Work Site-Based Cancer Prevention:
Primary Results from the
Working Well Trial
Glorian Sorensen, PhD, MPH, Beti Thompson, PhD, Karen Glanz, PhD, MPH,
Ziding Feng, PhD, Susan Kinne, PhD, Carlo DiClemente, PhD, Karen Emmons,
PhD, Jerianne Heimendinger, ScD, MPH, RD, Claudia Probarti, PhD, RD,
and Edward Lichtenstein, PhD, for the Working Well Trial
Introduction This study was conducted in 111 work
sites by four study centers, a coordinating
Each year approximately a half- center, and the National Cancer Institute.
million Americans die from cancer.' It is The large number of work sites permitted
estimated that up to 70% of these cancers assessment of change at the work site
are associated with lifestyle or environ- level. The primary hypothesis of the
mental exposures and are therefore pre- Working Well Trial was that a sustained
ventable.2 The National Cancer Institute 2-year comprehensive cancer control work
has taken aggressive steps to implement site health promotion intervention ad-
cancer prevention and control measures dressing dietary change and smoking
to reduce cancer risk. Among its objec- cessation, delivered by a participatory
tives are the reduction of average con- strategy that targeted individuals and the
sumption of fat to 30% of calories or less, work site environment, would be more
the increase in average consumption of
fiber to 20 g to 30 g per day, the increase in effective than a minimal intervention in
servings of fruits and vegetables to five or achieving both individual behavioral and
more per day, and the reduction in the environmental changes. This paper re-
percentage of adults who smoke to 15% ports findings on individual behavior
or less.3 changes.
Work sites have been targeted as a
priority location for intervention efforts
aimed at these objectives. They provide
ready access to working populations, the Glorian Sorensen is with the Dana-Farber
opportunity for promoting environmental Cancer Institute and the Harvard School of
supports for behavior change, and natural Public Health, Boston. Beti Thompson, Ziding
structures for social support.45 Few stud- Feng, and Susan Kinne are with the Fred
ies, however, have employed a random- Hutchinson Cancer Research Center, Seattle,
Wash. Karen Glanz is with the Cancer Re-
ized design; often, even those with a search Center of Hawaii, Honolulu. Carlo
randomized design have compared too DiClemente is with the University of Houston,
few work sites to have adequate statistical Tex. Karen Emmons is with Brown University/
power.6 Interventions have typically tar- The Miriam Hospital, Providence, RI. Jerianne
Heimendinger is with the National Cancer
geted only the individual, ignoring the Institute, Rockville, Md. Claudia Probart is
organizational context. Only rarely have with the Nutrition Department, Penn State
results been based on change in the entire University, University Park. Edward Lichten-
work site population rather than the stein is with the Oregon Research Institute,
subset of employees participating in the Eugene. For a list of the centers and research-
ers in the Working Well Trial, see the
program.7-9 The study reported here was Acknowledgments.
designed to address many of these meth- Requests for reprints should be sent to
odological problems. Glorian Sorensen, PhD, MPH, Dana-Farber
The purpose of this paper is to Cancer Institute, Division of Cancer Epidemi-
ology and Control, 44 Binney St, Boston MA
present the primary outcomes of the 02115.
Working Well Trial, the largest work-site This paper was accepted November 1,
cancer control trial in the United States.'0 1995.
American Journal of Public Health 939
Sorensen et aL
Methods represented manufacturing, communica- The protocol defined core interven-
tions, public service, and utilities. Work tions directed toward individuals; these
The Working Well Trial used a sites ranged in size from 49 to 1700 included a kickoff event, interactive activi-
randomized, matched-pair research de- workers (mean = 316). As a condition of ties, posters and brochures, self-assess-
sign, with the work site as the unit of work-site participation, managers of all ments, self-help materials, campaigns and
assignment and analysis.10 Because the work sites agreed on random assignment contests, and direct education through
work site was the unit of both randomiza- of their work site to the intervention or classes and groups. Core interventions
tion and analysis, data from the 111 control group and also agreed to adminis- aimed at environmental change included
participating work sites were pooled to ter employee and organizational surveys consultation on the formation and imple-
test the hypotheses. Cross-sectional sur- and to deliver the intervention based on mentation of smoking policy, changes in
veys of individuals and surveys of key their assignment to group. Although the food offerings and/or nutrition education
informants were conducted in each work work sites were a convenience sample and in cafeterias and vending machines, and
site at baseline and follow-up. After were recruited by the use of different catering policies. Additional information
baseline assessments, work sites were strategies within each study center,'0 a on the Working Well intervention is
stratified, matched into pairs, and ran- variety ofwork site environments, types of provided by Abrams et al.'0
domly assigned within pairs to the inter- business, and geographic regions was Control sites received summary re-
vention or control group. Stratification represented. As reported previously, there sults from the employee survey for distri-
factors were the presence of a cafeteria, were no demographic differences at base- bution to employees and were asked to
work site size, type of smoking policy, line or follow-up between intervention document health promotion activities.
company type, sex distribution, distribu- and control work sites.11 Three of the four study centers provided
tion of blue- and white-collar jobs, and an optional minimal intervention at con-
response rate to the baseline survey.'0 The Working Well Intervention trol sites, following a standardized proto-
Evaluation of the effects of the interven- col that included the distribution of
tion was based on the difference between The Working Well intervention was printed materials such as posters and
intervention and control work site means based on a theoretical model derived newsletters.
