The Relationship of Smoking Cessation to Coronary Heart Disease and Lung Cancer in the Multiple Risk Factor Intervention Trial (MRFIT) JUDITH K. OCKENE, PHD, LEWIS H. KULLER, MD, DRPH, KENNETH H. SVENDSEN, MS, AND ELAINE MEILAHN, MPH, DRPH Abstract: The impact of smoking cessation on coronary heart disease (CHD) and lung cancer was assessed after 10.5 years of follow-up in the 12,866 men in the Multiple Risk Factor Intervention Trial (MRFIT). Those men who died of lung cancer (n = 119) were either cigarette smokers at entry or ex-smokers; no lung cancer deaths occurred among the 1,859 men who reported never smoking cigarettes. The risk of lung cancer for smokers, adjusted for selected baseline variables using a Cox proportional hazards model, increased as the number of cigarettes smoked increased (B = 0.0203, SE = 0.0076). There was not the same graded response for CHD among smokers at entry. The risk of CHD death was greater among smokers than nonsmokers (RR = 1.57) (B = -0.0034, S.E. = 0.0048). After
one year of cessation, the relative risk of dying of CHD for the quitters as compared to non-quitters (RR = 0.63) was significantly lower even after adjusting for baseline differences and changes in other risk factors. The relative risk for smokers who quit for at least the first three years of the trial was even lower compared to non-quitters (RR = 0.38). However, the relative risk for lung cancer for quitters versus non-quitters was close to 1 both for quitters at 12 months and at three years. These data support the benefits of cessation in relation to CHD and are consistent with other epidemiologic studies which suggest that the lag time for a beneficial effect of smoking cessation on lung cancer may be as long as 20 years. (Am J Public Health 1990; 80:954-958.)
Introduction In the last two decades there have been continuing decreases in coronary heart disease (CHD) mortality rates in all United States populations'-3 while lung cancer mortality rates continue to increase in certain populations.4,5 Epidemiologic studies have consistently demonstrated a significantly higher relative risk for smokers compared to former smokers for each of these diseases." 26-8 The beneficial effect of smoking cessation on CHD is rapid; as the period of cessation increases, the ex-smoker's relative risk approaches that of the never smoker.3 However, in one study this return occurred gradually over a 20-year period.8 The relative risk of developing lung cancer has been demonstrated to decrease as the period of cessation increases for smokers who have stopped smoking,6-8 but their relative risk is still significantly elevated after 10 to 20 years of cessation compared to lifetime never smokers and is related to both the duration of smoking and length of cessation.6-8 In the available studies, the effect of smoking cessation on the risk of developing CHD and lung cancer was estimated either using data gathered retrospectively6-8 or by considering that smoking habits were unchanged over the period of follow-up.6,7 It is likely that during the course of these studies some ex-smokers became current smokers and vice versa. In addition, use of retrospective data entails problems of recall, and reports of cessation during the course of follow-up of these studies were not validated with biochemical measures. As a result, the mortality ratios reported for ex-smokers may be too high, and the benefits for cessation may actually be greater than those previously demonstrated for lung cancer and for CHD. Using the detailed smoking behavior data gathered
prospectively in the Multiple Risk Factor Intervention Trial (MRFIT), biochemically validated with serum thiocyanate measurements, we evaluated the relationship of smoking cessation to the development of CHD and lung cancer among middle-aged men, after 10.5 years of follow-up.
