Drug and Alcohol Dependence,
Elsevier Scientific Publishers
Reinforcing
29 (1991) 47 - 62
47
Ireland Ltd.
breath carbon monoxide reductions obstructive pulmonary disease
in chronic
Thomas J. Crowleya, Marilyn J. MacDonald”, Gary 0. Zerbeb and Thomas L. Petty” ‘Addiction
Research and Treatment Service, Department
of Psychiatry,
Box C-268, bDepartment of Biometrics.
University of
Colorado. School of Medicine, 4200 East Ninth Avenue, Denver, Colorado 80262 and rPresbyterianlSt, Lukea Hospital, Center for Health Sciences Education, 1719 East 19th Avenue, Denver, Colorado 80218 (U.S.A.)
(Received March 19th, 1991)
Chronic Obstructive Pulmonary Disease (COPD) usually results from tobacco smoking. Smoking cessation slows COPD’s progression, but few have studied anti-smoking treatments in COPD. In S-month trials we paid lottery tickets during daily home visits to still-smoking COPD patients for reductions in breath carbon monoxide (CO), a measure of smoke intake. In our first protocol experimental patients received 0 - 3 tickets per day, depending upon the extent of CO reduction below pre-treatment baselines; yoked controls received the same number of tickets, but not contingent on CO. The protocol produced no change. In a second study patients were assigned a post-baseline quit-date, received nicotine gum, and were paid up to 5 tickets per night, but only for CO < 10 parts per million (ppm). CO fell sharply as the intervention began, but gradually rose again. A third protocol added special reinforcement schedules for those who did not quit or relapsed (up to 20 tickets per night for CO < 10 ppm). Daily CO concentrations fell from 27.1 parts per million (baseline mean) to 12.7 (intervention mean), but rapid increases followed the intervention. Few patients stopped smoking, but CO and cigarettes used per day significantly fell during Studies 2 and 3. Post-hoc analysis suggested only a small effect from gum. Key words: carbon monoxide; smoking; pulmonary disease
Introduction ‘All those men have their price’. The Memoirs of Walpole, 1798 Chronic obstructive pulmonary disease (COPD) is America’s fifth leading source of death and causes 13 percent of the nation’s hospitalizations (Feinleib et al., 1989). Tobacco smoking produces most cases of COPD, and these cases yield nearly 60 000 deaths annually (U.S. Surgeon General, 1984). Smoking cessation slows COPD’s progression (Fletcher et al., 1976; Boss? et al., 1980; Boss? et al., 1981; Correspondence
to: Dr. T.J. Crowley, Addiction Research and Treatment Service, Department of Psychiatry, Box C-268, University of Colorado School of Medicine, 4200 East Ninth Avenue, Denver, CO 80262, U.S.A.
0376~8716/91/$03.50 0 1991 Elsevier Scientific Publishers Printed and Published in Ireland
Hughes et al., 1982; Tashkin et al., 1984), so effective anti-smoking treatment should help prevent pulmonary impairment. Some COPD patients do stop smoking (Crowley et al., 1989); those who persist despite pulmonary impairment and physician advice have seemed unanti-smoking for candidates promising treatment (Hall et al., 1983; Dudley et al., 1977; Daughton et al., 1980; Pederson, 1982; Pederson et al., 1982). Cotinine, breath carbon monoxide (CO) and thiocyanate assess smoke intake and self-report veracity; ‘No firm conclusions can be drawn until patient misrepresentation of smoking status’ is considered (Pederson, 1982). But we find no controlled smoking-treatment studies which used these measures among COPD patients; two such studies did include COPD patients along Ireland Ltd.
48
with others (Hall et al., 1983; Subcommittee of the Research Committee of the British Thoracic Society, 1984). Cotinine offers the best sensitivity and specificity, but CO assays are almost as accurate, give immediate results, and are cheaper; thiocyanate is less valuable (Jarvis et al., 1987). For behavioral reinforcement CO’s shorter half-life, 6-7 h in COPD patients (Crowley et al., 1989), may be an advantage. Learning theory suggests that new behavior is best ‘shaped’ by reinforcement for initial small or brief steps; several hours of tobacco abstinence may earn initial reinforcement for reduced CO, while the same behavioral change may not influence the more persistent cotinine. Smokers generally smoke less when paid for breath CO reductions; CO measures are objective and convenient, responsive to reinforcement, and correlate with smoking self-reports (Stitzer, Bigelow, 1982; Stitzer, Bigelow, 1983; Stitzer, Bigelow, 1984; Stitzer, Bigelow, 1985; Rand et al., 1989). After a pilot study of such payments in COPD patients (Crowley et al., 1989), we began the present studies. Our questions were, first, could payments reduce or stop smoking in patients who smoked on despite advancing COPD, and second, could such behavioral procedures produce a practical treatment. We did not recruit with newspaper advertisements, and patients were not charged for participation (Hall et al., 1983), procedures which may not attract the less-motivated. Instead, we identified still-smoking COPD patients in pulmonary clinics, inviting all of them to be paid participants. Methods and Results General methods Recruitment, We reviewed
pulmonary clinic charts at the Denver VA Hospital, University Hospital, Denver General Hospital, a Kaiser outpatient clinic and a private pulmonary office, for diagnoses of COPD. Upon arrival these patients blew into a MiniCo monitor (Catalyst Research, Baltimore), which showed breath CO in parts per million (ppm) (Crowley et al., 1989). Ambient air CO concentrations were subtracted
from breath concentrations, providing a Corrected Carbon Monoxide (CCO) value. Patients with COPD and CCOs 2 15 were told that they might be candidates for a study aimed at developing treatment to reduce patients’ CO with lottery-ticket payments. Interested patients were asked about demography and current smoking. Subsequently, a physician described the project and obtained signed informed consent. Admission criteria. Volunteers could participate if they had a diagnosis of COPD, a postbronchodilation l-second forced expiratory volume (FEV,) < 70% of predicted normal for age and height, and CO 2 15 ppm. A useful smoker/non-smoker cut-off among COPD patients is approximately 10 ppm CC0 (Crowley et al., 1989); however, patients just above the cutoff have little room to improve, so we required 15. Exclusion criteria were alphal-antitrypsin deficiency (a non-cigarette cause of COPD), recent myocardial infarction, active cancer, bilirubin > 2 mg/dl, BUN > 40 mg/dl, or other major disease. Patients residing outside the metro area, expecting to move within three months, completely denying recent smoking, or reporting more than one-quarter of tobacco intake from cigars or pipes, also were excluded. ISvaLuations. Patients had blood evaluations and spirometry (with bronchodilation) if not done within the last year. (Spirometry values for three Study 2 patients were approximately 2 years old. Because FEV, declines in COPD, their recorded FEV1’s conservatively estimate disease severity.) Patients had alphal-antitrypsin determinations, and did a nicotinedependence questionnaire (Fagerstrom, 1978), a modified American Lung Association intake questionnaire, and a pulmonary Environmental Effects Questionnaire (Tashkin et al., 1984). They consented to our contacting a friend or relative for locating patients lost to follow-up. Monitoring visits. We made home visits seven evenings per week between 16.00 and 20.00 h. Technicians first presented the patient with a Colorado lottery ticket (purchase value, $1) as compensation for the inconvenience of the visit. The technician asked now many cigarettes had
been smoked since the last visit and the time since the last cigarette, and then measured ambient and breath CO as described above, recording the difference as CCO. Some patients at some times then were reinforced with additional lottery tickets, based on CC0 (see below). Brochures.