within each work site pair, with adjust- from individual, organizational, and com-
ment for the baseline work site mean as a munity activation theories.10 Based on Data Collection
covariate. Calculations of sample size these theories, the intervention focused
were based on the differences thought to on (1) promotion and building awareness, Data were collected from individual
be important to detect between interven- (2) action and skills training, and (3) employees with self-administered surveys
tion and control sites: 2 percentage points maintenance of behavior and preventing containing standard items in all study
for the percentage of energy obtained relapse.'2'13 Participatory strategies fol- centers. Baseline data were collected
from fat consumption; 3 g of fiber per day lowed Rothman's community activation from September to December 1990, and
(1.5 g of fiber per 1000 kilocal); one principles.14 This literature indicates that follow-up data, from September to De-
serving of fruits and vegetables per day; participation in activities is enhanced cember 1993. Eligible employees were
and a 6-month smoking abstinence rate of when people are involved in planning and permanent employees working at least
6%. The power to detect these differences implementation.15 50% of the work time. The methods of
was at least 80%. The sample size was A common intervention protocol survey distribution varied by study cen-
determined primarily by the smoking specifying strategies and process objec- ter." Briefly, Florida and Brown mailed
outcomes, and excess power was there- tives was implemented in the four study surveys to each employee in the work site,
fore available for the dietary outcomes. In centers. The common intervention was Dana-Farber mailed surveys to a random
addition, analyses were conducted to targeted at eating patterns in all four sample of employees in each work site,
examine work-site smoking prevalence. study centers and smoking in three of the and MD Anderson administered question-
four study centers. Florida did not include naires to employees at mandatory work
Description of the Sample a smoking intervention, since smoking
site meetings. Follow-up reminders were
was banned at all participating work sites, sent to maximize response rates. No
The study was conducted in four follow-up surveys of nonrespondents were
study centers: the Brown University School but did target cancer screening practices. conducted, owing to constraints imposed
of Medicine/Miriam Hospital, the Dana- The other three sites targeted smoking
and nutrition plus one additional risk
by the work sites.
Farber Cancer Institute/University of
Massachusetts, the University of Florida, factor (occupational exposures to carcino-
and the MD Anderson Cancer Center. gens, Dana-Farber; exercise, Brown; and Pimnary Outcomes
The sample contained 111 work sites that smokeless tobacco, MD Anderson). Nutrition outcomes. The primary
employed over 28 000 workers in 16 All study centers relied on an inter- evaluation of dietary change among indi-
states. (As noted in Abrams et al.,'0 114 vention model that used participatory viduals was based on assessment of
work sites were initially recruited to this strategies. An employee from each work nutrient intakes of fat, fiber, and fruits
study. Because of economic dislocations, site was appointed as the work-site coordi- and vegetables, using an 88-item semi-
three work sites located at Brown did not nator and served as the gatekeeper to the quantitative food-frequency question-
participate in the final survey, leaving a work site. In addition, employee advisory naire with portion sizes (176 items to-
total of 111 work sites. Two of these sites boards were formed as a way to incorpo- tal).'6"7 This questionnaire was based on
were from the intervention condition, one rate employee input and concerns. These the Block food-frequency questionnaire,
from the control. For pairwise analyses, boards had from 4 to 12 members, who which has been validated in previous
three pairs were therefore excluded, for a were trained in the goals and content studies.'8 The analysis software for the
total of 108 work sites.) The companies areas of the project.'0 Working Well food-frequency question-
940 American Journal of Public Health July 1996, Vol. 86, No. 7
Work-Site Cancer Prevention
naire was based on a nutrient database day, or who defined themselves as current not used, since there is no theoretical
developed by the University of Minnesota smokers. approach or literature to justify differen-
Nutrition Coordinating Center.'9 This tial weighting of particular items.
instrument was pretested prior to use in Process Evaluation
this study, and minor modifications were A process evaluation was designed to StatisticalAnalyses
made to reflect regional dietary differ- (1) assess the extent to which the interven- The primary analyses covered 108
ences. The food-frequency questionnaire tion was delivered, based on data from the work sites since the 3 work sites not
was selected for use in this study because "senders" of the intervention (i.e., project completing the study represented 3 sepa-
it was able to estimate total dietary habits staff), and (2) assess the extent to which rate pairs (see "Description ofthe Sample"
and was feasible for such a large-scale, the intervention was received; using data under "Methods"). Analyses conducted
population-based study. from the "receivers" of the intervention take into account the work site as the unit
The outcome variables calculated (i.e., work-site employees).10 These mea- of randomization. For continuous vari-
from the food frequency questionnaire sures were included to assess indepen- ables (consumption of fat, fiber, and fruits
included the percentage of energy that dently whether the independent variable and vegetables at the final employee
came from fat, grams of fiber per 1000 was in fact differentially manipulated survey), mixed linear models were used,22
kilocalories, and daily servings of fruits between the intervention and control where the study center and intervention
and vegetables. The fat and fiber densities conditions. condition (or treatment arm) were fixed
were chosen as superior to measures of Assessment of delivery of the interven- effects, and the pair (or block) and
total grams of fat and fiber because the tion. To ascertain the extent to which the treatment-arm-by-block interaction were
densities control for total energy intake. intervention was delivered to the work random effects. The linear effect of the
Because grams of fiber per 1000 calories sites randomized to the intervention con- work site baseline mean for the variable
and servings of fruits and vegetables were dition, a process-tracking system was being analyzed was included as a covari-
skewed toward higher values, these vari- developed to monitor achievement of the ate. The effect of the intervention was
ables were transformed to a logarithmic process objectives specified in the interven- evaluated by the square root of the ratio
scale (ln(x) for fiber and ln(1 + x) for tion protocol.'0 An a priori listing of the of the mean square for treatment to the
fruits and vegetables) in order to make number and type of interventions ex- mean square for treatment by block
the distribution of the data approximately pected at each work site yielded 15 interaction and was compared to a t
normal. The observed means and differ- process objectives aimed at individual distribution. The degrees of freedom for
ences as well as the covariate adjusted change in the two risk factors plus the significance levels presented are based
differences are presented here trans- attendance at the kickoff; additional pro- on the numbers of work-site pairs. There-
formed back into original units. Servings cess objectives targeted change in the fore, the analysis can be regarded as a
of fruits and vegetables were calculated worksite environment.10 A computerized weighted paired t test, made more effi-
on the basis of two questions about usual relational database management system cient since the work sites vary in size. For
intakes of fruit (excluding juice) and documented the types of activities imple- binary response variables (6-month smok-
vegetables (excluding potatoes and sal- mented, the materials distributed, the ing abstinence rate and smoking preva-
ads). The number obtained was added to time and resources expended, and other lence at final survey), mixed model logistic
the responses to items about salad, po- pertinent factors. To assess the delivery of regression was used,23'24 where the center
intervention, the mean proportion of and treatment arm were fixed effects, and
tato, and fruit juice servings (weighted for
serving size).'8 process objectives achieved in each work the block and arm-by-block interaction
site was summed and was divided by the were random effects. The effect of the
Smoking outcomes. Analyses of two number of work sites. intervention was evaluated by the ratio of
smoking outcomes were conducted with In general, process data were re- the restricted maximum likelihood estima-
data from only the three study centers at tion of the regression coefficient for
corded by research intervention staff, with
which smoking interventions were con- the exception of MD Anderson. Because treatment arm to its standard error and
ducted (Brown, Dana-Farber, and MD of its unique use of participatory strate- was compared with a normal distribu-
Anderson): gies and its widely dispersed work sites, tion.23 Secondary analyses were done by
(1) The 6-month abstinence rate MD Anderson relied on work-site employ- adding age, gender, and education level
was measured by self-reported abstinence ees to function as intervention coordina- into the models as covariates.