Address reprint requests to Judith K. Ockene, PhD, Department of Medicine, Division of Preventive and Behavioral Medicine, University of Massachusetts Medical School, 55 Lake Avenue North, Worcester, MA 01655. Drs. Kuller and Meilahn are with the Department of Epidemiology, University of Pittsburgh, Graduate School of Public Health, Pittsburgh, PA; Mr. Svendsen is with the Division of Biostatistics, University of Minnesota, School of Public Health, Minneapolis, MN. This paper, submitted to the Journal August 22, 1989, was revised and accepted for publication March 6, 1990. © 1990 American Journal of Public Health 0090-0036/90$1.50
954
Methods The MRFIT9-2' was sponsored by the National Heart, Lung and Blood Institute (NHLBI) as a clinical trial, to test the effect of a multifactor intervention program on mortality from CHD. Other objectives included examination of decreases in other mortality and morbidity end points including lung cancer. In summary, the participants were 12,866 men ages 35-57 years at increased risk for CHD and free of overt CHD on entry. Men were designated at increased risk if their levels of the three major CHD risk factors-serum cholesterol, cigarette smoking, and blood pressure-were high enough at first screening to place them in the top 15 percent (later changed to 10 percent for 72 percent of the men screened) of a risk score distribution based on data from the Framingham Heart Study. Men with diastolic blood pressure of 115 mmHg or higher and those with total cholesterol of 350 mg/dl or higher were excluded. Of the 12,866 men randomized to the special intervention (SI) and usual care (UC) groups, 64 percent (n = 8194) reported smoking cigarettes at the initial screening examinations, an average of 33.9 cigarettes/day. Randomization occurred from December 1973 to February 28, 1976. Thus, on February 28, 1982, when active intervention was terminated, each surviving man had been in the trial for at least six years, with a mean duration of 6.9 years of follow-up from randomization. All SI men were encouraged and helped to make dietary changes directed at lowering their blood cholesterol levels. The smoking intervention program used an initial intensive group approach as well as individual counseling sessions when indicated. Both approaches included the use of behavioral and cognitive interventions. SI men who were hypertensive (mean diastolic blood pressure greater than or equal to 90 mm Hg on two consecutive visits, or on antihypertensive medication from their personal physicians) were treated with a stepped care approach.'318 UC men were not offered AJPH August 1990, Vol. 80, No. 8
SMOKING CESSATION, CHD AND LUNG CANCER
intervention at the clinical centers and were referred back to their usual source of medical care. They were asked to return to the MRFIT clinical center once a year for a comprehensive evaluation, including assessment of risk factors and morbidity status. Baseline and annual data collection for both groups during the trial included a medical history and examination, information on past smoking patterns, and reported present cigarettes per day. Serum thiocyanate (Scn) was also measured at baseline and annually to validate reported smoking status. In addition to cigarette smoking, Scn may be elevated by the use of pipes, cigars and cigarillos, certain foods and diuretics.22 Subjects with an Scn level of more than 100 ,u mols/L were assessed to be smoking. Mortality Ascertainment During the trial and at termination of active intervention on February 28, 1982, deaths were ascertained by clinic staff through contact with family or friends of the deceased, routine follow-up of missed clinic visits, response to postcards requesting change-of-address information sent twice yearly to UC participants, and searches ofpublicly accessible files of deceased persons. The main source of vital status information after completion of participation in February 1982 has been the Social Security Administration and the National Death Index. Mortality ascertainment through December 31, 1985 was estimated to be approximately 100 percent complete using these data sources.'8 The cause of death used in this report is based on death certificate reports. Cause-specific death rates are based on coding of death certificates by trained nosologists, according to the Ninth Revision of the International Classification of Diseases (ICD-9).23 Each death certificate was independently coded by two nosologists and disagreements were adjudicated by a third nosologist. ICD-9 codes 410 to 414 and 429.2 were used to identify CHD deaths; ICD-9 code 162 was used to identify lung cancer deaths. Statistical Methods The SI-UC differences for the end points CHD and lung cancer mortality were tested for statistical significance using the log-rank test.24 As in other MRFIT publications, 90 percent confidence intervals for the percentage difference in death