During a baseline period while patients continued smoking, a physician visited their homes to discuss a brochure which described the program’s phases and gave smoking-cessation tips (adapted from American Cancer Society, 1987; National Cancer Institute, 1989).
have returned to their own mean baseline values for CCO, and for cigarette numbers and timing. Statistical procedures. Analyses of variance (ANOVA) and repeated-measures multivariate analyses of variance (MANOVA) were utilized (see SPSS Inc., 1990; Eckstrom et al., 1990; Lavori, 1990). In MANOVA’s we sometimes used as data each patient’s block-average in blocks of 10, 14, or 30 days, and we assessed homogeneity of variance with Box’s M statistic (SPSS Inc., 1990). Statistical details are ab breviated here, but a more detailed report is available (NAPS 1991).
Nicotine polacrilex gum.
Study
In Studies 2 and 3 the brochure also instructed patients on using nicotine gum. Alter the last baseline CC0 measurement in those studies, we delivered free, physician-prescribed nicotine gum (2 mg per piece) to patients. At subsequent visits we counted remaining pieces, always leaving a 2-day supply (up to 24 pieces per day). Standardizing staff-patient interactions. A technicians’ manual covered CO-monitor operation and measuring of breath and ambient CO, data recording, when patients were to be informed of CO values, responses to patients’ questions, acceptable and unacceptable kinds of encouragement, and handling of saliva samples. Technicians were supervised by a research coordinator and also met with the senior investigator. Saliva thiocyanate. Since some patients received nicotine gum, we assayed saliva thiocyanate rather than cotinine (a nicotine
tients received lottery tickets for gradual CC0 reductions, while control patients received tickets which were not contingent on CCO. We formed same-sex pairs of patients stratified for FEVl (either 41-70% of predicted normal, or < 41%). Each first-admitted pair-mate received experimental behavioral-reinforcement the and the second received the treatment, sham-reinforcement control treatment. After a 14-day ‘normal smoking’ baseline (when patients were not informed of their CC0 readings), experimental patients earned each night for 3 months 0 - 3 lottery tickets, depending on their ‘abstinence percent score’. This score on any day represented the percent reduction from that patient’s mean baseline CC0 toward 5 ppm (the average CC0 among nonsmoking COPD patients; Crowleg et al., 1989):
Percent
abstinence
= 100 -
1. Reinforcing gradual smokirq re&w tion, a ‘yoked-control’ clinical triaL. Study 1 methods. Study 1 experimental pa-
100 x (current ppm - current ambient ppm - 5 ppm) ~~~ -~~ (mean baseline ppm - mean baseline ambient ppm - 5 ppm)
metabolite). But thiocyanate data were quite variable, as in others’ studies (Jarvis et al., 1987), and so are not reported. Dropouts. In each study a baseline period (10 - 14 days) preceded an experimental period. Patients dropping out before the experimental period are excluded from analyses; those leaving during the experimental period are assumed to
Higher scores show greater CC0 reduction, and patients had to improve each month to continue their daily earning rate. One ticket was earned by scores 10 - 24 in Month 1, 25 - 39 in Month 2 and 40 - 64 in Month 3; two tickets were earned by scores 25-39,40-64 and 65 -89, in those respective months; three tickets were earned by 40 - 100, 65 - 100 and 90 - 100, in those months.
Scheduled visits Protocol
Special admission criteria
Procedures
80 60 (2.7) 40 (4.4) 25.8 (2.3) 70 26.1 (4.7) 6.3 (0.7)
X0 64 (2.6) 43 (3.3) 24.9 (1.6)
60 21.8 (2.3)
5.1 (0.4)
104 (A) 14-day baseline. (B) 3 mo. ‘yoked’ to experimental pairmate, same no. of lottery tickets as pairrnate each night, not contingent on CCO. No feedback of CC0 levels. No gum.
10
10
104 (A) Il-day baseline. (B) 3 months, 0- 3 tickets, sliding scale. Lower CC0 level required each month. No gum.
10
Sham shaping of CC0 reduction
10
Shaping CC0 reduction
Intervention
Subjects No. completing baseline % Dropout after baseline % Male Age (S.E.M.) FEVi (S.E.M.) % CC0 1st clinic contact (S.E.M.) % Receiving 0s Mean baseline CC0 (S.E.M.) Fagerstrom (S.E.M.)
Study 1 Yoked clinical trial
Summary of studies.
Design
Table I.
Gum criteria: dental or medical contraindications a4 (A) 14-day baseline, then Quit Date. (B) 42 days, 5 tickets if CC0 < 10. First on each night, then on random, unpredictable nights with decreasing frequency. Gum. (C) 1Cday wean gum. No CC0 reinforcement. (D) 14-day baseline. No gum or reinforcement.
6.3 (0.4)
50 31.2 (4.5)
69 63 (2.0) 45 (4) 31.5 (4.0)
25
16
Reinforce ‘Non-Smoker’ CC0 + Gum
Study 2 ABA intervention
(A) lo-day baseline, then Quit Date. (B) Day 11- 65. Multiple schedules with changing contingencies. Special high reinforcement for relapsers, non-quitters. Gum. (C) Day 66-75, wean gum, reinforcement continues. (D) IO-day baseline. No gum or tickets.
85
Gum criteria
5.6 (0.7)
50 27.8 (3.1)
57 63 (2.0) 44 (4) 21.5 (1.8)
29
14
Multipath reinforcement of ‘Non-Smoker’ CC0 + GUM
Study 3 ABA intervention
N.S. N.S.
N.S. N.S. N.S. N.S,
N.S.