for the 6 months prior to the survey.10 The tors to implement and document interven- Two alternative analyses were con-
denominator included all individuals who tions, and as a result, some intervention ducted to examine the robustness of the
had been employed by the work site for a activities were underreported. results: (1) the generalized estimating
minimum of 6 months and who either Assessment of receipt of the interven- equation,25 and (2) an analysis of the
were current smokers or had quit smoking tion. With data from the individual em- mean changes within each work site pair,
during the 2-year intervention period. A ployee survey, two indices for each risk by means of a bootstrap t test.26 Although
6-month abstinence rate has been used by factor were created to calculate receipt of data are not presented, the two analyses
many trials as a reasonable approximation the intervention. The first index included had results similar to those of the mixed
of continuous, long-term cessation.20'21 items that measured awareness of inter- models, with the analyses at the work-site
(2) Work site smoking prevalence vention activities. The second index as- level providing somewhat more conserva-
was also measured at baseline and in the sessed activities that were directed toward tive results owing to equal weighting for
final survey. Current smokers were de- behavior change. For both indices, items all work sites, which are of substantially
fined as individuals who had smoked at were scored 1 or 0; the items were added different sizes.
least 100 cigarettes in their lives and and were divided by the total number of Analyses were also conducted to
currently smoked at least 1 cigarette per items. Weighting was considered but was assess differences in the intervention
July 1996, Vol. 86, No. 7 American Journal of Public Health 941
Sorensen et al.
TABLE 1-Nutrition Outcomes at Baseline and Follow-Up at Working Well Trial Work Sites, by Study Center
Brown Dana-Farber Florida MD Anderson All Centers
(n = 20 (n = 24 (n = 24 (n = 40 (n = 108
Work Sites) Work Sites) Work Sites) Work Sites) Work Sites)
% energy from fat
Baseline 35.42 35.83 36.81 38.43 36.71
Follow-up 33.30 33.83 34.36 36.29 34.64
Difference (follow-up minus baseline) -2.12 -2.00 -2.45 -2.14 -2.07
Baseline 35.26 35.54 36.66 39.00 36.70
Follow-up 33.67 34.17 34.71 36.90 35.00
Difference (follow-up minus baseline) -1.59 -1.37 -1.95 -2.10 -1.70
Difference (intervention minus control)
Baseline 0.16 0.29 0.15 -0.57 0.01
Follow-up -0.37 -0.34 -0.35 -0.61 -0.36
Difference (follow-up minus baseline) -0.53 -0.63 -0.50 -0.04 -0.37*
Adjusted differencea (SE) -0.42 (0.27) -0.43 (0.36) -0.56 (0.47) -0.09 (0.30) -0.35* (0.16)
Dietary fiber, g/1000 kcals
Baseline 7.92 8.02 8.72 7.83 8.03
Follow-up 8.76 8.59 9.20 8.33 8.61
Difference (follow-up minus baseline) 0.84 0.57 0.48 0.50 0.58
Baseline 7.89 8.06 8.35 7.80 7.96
Follow-up 8.44 8.45 8.69 8.25 8.41
Difference (follow-up minus baseline) 0.55 0.39 0.34 0.45 0.45
Difference (intervention minus control)
Baseline 0.03 -0.04 0.37 0.03 0.07
Follow-up 0.32 0.14 0.51 * 0.08 0.20*
Difference (follow-up minus baseline) 0.29 0.18 0.14* 0.05 0.13
Adjusted percent increasea (SE) 0.95 (2.1) 2.1 (1.2) 5.6* (2.2) 1.4 (1.5) 1.7 (0.87)
Servings of fruits and vegetables per day
Baseline 2.69 2.71 2.66 2.40 2.60
Follow-up 2.82 2.99 2.97 2.55 2.80
Difference (follow-up minus baseline) 0.13 0.28 0.31 0.15 0.20
Baseline 2.66 2.74 2.60 2.37 2.58
Follow-up 2.66 2.83 2.58 2.38 2.60
Difference (follow-up minus baseline) 0.00 0.09 -0.02 0.01 0.02
Difference (intervention minus control)
Baseline 0.03 -0.03 0.06 0.03 0.02
Follow-up 0.16 0.16* 0.39** 0.17* 0.20***
Difference (follow-up minus baseline) 0.13 0.19* 0.33** 0.14* 0.1 8***
Adjusted percent increasea (SE) 3.9 (3.7) 5.3* (1.7) 1 1.7** (3.5) 5.8* (2.3) 5.6*** (1.3)
Note. SE = standard error.
aBaseline work site mean value is added as a covariate.