rates are presented. Mortality rates are presented as number of cases per 1,000 person-years. The relationship between smoking-related variables and end points was assessed for the combined MRFIT SI and UC smokers at entry into the study using the cox proportional hazards model.25,26 In one analysis, smokers at entry who attended the first annual examination were classified as quitters if they reported not smoking at the first annual examination and had serum Scn levels that were less than 100 ,u mols/L, or as smokers. In other analyses, smokers at entry who attended the first three annual examinations were classified as persistent quitters, if at each of the three annual examinations they reported not smoking and had serum Scn levels which were less than 100 p. mols/L at each visit, or as other smokerslquitters. Relative risk (RR) estimates and 95 percent confidence intervals for quitters compared to non-quitters were calculated using the proportional hazards regression model. RRs are presented both unadjusted and adjusted for age, baseline blood pressure, cholesterol, number of cigarettes smoked per day, and years of education (as an index of socioeconomic status) added to the model as continuous variables. Differences in baseline characteristics and changes in risk factor AJPH August 1990, Vol. 80, No. 8
levels from baseline to the third annual examination for persistent quitters compared to other smokers/quitters were tested for statistical significance using Student's t test. Results As of December 31, 1985, after an average period of follow-up of 10.5 years, there were 1,033 deaths in the MRFIT. There were 146 CHD deaths (3.42/1,000 personyears) among the SI smokers at entry and 165 (3.89/1,000 person-years) among the UC smokers at entry, a relative difference of -12.1 percent (90% CI = -27.1, 5.9). The CHD mortality rates were 2.26/1,000 person-years and 2.44/1,000 person-years for SI and UC non-smokers at entry, respectively. There were 56 lung cancer deaths (1.31/1,000 personyears) among the SI smokers at entry and 50 (1.18/1,000 person-years) among the UC smokers at entry, a relative difference of + 11.3 percent (90% CI = -19.2, 53.3). The nine SI and four UC non-smokers at entry who died of lung cancer had been smokers at some point prior to entry into the study. Mortality in Subgroups Defined after Randomization Although smoking intervention in the MRFIT resulted in a significantly higher cessation rate for SI compared to UC men (p < .001), the lung cancer and CHD mortality rates were not significantly different for these two groups. (Smoking cessation results in the MRFIT will be presented in detail in a subsequent section.) The within group analyses presented in Tables 1-5 are observational, combining data for SI and UC participants. No lung cancer deaths occurred among the 1,859 MRFIT men who had never smoked cigarettes. Among cigarette smokers there appeared to be a graded relationship between the number of cigarettes smoked per day at baseline, and the lung cancer mortality rate. However, number of cigarettes smoked per day at baseline and risk of CHD mortality did not appear to have as strong a relationship as for lung cancer. Ex-smokers appeared to have a risk of dying from CHD that approached that of never smokers (Table 1). In the MRFIT study group, never smokers had higher cholesterol and/or blood pressure levels than smokers at baseline in order to be in the top 10 percent of CHD risk. Regression analyses to account for confounders shown in Tables 2 and 3 indicate that only age, the screen 1 cigarettes/ day, and baseline Scn are of importance for lung cancer; for CHD mortality age, baseline Scn, DBP, and cholesterol were all significant risk factors (p < 0.01). Removing Scn from the regression analysis did not result in a significant relationship between number of cigarettes smoked and CHD risk among smokers. TABLE 1-CHD and Lung Cancer Deaths and Mortality Rates per 1,000 Person-Years by Number of Cigarettes Smoked per Day at Baseline for MRFIT SI and UC Men Combined CHD Deaths
Never-smokers Ex-smokers Smokers* 1-19 20-39
.40
Lung Cancer Deaths
Total
N
Rate/i 000 Person-Years
N
Rate/i 000 Person-Years
1859 2813
44 73
2.22 2.44
0 13
0 0.43
856 3747 3591
23 173 115
2.56 4.45 3.08
2 50 54
0.22
1.29 1.45
*Number of cigarettes/day smoked
955
OCKENE, ET AL. TABLE 2-Regression Coefficients from Cox Proportional Hazards Model for the End Point Lung Cancer Death for MRFIT Si and UC Men Who Reported Smoking Cigarettes at Entry
Variables
Coefficient
SE
Age (years) Screen 1 cigarettes/day Serum Thiocyanate (p.moles/l) Age began smoking (years) Filter/non-filter (1 = filter, 0 = non-filter) Tar Nicotine Alcoholic drinks/week Screen 1 DBP (mmHg) Screen 1 cholesterol (mg/dl)
0.1239 0.0203 0.0041 -0.0147
0.0198 0.0076 0.0019 0.0270
-0.6440 -0.1334 1.9110 -0.0120 0.0153 -0.0025
0.4076 0.0952 1.3438 0.0090 0.0155 0.0030
TABLE
