Longitudinal Relationship between Elapsed ... - Wiley Online Library

Longitudinal Relationship between Elapsed Time in the Action Stages of Change and Weight Loss Everett E. Logue,* David G. Jarjoura,† Karen S. Sutton,‡ William D. Smucker,* Kristin R. Baughman,§ and Cynthia F. Capers¶

Abstract LOGUE, EVERETT E., DAVID G. JARJOURA, KAREN S. SUTTON, WILLIAM D. SMUCKER, KRISTIN R. BAUGHMAN, AND CYNTHIA F. CAPERS. Longitudinal relationship between elapsed time in the action stages of change and weight loss. Obes Res. 2004;12:1499-1508. Objective: The objective of this study was to examine the longitudinal relationship between the elapsed time in the action and maintenance stages of change for multiple target behaviors and weight loss or gain. Research Methods and Procedures: The research design was a prospective cohort study of overweight and obese primary care patients randomized to an obesity management intervention based on the Transtheoretical Model and a chronic disease paradigm. The target behaviors included increased planned exercise and usual physical activity, decreased dietary fat, increased fruit and vegetable consumption, and increased dietary portion control. The participants were 329 middle-aged men and women with elevated body mass indices recruited from 15 primary care practices in Northeastern Ohio; 28% of the participants were African Americans. The main outcomes were weight loss (5% or more) or weight gain (5% or more) after 18 or 24 months of follow-up. Results: There were significant (p ⬍ 0.05) longitudinal relationships between the number of periods (0 to 4) in

Received for review October 8, 2003. Accepted in final form July 8, 2004. The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. *Department of Family Practice, Summa Health System, Akron, Ohio; †Office of Biostatistics, Northeastern Ohio Universities College of Medicine, Rootstown, Ohio; ‡Department of Psychology, Oberlin College, Oberlin, Ohio; §Louis Stokes Veteran Affairs Medical Center, Brecksville, Ohio; and ¶College of Nursing, University of Akron, Akron, Ohio. Address correspondence to Everett Logue, Department of Family Practice, Summa Health System, 525 East Market Street, Suite 290, Akron, OH 44309-2090. E-mail: [email protected] Copyright © 2004 NAASO

action or maintenance for each of the five target behaviors, or a composite score taken across the five target behaviors, and weight loss. In all cases, there was a significant (p ⬍ 0.05) stepped (graded) relationship between the time in action or maintenance and weight loss (or gain). Discussion: The data support the concept of applying the Transtheoretical Model to the problem of managing obesity in primary care settings. The remaining challenge is to identify those factors that reliably move patients into the action and maintenance stages for long periods. Key words: Transtheoretical Model, weight loss, longitudinal study, primary care

Introduction Obesity is epidemic in the general population because of a widespread polygenetic susceptibility and an obesogenic environment that encourages multiple behaviors that frequently produce a small positive energy balance in many individuals (1,2– 4). The Transtheoretical Model has previously been applied to weight-loss related behaviors, but most studies have been cross-sectional (5–23). There are relatively few longitudinal data describing the empirical relationship between the elapsed time in the action and maintenance stages of change for multiple target behaviors and weight loss or gain (24 –29). Thus, the objective of this study was to examine the longitudinal relationship between stages of change (SOC)1 for increased planned exercise, usual physical activity, portion control, fruit and vegetable consumption, or decreased dietary fat and weight loss or gain. We hypothesized that overweight or obese primary care patients who spend more time in the action or mainte-

1 Nonstandard abbreviations: SOC, stage(s) of change; REACH, Reasonable Eating and Activity to Change Health; RR, risk ratio.

OBESITY RESEARCH Vol. 12 No. 9 September 2004

1499

Time in Action and Weight Loss, Logue et al.

nance SOC for weight loss-related behaviors would be more likely to lose weight, or would be less likely to gain weight, than patients who spend more time in the preaction SOC. (18,30) The data for this study were obtained from 329 overweight or obese primary care patients who were assigned to a minimal intervention in a randomized clinical trial. Staging data were not available for 336 patients assigned to the comparison group.