*P < .05; **P < .01; ***P < .001.
effect for subgroups with high vs low the statistical significance of the effect of intervention and control work-site means
response rates. MD Anderson's data were the interaction of treatment arm by within blocks. These pairwise differences
not used for this analysis since this study response group. In this analysis, the were regressed against experimental de-
center used a different method of survey blocking effect was dropped from the sign covariates employing bootstrap regres-
administration, which resulted in high model, since some pairs were broken sion methods.26 Two sets of predictor
response rates, thus confounding the when the work sites were classified as variables were employed: an intercept-
response rate with the center effect. Two having high or low response rates. only model and a model with center
response-rate subgroups were created Analyses of the receipt of interven- effects added to the regression. The
( < 65% and > 65%). Analyses examined tion examined the difference between intercept-only model was fit for each
942 American Journal of Public Health July 1996, Vol. 86, No. 7
Work-Site Cancer Prevention
TABLE 2-Observed Proportion of 6-Month Smoking Abstinence and Smoking Prevalence at Final Employee Survey,
by Study Center
Brown Dana-Farber MD Anderson All Centers
(n= 20 (n = 24 (n = 40 (n = 84
Work Sites), % Work Sites), % Work Sites), % Work Sites), %
6-month abstinence rate (% of quitters in total)
Intervention sites 12.3 17.3 11.5 13.8
Control sites 11.2 12.7 12.9 12.3
Dffference (intervention minus control) (95% Cl) 1.04 (-2.5, 6.1) 4.61 * (0.25, 9.6) -1.49 (-4.8, 2.1) 1.53 (-1.0, 3.7)
Smoking prevalence (% of smokers in total)
Intervention sites 24.8 20.3 19.5 21.2
Control sites 24.5 21.4 19.9 21.8
Difference (intervention minus control) (95% Cl) 0.37 (-7.2, 5.5) -1.06 (-3.8, 2.7) -0.42 (-2.5,1.7) -0.66 (-3.0,1.2)
Note. Cl = confidence interval.
alntervention minus control.
*P < .05.
individual study center and for all centers
combined. The model with center effects
added was used to examine whether the TABLE 3-Work Sites' Achievement of Process Objectives, by Study Center
treatment effect differed among study
centers. % Process Objectives Achieved
In all analyses, two-sided tests were Dana- MD All Centers
used and no multiple comparison adjust- Process Objectivea Brown Farber Florida Anderson Combined
ments were made.
Kickoff participation (50% of 48 60 69 84 68
Response Rate Analyses Nutritionb
At baseline, the overall response rate No. work sites 11 12 12 20 55
Interactive kickoff activity (1) 100 92 100 95 96
to the individual survey was 69% (average Posters (4) 100 98 100 51 82
work-site response rate, 72%; study cen- Video/single session 97 83 100 68 84
ter mean range, 61% to 89%). The overall presentation (3)
response rate at the follow-up survey was Self-assessment activity (2) 100 100 100 68 88
Self-help program (2) 100 96 96 45 78
71% (average work-site response rate, Multisession direct education (2) 100 92 100 20 69
75%; study center mean range 68% to Campaign (1) 100 92 92 50 78
86%). The interaction of the response- Total 100 93 98 57 82
rate subgroup (cutpoint, 65%) and the
intervention group indicated no relation- Smokingb
ship between the intervention effects and No. work sites 11 12 NA 20 43
the work site's response rate to the Interactive kickoff activity (1) 100 92 NA 45 72
individual survey (smallest P = 0.24). Posters (4) 98 92 NA 66 81
Video/single session 91 56 NA 62 67
Nutrition presentation (3)
Self-assessment activity (2) 91 96 NA 75 85
For percentage of energy obtained Self-help program (2) 100 100 NA 60 81
from fat consumption, there was a net Multisession direct education (2) 100 58 NA 35 58
Campaign (1) 100 83 NA 50 72
decrease of 0.37 percentage points 97 82 NA 56 74
(P = .033) (see Table 1). Results for each
of the four study centers showed a trend Note. NA = not applicable.
in the desired direction, although only the aExcludes process objectives directed toward environmental change; numbers in parentheses
combined results were statistically signifi- indicate the number of times an activity was to be done.
bConducted in each intervention work site.
Also as shown in Table 1, the net
increase in fiber consumption was only
0.13 g per 1000 kcal (P = .056), since increase of 0.14, P = 0.024) were statisti- intervention sites and was negligible in
control-site employees also increased fi- cally significant. The intake of fruits and most control sites.
ber intake an average of 0.45 g. Results vegetables increased a net average of 0.18 Smoking
for three of the four study centers showed servings per day for all study centers
a trend in the desired direction, although (P .0001). Increased fruit and vegetable
= For the trial overall, there was a
only the results for Florida work sites (net consumption was consistently higher in nonsignificant difference of 1.53% in the
July 1996, Vol. 86, No. 7 American Journal of Public Health 943
Sorensen et al.
behavior after a sustained 2-year interven-
TABLE 4-Bootstrap Regression Estimates for Awareness and Action Indexes, tion. A common intervention protocol
by Study Center was applied to the four study centers,
covering 111 work sites that were ran-
Intervention-Control domly assigned to intervention or control
Awareness/ Action Index Difference SE P 95% Cl conditions. The study had sufficient power
for study center-specific evaluations as
Smoking awareness well as for data from all sites combined.
Brown 0.15 0.09 0.15 -0.05, 0.36
Dana-Farber 0.09 0.06 0.15 -0.02, 0.28 Results presented here focused on the
MD Anderson 0.16 0.04 0.00 0.08, 0.27 combined data reflecting individual behav-
All centers combined 0.14 0.03 0.00 0.08, 0.22 ior changes. The work site was the unit of
Nutrition awareness randomization, intervention, and analysis.
Brown 0.22 0.07 0.02 0.05, 0.40 Analyses were also conducted to elimi-
Dana-Farber 0.22 0.05 0.00 0.10, 0.33 nate the possibility of response-rate bias.