3-Regression Coefficients from the Cox Proportional Hazards Model for the End Point CHD Death for MRFIT Si and UC Men who Smoked Cigarettes at Entry
Variables
Coefficient
SE
Age (years) Screen 1 cigarettes/day Serum Thiocyanate (>moles/l) Age began smoking Filter/non-filter (1 = filter, 0 = non-filter) Tar Nicotine Alcoholic drinks/week Screen 1 DBP (mmHg) Screen 1 cholesterol (mg/dl)
0.0626 -0.0034 0.0029 -0.0127
0.0104 0.0048 0.0011 0.0154
-0.3337 0.0489 -0.9942 -0.0150 0.0317 0.0047
0.2447 0.0567 0.7927 0.0055 0.0091 0.0017
CHD and lung cancer mortality rates for all men who quit smoking during the first year were compared with the rates of those who continued to smoke (Table 4). The total group of smokers who reported quitting in year one and had Scn < 100 ,. mols/L had a relative risk of dying from CHD .63 that of non-quitters. The relative risk estimate adjusted for entry risk variables were similar to unadjusted estimates (Table 4). The quitters had the same lung cancer mortality rate from year one to the end of the trial as those who were non-quitters at year one (Table 4). (This was true regardless of baseline levels of smoking or whether they were in the UC or SI group.) TABLE 4-Number of CHD and Lung Cancer Deaths and Mortality Rate (per 1,000 person-years) from the 12-month visit through 12/31/85 for Baseline Smokers by Smoking Status at 12 months, and Relative Risk Estimates (Quitters/Not Quitters) for MRFIT SI and UC Men # of
Smoking Status
Quit"
Participants SI + UC
Deaths
SI
+
UC
Relative Risk Quit/Not Quit (95% Cl)
Relative Riskd (95% Cl)
Coronary Heart Disease 1365
Not Quit
6298
Quit"
1365
239 (4.04) Lung Cancer
.63 (.44,.90)
6298
TABLE 5-Number of CHD and Lung Cancer Deaths and Mortality Rate (per 1,000 person-years) from the 36-month visit through 12/31/855 for Baseline Smokers by Smoking Status at 36 months, and Relative Risk Estimates (Quitters/Not Quitters) for MRFIT SI and UC Men
Smoking Status
.58 (.40,.84)
1.05 (.62,1.79)
78 (1.32)
Not Quit
956
6035 908
Not Quit a
This represents an average of 9.5 years of followup bQuitters are S, smokers who reported no smoking at year 1 with SCN moles/L. c Rates per 1000 person-years. d Adjusted for age, baseline cigarettes/day, cholesterol, blood pressure, education a
Relative Risk Deaths SI + UC
QuiVNot Quit (95% CI)
Adjusted
Relative Riskd (95% Cl)
12 (1.72)c
908
Quit"
17 (1.31)c
# of Participants SI + UC
Coronary Heart Disease
Quit"
33 (2.54)c
0.99 (.59,1.67) Not Quit
Adjusted
We also looked at persons who had quit smoking for at least the first three years of the trial (persistent quitters). Compared to the continuing smokers, the persistent quitters experienced a reduced risk of CHD death (RR = .38) but no change in risk of dying of lung cancer (Table 5). These comparisons are limited by the small number of quitters who persisted for three years which was less than the number who persisted for one year. (Participants who died in the first three years of the trial are also not included.) Persistent quitters were also lighter smokers than those who did not persist (average 30.1 vs 34.2 cigarettes per day). Persistent quitters and continuing smokers had differences in baseline risk factors and in changes in other risk factors through the three years the two groups were compared. On average, persistent quitters smoked four fewer cigarettes per day and drank 2.2 fewer alcoholic drinks per week at baseline than did other smokers or short-term quitters. Persistent quitters experienced significantly greater weight gain by three years than other smokers or quitters. Adjusting for the differences in risk factors at baseline did not change the relative risk estimates (0.38 versus 0.35), nor did adjusting for the differences in risk factor changes (RR = 0.36, 95% CI = 0.20, 0.65). Likewise, none of these models produced changes in relative risk estimates for lung cancer. The distribution of lung cancer deaths throughout the span of the trial does not indicate any difference in early versus late mortality by randomization status. Therefore, there is no evidence to date for a trend toward fewer SI as compared to UC lung cancer deaths as time from entry becomes longer, or closer to the 10.5 year maximum followup. Because of the relatively small number of participants who were long-term quitters of three or more years and died of lung cancer, it is not possible to evaluate whether the lung cancer death rates might be declining in the latter years of the trial among persistent quitters. Cessation and Relapse Patterns Although CHD and lung cancer mortality rates were not significantly different between the randomized groups, smoking cessation rates were significantly different for SI compared to UC men (Figure 1). The cross-sectional data indicate that, throughout the six years of follow-up in the trial, cessation rates for SI smokers were approximately 17 percentage points greater than they were for UC smokers.