Significance
51
and 2 to review the brochure and encourage smoking reduction. Thirty patients provided consent. Eight dropped out before the baseline ended. We discontinued two others during baseline; one had stopped smoking, and the other got nicotine gum elsewhere. These all were discounted from analyses. Then, two dropped out after the baseline; for analyses we assumed that they restarted smoking at their baseline levels, and for payment-matching we paired their pairmates with remaining patients. One patient, hospitalized on Day 95 of the 104-day study, died there of apparently unrelated problems. We analyzed his data by assigning to each missing day values equal to the mean of his last 10 days
Control patients always began at least a few days after their experimental partners. They were unaware of the pairing. On post-baseline evenings control patients, regardless of CC0 or abstinence score, received the same number of tickets as their pair-mates had earned on that experimental (not calendar) night. Controls, receiving the ‘free’ tickets before the CO measurement, knew that payments were not CO-contingent. We simply said that we could dispense tickets sometimes, based on how many had been given to others that night. Each night we told experimental patients, but not controls, their abstinence scores. No nicotine gum was dispensed. A physician visited patients’ homes at the end of Months 1
STUDY
1: CONTROLS
STUDY
1: EXPERIMENTALS
9BSLN
-14
0
14
28
42
STUDY
56
70
84
o-
’
-14
0
14
2
28
42
STUDY
56
70
84
56
70
64
3
36 8 2
27
6 ii K
18
8
-14
0
14
42
26 DAY
56
70
64
-14
0
14
28
42 DAY
Fig. 1. Daily mean ambient-corrected carbon monoxide (CCO) in parts per million, four patient groups. periods (14 days in Studies 1 and 2, 10 days in Study 3) without CC0 feedback, reinforcement, or nicotine
BSLN: gum.
baseline
52
of Study 1, we used each patient’s monthly average value in repeated-measures MANOVAs with three different analytic strategies (details in NAPS 1991). No analyses showed significant differences between groups for CC0 (Fig. l), cigarettes per day (Fig. 4), or time to last cigarette (Fig. 5), and no significant effect of month occurred in these variables. The baseline in Fig. 2 shows that one to three control patients in Study 1 achieved CC0 concentrations < 10 ppm on most baseline nights; this was rare in the baseline for other patient groups. Interestingly, only the Study 1 controls subsequently had a lengthy period (18 days) when none achieved CC0 < 10 ppm. Also, no controls ever achieved the double criteria of Fig. 3. Once again, multiple analytic approaches with MANOVA’s show-
in the study. Thus, we present data for all 20 who began the experimental period. Study 1 results. These two groups (and the other two discussed below) did not differ by ANOVA on admission evaluations (Table I). Figures l-6 respectively show for the two groups of Study 1 (and the two groups discussed below) for each night: CC0 concentrations, proportion of patients with CC0 < 10, proportion of patients with both CC0 < 10 and selfreported 24-h tobacco abstinence, mean selfreported number of cigarettes for the last 24 h, mean self-reported hours since last cigarette (the maximum daily entry per patient is 24 h, although some patients abstained longer), and mean tickets dispensed per patient per visit. To compare experimental and control groups
STUDY
1: CONTROLS
STUDY
801
I
60.
60
I
I
-14
0
L
1: EXPERIMENTALS
80
14
28
42
STUDY
56
70
84
-14
0
14
2
28
42
STUDY
80
801
60
6Oi
56
70
84
3
40.
20.
-14
0
14
28
42
56
70
84
DAY
Fig. 2. Fig. 1.
Daily percent of patients with ambient-corrected
114
0
14
28
42
56
70
84
DAY
carbon monoxide below 10 ppm, four patient groups. BSLN, as in
ed no significant (and technically valid) differences between the Study 1 controls and experimentals in the data of Figs. 2 or 3. Neither did the groups differ significantly in daily tickets received (as expected from the design). But since CC0 abstinence scores did not improve, and since improvement was required to maintain earnings, ticket awards declined significantly month by month (Fig. 6; ANOVA, F =101.2, P < 0.0005). In summary, Study 1 showed no significant differences between reinforcement and sham reinforcement.
differences among the three studies. With such poor response to gradual shaping in Study 1, we encouraged Study 2 patients to set a ‘quit date’ for immediately after the baseline, and then we reinforced only CC0 < 10 ppm, a useful smoker/non-smoker cutoff in COPD (Crowley et al., 1989). First offering higher reinforcement in an effort to influence more patients, we later ‘weaned down’ reinforcement by giving tickets only on random nights (a variable reinforcement schedule). We also prescribed nicotine gum, excluding from the study patients with dental conditions or recent myocardial infarctions who could not use gum. The control for Study 2 lay in the ABA design of its six 2-week segments. Patients received a flyer describing each segment as it began.
Study 2. Reinforce only ‘Non-Smoker’ CCO’s, nice tine gum Study 2 methods. Table I (‘Protocol’) outlines
STUDY
-14
0
14
STUDY
1: CONTROLS
28
42
STUDY
56
70
84
0
14
2
28
42
STUDY
4!
56
70
I
I BSLN
1
0
14
28
42
56
70
84
DAY
Daily percent of patients with ambient-corrected in last 24 h. BSLN, as in Fig. 1.
-14
carbon
0
monoxide
14
below
28
DAY
84
3
45
OL -14
Fig. 3. smoking
-14
1: EXPERIMENTALS
42
10 parts
56
I
1
70
per million
64
and self-report
of no
54
First
baseline
segment.
The third 14 days were like the second, but CC0 < IO ppm earned payment only on randomly determined nights (with a 35.7% chance that any night would be a pay-eligible night). Patients did not know in advance which nights could result in reinforcement.
During nightly CO monitoring there was no feedback of CC0 level. A physician visited patients’ homes, reviewing the brochure which now also explained nicotine gum use, set a quit date immediately after the baseline, and advised throwing cigarettes away then. Patients received a lottery ticket for each breath sample (regardless of CCO) here and throughout the project. They received a 2-day supply of gum after this segment’s last breath sample. Continuous reinforcement segment. In these 14 days patients were informed of each daily CC0 value and received five tickets for each daily sample < IO ppm. Gum was available (up to 24 pieces per day) in the second to fourth segments. Variable
reinforcement
(one-third)
Variable
reinforcement
(one-f%Ith) segment.
The fourth 14-day segment was like the third, but the chance of pay-eligibility fell to 21.4% on any given evening. Gum-weaning, no-reinforcement segment. The fifth 14 days included neither CC0 feedback nor CCO-contingent payment, and we gradually reduced gum pieces to zero by the last day. A physician visited patients’ homes as this weaning began to explain it, to review the brochure, and to encourage tobacco abstinence. Second baseline segment. Finally, a 14-day se-
segment.
STUDY 1: CONTROLS
STUDY
1: EXPERIMENTALS
30
30!
qSLN
Q
BSLN
Q
-~-l-r---r-‘.’
114
Q
14
28
42
56
70
84
-14
0
14
STUDY 2
28
42
56
70
64
STUDY 3
30-
20
qSLN
Q-t -14 Fig. 4.