Florida 0.14 0.03 0.01 0.09, 0.24 For the trial as a whole, significant
MD Anderson 0.15 0.04 0.00 0.07, 0.24
All centers combined 0.17 0.02 0.00 0.13, 0.22 results were observed for two of three
Smoking action individual nutrition outcomes. Although
Brown 0.18 0.04 0.01 0.11, 0.30 the percentage of energy obtained from
Dana-Farber 0.12 0.01 0.00 0.10, 0.17 fat consumption decreased by 2.07 per-
MD Anderson 0.18 0.03 0.00 0.13, 0.24 centage points between baseline and
All centers combined 0.13 0.02 0.00 0.10, 0.17 follow-up, the percentage of energy from
Nutrition action fat decreased 1.70 percentage points
Brown 0.31 0.03 0.00 0.25, 0.40 among employees in the control sites. The
Dana-Farber 0.24 0.03 0.00 0.18, 0.29
Florida 0.32 0.02 0.00 0.26, 0.37 level of change observed in control sites
MD Anderson 0.20 0.04 0.00 0.13, 0.29 suggests a modest secular trend in the
All centers combined 0.26 0.02 0.01 0.22, 0.29 reduction of fat consumption. The largest
net effect for nutrition was change in the
Note. SE = standard error; Cl = confidence interval. consumption of fruits and vegetables.
Intake of fruits and vegetables increased
an average of 0.18 servings for all study
centers; one study center obtained a
6-month quit rates between intervention centers. For the trial overall, process difference of approximately a third of a
and control work sites (see Table 2). objective attainment in this risk factor was serving.
We also examined changes in smok- high, with an overall 82% of process The increased intake of fruits and
ing prevalence, although sufficient power objectives attained. Process objectives vegetables may be interpreted, for ex-
for this outcome was not part of the attained for smoking-control activities ample, as a change in one fifth of a serving
design of the Working Well Trial. As were not as high as for nutrition, with an for every individual or as one person in
shown in Table 2, smoking prevalence overall trial attainment of 74%. five having increased consumption by a
dropped considerably in both the interven- full serving. Results reported here repre-
tion (from 24.5% to 21.2%) and control Receipt of Intervention sent changes occurring in the entire
(from 25.8% to 21.8%) conditions in the 2 Cronbach's alpha was computed for work-site population, among employees
years of the trial. each receipt index, and values were found in intervention sites who actively partici-
to be uniformly high (range = .78 to .85). pated in the intervention as well as those
Delivery of the Intervention Table 4 shows the treatment effects on the who were unaware of the program. Addi-
The duration of the intervention receipt of the intervention activities by tional analyses are needed to assess
varied, with the median (and range) study center and for the trial overall. The whether small changes were made by a
number of weeks between the kickoff center-by-treatment-condition interac- majority of respondents or whether there
event and the final survey at the four study tion was tested for each index, and no were larger changes concentrated among
centers as follows: at Brown, 120 (117 to significant interactions were observed. fewer individuals. Although such changes
125); at Dana-Farber, 97 (82 to 120); at For each of the receipt indices, an are small in clinical terms, they may be
Florida, 121 (102 to 123); and at MD intervention-minus-control difference was indicative of a potentially important pub-
Anderson, 97 (80 to 104). significant (P < .001). The data provide lic health impact if they are maintained
Table 3 shows the percentage of the evidence that intervention materials and and are cumulative, when we consider the
process objectives achieved by study cen- activities reached employees in the work large numbers of workers represented by
ter and for all study centers combined. sites; furthermore, these materials and this trial.
The process objective that required that activities were utilized to a greater extent The intervention failed to produce
50% of employees attend the overall than any programs or materials available statistically significant differences be-
kickoff event was met by all study centers in control work sites. tween intervention and control sites for
except Brown, where average work-site measures of smoking. Only at one study
kickoff participation was 48%. Overall center, Dana-Farber, was the difference
Discussion in 6-month quit rates statistically signifi-
participation in the kickoff averaged 68%.
For nutrition, there was high delivery The Working Well Trial measured cant. This significant finding in one site
of the intervention in three of the four changes in eating patterns and smoking suggests that effective work-site smoking
944 American Journal of Public Health July 1996, Vol. 86, No. 7
Work-Site Cancer Prevention
cessation interventions may be possible; were significantly higher in the interven- impact of variations in levels of delivered
there still remains the important chal- tion sites than in control sites. Respon- and received interventions, policy changes,
lenge of determining how best to inter- dents were more likely to report aware- and participatory strategies on the ob-
vene in work sites to promote smoking ness of nutrition- than smoking-related served outcomes will be explored in future
cessation. Additional analyses will exam- intervention activities, perhaps because papers. The interaction of multiple risk
ine the role of work site characteristics in nutrition is important to more people factor interventions, as implemented and
the observed changes. than is smoking. Similarly, a greater sequenced at various work sites, may have
The changes in smoking behavior percentage of nutrition- than smoking- contributed to center differences in ob-
observed in these intervention work sites related process objectives were achieved; served outcomes.
compare favorably with abstinence rates this probably contributed to the observed Finally, the 2-year duration of the
reported in previous minimal intervention differences in significant outcomes for Working Well Trial intervention may not
trials'9 and with reductions in smoking nutrition. The analyses of the receipt of have been long enough to observe the
prevalence reported by similar work site- intervention indices also underline the intended effects, owing to several features
based interventions.27 However, the con- high level of intervention activity in of the intervention design. First, the
trol group's 6-month smoking abstinence control work sites, reflecting the secular Working Well Trial intervention used the
rate (11.2%) is somewhat higher than trend toward increasing health promotion stage-of-change model to develop an
expected, based on several comparable efforts at the work site. intervention that emphasized awareness,
prior reports. For example, a recent The intervention protocol also aimed active change, and maintenance activities
examination of 10 prospective studies of to promote the adoption of work-site at the work site, with the understanding
self-quitters found a median 6-month smoking bans and increase the availability that change occurs in small increments in
abstinence rate of 6.0%.21 Similarly, the of healthy foods at the work place since a cyclical pattern over time.41 A longer
1987 National Health Interview Survey behavior change and its maintenance intervention period may be needed to
found that only 6.0% of ever-smokers who requires a supportive social environ- observe the movement of individuals
attempted to quit in the past 12 months ment.31-33 It was beyond the scope of this through the varying stages of readiness for
were continuously abstinent for more paper to present changes in the work site change. Second, following a participatory
than 3 months.28 At the Working Well environment resulting from this interven- strategies model, employee advisory
baseline, the 4- to 12-month continuous tion. boards were formed to provide worker
abstinence rate was 6.0%."1 This increase Several limitations must be noted in input into intervention planning and
in quit rates within the control group may the interpretation of these results. For implementation. Organization and main-
reflect the high level of smoking-related both nutrition and smoking outcomes, tenance of these boards required substan-
intervention activities at control sites, as self-reports were used to assess change in tial investments of time and in some cases
indicated by the data on the receipt of the the outcome variables. For nutrition, the may have delayed the start-up of interven-
intervention. Some control sites also re- food-frequency questionnaire was previ- tion delivery.