6035
0.38 (.21, .68) 205 (4.52) Lung Cancer 9 (1.29)c .96 (.48,1.93) 61 (1.34)
0.35 (.20,.63)
1.02 (.50,2.05)
This represents an average of 7.5 years of followup
bQuitters are SI smokers who reported no smoking at years 1, 2 and 3 with Scn
one year of cessation, the relative risk of dying of CHD for the quitters as compared to non-quitters (RR = 0.63) was significantly lower even after adjusting for baseline differences and changes in other risk factors. The relative risk for smokers who quit for at least the first three years of the trial was even lower compared to non-quitters (RR = 0.38). However, the relative risk for lung cancer for quitters versus non-quitters was close to 1 both for quitters at 12 months and at three years. These data support the benefits of cessation in relation to CHD and are consistent with other epidemiologic studies which suggest that the lag time for a beneficial effect of smoking cessation on lung cancer may be as long as 20 years. (Am J Public Health 1990; 80:954-958.)
Introduction In the last two decades there have been continuing decreases in coronary heart disease (CHD) mortality rates in all United States populations'-3 while lung cancer mortality rates continue to increase in certain populations.4,5 Epidemiologic studies have consistently demonstrated a significantly higher relative risk for smokers compared to former smokers for each of these diseases." 26-8 The beneficial effect of smoking cessation on CHD is rapid; as the period of cessation increases, the ex-smoker's relative risk approaches that of the never smoker.3 However, in one study this return occurred gradually over a 20-year period.8 The relative risk of developing lung cancer has been demonstrated to decrease as the period of cessation increases for smokers who have stopped smoking,6-8 but their relative risk is still significantly elevated after 10 to 20 years of cessation compared to lifetime never smokers and is related to both the duration of smoking and length of cessation.6-8 In the available studies, the effect of smoking cessation on the risk of developing CHD and lung cancer was estimated either using data gathered retrospectively6-8 or by considering that smoking habits were unchanged over the period of follow-up.6,7 It is likely that during the course of these studies some ex-smokers became current smokers and vice versa. In addition, use of retrospective data entails problems of recall, and reports of cessation during the course of follow-up of these studies were not validated with biochemical measures. As a result, the mortality ratios reported for ex-smokers may be too high, and the benefits for cessation may actually be greater than those previously demonstrated for lung cancer and for CHD. Using the detailed smoking behavior data gathered
prospectively in the Multiple Risk Factor Intervention Trial (MRFIT), biochemically validated with serum thiocyanate measurements, we evaluated the relationship of smoking cessation to the development of CHD and lung cancer among middle-aged men, after 10.5 years of follow-up.
Address reprint requests to Judith K. Ockene, PhD, Department of Medicine, Division of Preventive and Behavioral Medicine, University of Massachusetts Medical School, 55 Lake Avenue North, Worcester, MA 01655. Drs. Kuller and Meilahn are with the Department of Epidemiology, University of Pittsburgh, Graduate School of Public Health, Pittsburgh, PA; Mr. Svendsen is with the Division of Biostatistics, University of Minnesota, School of Public Health, Minneapolis, MN. This paper, submitted to the Journal August 22, 1989, was revised and accepted for publication March 6, 1990. © 1990 American Journal of Public Health 0090-0036/90$1.50
954
Methods The MRFIT9-2' was sponsored by the National Heart, Lung and Blood Institute (NHLBI) as a clinical trial, to test the effect of a multifactor intervention program on mortality from CHD. Other objectives included examination of decreases in other mortality and morbidity end points including lung cancer. In summary, the participants were 12,866 men ages 35-57 years at increased risk for CHD and free of overt CHD on entry. Men were designated at increased risk if their levels of the three major CHD risk factors-serum cholesterol, cigarette smoking, and blood pressure-were high enough at first screening to place them in the top 15 percent (later changed to 10 percent for 72 percent of the men screened) of a risk score distribution based on data from the Framingham Heart Study. Men with diastolic