

PANA NA

Fig 7.

Anger

Fig 8.

Anxiety

Z-score changes in smoking abstinence symptoms (SAS) were analyzed using SPSS v.24 (IBM Corp, Armonk, NY) mixed-design multivariate MANCOVAs and ANCOVAS with treatment Group as the between-subjects variable and Time as the within-subjects variable. Initial MANCOVAs included Gender as a between-subjects variable and FTND score as a covariate. However, neither Gender nor FTND scores interacted with Treatment or Treatment × Time interactions and, thus, reported analyses excluded Gender and FTND. The 67-day assessment MANCOVAs were conducted using the above-noted composite Depression, composite Anger, and composite Anxiety composites as dependent variables (DVs) with NEO-PI Neuroticism, Depression, Anger, and Anxiety as covariates in separate analyses. Partial correlations controlling for treatment group were also used to assess the associations of nicotine dependence and NEO-PI traits with SAS severity at days 3, 24, and 67 of abstinence. Separate ANCOVAs also run for each of the three NA SAS composites using the same covariates as the MANCOVAs. As noted in the corresponding tables notes, we used Benjamini and Hochberg False Discovery Rate (BH FDR) significance values to correct for the multiple MANCOVA and ANCOVA analyses. Values achieving BH FDR significance are in bolded in the tables.

To ensure that findings were not driven by outliers, the two individuals with the largest z - scores and the individuals with the smallest SAS z-scores on the 24^th day of abstinence (corresponding day for DQG) within each of the four treatment groups were eliminated from the statistical analyses. These analyses resulted in the same main effects and simple interactions as the full sample, so the findings of the full sample were minimally influenced by outliers. The 24^th day of abstinence was chosen because it was the last day during which individuals in the NRT group were on full strength (21mg) patches and the 7^th of the 14 post-quit assessment sessions.

Results

Sample Characteristics

Table 1 provides demographic details for the four treatment groups.

Abstainer-Relapser differences.

Compared to the participants who maintained abstinence (or had fewer than 4 cigarettes) for the course of the study (subsequently referred to as Abstainers), the combined group of relapsers (4+ cigarettes during the abstinence period) and drop-outs (subsequently assumed to be and referred to as Relapsers) did not differ as a function of Group, F(2,100) =0.060, p = .946, and there was no Group × Gender interaction, F(2,100) =0.166, p = .847, or main effect of Gender, F(1,100) = 0.776, p = .380. Relapsers differed from Abstainers in terms of a younger age of onset of daily smoking (16.03 vs. 17.76), F(1,101) = 5.271, p = .024, and nonsignificant tendencies for lower education (3.43 vs. 4.04), F(1,105) = 3.708, p = .057, and higher scores on the FTND (4.00 vs. 3.97), F(1,105) = 2.445, p = .121. Relapsers did not differ from Abstainers in terms of prequit baseline CO concentration (27.97 ppm vs. 28.62 ppm), F(1,105) = 0.034, p = .855, mean pre-quit baseline smoking (17.72 vs. 18.21 cigarettes/day), F(1,105) = 0.110, p = .741, or NEO-PI Neuroticism, Anger, Anxiety, or Depression (all ps > .10). Baseline SAS and symptoms on the third days of abstinence (when all but 9 of the quitters remained in the sample) did not predict abstinence vs. relapse during the 67 days of reinforced abstinence.

Abstinence Symptom Trajectories

Delayed-Quit Group vs. Treatment Group SAS across 67 Days.

Relative to DQG levels, quitting in all groups (NRT, BUP, and Placebo) was associated with significant Benjamini and Hochberg False Discovery Rate-corrected increases in SAS severity in each of the three NA domains (Table 2) that lasted across the 67 assessed days (Figs. 4–8). Multiple comparisons at Day 67 showed the DQG to have significantly lower change scores from baseline for all NAs at abstinence day 67 (all ps <.05), as well as at earlier points in time. PANAS PA was not influenced by quitting, as indicated by the lack of a Treatment or Treatment × Time interaction effects (Table 2). However, there was also a significant interaction of Craving with Time such that it was elevated in the three quit groups relative to the DQG and baseline levels through day 3 and below DQG and baseline levels by day 31 (Fig. 3, Table 2).

Table 2.

Effects of Treatment, Time, and Treatment × Time on Abstinence Symptoms (including Delayed-Quit Control Group) Across 67 Days.