BSLN
Q
14
26 42 DAY
Mean self-reported
cigarettes
56
BSLN
70
64
OJ -14
I,I
I
0
14
26
DAY
42
per day, four patient groups. BSLN, as in Fig. 1
56
,
70
84
cond baseline repeated the first baseline conditions: daily monitoring, no CC0 feedback or reinforcement, and no gum. Twenty-one patients consented to participate. Four dropped out without finishing the first baseline. We dropped one more during the baseline because of his mental confusion. Of the 16 who began the intervention, four dropped out before completion, and for analyses we assumed that each smoked at his or her baseline level from dropout until the study ended. Study 2 results. Study 2 patients, receiving gum and 5-ticket payments (but only for CC0 < 10) abruptly reduced CC0 (Fig. 1) and cigarettes per day (Fig. 4) after their quit-date. We analyzed Study 2 in 14-day time blocks, using each patient’s mean value in each block. Differences between the baseline period and the first 14 intervention days (linear contrast in a repeated measures MANOVA) were highly significant for CC0 (F = 30.3; df 15,l; P < 0.0005) and for cigarettes (F = 60.4; df 15,l; P < 0.0005). Patients also rescheduled (as well as reduced) their smoking; mean time from last cigarette to daily monitoring visit increased significantly from First Baseline to the next 14 days (Fig. 5; F = 21.7; df 15,l; P < 0.0005). There were significant increases from the First Baseline to the next 14 days in the proportion of patients with CC0 < 10 (Fig. 2; F = 19.82, df 15,l; P < 0.0005), and in the proportion with both CC0 < 10 and seh-report of no smoking (change from zero in baseline, Fig. 3; F = 12.63, df 15,l; P = 0.003). In this ABA design Second Baseline conditions replicated the First. We assessed their differences with repeated measures MANOVA’s, finding significance for CC0 (Fig. 1, F = 7.44; df 15,l; P = 0.016), cigarettes per day (Fig. 4, F = 8.18; df 15,l; P = O.Ol), and hours since last cigarette (Fig. 5, F = 6.26; df 15,l; P = 0.02). Similar analyses for proportion of patients with CC0 < 10 (Fig. 2) were not significant. Unequal variances invalidated such analyses for proportion of patients with both CC0 < 10 and selfreport of no smoking (Fig. 3). To summarize, during 14 days after the intervention ended, patients smoked fewer cigarettes, inhaled less CO,
and abstained longer before monitoring than they had before the intervention.
visits
3. Multipath reinforcement of ‘NonSmoker’ CC0 Is, gum Study 3 methods. The design of Study 2 Study
prevented individualized treatment of nonquitters or those who relapsed, and also response may have weakened as reinforcement’s frequency and novelty diminished. Study 3 changed reinforcement programs frequently and scheduled both to sustain interest, diminishing payments for abstaining patients and larger payments for non-quitters and relapsers. To gain social support for abstinence efforts, patients were allowed (not required) to choose another person to receive one ticket whenever the patient earned tickets, if that person were present during monitoring. Providing nicotine gum, we again excluded patients with medical or dental contraindications to it. The control for Study 3 was in its ABA design. Figure ‘7 outlines the complex protocol. Patients received flyers describing each reinforcement schedule as they began it. In a lo-day First Baseline they were encouraged to smoke ‘normally’ and were not informed of their daily CC0 values. A physician visited their homes to review the brochure, explaining the use of nicotine gum and setting a quit date for just after the baseline. After the last baseline CO determination, patients received a a-day supply of gum and encouragement to throw away all cigarettes. The two 5-day Training Schedules (Fig. 7) were identical, but were separated to allow early shunting of non-responders from the main path to the No-Quit Schedule. Similarly, patients who relapsed shunted to the Relapse Schedule from any main-path schedule after First Training. In the No-quit Schedule (Fig. 7) a physician visited again, set another quit date, and reviewed the brochure. Patients could remain in that schedule thereafter, but returned to the First Training Schedule if they received payment on 5 nights. They also could oscillate repeatedly between these schedules. The two Training Schedules each were design-
56
In the Relapse Schedule (Fig. 7) patients were encouraged by the monitor to set new quit dates, and for each CC0 < 10 ppm they received ten tickets until they accumulated five such payments, not necessarily consecutively. But a CC0 > 9 on three consecutive nights moved these patients to the No-Quit Schedule. Regardless of their current schedule, patients’ gum supplies were tapered to zero during Days 66-75, and all patients were in a Second Baseline (Fig. 7) from Days 76-85. These days replicated the First Baseline, except that no physician’s home-visit occurred. The various schedules reinforced only ‘nonsmoker’ CC0 levels, first daily and then at longer intervals, while applying special, larger reinforcements for non-quitters or those who relapsed. Patients were blind to schedules in
ed to produce ‘non-smoker’ CCO’s on at least 2 out of 5 nights; the Stepped Reinforcement Schedule (Fig. 7) was intended to sustain abstinence for 5 nights. Reinforcement increased each night for CC0 5 9 ppm, but one CC0 > 9 returned the patient to the start (2-ticket reinforcement) of the progression. Since Stepped Reinforcement lasted only five days, one relapse ended the opportunity for maximum payment. In the First Variable Reinforcement period (Fig. 7) patients could earn 10 tickets for CC0 < 10 on a random, unannounced, one-third of nights. The Second Variable Reinforcement period was like the first, except that patients could earn tickets on only 20% of nights. These schedules aimed to suppress smoking with large, infrequent payments until the intervention ended. STUDY
1: CONTROLS
STUDY
16
1: EXPERIMENTALS
16 BSLN
BSLN
8
0 -14
.“‘.‘.T.‘.-.’
0
14
28
42
STUDY
56
70
I
I
I
84
0
14
2
28
42
STUDY
56
70
84
56
70
84
3
16
8
-14
0
14
28
42 DAY
56
70
84
-14
0
14
28
42 DAY
Mean self-reported interval from last cigarette to evening monitoring visit, four patient groups. Maximum value per patient per day is 24 h, even if patient had abstained longer. BSLN, as in Fig. 1.
Fig. 5.
which they had not participated, so that they could not, for example, plan relapses to get to higher-paying schedules. Twenty patients consented to participate, but five dropped out before completing First Baseline; we also dropped one who stopped smoking during that baseline. Of 14 beginning the intervention, 4 dropped out before completing the study. For data analyses those four were presumed to maintain their baseline values from drop-out until the study ended. Study 3 results. Patients abruptly reduced CC0 (Fig. 1) and daily cigarette numbers (Fig. 4) immediately after the baseline. For example, the baseline mean CC0 was 27.1 ppm; the intervention mean was 12.7 ppm. We analyzed
STUDY
lo-day time blocks, using each patient’s mean value from each block. Comparing the lo-day baseline with the first 10 intervention days (via linear contrast in repeated measures MANOVA’s) the differences were highly significant for CC0 (8’ = 43.6; df 13,l; P < 0.0005 ) and for daily cigarettes (F = 43.4; df 13,l; P < 0.0005). Patients rescheduled (as well as reduced) their smoking. The mean time from last cigarette to daily visit increased significantly after the baseline (baseline vs. next 10 days, Fig. 5 (F = 14.6; df 13,l; P < 0.002)). The proportion of patients with CC0 < 10 (Fig. 2) rose significantly from First Baseline to the first 10 intervention days (F = 29.13, df 13,1, P
Elsevier Scientific Publishers
Reinforcing
29 (1991) 47 - 62
47
Ireland Ltd.
breath carbon monoxide reductions obstructive pulmonary disease
in chronic
Thomas J. Crowleya, Marilyn J. MacDonald”, Gary 0. Zerbeb and Thomas L. Petty” ‘Addiction
Research and Treatment Service, Department
of Psychiatry,
Box C-268, bDepartment of Biometrics.