blood pressure of 115 mmHg or higher and those with total cholesterol of 350 mg/dl or higher were excluded. Of the 12,866 men randomized to the special intervention (SI) and usual care (UC) groups, 64 percent (n = 8194) reported smoking cigarettes at the initial screening examinations, an average of 33.9 cigarettes/day. Randomization occurred from December 1973 to February 28, 1976. Thus, on February 28, 1982, when active intervention was terminated, each surviving man had been in the trial for at least six years, with a mean duration of 6.9 years of follow-up from randomization. All SI men were encouraged and helped to make dietary changes directed at lowering their blood cholesterol levels. The smoking intervention program used an initial intensive group approach as well as individual counseling sessions when indicated. Both approaches included the use of behavioral and cognitive interventions. SI men who were hypertensive (mean diastolic blood pressure greater than or equal to 90 mm Hg on two consecutive visits, or on antihypertensive medication from their personal physicians) were treated with a stepped care approach.'318 UC men were not offered AJPH August 1990, Vol. 80, No. 8
SMOKING CESSATION, CHD AND LUNG CANCER
intervention at the clinical centers and were referred back to their usual source of medical care. They were asked to return to the MRFIT clinical center once a year for a comprehensive evaluation, including assessment of risk factors and morbidity status. Baseline and annual data collection for both groups during the trial included a medical history and examination, information on past smoking patterns, and reported present cigarettes per day. Serum thiocyanate (Scn) was also measured at baseline and annually to validate reported smoking status. In addition to cigarette smoking, Scn may be elevated by the use of pipes, cigars and cigarillos, certain foods and diuretics.22 Subjects with an Scn level of more than 100 ,u mols/L were assessed to be smoking. Mortality Ascertainment During the trial and at termination of active intervention on February 28, 1982, deaths were ascertained by clinic staff through contact with family or friends of the deceased, routine follow-up of missed clinic visits, response to postcards requesting change-of-address information sent twice yearly to UC participants, and searches ofpublicly accessible files of deceased persons. The main source of vital status information after completion of participation in February 1982 has been the Social Security Administration and the National Death Index. Mortality ascertainment through December 31, 1985 was estimated to be approximately 100 percent complete using these data sources.'8 The cause of death used in this report is based on death certificate reports. Cause-specific death rates are based on coding of death certificates by trained nosologists, according to the Ninth Revision of the International Classification of Diseases (ICD-9).23 Each death certificate was independently coded by two nosologists and disagreements were adjudicated by a third nosologist. ICD-9 codes 410 to 414 and 429.2 were used to identify CHD deaths; ICD-9 code 162 was used to identify lung cancer deaths. Statistical Methods The SI-UC differences for the end points CHD and lung cancer mortality were tested for statistical significance using the log-rank test.24 As in other MRFIT publications, 90 percent confidence intervals for the percentage difference in death rates are presented. Mortality rates are presented as number of cases per 1,000 person-years. The relationship between smoking-related variables and end points was assessed for the combined MRFIT SI and UC smokers at entry into the study using the cox proportional hazards model.25,26 In one analysis, smokers at entry who attended the first annual examination were classified as quitters if they reported not smoking at the first annual examination and had serum Scn levels that were less than 100 ,u mols/L, or as smokers. In other analyses, smokers at entry who attended the first three annual examinations were classified as persistent quitters, if at each of the three annual examinations they reported not smoking and had serum Scn levels which were less than 100 p. mols/L at each visit, or as other smokerslquitters. Relative risk (RR) estimates and 95 percent confidence intervals for quitters compared to non-quitters were calculated using the proportional hazards regression model. RRs are presented both unadjusted and adjusted for age, baseline blood pressure, cholesterol, number of cigarettes smoked per day, and years of education (as an index of socioeconomic status) added to the model as continuous variables. Differences in baseline characteristics and changes in risk factor AJPH August 1990, Vol. 80, No. 8