ceived minimal interventions such as the ously validated.-3439 However, the need In conclusion, the Working Well
distribution of posters and brochures. In for biochemical validation of smoking Trial work-site intervention resulted in
particular, smoking policy awareness and cessation in field studies such as this has small but significant decreases in fat
implementation at all work sites was high; been increasingly challenged.28,40 consumption and increases in fruit and
contributing to this trend may have been Work sites were not randomly se- vegetable intake. The potential public
the release, midway through the trial, of lected for inclusion in this study; although health significance of such small changes
the Environmental Protection Agency a wide range of company types and must be debated within the context of the
report on environmental tobacco smoke geographic regions were included, these work site-wide nature of this intervention
as a carcinogen.29 results should be generalized only to and its evaluation. Although significant
More powerful intervention strate- similar work sites. Baseline results sug- smoking cessation differences were not
gies may be needed to increase smoking gested that both intervention and control observed trial-wide, the success of one
cessation rates beyond the current secular work sites were more likely to have had study center in achieving a significant
trend noted in the Working Well Trial health promotion activities than worksites difference in cessation suggests the oppor-
control work sites. This trial used a nationwide," a selection bias that may tunity for future initiatives if the compo-
state-of-the-art intervention model based have contributed to the high level of nents and attributes of successful cessa-
on participatory public health strategies.30 behavior change in the control work sites. tion interventions can be identified. For
Some possible ways of increasing the By necessity, only work sites where there both the nutrition and smoking outcomes,
potency of the intervention are to provide is interest in health promotion are likely sizable secular trends observed during the
programming of increased intensity and to be enrolled in studies such as these. study period may have accounted for
duration, to integrate the health promo- Centers and work sites differed in a some of the reduced magnitude of the
tion intervention with an intervention number of dimensions, which certainly observed differences between interven-
targeting occupational health and safety influenced uniform application of the tion and control groups. Process-tracking
or other health-related concerns of work- intervention. Although a common inter- data supported the overall integrity of the
ers, to incorporate the use of pharmaco- vention protocol was used, there were delivery of the intervention, and worker
logical aids, and to target specific types of variations in the combination of risk data showed significantly greater aware-
work sites or workers, such as disadvan- factors addressed and in the intensity of ness of and participation in nutrition and
taged or young smokers. the contacts with the work sites (e.g., work smoking-control activities in intervention
Consistently, the intervention receipt sites at MD Anderson were scattered sites. If more substantial changes are to be
indices, which compared employee re- across 11 states and therefore had less expected throughout the entire work-site
ports of the level of intervention activities, intensive contact with project staff). The population, future work-site interventions
July 1996, Vol. 86, No. 7 American Journal of Public Health 945
Sorensen et al.
may need to reexamine the intensity and Allan Williams. Programmers: Stephanie conceptual and practical considerations.
duration of interventions, the sequencing Stafford, Peter Dueber, Lynette Brown (1990- In: Richmond R, ed. Interventions for
and timing of environmental interven- 1991). Cost analysts: Shelly Hager, Addy Smokers: An International Perspective.
Tseng, Carole Shaw (1990-1993). Support Baltimore, Md: Williams & Wilkins, 1994:
tions, and the potential synergism of staff: Catherine Cohen, Ellen Powell (1991- 137-170.
multiple risk factor interventions. El 1992), Estella Bennett (1990-1992), Jenny 6. Erfurt JC, Foote A, Heirick MA. Worksite
Devaney (1990-1991). wellness programs: incremental compari-
Working Groups and Subcommittees son of screening and referral alone, health
Acknowledgments Steering Committee: Ed Lichtenstein education, follow-up counseling, and plant
This work was supported by a Cooperative (Chair), David B. Abrams, Ellen R. Gritz, organization.Am JHealthPromotion. 1991;
Agreement from the National Cancer Institute, James Grizzle, Jerianne Heimendinger, Glo- 5:438-448.
Grants UOI CA51687, UO1 CA61771, UO1 rian Sorensen, Jill Varnes, W. Bryant Boutwell 7. Jeffery RW, Forster JL, French SA, et al.
CA51686, UO1 CA516888, and P01 CA50087. (1991-1993), Michael Ericksen (1989-1991), The Healthy Worker Project: A work-site
The investigators and committee mem- Ellen Gritz (1993-1994). intervention for weight control and smok-
bers for the Working Well Trial are as follows: Data Managers/ Tracking Coordinators: Al- ing cessation. Am J Public Health. 1993;83:
National Cancer Institute (NCI), Division lan Williams (Chair), Peter Dueber, James 395-401.
of Cancer Prevention and Control. Program Eldridge, Kathy Galloway, Susan Gucken- 8. Sorensen G, Morris D, Hunt MK, et al.
director: Jerianne Heimendinger. Statisticians: berger, Jean Hsieh, Cynthia Lostoski, Steve The effects of a worksite nutrition interven-
Charles Brown, Don Corle. Fellows: Nancy Potter, Stephanie Stafford, David Vampola, tion on employees' dietary habits: the
Cotugna (1990), Susan Sullivan (1990-1991). Kate Lapane (1989-1992). Treatwell program. Am J Public Health.