Effect	Abstinence Symptom	F(df)	P	pη2	ePo
Treatment	Multi-Var Ang+Anx+Dep	(3,91) = 7.292	<.001	.194	.980
	Anger	(3,91) = 6.740	<.001	.182	.971
	Anxiety	(3,91) = 5.120	.003	.144	.911
	Depression	(3,91) = 5.939	.001	.164	.949
	PANAS Negative Affect	(3,91) = 7.832	<.001	.205	.987
	PANAS Positive Affect	(3,91) = 1.172	.325	.037	.306
	Craving	(3,91) = 1.118	.346	.036	.293
Time	Multi-Var Ang+Anx+Dep	(13,79) = 3.411	<.001	.360	.995
	Anger	(13,79) = 2.640	.004	.303	.973
	Anxiety	(13,79) = 3.238	.001	.384	.993
	Depression	(13,79) = 2.621	.004	.301	.972
	PANAS Negative Affect	(13,79) = 2.750	.003	.312	.979
	PANAS Positive Affect	(13,79) = 2.942	.002	.326	.986
	Craving	(13,79) = 8.228	<.001	.575	1.0
Treatment x Time	Multi-Var Ang+Anx+Dep	(39,243) = 1.411	.015	.435	.978
	Anger	(39,243) = 1.371	.081	.180	.986
	Anxiety	(39,243) = 1.162	.246	.157	.960
	Depression	(39,243) = 1.489	.039	.193	.993
	PANAS Negative Affect	(39,243) = 0.869	.693	.122	.858
	PANAS Positive Affect	(39,243) = 0.012	.457	.14	.920
	Craving	(39,243) = 1.783	.005	.223	.999

Note. Multi-Var Ang+Anx+Dep = Within-subject multivariate combination of Anger, Anxiety, and Depression composite scores. Anger, Anxiety, Depression, and Craving = z-score composites of corresponding scales (e.g., Anger = Mean of the Z-score for POMS anger, STPI anger, and SJ irritability). Valence has 3 levels, corresponding to different z-score composite scores pre-post quit change values in Anger vs. Anxiety vs. Depression. Benjamini and Hochberg False Discovery Rate (BH FDR) p < .05 = p < .0190; BH FDR p < .01 = p < .0038.

Fig 3.

Craving

Treatment main effects were non-significant for analyses limited to the three quit groups and the first 24 days of treatment when the NRT dose was 21 mg (See Supplemental material for MANOVA tables and discussion). Though there was no main effect of Treatment for Craving, there was highly significant effect of Time, such that Craving decreased across the 24 days of abstinence, Time, F(5,70) = 9.163, p = .001, _pη² =.396, est. power = 1.0.

Follow-up analyses of the significant effect of Treatment reflected lower NA-SAS scores in the DQG relative to each of the three immediate-quit groups (Figures 4–8). NA-SAS in the DQG remained significantly lower than those in any of the quit groups throughout the 67-day abstinence period. Additionally, there was a significant downward linear trend in NA SAS in the DQG as assessed by the Time factor in repeated measures MANOVAs across the four pre-quit points in time and subsequent 14 post-quit time points. The downward linear slopes were significant for the overall MANOVA across the Anxiety, Anger, and Depression SAS, F(1,17) = 8.202, p = .011; as well as for each of the composite scores for Anger, F(1,17) = 7.970, p = .002; and Depression, F(1,17) = 9.346, p = .007, but only achieved one-tailed significance for Anxiety, F(1,17) = 3.804, p = .048 (one-tailed), due to no further downward drift in Anxiety after the day corresponding to the ninth day of abstinence in the quit groups. However, there was a significant two-tailed downward drift for anxiety across the first nine days of abstinence, F(1,17) = 9.363, p = .003.

In summary, NA SAS in the three quit groups did not resolve to DQG levels within 67 days of abstinence, but craving SAS did, and the NA SAS in the DQG decreased linearly across time and positive affect was not influenced by abstinence. The importance of using the DQG contrasts for NA SAS is seen by contrasting these findings with the traditional use of using one’s own prequit baseline as the criterion for recovery. We turn to this point in the next paragraph.

Estimated NA-Related SAS Resolution Duration in Quitters Using Participant’s Own Baseline vs. Delayed-Quit Controls.