University of
Colorado. School of Medicine, 4200 East Ninth Avenue, Denver, Colorado 80262 and rPresbyterianlSt, Lukea Hospital, Center for Health Sciences Education, 1719 East 19th Avenue, Denver, Colorado 80218 (U.S.A.)
(Received March 19th, 1991)
Chronic Obstructive Pulmonary Disease (COPD) usually results from tobacco smoking. Smoking cessation slows COPD’s progression, but few have studied anti-smoking treatments in COPD. In S-month trials we paid lottery tickets during daily home visits to still-smoking COPD patients for reductions in breath carbon monoxide (CO), a measure of smoke intake. In our first protocol experimental patients received 0 - 3 tickets per day, depending upon the extent of CO reduction below pre-treatment baselines; yoked controls received the same number of tickets, but not contingent on CO. The protocol produced no change. In a second study patients were assigned a post-baseline quit-date, received nicotine gum, and were paid up to 5 tickets per night, but only for CO < 10 parts per million (ppm). CO fell sharply as the intervention began, but gradually rose again. A third protocol added special reinforcement schedules for those who did not quit or relapsed (up to 20 tickets per night for CO < 10 ppm). Daily CO concentrations fell from 27.1 parts per million (baseline mean) to 12.7 (intervention mean), but rapid increases followed the intervention. Few patients stopped smoking, but CO and cigarettes used per day significantly fell during Studies 2 and 3. Post-hoc analysis suggested only a small effect from gum. Key words: carbon monoxide; smoking; pulmonary disease
Introduction ‘All those men have their price’. The Memoirs of Walpole, 1798 Chronic obstructive pulmonary disease (COPD) is America’s fifth leading source of death and causes 13 percent of the nation’s hospitalizations (Feinleib et al., 1989). Tobacco smoking produces most cases of COPD, and these cases yield nearly 60 000 deaths annually (U.S. Surgeon General, 1984). Smoking cessation slows COPD’s progression (Fletcher et al., 1976; Boss? et al., 1980; Boss? et al., 1981; Correspondence
to: Dr. T.J. Crowley, Addiction Research and Treatment Service, Department of Psychiatry, Box C-268, University of Colorado School of Medicine, 4200 East Ninth Avenue, Denver, CO 80262, U.S.A.
0376~8716/91/$03.50 0 1991 Elsevier Scientific Publishers Printed and Published in Ireland
Hughes et al., 1982; Tashkin et al., 1984), so effective anti-smoking treatment should help prevent pulmonary impairment. Some COPD patients do stop smoking (Crowley et al., 1989); those who persist despite pulmonary impairment and physician advice have seemed unanti-smoking for candidates promising treatment (Hall et al., 1983; Dudley et al., 1977; Daughton et al., 1980; Pederson, 1982; Pederson et al., 1982). Cotinine, breath carbon monoxide (CO) and thiocyanate assess smoke intake and self-report veracity; ‘No firm conclusions can be drawn until patient misrepresentation of smoking status’ is considered (Pederson, 1982). But we find no controlled smoking-treatment studies which used these measures among COPD patients; two such studies did include COPD patients along Ireland Ltd.
48
with others (Hall et al., 1983; Subcommittee of the Research Committee of the British Thoracic Society, 1984). Cotinine offers the best sensitivity and specificity, but CO assays are almost as accurate, give immediate results, and are cheaper; thiocyanate is less valuable (Jarvis et al., 1987). For behavioral reinforcement CO’s shorter half-life, 6-7 h in COPD patients (Crowley et al., 1989), may be an advantage. Learning theory suggests that new behavior is best ‘shaped’ by reinforcement for initial small or brief steps; several hours of tobacco abstinence may earn initial reinforcement for reduced CO, while the same behavioral change may not influence the more persistent cotinine. Smokers generally smoke less when paid for breath CO reductions; CO measures are objective and convenient, responsive to reinforcement, and correlate with smoking self-reports (Stitzer, Bigelow, 1982; Stitzer, Bigelow, 1983; Stitzer, Bigelow, 1984; Stitzer, Bigelow, 1985; Rand et al., 1989). After a pilot study of such payments in COPD patients (Crowley et al., 1989), we began the present studies. Our questions were, first, could payments reduce or stop smoking in patients who smoked on despite advancing COPD, and second, could such behavioral procedures produce a practical treatment. We did not recruit with newspaper advertisements, and patients were not charged for participation (Hall et al., 1983), procedures which may not attract the less-motivated. Instead, we identified still-smoking COPD patients in pulmonary clinics, inviting all of them to be paid participants. Methods and Results General methods Recruitment, We reviewed
pulmonary clinic charts at the Denver VA Hospital, University Hospital, Denver General Hospital, a Kaiser outpatient clinic and a private pulmonary office, for diagnoses of COPD. Upon arrival these patients blew into a MiniCo monitor (Catalyst Research, Baltimore), which showed breath CO in parts per million (ppm) (Crowley et al., 1989). Ambient air CO concentrations were subtracted
from breath concentrations, providing a Corrected Carbon Monoxide (CCO) value. Patients with COPD and CCOs 2 15 were told that they might be candidates for a study aimed at developing treatment to reduce patients’ CO with lottery-ticket payments. Interested patients were asked about demography and current smoking. Subsequently, a physician described the project and obtained signed informed consent. Admission criteria. Volunteers could participate if they had a diagnosis of COPD, a postbronchodilation l-second forced expiratory volume (FEV,) < 70% of predicted normal for age and height, and CO 2 15 ppm. A useful smoker/non-smoker cut-off among COPD patients is approximately 10 ppm CC0 (Crowley et al., 1989); however, patients just above the cutoff have little room to improve, so we required 15. Exclusion criteria were alphal-antitrypsin deficiency (a non-cigarette cause of COPD), recent myocardial infarction, active cancer, bilirubin > 2 mg/dl, BUN > 40 mg/dl, or other major disease. Patients residing outside the metro area, expecting to move within three months, completely denying recent smoking, or reporting more than one-quarter of tobacco intake from cigars or pipes, also were excluded. ISvaLuations. Patients had blood evaluations and spirometry (with bronchodilation) if not done within the last year. (Spirometry values for three Study 2 patients were approximately 2 years old. Because FEV, declines in COPD, their recorded FEV1’s conservatively estimate disease severity.) Patients had alphal-antitrypsin determinations, and did a nicotinedependence questionnaire (Fagerstrom, 1978), a modified American Lung Association intake questionnaire, and a pulmonary Environmental Effects Questionnaire (Tashkin et al., 1984). They consented to our contacting a friend or relative for locating patients lost to follow-up. Monitoring visits. We made home visits seven evenings per week between 16.00 and 20.00 h. Technicians first presented the patient with a Colorado lottery ticket (purchase value, $1) as compensation for the inconvenience of the visit. The technician asked now many cigarettes had
been smoked since the last visit and the time since the last cigarette, and then measured ambient and breath CO as described above, recording the difference as CCO. Some patients at some times then were reinforced with additional lottery tickets, based on CC0 (see below). Brochures.