Research Methods and Procedures The Reasonable Eating and Activity to Change Health (REACH) trial was a 24-month randomized, parallel group, weight management trial conducted at 15 primary care practices in Northeastern Ohio between July 1998 and December 2002. The principle objective of the trial was to test a hypothesis concerning the effectiveness of a minimal cognitive-behavioral intervention based on the chronic disease paradigm (18). Participants were recruited when they inquired about the study after either talking to their physician or reading study brochures, posters, or informational letters. All participants provided written informed consent according to a procedure approved by multiple institutional review boards. Primary care patients, 40 to 69 years old, with elevated BMIs (⬎27) or elevated waist-to-hip ratios (⬎0.950 for men or 0.800 for women) were eligible for the study. Exclusion criteria were no access to a telephone, low literacy, pregnancy, lactation, less than 6 months postpartum, use of a wheel chair, or severe heart or lung disease. After the informed consent process was completed and baseline data were collected, participants were randomized into permuted blocks of 10 patients. Participants assigned to the experimental group had the same basic care as the comparison group (31,32). In addition, the experimental group completed periodic SOC assessments for five target behaviors and received SOC mailings and telephone calls from a weight loss advisor (30). The target behaviors were increased planned exercise (deliberately walking for exercise), increased usual physical activity (walking incidental to paid employment or domestic chores), increased dietary portion control, decreased dietary fat, and increased fruit and vegetable consumption. SOC for the five target behaviors was assessed by a set of multi-item algorithms that included a decision rule based on order statistics (23). Instead of the mean response to the items, the response at the 33rd percentile from the lowest stage was used as the SOC estimate. The rationale for this approach has been previously described in detail (23). In brief, the measurement goal was to assess SOC for multiple dietary (e.g., regularly drinking skim milk instead of whole milk) and exercise-related (e.g., regularly using the stairs instead of elevators) behaviors that are related to weight loss. We chose the 33rd percentile to highlight problematic target areas in which intervention could be made (see Appendix A, 1500


available at www.obesityresearch.org). The entire 24-month follow-up period was first divided into four subperiods: 0 to 6 months, ⬎6 to 12 months, ⬎12 to 18 months, and ⬎18 to 24 months. Then, the 6-, 12-, 18-, and 24-month SOC data for each behavior and patient were examined to ascertain whether the patient was largely in action or maintenance for a given target behavior for the preceding 6-month period. General physical and mental health, self-efficacy, social desirability, and psychopathology data were also collected (33,34). Anthropometric assessments were scheduled at 0 (baseline), 6, 12, 18, and 24 months. Longitudinal weight change was ascertained by comparing baseline and 24-month weights, or baseline and 18-month weights, if the 24-month weight was not available. Weight was measured with portable electronic scales that were periodically checked against a standard balance beam scale. Patients were weighed in their street clothes but without heavy outer garments. Measured weights were adjusted downward by 1.0 lb when the patients did not take off their shoes as requested. Weights from primary care records were substituted for some measured weights when the latter were missing. Seventy percent of the participants had a measured weight at 18 or 24 months of follow-up, but an additional 20% of participants had a chart weight at 18 or 24 months of follow-up, resulting in only 10% of our patients with missing weights at 18 or 24 months. Pearson correlations between measured weights and chart weights, when both were available, averaged 0.99 across the four postbaseline measurement points. Statistical Analysis The empirical distribution of relative weight change was divided into three groups: weight loss (5% or more), no weight change (⬍5% loss or gain), or weight gain (5% or more) for analysis purposes. We used these cut-off points because weight change of ⬍5% over 24 months is not likely to affect clinically important risk factors such as blood pressure, blood lipids, or blood glucose (35). In addition, a 5% or greater change cannot be explained away by natural variation or instability of weight, whereas a change of ⬍5% over 24 months may not be meaningful. We derived a time in action or maintenance SOC variable by counting the number of periods a subject was in action or maintenance, thus yielding scores that ranged from 0 to 4 for each target behavior. We also derived a composite score by summing the number of periods in action or maintenance across target behaviors. The composite score ranged from 0 (zero periods ⫻ five behaviors) to 20 (four periods ⫻ five behaviors). The relationship between time in action or maintenance for each target behavior and weight change was summarized in five ⫻ three contingency tables (Proc Freq, SAS version 8.2; SAS Institute Inc., Cary, NC). Ordinal logistic regres-


sion analysis was used to model the relation between the composite score and the weight change triplet (Proc Logistic with an extra intercept term) while adjusting for baseline covariables. Statistical significance was set at ␣ ⫽ 0.05.