In contrast to the above-noted failure of NA-related SAS (with the exception of Anxiety) to recover to DQG levels within 67 days, when compared to their own pre-quit baseline, each of the four NA-SAS indices recovered to baseline levels within 3–31days of abstinence (Table 3) if the criterion of not statistically significantly greater than zero (baseline) were used as the criterion of symptom resolution as detailed in the analytic plan above. In contrast, comparisons of SAS for each of the three quit groups with the DQG control group showed that SAS were significantly greater (two-sided t tests with p < .05) in each of the three groups for the full 67-day assessment period with one exception—Anxiety in the NRT group was no longer significantly greater than the DQG beginning on the 31^st day of abstinence. Craving also took slightly longer to resolve when using the DQG comparisons, relative to the own baseline resolution (Table 3). The most extreme difference in interpretations of duration of symptom resolution can be seen in Figure 6, where, in the NRT group recovery to baseline level of depressive symptoms occurred by day six or eight after quitting, in contrast, with the use of the delayed quit controls there are still highly significant elevation of depressive symptoms through the end of 67 days of abstinence. It should also be noted that, except for depression in the NRT group and PANAS NA in the placebo group, none of the SAS in the quit groups recovered to baseline absolute level across the 67 days of abstinence (Figures 4–8).

Table 3.

Smoking Abstinence Symptom Recovery Times Indexed from Personal Baseline with and without No-Quit Control Comparisons.

Smoking Abstinence Symptom	1st Day SAS Mean No Longer Significantly Greater than Pre-Quit Baseline			1st Day SAS Mean No Longer Significantly Greater than DQG Controls
	Placebo	NRT	Bupropion	Placebo	NRT	Bupropion
Anger	31	9	31	67	67	67
Anxiety	12	6	31	67	31	67
Depression	6	3	6	67	67	67
PANAS-NA	6	3	6	67	67	67
Craving	6	3	3	6	18	6

Note. The baseline comparisons indicate the 1^st post-quit day on which the SAS was no longer statistically significantly greater than 0 without return to significance at two or more later points in time; the mean of the last two pre-quit baseline levels as assessed by a two-sided t tests with p < .05. DQG control comparisons indicate the 1st Day SAS post-quit means for the treatment group are no longer significantly greater than the DQG controls without return to significance at two or more later points in time; using two-sided t tests with p < .05.

Fig 6.

Depression

Statistical Comparisons of the Relative Effects of Specific NA-SAS Valence.

A MANOVA revealed a main effect of valence on average across the 67 days of abstinence such that Anger SAS generally increased to a greater extent than Anxiety SAS, which in turn were greater than Depressive SAS (Figures 6–8), Valence main effect, F(2,73) = 4.162, p = .019; Anger vs. Depression, p = .005; Anger vs. Anxiety, p = .114; Anxiety vs. Depression, p = .147. There was also a significant Valence × Time interaction, F(2,1976) = 5.393, p < .001, such that significant differences between valences existed only during the first 38 days of abstinence, and were strongest at abstinence day 6 when each group was significantly different from the other, Anger >Depression, p < .001, Anger >Anxiety, p = .039, Anxiety >Depression, p = .005. Given a priori planned analyses and below-noted Treatment × Valence interactions, all further analyses were based on separate SAS scores for Anger, Anxiety, and Depression, as well as for PANAS NA and PANAS PA.

Negative Affect-Related Traits and Dependence Predictors of SAS.

Partial correlations controlling for treatment group and excluding DQG individuals demonstrated that individuals scoring higher in NA-related personality traits (especially Neuroticism) were significantly more likely to experience larger increases in both NA SAS and in craving, especially at latter two of the three a priori periods of assessment, abstinence day 24-day when on maximal medication and at day 67, when off medication (Table 4). However, there was not a high degree of specificity in associations between trait and state measures of NA. For example, trait depression correlated more highly with changes in SAS Anger than it did with SAS Depression (Table 4). MANCOVAs using NA personality traits as covariates and using Treatment × Time assessments of SAS also indicated that NA traits moderated the effects of treatment on both NA SAS and craving such that greater NA traits tended to correlate positively with larger abstinence-associated NA SAS and craving in a manner that varied as a function of treatment (Supplemental Tables 2S and 3S).

Table 4.

Partial correlations of trait variables with changes in smoking abstinence symptoms from baseline in abstainers.