levels from baseline to the third annual examination for persistent quitters compared to other smokers/quitters were tested for statistical significance using Student's t test. Results As of December 31, 1985, after an average period of follow-up of 10.5 years, there were 1,033 deaths in the MRFIT. There were 146 CHD deaths (3.42/1,000 personyears) among the SI smokers at entry and 165 (3.89/1,000 person-years) among the UC smokers at entry, a relative difference of -12.1 percent (90% CI = -27.1, 5.9). The CHD mortality rates were 2.26/1,000 person-years and 2.44/1,000 person-years for SI and UC non-smokers at entry, respectively. There were 56 lung cancer deaths (1.31/1,000 personyears) among the SI smokers at entry and 50 (1.18/1,000 person-years) among the UC smokers at entry, a relative difference of + 11.3 percent (90% CI = -19.2, 53.3). The nine SI and four UC non-smokers at entry who died of lung cancer had been smokers at some point prior to entry into the study. Mortality in Subgroups Defined after Randomization Although smoking intervention in the MRFIT resulted in a significantly higher cessation rate for SI compared to UC men (p < .001), the lung cancer and CHD mortality rates were not significantly different for these two groups. (Smoking cessation results in the MRFIT will be presented in detail in a subsequent section.) The within group analyses presented in Tables 1-5 are observational, combining data for SI and UC participants. No lung cancer deaths occurred among the 1,859 MRFIT men who had never smoked cigarettes. Among cigarette smokers there appeared to be a graded relationship between the number of cigarettes smoked per day at baseline, and the lung cancer mortality rate. However, number of cigarettes smoked per day at baseline and risk of CHD mortality did not appear to have as strong a relationship as for lung cancer. Ex-smokers appeared to have a risk of dying from CHD that approached that of never smokers (Table 1). In the MRFIT study group, never smokers had higher cholesterol and/or blood pressure levels than smokers at baseline in order to be in the top 10 percent of CHD risk. Regression analyses to account for confounders shown in Tables 2 and 3 indicate that only age, the screen 1 cigarettes/ day, and baseline Scn are of importance for lung cancer; for CHD mortality age, baseline Scn, DBP, and cholesterol were all significant risk factors (p < 0.01). Removing Scn from the regression analysis did not result in a significant relationship between number of cigarettes smoked and CHD risk among smokers. TABLE 1-CHD and Lung Cancer Deaths and Mortality Rates per 1,000 Person-Years by Number of Cigarettes Smoked per Day at Baseline for MRFIT SI and UC Men Combined CHD Deaths
Never-smokers Ex-smokers Smokers* 1-19 20-39
.40
Lung Cancer Deaths
Total
N
Rate/i 000 Person-Years
N
Rate/i 000 Person-Years
1859 2813
44 73
2.22 2.44
0 13
0 0.43
856 3747 3591
23 173 115
2.56 4.45 3.08
2 50 54
0.22
1.29 1.45
*Number of cigarettes/day smoked
955
OCKENE, ET AL. TABLE 2-Regression Coefficients from Cox Proportional Hazards Model for the End Point Lung Cancer Death for MRFIT Si and UC Men Who Reported Smoking Cigarettes at Entry
Variables
Coefficient
SE
Age (years) Screen 1 cigarettes/day Serum Thiocyanate (p.moles/l) Age began smoking (years) Filter/non-filter (1 = filter, 0 = non-filter) Tar Nicotine Alcoholic drinks/week Screen 1 DBP (mmHg) Screen 1 cholesterol (mg/dl)
0.1239 0.0203 0.0041 -0.0147
0.0198 0.0076 0.0019 0.0270
-0.6440 -0.1334 1.9110 -0.0120 0.0153 -0.0025
0.4076 0.0952 1.3438 0.0090 0.0155 0.0030
TABLE
3-Regression Coefficients from the Cox Proportional Hazards Model for the End Point CHD Death for MRFIT Si and UC Men who Smoked Cigarettes at Entry
Variables
Coefficient
SE
Age (years) Screen 1 cigarettes/day Serum Thiocyanate (>moles/l) Age began smoking Filter/non-filter (1 = filter, 0 = non-filter) Tar Nicotine Alcoholic drinks/week Screen 1 DBP (mmHg) Screen 1 cholesterol (mg/dl)
0.0626 -0.0034 0.0029 -0.0127
0.0104 0.0048 0.0011 0.0154
-0.3337 0.0489 -0.9942 -0.0150 0.0317 0.0047
0.2447 0.0567 0.7927 0.0055 0.0091 0.0017