During a baseline period while patients continued smoking, a physician visited their homes to discuss a brochure which described the program’s phases and gave smoking-cessation tips (adapted from American Cancer Society, 1987; National Cancer Institute, 1989).
have returned to their own mean baseline values for CCO, and for cigarette numbers and timing. Statistical procedures. Analyses of variance (ANOVA) and repeated-measures multivariate analyses of variance (MANOVA) were utilized (see SPSS Inc., 1990; Eckstrom et al., 1990; Lavori, 1990). In MANOVA’s we sometimes used as data each patient’s block-average in blocks of 10, 14, or 30 days, and we assessed homogeneity of variance with Box’s M statistic (SPSS Inc., 1990). Statistical details are ab breviated here, but a more detailed report is available (NAPS 1991).
Nicotine polacrilex gum.
Study
In Studies 2 and 3 the brochure also instructed patients on using nicotine gum. Alter the last baseline CC0 measurement in those studies, we delivered free, physician-prescribed nicotine gum (2 mg per piece) to patients. At subsequent visits we counted remaining pieces, always leaving a 2-day supply (up to 24 pieces per day). Standardizing staff-patient interactions. A technicians’ manual covered CO-monitor operation and measuring of breath and ambient CO, data recording, when patients were to be informed of CO values, responses to patients’ questions, acceptable and unacceptable kinds of encouragement, and handling of saliva samples. Technicians were supervised by a research coordinator and also met with the senior investigator. Saliva thiocyanate. Since some patients received nicotine gum, we assayed saliva thiocyanate rather than cotinine (a nicotine
tients received lottery tickets for gradual CC0 reductions, while control patients received tickets which were not contingent on CCO. We formed same-sex pairs of patients stratified for FEVl (either 41-70% of predicted normal, or < 41%). Each first-admitted pair-mate received experimental behavioral-reinforcement the and the second received the treatment, sham-reinforcement control treatment. After a 14-day ‘normal smoking’ baseline (when patients were not informed of their CC0 readings), experimental patients earned each night for 3 months 0 - 3 lottery tickets, depending on their ‘abstinence percent score’. This score on any day represented the percent reduction from that patient’s mean baseline CC0 toward 5 ppm (the average CC0 among nonsmoking COPD patients; Crowleg et al., 1989):
Percent
abstinence
= 100 -
1. Reinforcing gradual smokirq re&w tion, a ‘yoked-control’ clinical triaL. Study 1 methods. Study 1 experimental pa-
100 x (current ppm - current ambient ppm - 5 ppm) ~~~ -~~ (mean baseline ppm - mean baseline ambient ppm - 5 ppm)
metabolite). But thiocyanate data were quite variable, as in others’ studies (Jarvis et al., 1987), and so are not reported. Dropouts. In each study a baseline period (10 - 14 days) preceded an experimental period. Patients dropping out before the experimental period are excluded from analyses; those leaving during the experimental period are assumed to
Higher scores show greater CC0 reduction, and patients had to improve each month to continue their daily earning rate. One ticket was earned by scores 10 - 24 in Month 1, 25 - 39 in Month 2 and 40 - 64 in Month 3; two tickets were earned by scores 25-39,40-64 and 65 -89, in those respective months; three tickets were earned by 40 - 100, 65 - 100 and 90 - 100, in those months.
Scheduled visits Protocol
Special admission criteria
Procedures
80 60 (2.7) 40 (4.4) 25.8 (2.3) 70 26.1 (4.7) 6.3 (0.7)
X0 64 (2.6) 43 (3.3) 24.9 (1.6)
60 21.8 (2.3)
5.1 (0.4)
104 (A) 14-day baseline. (B) 3 mo. ‘yoked’ to experimental pairmate, same no. of lottery tickets as pairrnate each night, not contingent on CCO. No feedback of CC0 levels. No gum.
10
10
104 (A) Il-day baseline. (B) 3 months, 0- 3 tickets, sliding scale. Lower CC0 level required each month. No gum.
10
Sham shaping of CC0 reduction
10
Shaping CC0 reduction
Intervention
Subjects No. completing baseline % Dropout after baseline % Male Age (S.E.M.) FEVi (S.E.M.) % CC0 1st clinic contact (S.E.M.) % Receiving 0s Mean baseline CC0 (S.E.M.) Fagerstrom (S.E.M.)
Study 1 Yoked clinical trial
Summary of studies.
Design
Table I.
Gum criteria: dental or medical contraindications a4 (A) 14-day baseline, then Quit Date. (B) 42 days, 5 tickets if CC0 < 10. First on each night, then on random, unpredictable nights with decreasing frequency. Gum. (C) 1Cday wean gum. No CC0 reinforcement. (D) 14-day baseline. No gum or reinforcement.
6.3 (0.4)
50 31.2 (4.5)
69 63 (2.0) 45 (4) 31.5 (4.0)
25
16
Reinforce ‘Non-Smoker’ CC0 + Gum
Study 2 ABA intervention
(A) lo-day baseline, then Quit Date. (B) Day 11- 65. Multiple schedules with changing contingencies. Special high reinforcement for relapsers, non-quitters. Gum. (C) Day 66-75, wean gum, reinforcement continues. (D) IO-day baseline. No gum or tickets.
85
Gum criteria
5.6 (0.7)
50 27.8 (3.1)
57 63 (2.0) 44 (4) 21.5 (1.8)
29
14
Multipath reinforcement of ‘Non-Smoker’ CC0 + GUM
Study 3 ABA intervention
N.S. N.S.
N.S. N.S. N.S. N.S,
N.S.