Trait with SAS		Partial Correlations Controlling for Treatment Group
Trait	SAS
		D3	D24	D67
NEO-Anger	Anger	−.154	.068	.187
	Anxiety	−.119	.000	.140
	Depression	−.108	.102	.240*
	PANAS-NA	−.024	.133	.241*
	PANAS-PA	.185	.022	.053
	Craving	.125	.353**	.410**
NEO-Anxiety	Anger	.006	.179	.244*
	Anxiety	.029	.161	.241*
	Depression	.066	.197	.353**
	PANAS-NA	.070	.331**	.257*
	PANAS-PA	.063	.026	−.040
	Craving	.284*	.391**	.356**
NEO-Depression	Anger	.133	.228	.316**
	Anxiety	.096	.128	.212
	Depression	.018	.159	.263*
	PANAS-NA	.133	.193	.237*
	PANAS-PA	.172	.108	.116
	Craving	.283*	.332**	.370**
NEO-Neuroticism	Anger	.065	.245*	.312**
	Anxiety	.051	.185	.290*
	Depression	.033	.189	.315**
	PANAS-NA	.080	.264*	.269*
	PANAS-PA	.149	.052	.038
	Craving	.302**	.432**	.438**
FTND	Anger	−.060	.006	−.026
	Anxiety	−.018	−.019	−.092
	Depression	.064	.003	.002
	PANAS-NA	−.016	.015	.038
	PANAS-PA	−.096	.010	−.112
	Craving	−.112	−.038	−.122

Note. AS = abstinence symptom z-score change from baseline for a given dependent variable (DV). Partial correlations are across the three treatment groups (N = 77) controlling for treatment type.

^*= p < .05;

^**= p < .01.

On the third day of abstinence change in craving was the only variable to correlate with NA-related traits; while on day 67 (when off medications) 13 of the 16 NA trait correlations with NA-related SASs were significant and Craving was significantly correlated with all NA traits after 67 days of abstinence. FTND scores did not correlate with any SASs at any point in time.

Baseline Symptom Score Associations with Treatment and Traits.

There were no pre-quit baseline differences between treatment groups for any SAS scores (all ps > .05). Baseline PANAS-NA and the three composite NA scores were positively correlated with NEO-PI Neuroticism, Depression, Anger, and Anxiety, r’s .35-.60, p’s< .001. FTND nicotine dependence score did not correlate with the baseline NA mood scores but did correlate with breath CO concentration, r = .34, p < .001. More detailed presentations of correlations of baseline scores with trait indices are presented in Table 3S in the Supplemental Materials.

Discussion

The current findings provide a step toward better characterization of the duration of SAS and their moderation by NA-related traits. NA-related SAS did not resolve to delayed-quit control group levels across 67 days and higher scores on NA-related trait measures tended to predict greater NA symptom severity at 24 and especially at 67 days of abstinence (when off medications). These two findings are consistent with the hypothesis that NA-related SAS may substantially reflect the unmasking of NA-related traits that were otherwise minimized by smoking. The unmasking hypothesis is closely related to the self-medication hypothesis that suggests that many smokers use tobacco/nicotine to self-medicate distressing affective traits and states (Audrain-McGovern et al., 2009; Gilbert, 1995; Eysenck, 1980). Others have provided findings that support the unmasking/self-medication hypothesis of NA-related SAS (Gilbert, 1997; Khantzian, 1997), but the present findings offer particularly robust evidence for this hypothesis given that participants were randomized to smoking versus abstinence over a prolonged period with substantially lower dropout and relapse rates than previous studies. The inclusion of a delayed-quit group also helps account for potential testing effects, as well as the possibility that SAS escalate during the few days before initiation of quitting (Gilbert & Pergadia, 2017).

The present findings provide strong support for the view that randomized no-quit control groups that continue to smoke yet are repeatedly assessed at the same time as quitters, are essential for the accurate characterization of abstinence symptom trajectories because the use of pre-quit baseline levels of SAS are biased due to the downward drift in SAS from repeated testing subsequent to quitting. The statistically significant downward drift was seen in our DQG NA-SAS scores replicates our previous studies using DQGs (Gilbert et al., 1998; Gilbert et al., 2002) and is consistent with downward drifts with NA scores observed in non-smoking samples (Ahava et al.,1998; Arrindell, 2001; Atkeson et al., 1982; Choquette & Hesselbrock, 1987; Hatzenbuehler et al., 1983; Sharpe & Gilbert, 1998).

NA-related traits also predicted craving at 3, 24, and 67 days of abstinence, while FTND dependence did not (with one exception) correlate with craving nor with NA-related SAS. Thus, the associations of NA-related trait neuroticism, anger, anxiety, and depression to SAS severity cannot be attributed to a correlation of dependence with SAS severity. The failure of dependence to correlate significantly with craving and NA-related SAS severity may reflect our exclusionary criterion of 7+ cpd and/or the relatively modest number of very heavy smoking participants in our sample. Contrary to the Trait Adaptive Response portion of the Situation by Trait Adaptive Response (STAR) model (Gilbert, 1995; Gilbert, 1997), there was not a high degree of specificity in associations between trait and state measures of NA. For example, trait depression correlated more highly with changes in SAS Anger than it did with SAS Depression (Table 4).