Results The SOC and weight change recodes yielded nonmissing data for 294 (89%) of 329 patients. Missing baseline data were due to incomplete questionnaires. The majority of the missing follow-up weight data were due to patient refusals to attend the scheduled assessment and missing chart weights. However, there were no important demographic or medical history differences between the 294 patients with nonmissing data and the 329 randomized patients. Participant Characteristics Table 1 summarizes the characteristics of the REACH trial participants who were randomized to the experimental intervention. Eighty-four percent of the participants were between the ages of 40 and 59 years, 30% were men, 28% identified themselves as African American, and 45% had BMIs over 34.9 kg/m2. Mean SF12 physical health and mental health scores appeared to be slightly lower than the norms for similarly aged individuals from the general population (36). Table 2 shows the baseline SOC distribution for the five target behaviors. Preparation was the modal stage for all behaviors, which is consistent with the status of the participants as volunteers for a weight management trial. The 24-month weight change distribution was symmetric; 23% experienced a weight loss of 5% or more of their baseline weight, and 18% experienced a weight gain of 5% or more of their baseline. The fact that 60% of the patients were within 5% of their baseline weight at the end of follow-up is presumably a function of the obesogenic environment interacting with participant characteristics and the nature of the experimental REACH trial intervention (1,2). The regaining of initial weight losses over time is a wellrecognized problem (37). Moreover, as noted earlier, the experimental intervention was minimal by design (31). SOC for Dietary Targets and Weight Change Table 3 describes the longitudinal relationship between SOC for increased portion control, less dietary fat, or more fruits and vegetables and weight change. In particular, the top of the table shows the significant [␹2 (df ⫽ 8) ⫽ 35.0; p ⬍ 0.0001) nominal relationship when the ordering across periods is ignored. The linear trend in weight loss incidence, which considers the ordering across periods in action for portion control, explains one-half of this, and was also significant (␹2 ⫽ 18.3; p ⫽ 0.00002). When the number of periods in the action or maintenance stage increased from 0 to 4, the incidence of weight loss (5% or more of baseline

Table 1. Characteristics of patients randomized to the stage of change intervention N ⴝ 329 Age, gender, ethnicity 40 to 49 years 50 to 59 years 60 to 69 years Men African Americans Baseline BMI Group 25 to 29.9 kg/m2 30 to 34.9 35 to 39.0 40.0 or more Medical History Hypertension Elevated blood cholesterol (Osteo) arthritis Stomach problems Diabetes mellitus Prior/current psychotropic meds Depression (PRIME-MD) Anxiety Eating disorder Prior attempts to lose weight MD said to lose weight Prior commercial program Physical health score (SF12) Mental health score (SF12) Self-efficacy for healthy eating Self-efficacy for exercise Social desirability

Frequency

(%)

139 138 52 97 88

(42.3) (42.0) (15.8) (29.5) (27.8)

59 119 69 79

(18.1) (36.5) (21.2) (24.2)

138 107 106 73 41 85 12 26 64 306 246 147 Mean 45.3 47.9 136.9 28.9 20.0

(44.1) (35.6) (35.3) (24.7) (13.9) (25.9) (3.7) (8.0) (19.6) (96.8) (78.9) (47.0) (SD) (9.8) (10.9) (35.2) (11.0) (5.8)

PRIME-MD, Primary Care Evaluation of Mental Disorders.

weight) increased from 9.9% to 37.9 [risk ratio (RR) ⫽ 3.84; 95% confidence interval ⫽ (1.9, 7.9)]. Conversely, the incidence of weight gain (5% or more of baseline) decreased from 33.3% to 9.1% [RR ⫽ 0.27; 95% confidence interval ⫽ (0.1, 0.6)] in a similar comparison. Figure 1 shows the incidence of weight loss or gain by the number of periods in action or maintenance for portion control. The middle of Table 3 shows that the nominal test for weight change and the dietary fat variable just missed being significant [␹2 (df ⫽ 8) ⫽ 13.2; p ⫽ 0.10], but the linear trend explained one-half of this and was significant (␹2 ⫽ 6.4; p ⫽ 0.01). The average incidence of weight loss inOBESITY RESEARCH Vol. 12 No. 9 September 2004