Study Centers Evaluation Working Group: James Grizzle 1992;82:877-880.
Brown University School of MedicinelThe (Chair), Don Corle, James Eldridge, David 9. Glasgow RE, Terborg JR, Hollis JF,
Miriam Hospital. Principal investigator: David Miller, Anne Stoddard, Wayne Velicer. Severson HH, Boles SM. Take Heart:
B. Abrams. Co-principal investigator: Lois Nutrition Working Group: Karen Glanz results from the initial phase of a work-site
Biener. Co-investigators: Karen M. Emmons, (Chair), James Hebert, Jerianne Heimend- wellness program. Am J Public Health,
Laura Linnan. Statisticians: Joe Fava, Wayne inger, Mary Kay Hunt, Alan Kristal, Laura 1995;85:209-216.
Velicer. Project director: Laura Linnan. Inter- Linnan, Ruth Patterson, Claudia Probart. 10. Abrams DB, Boutwell WB, Grizzle J,
vention specialists: Edward Galuska, Mary Additional Nutrition Intervention Subcommittee Heimendinger J, Sorensen G, Varnes J.
Lynne Hixson, Sheila Jacobs, Suzanne Mori- Participants: Mary Lynne Hixson, Judith Phill- Cancer control at the workplace: the
arty. Data specialist: David Vampola, Kate ips, Patricia Pillow. Working Well Trial. Prev Med. 1994;23:
Lapane (1989-1992). Research associate: Ja- Organizational Variables Working Group: 1-13.
nine Costa. Administrative assistant: Mark Lois Biener (Chair), Susan Kinne (Cochair), 11. Heimendinger J, Feng Z, Emmons K, et al:
Morgenstern. Karen Basen-Engquist, Jerianne Heimend- The Working Well Trial: Baseline dietary
Dana-Farber Cancer Institute! University of inger, Glorian Sorensen, Jill Varnes. and smoking behaviors of employees and
Massachusetts Medical SchooL Principal investi- Process Evaluation Working Group: Glo- related worksite characteristics. Prev Med.
gator: Glorian Sorensen. Co-principal investiga- rian Sorensen (Chair), Karen Basen-Engquist, 1995;24:180-193.
tors: Jay Himmelstein, Judith Ockene. Co- Shelly Hager, Jean Hsieh, Mary Kay Hunt, 12. Abrams DB, Elder JH, Carleton RA, et al.
investigators: Katharine Hammond, James Laura Linnan, Cynthia Lostoski, David Miller, Social learning principles for organiza-
Hebert, Ruth Palumbo, Anne Stoddard. Project Stephanie Stafford, Addy Tseng, Nicole Urban tional health promotion: an integrated
director: Mary Kay Hunt. Evaluation coordina- (1990-1992). approach. In: Cataldo MF, Coates TJ, eds.
tor: Jean Hsieh. Intervention coordinators: Participatory Strategies Working Group: Health and Industry: A Behavioral Medicine
Lynda Graham-Meho, Elizabeth Harden, Jane Claudia Probart (Chair), Karen Basen-En- Perspective. New York, NY: Wiley; 1986:
Ellen Thompson. Nutrition coordinator: Judith gquist, Karen Emmons, Jerianne Heimend- 28-51.
Phillips. Industrial hygienist: Richard Young- inger, Mary Kay Hunt, Laura Linnan, Ingrid 13. Abrams DB. Conceptual models to inte-
strom. Process tracking coordinator: Steve Nielsen, Beti Thompson. grate individual and public health interven-
Potter. Administrative assistant: Elizabeth Farr. Publications and Presentations Subcommit- tions: the example of the workplace. In:
University of Florida at Gainesville. Princi- tee: Jill Varnes (Co-Chair), Karen Emmons Henderson M, chair. Proceedings of the
pal investigator: Jill Varnes. Co-principal inves- (Cochair 1991-1992, 1994), Ziding Feng, Ellen International Conference on Promoting Di-
tigator: Claudia Probart. Co-investigator, medi- R. Gritz, Jerianne Heimendinger, Anne Stod- etary Change in Communities. Seattle, Wash:
cal consultant: David Schapira. Statistician: dard, David B. Abrams (Chair 1989-1991), W. The Fred Hutchinson Cancer Research
David Miller. Intervention coordinators: Jean- Bryant Boutwell (Cochair, 1992-1993). Center, 1991:173-194.
ine Lahey, Steve Livesay, Valerie Studnick Smoking Working Group: Karen Emmons 14. Rothman J. Three models of community
(1990-1992). Interventionist/ tracking coordina- (Chair), Carlo DiClemente (Chair 1989-1991), organizational practice. In: Cox FM, ed.
tor: Susan Guckenberger. Office manager, data Jerianne Heimendinger, Judith Ockene, Beti Strategies of Community Organization. Itasca,
manager: Kathy Galloway. Thompson, David B. Abrams (1989-1992). Ill: FE Peacock Publishers, 1970.
University of Texas M.D. Anderson Cancer 15. Bracht JF, ed. Health Promotion at the
Center. Principal investigator: Ellen R. Gritz, References Community Level. Newbury Park, Calif:
W. Bryant Boutwell (1991-1993), Michael 1. American Cancer Society. Cancer Facts Sage Publications, 1990.
Ericksen (1989-1991). Co-principal investiga- and Figures: 1993. Atlanta, Ga: American 16. Kristal AR, Shattuck AL, Williams AE.
tor: Carlo DiClemente. Co-investigators: Rob- Cancer Society; 1993. Food frequency questionnaires for diet
ert Chamberlain, Guy Parcel, Karen Basen- 2. Doll R, Peto R. The causes of cancer: intervention research. In: Proceedings of the
Engquist, John Foryet, Larry Whitehead, Alan Quantitative estimates of avoidable risks of 17th National Nutrient Data Bank Confer-
Herd, Jane Mayfield. Project coordinator: cancer in the United States today. JNCI ence, June 7-10, 1992. Washington, DC:
Scott R. Cummings. Intervention specialists: 1981;66:1191-1308. International Life Sciences Institute; 1994:
Scott R. Cummings, Ingrid Nielsen, James 3. Greenwald P, Sondik E, eds. Cancer 110-125.