CHD and lung cancer mortality rates for all men who quit smoking during the first year were compared with the rates of those who continued to smoke (Table 4). The total group of smokers who reported quitting in year one and had Scn < 100 ,. mols/L had a relative risk of dying from CHD .63 that of non-quitters. The relative risk estimate adjusted for entry risk variables were similar to unadjusted estimates (Table 4). The quitters had the same lung cancer mortality rate from year one to the end of the trial as those who were non-quitters at year one (Table 4). (This was true regardless of baseline levels of smoking or whether they were in the UC or SI group.) TABLE 4-Number of CHD and Lung Cancer Deaths and Mortality Rate (per 1,000 person-years) from the 12-month visit through 12/31/85 for Baseline Smokers by Smoking Status at 12 months, and Relative Risk Estimates (Quitters/Not Quitters) for MRFIT SI and UC Men # of
Smoking Status
Quit"
Participants SI + UC
Deaths
SI
+
UC
Relative Risk Quit/Not Quit (95% Cl)
Relative Riskd (95% Cl)
Coronary Heart Disease 1365
Not Quit
6298
Quit"
1365
239 (4.04) Lung Cancer
.63 (.44,.90)
6298
TABLE 5-Number of CHD and Lung Cancer Deaths and Mortality Rate (per 1,000 person-years) from the 36-month visit through 12/31/855 for Baseline Smokers by Smoking Status at 36 months, and Relative Risk Estimates (Quitters/Not Quitters) for MRFIT SI and UC Men
Smoking Status
.58 (.40,.84)
1.05 (.62,1.79)
78 (1.32)
Not Quit
956
6035 908
Not Quit a
This represents an average of 9.5 years of followup bQuitters are S, smokers who reported no smoking at year 1 with SCN moles/L. c Rates per 1000 person-years. d Adjusted for age, baseline cigarettes/day, cholesterol, blood pressure, education a
Relative Risk Deaths SI + UC
QuiVNot Quit (95% CI)
Adjusted
Relative Riskd (95% Cl)
12 (1.72)c
908
Quit"
17 (1.31)c
# of Participants SI + UC
Coronary Heart Disease
Quit"
33 (2.54)c
0.99 (.59,1.67) Not Quit
Adjusted
We also looked at persons who had quit smoking for at least the first three years of the trial (persistent quitters). Compared to the continuing smokers, the persistent quitters experienced a reduced risk of CHD death (RR = .38) but no change in risk of dying of lung cancer (Table 5). These comparisons are limited by the small number of quitters who persisted for three years which was less than the number who persisted for one year. (Participants who died in the first three years of the trial are also not included.) Persistent quitters were also lighter smokers than those who did not persist (average 30.1 vs 34.2 cigarettes per day). Persistent quitters and continuing smokers had differences in baseline risk factors and in changes in other risk factors through the three years the two groups were compared. On average, persistent quitters smoked four fewer cigarettes per day and drank 2.2 fewer alcoholic drinks per week at baseline than did other smokers or short-term quitters. Persistent quitters experienced significantly greater weight gain by three years than other smokers or quitters. Adjusting for the differences in risk factors at baseline did not change the relative risk estimates (0.38 versus 0.35), nor did adjusting for the differences in risk factor changes (RR = 0.36, 95% CI = 0.20, 0.65). Likewise, none of these models produced changes in relative risk estimates for lung cancer. The distribution of lung cancer deaths throughout the span of the trial does not indicate any difference in early versus late mortality by randomization status. Therefore, there is no evidence to date for a trend toward fewer SI as compared to UC lung cancer deaths as time from entry becomes longer, or closer to the 10.5 year maximum followup. Because of the relatively small number of participants who were long-term quitters of three or more years and died of lung cancer, it is not possible to evaluate whether the lung cancer death rates might be declining in the latter years of the trial among persistent quitters. Cessation and Relapse Patterns Although CHD and lung cancer mortality rates were not significantly different between the randomized groups, smoking cessation rates were significantly different for SI compared to UC men (Figure 1). The cross-sectional data indicate that, throughout the six years of follow-up in the trial, cessation rates for SI smokers were approximately 17 percentage points greater than they were for UC smokers.
6035
0.38 (.21, .68) 205 (4.52) Lung Cancer 9 (1.29)c .96 (.48,1.93) 61 (1.34)
0.35 (.20,.63)
1.02 (.50,2.05)
This represents an average of 7.5 years of followup
bQuitters are SI smokers who reported no smoking at years 1, 2 and 3 with Scn