Significance
51
and 2 to review the brochure and encourage smoking reduction. Thirty patients provided consent. Eight dropped out before the baseline ended. We discontinued two others during baseline; one had stopped smoking, and the other got nicotine gum elsewhere. These all were discounted from analyses. Then, two dropped out after the baseline; for analyses we assumed that they restarted smoking at their baseline levels, and for payment-matching we paired their pairmates with remaining patients. One patient, hospitalized on Day 95 of the 104-day study, died there of apparently unrelated problems. We analyzed his data by assigning to each missing day values equal to the mean of his last 10 days
Control patients always began at least a few days after their experimental partners. They were unaware of the pairing. On post-baseline evenings control patients, regardless of CC0 or abstinence score, received the same number of tickets as their pair-mates had earned on that experimental (not calendar) night. Controls, receiving the ‘free’ tickets before the CO measurement, knew that payments were not CO-contingent. We simply said that we could dispense tickets sometimes, based on how many had been given to others that night. Each night we told experimental patients, but not controls, their abstinence scores. No nicotine gum was dispensed. A physician visited patients’ homes at the end of Months 1
STUDY
1: CONTROLS
STUDY
1: EXPERIMENTALS
9BSLN
-14
0
14
28
42
STUDY
56
70
84
o-
’
-14
0
14
2
28
42
STUDY
56
70
84
56
70
64
3
36 8 2
27
6 ii K
18
8
-14
0
14
42
26 DAY
56
70
64
-14
0
14
28
42 DAY
Fig. 1. Daily mean ambient-corrected carbon monoxide (CCO) in parts per million, four patient groups. periods (14 days in Studies 1 and 2, 10 days in Study 3) without CC0 feedback, reinforcement, or nicotine
BSLN: gum.
baseline
52
of Study 1, we used each patient’s monthly average value in repeated-measures MANOVAs with three different analytic strategies (details in NAPS 1991). No analyses showed significant differences between groups for CC0 (Fig. l), cigarettes per day (Fig. 4), or time to last cigarette (Fig. 5), and no significant effect of month occurred in these variables. The baseline in Fig. 2 shows that one to three control patients in Study 1 achieved CC0 concentrations < 10 ppm on most baseline nights; this was rare in the baseline for other patient groups. Interestingly, only the Study 1 controls subsequently had a lengthy period (18 days) when none achieved CC0 < 10 ppm. Also, no controls ever achieved the double criteria of Fig. 3. Once again, multiple analytic approaches with MANOVA’s show-
in the study. Thus, we present data for all 20 who began the experimental period. Study 1 results. These two groups (and the other two discussed below) did not differ by ANOVA on admission evaluations (Table I). Figures l-6 respectively show for the two groups of Study 1 (and the two groups discussed below) for each night: CC0 concentrations, proportion of patients with CC0 < 10, proportion of patients with both CC0 < 10 and selfreported 24-h tobacco abstinence, mean selfreported number of cigarettes for the last 24 h, mean self-reported hours since last cigarette (the maximum daily entry per patient is 24 h, although some patients abstained longer), and mean tickets dispensed per patient per visit. To compare experimental and control groups
STUDY
1: CONTROLS
STUDY
801
I
60.
60
I
I
-14
0
L
1: EXPERIMENTALS
80
14
28
42
STUDY
56
70
84
-14
0
14
2
28
42
STUDY
80
801
60
6Oi
56
70
84
3
40.
20.
-14
0
14
28
42
56
70
84
DAY
Fig. 2. Fig. 1.
Daily percent of patients with ambient-corrected
114
0
14
28
42
56
70
84
DAY
carbon monoxide below 10 ppm, four patient groups. BSLN, as in
ed no significant (and technically valid) differences between the Study 1 controls and experimentals in the data of Figs. 2 or 3. Neither did the groups differ significantly in daily tickets received (as expected from the design). But since CC0 abstinence scores did not improve, and since improvement was required to maintain earnings, ticket awards declined significantly month by month (Fig. 6; ANOVA, F =101.2, P < 0.0005). In summary, Study 1 showed no significant differences between reinforcement and sham reinforcement.
differences among the three studies. With such poor response to gradual shaping in Study 1, we encouraged Study 2 patients to set a ‘quit date’ for immediately after the baseline, and then we reinforced only CC0 < 10 ppm, a useful smoker/non-smoker cutoff in COPD (Crowley et al., 1989). First offering higher reinforcement in an effort to influence more patients, we later ‘weaned down’ reinforcement by giving tickets only on random nights (a variable reinforcement schedule). We also prescribed nicotine gum, excluding from the study patients with dental conditions or recent myocardial infarctions who could not use gum. The control for Study 2 lay in the ABA design of its six 2-week segments. Patients received a flyer describing each segment as it began.
Study 2. Reinforce only ‘Non-Smoker’ CCO’s, nice tine gum Study 2 methods. Table I (‘Protocol’) outlines
STUDY
-14
0
14
STUDY
1: CONTROLS
28
42
STUDY
56
70
84
0
14
2
28
42
STUDY
4!
56
70
I
I BSLN
1
0
14
28
42
56
70
84
DAY
Daily percent of patients with ambient-corrected in last 24 h. BSLN, as in Fig. 1.
-14
carbon
0
monoxide
14
below
28
DAY
84
3
45
OL -14
Fig. 3. smoking
-14
1: EXPERIMENTALS
42
10 parts
56
I
1
70
per million
64
and self-report
of no
54
First
baseline
segment.
The third 14 days were like the second, but CC0 < IO ppm earned payment only on randomly determined nights (with a 35.7% chance that any night would be a pay-eligible night). Patients did not know in advance which nights could result in reinforcement.
During nightly CO monitoring there was no feedback of CC0 level. A physician visited patients’ homes, reviewing the brochure which now also explained nicotine gum use, set a quit date immediately after the baseline, and advised throwing cigarettes away then. Patients received a lottery ticket for each breath sample (regardless of CCO) here and throughout the project. They received a 2-day supply of gum after this segment’s last breath sample. Continuous reinforcement segment. In these 14 days patients were informed of each daily CC0 value and received five tickets for each daily sample < IO ppm. Gum was available (up to 24 pieces per day) in the second to fourth segments. Variable
reinforcement
(one-third)
Variable
reinforcement
(one-f%Ith) segment.
The fourth 14-day segment was like the third, but the chance of pay-eligibility fell to 21.4% on any given evening. Gum-weaning, no-reinforcement segment. The fifth 14 days included neither CC0 feedback nor CCO-contingent payment, and we gradually reduced gum pieces to zero by the last day. A physician visited patients’ homes as this weaning began to explain it, to review the brochure, and to encourage tobacco abstinence. Second baseline segment. Finally, a 14-day se-
segment.
STUDY 1: CONTROLS
STUDY
1: EXPERIMENTALS
30
30!
qSLN
Q
BSLN
Q
-~-l-r---r-‘.’
114
Q
14
28
42
56
70
84
-14
0
14
STUDY 2
28
42
56
70
64
STUDY 3
30-
20
qSLN
Q-t -14 Fig. 4.