We believe that our findings provide the least biased estimates of long-term SAS trajectories to date. The relapse and study dropout prevalence in the present study was substantially lower than in studies in this area, even well-designed studies (e.g., Piasecki et al., 2003a), presumably because of the large financial incentives to maintain abstinence. If everyone that experienced NA-related symptoms had relapsed there would be no observed increase in NA subsequent to quitting smoking, which may partially account for why clinical studies with their higher dropout rates report different SAS trajectories than those found in the current study. Similarly, if only those who experienced the most severe stress and NA relapsed, the estimated severity of symptoms would be attenuated. Thus, the high degree of sustained abstinence in the present sample is a strength of the present investigation, though the observed severity of symptoms was still likely significantly underestimated as 26–29% of the quitters failed to maintain abstinence and/or dropped out and relapsers likely had more severe SAS and/or life stressors than the abstainers (Shiffman & Paty, 2006). Our findings that NA-related SAS take longer than 67 days to recover is consistent with literature demonstrating a prolonged time course of normalization of nicotinic cholinergic receptors (Cosgrove Batis, Bois, Maciejewski, Esterlis, Kloczynski, et al. (2009).

The failure of BUP to reduce NA-SAS, like the inconsistent effects of NRT to do so, may reflect the limited power of the study to detect these effects. However, the effects of BUP on NA-SAS in larger studies have been inconsistent, frequently showing no beneficial effects (Lerman et al., 2002; Jorenby et al., 1999; McCarthy et al., 2008; Hurt et al., 1997; Piper et al., 2008). In contrast, consistent with a trend in our findings, larger studies have consistently found that NRT (Gilbert et al., 2009; Shiffman et al., 2000) reduces NA-SAS. Our sample size was limited because the primary focus of the larger study on which the current findings are based was on SAS duration and the effects of treatment on electrocortical (EEG) measures during abstinence which tend to be more sensitive to abstinence than self-report indices (Gilbert et al., 1999; Gilbert et al., 2004). Treatment effects on EEG during smoking abstinence will be presented in a forthcoming publication, though BUP was found to reduce a parietal hemispheric EEG asymmetry index of depression during the pre-quit baseline of the current study but had no effect on baseline self-reported Beck Depression score (Zhu, Coppens, Rabinovich, & Gilbert, 2017).

Our finding of a large decrease in craving across abstinence to below prequit and DQG levels within a month time or so stands in sharp contrast to the failure of NA symptoms to recover to DQG levels but is consistent with virtually all of the preceding studies in this area (Shiffman et al., 2004), including our previous work comparing DQG controls (Gilbert et al., 1998; Gilbert et al., 2002). This discrepancy in the time trajectories between NA and craving is theoretically interesting and may have clinical implications. We found positive correlations of craving with NA-related traits (especially neuroticism) both while on medication and at the conclusion of the study when (67^th day of abstinence) when off of medication, but craving did not correlate significantly with the FTND index of nicotine dependence at any time. The lack of a correlation between craving and FTND may reflect the fact that our sample did not include individuals who smoked fewer than 7 cigarettes per day. One could hypothesize that craving is related to nicotine dependence in a somewhat stepwise manner such that those with relatively little daily smoking experience relatively little craving when abstaining, while those consuming nicotine above some critical value tend to experience craving that is more related to NA traits and associated elevated levels of chronic and acute NA, and associated self-medication than to nicotine dependence (Gilbert, 1997). Smokers report increased urges to smoke in response to increased negative affect (Shiyko, Naab, Shiffman, & Li, 2014).

Clinical Implications

The fact that NA symptoms did not resolve within 67 days of quitting has important clinical implications, while craving decreases to below prequit levels within two weeks is theoretically interesting and clinically important information concerning how to frame accurate messages about the typical course of withdrawal for smokers. The results suggest that either: 1) abstinence-related increases in NA take longer than 67 days to recover; and/or 2) abstinence unmasks vulnerabilities to NA, resulting in elevations of NA that do not resolve in at least some individuals. In either case, the findings suggest that successful smoking cessation may require treatment that lasts more than two months or potentially much longer, and that addresses the use of cigarettes to self-medicate affective symptoms. This is consistent with previous work finding that NRT continued to decrease withdrawal symptoms between weeks 6 and 10 (Shiffman, Khayrallah, and Nowak, 2000) and that extended NRT beyond 8 weeks (to 26 weeks) prevented relapse (Schnoll, Patterson, Wileyto, Heitjan, Shields, Asch, & Lerman, 2010). These findings support Shiffman and colleagues’ suggestion that “smokers may be best served by treatment that lasts at least 10 weeks and should be discouraged from discontinuing treatment prior to this time.” (pg. 371).