1501


Table 2. Baseline stage of change for the five target behaviors Baseline stage of change frequency (%) Target behavior

Precontemplation

Contemplation

Preparation

Action*

Total

Portion control Dietary fat Fruits and vegetables Planned exercise Usual physical activity†

41 (12.5) 45 (13.8) 17 (5.2) 6 (1.8) 75 (22.9)

70 (21.4) 81 (24.9) 59 (18.1) 30 (9.2) 68 (20.8)

174 (53.2) 148 (45.4) 159 (48.8) 226 (69.1) 104 (31.8)

42 (12.8) 52 (15.9) 91 (27.9) 65 (19.9) 80 (24.5)

327 (100.0) 326 (100.0) 326 (100.0) 327 (100.0) 327 (100.0)

* Patients indicating that they were in maintenance were counseled as if were in action because they were starting a weight loss program. † Walking and stair climbing at home, at work, or while shopping.

creased from 17.9% to 35.5% [RR ⫽ 1.98; 95% confidence interval ⫽ (1.3, 3.0)] as the average periods in action or maintenance for dietary fat increased from ⬃1.5 to 4, whereas the average incidence of weight gain decreased from 20.2% to 10.5% (p ⫽ 0.06). The bottom of Table 3 shows that weight change was also associated with the number of periods in action or mainte-

nance for increased servings of fruits and vegetables. Both the test statistic for the contingency table [␹2 (df ⫽ 8) ⫽ 18.2; p ⫽ 0.02] and the linear trend that explains more than one-half of the nominal result were significant (␹2 ⫽ 10.8; p ⫽ 0.001). The incidence of weight loss increased from 9.7% to 28.7% as the number of periods in action or maintenance for fruits and vegetables increased from 0 to 4

Table 3. Action or maintenance stages for dietary targets and weight change at 24 months Periods in action or maintenance For increased portion control 0 1 2 3 4 For less dietary fat 0 1 2 3 4 For more fruits and vegetables 0 1 2 3 4

Weight gain (%)

No change (%)

Weight loss (%)

Total [%(n)]

33.3* 8.9 18.2 12.8 9.1

56.8 71.4 65.9 55.3 53.0

9.9 19.6 15.9 31.9 37.9

100.0 (81) 100.0 (56) 100.0 (44) 100.0 (47) 100.0 (66)

22.3† 21.3 12.9 19.6 10.5

59.6 61.7 71.0 60.9 54.0

18.1 17.0 16.1 19.6 35.5

100.0 (94) 100.0 (47) 100.0 (31) 100.0 (46) 100.0 (76)

25.0‡ 25.7 18.9 10.2 12.9

65.3 54.3 64.9 55.1 58.4

9.7 20.0 16.2 34.7 28.7

100.0 (72) 100.0 (35) 100.0 (37) 100.0 (49) 100.0 (101)

* ␹2 (df ⫽ 8) ⫽ 35.0; p ⬍ 0.0001. † ␹2 (df ⫽ 8) ⫽ 13.2; p ⫽ 0.10. ‡ ␹2 (df ⫽ 8) ⫽ 18.2; p ⫽ 0.02.

1502



Figure 1: Time in action stages for portion control and weight change. N ⫽ 294; Chi-square (df ⫽ 8) ⫽ 35.0; P value ⬍ .0001.