Eldridge, Bill Mann. Consultants: Karen Glanz, Control Objectives for the Nation, 1985- 17. Glanz K, Patterson RE, Kristal AR, Di-
Robert Butera. 2000. Washington, DC: National Cancer Clemente CC, Heimendinger J, Linnan L.
Coordinating Center: Fred Hutchinson Cancer Institute; 1986. Stages of change in adopting healthy diets:
Research Center. Principal investigator: James 4. Klesges RC, Cigrang JA. Worksite smok- fat, fiber, and correlates of nutrient intake.
Grizzle. Co-investigators: Ziding Feng, Susan ing cessation programs: clinical and meth- Health Educ Q. 1994;21(4):73-93.
Kinne, Alan Kristal, Beti Thompson, Nicole odological issues. In: Hersen M., Eisler 18. Block G. Health Habits and History Question-
Urban (199(}1992). Fellow: Ruth Patterson. R.M., Miller R.M., eds. Progress in Behav- naire: Diet History and Other Risk Factors.
Coordinating center manager: Sonya Olsen, ioral Modification. Vol 23. Newbury Park, Bethesda, Md: National Cancer Institute;
Gwen Glaefke (1989-1992). Statistical re- Calif: Sage Publications; 1988:37-61. 1988.
search associates: Cynthia Lostoski, Steve 5. Abrams D, Emmons K, Linnan L, Biener 19. Schakel SF, Sievert TA, Buzzard IM.
Thomson, Dale McLerren. Database manager: L. Smoking cessation at the workplace: Sources of data for developing and main-
946 American Journal of Public Health July 1996, Vol. 86, No. 7
Work-Site Cancer Prevention
taining a nutrient database. J Am Diet Centers for Disease Control; 1990. DHHS 35. Block G, Woods M, Potosky A, Clifford C.
Assoc. 1988;88:1260-1271. publication (CDC) 90-8416. Validation of a self-administered diet his-
20. COMMIT Research Group. Community 29. Respiratory Health Effects of Passive Smok- tory questionnaire using multiple diet
Intervention Trial for Smoking Cessation ing: Lung Cancer and Other Disorders: The records. J Clin Epidemiol. 1990;43:1327-
(COMMIT): summary of design and inter- Report of the Environmental Protection 1335.
vention. JNCI. 1991;83:1620-1628. Agency. Washington, DC: US Environmen- 36. Sobell J, Block G, Koslowe P, Tobin J,
21. Cohen S, Lichtenstein E, Prochaska JO, et tal Protection Agency; 1993. Andres R. Validation of a retrospective
al. Debunking myths about self-quitting. 30. Thompson B, Kinne S. Social change questionnaire assessing diet 10-15 years
Am Psychol. 1989;44:1135-1365. theory: applications to community health. ago.AmJEpidemioL 1989;130:173-187.
22. Searle R, Casella G, McCulloch CE. In: Bracht N, ed. Health Promotion at the 37. Block G, Thompson FE, Hartman AM,
Variance Components. New York, NY. Community Level. Newbury Park, CA: Sage Larkin FA, Guire KM. Comparison of two
John Wiley, 1992. Publications; 1990;45-65. dietary questionnaires validated against
23. Schall R. Estimation in generalized linear 31. Thompson B, Wallack L, Lichenstein E, multiple dietary records collected during a
models with random effects. Biometrika. Pechacek T, for the COMMIT Research 1-year period. JAAL4. 1992;92:686-693.
1991;78:719-727. Group. Principles of community organiza- 38. Coates RJ, Eley JW, Block G, et al. An
24. Wolfinger R. Laplace's approximation for tion and partnership for smoking cessation evaluation of a food frequency question-
nonlinear mixed models. Biometrika. 1993; in the Community Intervention Trail for naire for assessing dietary intake of specific
80:791-795. Smoking Cessation (COMMIT). Intl Q carotenoids and Vitamin E among low-
25. Liang KY, Zeger SL. Longitudinal data Community Health Educ. 1991;11:187-203. income Black women. Am J Epidemiol.
analysis using generalized linear models. 32. Green L, McAlister A. Macro-intervention 1991;134:658-671.
Biometrika. 1986;73:13-22. to support health behavior change: some 39. Mares-Perlman JA, Klein BEK, Klein R,
26. Efron B, Tibshirani RJ. An Introduction to theoretical perspectives and practical reflec- et al. A diet history questionnaire ranks
the Bootstrap. New York, NY. Chapman & tions. Health Educ Q. 1994;11:322-339. nutrient intakes in middle-aged and older
Hall, 1993. 33. Wallack L, Wallerstein N. Health educa- men and women similarly to multiple food
27. DiClemente CC, Prochaska JO, Fairhurst tion and prevention: designing community records.JNutrEduc. 1993;123:489-501.
S, Velicer WF, Velasquez M, Rossi J. The initiatives. Intemational Q Community 40. Velicer WF, Prochaska JO, Rossi JS, Snow
process of smoking cessation: an analysis of Health Educ. 1986-87;7:319-342. M. Assessing outcome in smoking cessa-
precontemplation, contemplation and 34. Block G, Hartman AM. Dietary assess- tion studies. Psychol Bull. 1992;1 11:23-41.
preparation.JConsultClinPsychol. 1991;59: ment methods. In: Moon TE, Micozzi MS, 41. Prochaska JO, DiClemente CC, Norcross
295-304. eds. Nutrition and Cancer Prevention, Inves- JC. In search of how people change:
28. The Health Benefits ofSmoking Cessation:A tigating the Role of Micronutrients. New applications to addictive behaviors. Am
Report of the Surgeon General. Atlanta, Ga: York, NY: Marcel Dekker, 1989:159. Psychol. 1992;47:1102-1114.
July 1996, Vol. 86, No. 7 American Journal of Public Health 947