At a minimum, the present findings suggest we should question whether to tell smokers who are considering quitting that they will inevitably feel better soon. While such information may motivate some individuals to try quitting, such a message may not have credibility with smokers who have previously experienced or will experience longer durations of SAS. If symptoms persist beyond expectations, the quitter may direct blame toward their basic traits or character (e.g., “I just don’t have what it takes to quit” or “My withdrawal symptoms are stronger and more long-lasting than they should be”) and reinforce the notion that smoking is necessary to manage unwanted affective symptoms. However, it will also be important to inform smokers that studies uniformly show that craving decreases dramatically across the first two months of abstinence to below pre-quit levels.

Study Limitations and Future Directions

Participant selection factors limit the generalizability of study findings. The study excluded individuals with current major psychiatric disorders and psychoactive drug use. The time-demanding nature of the study also resulted in self-selection bias, the sample was more educated than smokers in the general population, and 89.6% of participants were Caucasian. Additional studies with larger and more diverse samples are needed to replicate, extend, and better characterize SAS trajectories for longer durations. The modest sample size greatly limits any conclusions related to the effects of NRT and BUP on SAS severity. However, the tendency of NRT to reduce SAS during early (day-3 in in the present study) abstinence is consistent with other studies (e.g., Gilbert et al., 2009; Shiffman et al., 2000). Finally, the current analyses were somewhat censored (26–29%) by relapse and dropout, as relapsers tended (though not significantly) to be more dependent smokers, and the group that maintained abstinence had relatively moderate dependence.

Despite the above-noted limitations, the current study was highly controlled and thus provides strong evidence that NA-related SAS on average last for at least 67 days. The present findings are also consistent with our previous three well-controlled RCTs that demonstrated little if any tendency for symptoms resolution relative to the smoke control group across 31 and 45 days of abstinence (Gilbert et al., 1998; Gilbert et al., 2002; Gilbert et al., 2009). Taken together, these results suggest that controlled trials using randomized delayed-quit groups with incentivized abstinence are as close as one can get to a gold standard for scientific research on NA-SAS trajectories, and that returning to control-group levels, rather than to baseline symptoms levels is best practice when examining NA-SAS and mental health resolution. Yet, no control group is perfect and other factors related to being aware that one is relieved of the pressure to quit smoking could account for some of the downward drift in NA in the DQG. For example, one could suggest that participants in the DQG may have been motivated to show they can do as well as or better than those who were quitting or to meet some subtle conscious or nonconscious desire to meet social or experimental expectations. However, when positing such alternative possibilities, it is also important to note that the downward drift NA occurs in individuals not assigned to treatment or control groups and who are not changing their behavior, including undergraduates who simply take the same NA measures every week or two (Sharpe & Gilbert, 1998). It is also likely that the substantial monetary incentives for the maintenance of abstinence may have impacted positive or negative affect, though it is not clear how such an effect would differentially impact the DQG relative to the quit groups because each of the groups received similar compensation. Thus, we believe that the preponderance of evidence from this study, similar studies by our group, and the related literature suggest that the current findings have important implications for smoking cessation treatment and that DQG controls provide better NA SAS trajectory estimates that do pre-quit baselines, for example, encouraging extended treatment by clinicians and may also help inform the regulation of nicotine and tobacco products by the FDA. Finally, future research may show that the need for randomized no-quit controls may not be required if very extended pre-quit baselines using ecological momentary assessments or other means are used

Figure 9.

PANA PA

Public Health and Theoretical Significance:

The current findings showing that negative affect (NA)-related smoking abstinence symptoms did not return to control group levels while craving does, may have important implications for smoking cessation treatment, for example, encouraging extended treatment by clinicians.

Importantly, the results strongly suggest that controlled trials using randomized delayed-quit groups with incentivized abstinence contingencies are as close as one can reasonably expect to get to a gold standard for scientific research on NA SAS trajectories, and that returning to control-group levels, rather than to baseline symptoms levels is best practice when examining smoking abstinence-symptom resolution.