[RR ⫽ 2.95; 95% confidence interval ⫽ (1.4, 6.4)]. Similarly, the incidence of weight gain decreased from 25.0% to 12.9% as the number of periods increased [RR ⫽ 0.51; 95% confidence interval ⫽ (0.3, 0.98)]. SOC for Exercise and Activity Targets and Weight Change Figure 2 and the top of Table 4 show the incidence of weight loss or gain by the number of periods in action or maintenance for planned physical exercise. The nominal test statistic from the contingency table was significant [␹2 (df ⫽ 8) ⫽ 18.5; p ⫽ 0.02], as was the linear trend test (␹2 ⫽ 6.0; p ⫽ 0.01). The average incidence of weight loss increased from 15.8% to 31.0% as the average number of periods in action or maintenance for exercise increased from ⬃0.5 to ⬃3.0 [RR ⫽ 1.97; 95% confidence interval ⫽ (1.3, 3.0)]. Conversely, the average incidence of weight gain decreased from 21.8% to 12.4% in a similar comparison [RR ⫽ 0.57; 95% confidence interval ⫽ (0.3, 0.98)]. Finally, the second panel of Table 4 shows the incidence of weight loss (or gain) associated with the number of periods in action or maintenance for usual physical activity. Both the contingency table test [␹2 (df ⫽ 8) ⫽ 23.9; p ⫽ 0.002] and the linear trend test were significant (␹2 ⫽ 8.7; p ⫽ 0.003). The incidence of weight loss increased from 12.9% to 30.8% as the number of periods in action or maintenance for usual activity increased from 0 to 4 [RR ⫽ 2.4; 95% confidence interval ⫽ (1.2, 4.6)]. The incidence of weight gain decreased from 27.1% to 9.2% as the number of periods increased [RR ⫽ 0.34; 95% confidence interval ⫽ (0.2, 0.8)]. Multiple Target Behaviors Figure 3 and the bottom of Table 4 show the incidence of weight loss or gain by successive composite score groups (0, 1 to 5, 6 to 11, 12 to 16, 17 to 20). Patients with composite scores of 0 were not in action or maintenance for any target behavior during follow-up, whereas patients with

Figure 2: Time in action stages for exercise and weight change. Chi-square (df ⫽ 8) ⫽ 18.5; P value ⫽ .02.

composite scores of 17 to 20 were in action or maintenance for most behaviors during most periods. The contingency table test was significant [␹2 (df ⫽ 8) ⫽ 20.6; p ⫽ 0.008], and the linear trend explains almost one-half of this and was significant (␹2 ⫽ 9.9;p ⫽ 0.002). The incidence of weight loss increased from 7.5% to 32.7 [RR ⫽ 4.36; 95% confidence interval ⫽ (1.4, 13.8)], and the incidence of weight gain decreased from 35.0% to 10.9% [RR ⫽ 0.31; 95% confidence interval ⫽ (0.1, 0.7)] when the number of periods and behaviors in action or maintenance increased from 0 to ⬃18.5. Regression Analyses Ordinal logistic regression analysis (of weight loss, no change, and weight gain) was undertaken to address the possibility of confounding by different distributions of selected baseline characteristics and to allow for the possibility of interactions between these baseline characteristics and the composite score variable. Table 5 shows the results from a logistic model with the main effects of all of the Table 1 variables included in the model. According to this analysis, the estimated odds ratio for a four-unit increase (15th percentile to the 90th percentile) in the grouped composite SOC score [odds ratio ⫽ 5.7; 95% confidence interval ⫽ (2.4, 13.5)] was largely unaffected by adjustments for the main effects of 21 extraneous variables. The extraneous variables were age group, gender, ethnicity, BMI group, hypertension, high cholesterol, arthritis, stomach problems, diabetes, psychotropic medicines, depression, anxiety, eating disorder, prior weight loss attempts, physician advice, prior experience with commercial programs, physical or mental health scores, healthy eating or exercise self-efficacy, or social desirability. Anxiety (according to the PRIME-MD instrument) was the only variable that was significantly associated with weight loss (p ⫽ 0.02) other than an increase in the composite SOC score (p ⬍ 0.0001). OBESITY RESEARCH Vol. 12 No. 9 September 2004

1503


Table 4. Action or maintenance stages for exercise, activity, or a dietary and exercise composite score, and weight change at 24 months Periods in action or maintenance For more planned exercise 0 1 2 3 4 For more usual activity 0 1 2 3 4 Dietary and exercise composite score 0 1 to 5 6 to 11 12 to 16 17 to 20

Weight gain (%)

No change (%)

Weight loss (%)

Total [%(n)]

25.5* 14.6 8.2 21.9 10.4

60.0 67.3 55.1 50.0 62.5

14.6 18.2 36.7 28.1 27.1

100.0 (110) 100.0 (55) 100.0 (49) 100.0 (32) 100.0 (48)

27.1† 14.3 26.3 10.5 9.2

60.0 71.4 39.5 63.2 60.0

12.9 14.3 34.2 26.3 30.8

100.0 (85) 100.0 (49) 100.0 (38) 100.0 (57) 100.0 (65)

35.0‡ 23.8 10.6 14.3 10.9

57.5 60.3 63.6 60.0 56.4

7.5 15.9 25.8 25.7 32.7

100.0 (40) 100.0 (63) 100.0 (66) 100.0 (70) 100.0 (55)

* ␹2 (df ⫽ 8) ⫽ 18.5; p ⫽ 0.02. † ␹2 (df ⫽ 8) ⫽ 23.9; p ⫽ 0.002. ‡ ␹2 (df ⫽ 8) ⫽ 20.6; p ⫽ 0.008.