BSLN
Q
14
26 42 DAY
Mean self-reported
cigarettes
56
BSLN
70
64
OJ -14
I,I
I
0
14
26
DAY
42
per day, four patient groups. BSLN, as in Fig. 1
56
,
70
84
cond baseline repeated the first baseline conditions: daily monitoring, no CC0 feedback or reinforcement, and no gum. Twenty-one patients consented to participate. Four dropped out without finishing the first baseline. We dropped one more during the baseline because of his mental confusion. Of the 16 who began the intervention, four dropped out before completion, and for analyses we assumed that each smoked at his or her baseline level from dropout until the study ended. Study 2 results. Study 2 patients, receiving gum and 5-ticket payments (but only for CC0 < 10) abruptly reduced CC0 (Fig. 1) and cigarettes per day (Fig. 4) after their quit-date. We analyzed Study 2 in 14-day time blocks, using each patient’s mean value in each block. Differences between the baseline period and the first 14 intervention days (linear contrast in a repeated measures MANOVA) were highly significant for CC0 (F = 30.3; df 15,l; P < 0.0005) and for cigarettes (F = 60.4; df 15,l; P < 0.0005). Patients also rescheduled (as well as reduced) their smoking; mean time from last cigarette to daily monitoring visit increased significantly from First Baseline to the next 14 days (Fig. 5; F = 21.7; df 15,l; P < 0.0005). There were significant increases from the First Baseline to the next 14 days in the proportion of patients with CC0 < 10 (Fig. 2; F = 19.82, df 15,l; P < 0.0005), and in the proportion with both CC0 < 10 and seh-report of no smoking (change from zero in baseline, Fig. 3; F = 12.63, df 15,l; P = 0.003). In this ABA design Second Baseline conditions replicated the First. We assessed their differences with repeated measures MANOVA’s, finding significance for CC0 (Fig. 1, F = 7.44; df 15,l; P = 0.016), cigarettes per day (Fig. 4, F = 8.18; df 15,l; P = O.Ol), and hours since last cigarette (Fig. 5, F = 6.26; df 15,l; P = 0.02). Similar analyses for proportion of patients with CC0 < 10 (Fig. 2) were not significant. Unequal variances invalidated such analyses for proportion of patients with both CC0 < 10 and selfreport of no smoking (Fig. 3). To summarize, during 14 days after the intervention ended, patients smoked fewer cigarettes, inhaled less CO,
and abstained longer before monitoring than they had before the intervention.
visits
3. Multipath reinforcement of ‘NonSmoker’ CC0 Is, gum Study 3 methods. The design of Study 2 Study
prevented individualized treatment of nonquitters or those who relapsed, and also response may have weakened as reinforcement’s frequency and novelty diminished. Study 3 changed reinforcement programs frequently and scheduled both to sustain interest, diminishing payments for abstaining patients and larger payments for non-quitters and relapsers. To gain social support for abstinence efforts, patients were allowed (not required) to choose another person to receive one ticket whenever the patient earned tickets, if that person were present during monitoring. Providing nicotine gum, we again excluded patients with medical or dental contraindications to it. The control for Study 3 was in its ABA design. Figure ‘7 outlines the complex protocol. Patients received flyers describing each reinforcement schedule as they began it. In a lo-day First Baseline they were encouraged to smoke ‘normally’ and were not informed of their daily CC0 values. A physician visited their homes to review the brochure, explaining the use of nicotine gum and setting a quit date for just after the baseline. After the last baseline CO determination, patients received a a-day supply of gum and encouragement to throw away all cigarettes. The two 5-day Training Schedules (Fig. 7) were identical, but were separated to allow early shunting of non-responders from the main path to the No-Quit Schedule. Similarly, patients who relapsed shunted to the Relapse Schedule from any main-path schedule after First Training. In the No-quit Schedule (Fig. 7) a physician visited again, set another quit date, and reviewed the brochure. Patients could remain in that schedule thereafter, but returned to the First Training Schedule if they received payment on 5 nights. They also could oscillate repeatedly between these schedules. The two Training Schedules each were design-
56
In the Relapse Schedule (Fig. 7) patients were encouraged by the monitor to set new quit dates, and for each CC0 < 10 ppm they received ten tickets until they accumulated five such payments, not necessarily consecutively. But a CC0 > 9 on three consecutive nights moved these patients to the No-Quit Schedule. Regardless of their current schedule, patients’ gum supplies were tapered to zero during Days 66-75, and all patients were in a Second Baseline (Fig. 7) from Days 76-85. These days replicated the First Baseline, except that no physician’s home-visit occurred. The various schedules reinforced only ‘nonsmoker’ CC0 levels, first daily and then at longer intervals, while applying special, larger reinforcements for non-quitters or those who relapsed. Patients were blind to schedules in
ed to produce ‘non-smoker’ CCO’s on at least 2 out of 5 nights; the Stepped Reinforcement Schedule (Fig. 7) was intended to sustain abstinence for 5 nights. Reinforcement increased each night for CC0 5 9 ppm, but one CC0 > 9 returned the patient to the start (2-ticket reinforcement) of the progression. Since Stepped Reinforcement lasted only five days, one relapse ended the opportunity for maximum payment. In the First Variable Reinforcement period (Fig. 7) patients could earn 10 tickets for CC0 < 10 on a random, unannounced, one-third of nights. The Second Variable Reinforcement period was like the first, except that patients could earn tickets on only 20% of nights. These schedules aimed to suppress smoking with large, infrequent payments until the intervention ended. STUDY
1: CONTROLS
STUDY
16
1: EXPERIMENTALS
16 BSLN
BSLN
8
0 -14
.“‘.‘.T.‘.-.’
0
14
28
42
STUDY
56
70
I
I
I
84
0
14
2
28
42
STUDY
56
70
84
56
70
84
3
16
8
-14
0
14
28
42 DAY
56
70
84
-14
0
14
28
42 DAY
Mean self-reported interval from last cigarette to evening monitoring visit, four patient groups. Maximum value per patient per day is 24 h, even if patient had abstained longer. BSLN, as in Fig. 1.
Fig. 5.
which they had not participated, so that they could not, for example, plan relapses to get to higher-paying schedules. Twenty patients consented to participate, but five dropped out before completing First Baseline; we also dropped one who stopped smoking during that baseline. Of 14 beginning the intervention, 4 dropped out before completing the study. For data analyses those four were presumed to maintain their baseline values from drop-out until the study ended. Study 3 results. Patients abruptly reduced CC0 (Fig. 1) and daily cigarette numbers (Fig. 4) immediately after the baseline. For example, the baseline mean CC0 was 27.1 ppm; the intervention mean was 12.7 ppm. We analyzed
STUDY
lo-day time blocks, using each patient’s mean value from each block. Comparing the lo-day baseline with the first 10 intervention days (via linear contrast in repeated measures MANOVA’s) the differences were highly significant for CC0 (8’ = 43.6; df 13,l; P < 0.0005 ) and for daily cigarettes (F = 43.4; df 13,l; P < 0.0005). Patients rescheduled (as well as reduced) their smoking. The mean time from last cigarette to daily visit increased significantly after the baseline (baseline vs. next 10 days, Fig. 5 (F = 14.6; df 13,l; P < 0.002)). The proportion of patients with CC0 < 10 (Fig. 2) rose significantly from First Baseline to the first 10 intervention days (F = 29.13, df 13,1, P