The presence of anxiety appeared to increase, rather than decrease, weight loss. Given the number of tests, this may be a chance finding. A model with main effects and cross-product terms between the grouped composite score and gender, ethnicity, BMI group, and age group did not yield any strong evidence of interactions for these variables. However, there was some evidence of interaction between the grouped composite score variable and the dichotomized at the median (SF12) mental health variable (p ⫽ 0.004). The model included variables for the grouped composite score, mental health, age group, gender, ethnicity, BMI group, depression, anxiety, eating disorder, and (SF12) physical health, and crossproduct terms between the grouped composite scores and ethnicity, gender, and the two SF12 variables. The main effect of the dichotomized mental health variable was negative with respect to weight loss, whereas the cross-product term with the score variable was positive. This suggests that the trend between the SOC composite score and weight loss was stronger in those patients with better mental health. This finding emerged from an exploratory analysis, so its significance must be interpreted cautiously. 1504


In a sensitivity analysis, we modeled the continuous outcome of percentage weight change from baseline, instead of the trichotomized outcome with the ⫾5% cut-off points. The results from the linear regression analyses were almost identical to those from the logistic regression analyses. The composite SOC score was inversely related to continuous percentage weight change (p ⫽ 0.0009). Specifically, patients in the highest quintile of the composite SOC score showed a 4.6% greater negative weight change than the patients in the lowest quintile. The interaction between the SOC composite and the dichotomized (SF12) mental health scores was also significant (p ⫽ 0.01) and had the same general interpretation in the linear regression model; i.e., patients with mental health scores above the median showed a stronger negative relationship between the SOC composite score and percentage weight change.

Discussion This is the first report describing the longitudinal relationship between the number of periods in action or maintenance for multiple weight-related behaviors and weight


Figure 3: Composite score and weight change. Composite score includes five target behaviors (usual activity, exercise, dietary fat, fruits and vegetables, and portion control) and the four measurement periods (0 – 6, 6.1–12, 12.1–18, 18.1–24 months). Chi-square (df ⫽ 8) ⫽ 20.6; P value ⫽ .008.

change over 2 years in a primary care sample. Bivariate analyses of elapsed time in action or maintenance for five target behaviors and weight loss incidence were significant. In addition, bivariate analysis describing the relationship between a composite SOC variable and weight loss was supported by ordinal logistic regression analysis. Because longitudinal data are not compromised by the antecedentconsequent problem that limits inferences from cross-sectional data, longitudinal data constitute a higher level of evidence. Thus, the data from this longitudinal study provide strong support for the a priori hypothesis that there is a positive relationship between the elapsed time in action or maintenance stages for weight-related behaviors and weight loss. These data also seem to validate our application of the Transtheoretical Model to the problem of weight management in primary care. Increased portion control had the largest RR estimate (RR ⫽ 3.84) for weight loss among the five target behaviors. This result is consistent with prior research indicating

Table 5. Ordinal logistic regression models of weight change (loss, no change, gain) Independent variables

Model coefficient (SE)

Composite score only* Composite score* Age group Gender Ethnicity BMI group Hypertension High cholesterol Arthritis Stomach problems Diabetes Psychotropic medicines Depression Anxiety Eating disorder Prior weight loss attempts MD advice Commercial program Physical health Mental health Healthy eating self-efficacy Exercise self-efficacy Social desirability

0.37 (.091) 0.43 (0.11) ⫺0.09 (0.19) 0.057 (0.17) ⫺0.32 (0.35) ⫺0.072 (0.15) 0.26 (0.29) ⫺0.46 (0.28) ⫺0.029 (0.32) 0.13 (0.34) 0.10 (0.41) ⫺0.19 (0.28) ⫺0.19 (0.36) 0.67 (0.28) 0.010 (0.17) ⫺0.33 (0.72) 0.41 (0.37) ⫺0.24 (0.30) ⫺0.0096 (0.016) 0.00062 (0.015) 0.0047 (0.0048) ⫺0.019 (0.016) 0.019 (0.024)

Odds ratio (95% confidence interval) 4.3 (2.1, 8.9)† 5.7 (2.4, 13.5)†

3.80 (1.3, 11.3)

p ⬍0.0001 ⬍0.0001 0.65 0.73 0.36 0.63 0.38 0.10 0.93 0.71 0.80 0.51 0.59 0.02 0.95 0.65 0.26 0.43 0.55 0.97 0.32 0.22 0.43

* Grouped raw score (approximate quintiles) treated as a continuous variable with values 0, 1, 2, 3, and 4. † Four “quintile” change in grouped composite score.


1505


that dietary energy restriction is the most direct means of inducing weight loss (38). Other research indicates that increased exercise is important for weight loss and maintenance (39). In our data, fewer patients were in action for planned exercise during all four periods (n ⫽ 48) than for increased activity (n ⫽ 65), portion control (n ⫽ 66), decreased dietary fat (n ⫽ 76), and increased fruits and vegetables (n ⫽ 101). Portion control (decreased portion sizes) may be a behaviorally easier target than increased planned exercise in the current environment but not as easy as increased fruit and vegetable consumption. However, increased fruit and vegetable consumption had a weaker relationship (RR ⫽ 2.95) with weight loss than portion control. The internal validity of any prospective study can be compromised by a combination of selection, information, or confounding bias (40). Selection bias can occur when participant follow-up is ⬍100%, and there is a relationship between the study factor and participant retention (41). In this study, 35 of 329 (11%) patients were excluded from the longitudinal bivariate analyses (Figures 1 to 3) because of missing weight change data. Because (we believe that) these patients are more likely to have failed to lose weight early in follow-up and had low composite SOC scores, our RR estimates may be (conservatively) biased toward the null valve. This interpretation is consistent with Ware’s recommendation to make the conservative assumption that early dropouts have not lost weight (42). It is possible that the criteria for the action stage were insufficiently stringent to correctly classify patients who were actually in preparation, or in an earlier stage. Because our staging algorithm was specifically designed to minimize misclassification bias from insufficiently stringent action criteria for complex domains, this source of bias is probably small relative to other methods (23). However, more research in this area would be useful. Retrospective data from the National Weight Loss Registry suggest that successful long-term loss of 15 kg may require an hour of moderate to vigorous exercise each day and strict adherence to a portioncontrolled lower calorie (fat) diet (43). Patient cycling from action to earlier stages between assessments may partially account for the observation that ⬍40% of patients who ostensibly spent four periods in action (Tables 3 and 4) also loss 5% or more of their baseline weight. Weight measurement errors should be small because of the high correlation between measured and chart weights and the use of a standardized weight measurement protocol on calibrated scales. The ordinal logistic regression (and the linear regression) analyses suggest that confounding bias from multiple relevant extraneous factors is minimal, but unmeasured confounders are always an issue in observational studies. The external validity of our results is limited by the characteristics of the participants in the REACH trial 1506


(31,32,33). Because participants were recruited in a primary care setting, they are more likely to have diagnosed hypertension and other conditions associated with obesity than members of the general population. Participants may not be a representative sample of middle-aged primary care patients with weight problems seen in Northeastern Ohio or elsewhere. In terms of the baseline prevalence of (binge) eating disorder, the participants seem to be similar to participants who volunteer for other obesity treatment programs (44). Most observational studies that have applied the Transtheoretical Model to weight loss-related behaviors have had cross-sectional designs and have used different staging algorithms (5–23). However, the majority of these studies offer qualified support for the construct validity and potential clinical utility of the staging concept for multiple weight loss-related dietary and exercise behaviors (5,6,9,23). Skeptics have either challenged the merits of the underlying concept or highlighted the measurement or classification issues raised by the necessity of staging patients on multiple complex behaviors (45,46). The few studies with longitudinal data have had qualified or mixed results or have not assessed weight change (24 – 29). Anderson and Apovian reported that patients prescribed low-energy diets and in action stage for weight loss according to the University of Rhode Island Change Assessment Scale lost significantly more weight over 4 weeks than patients in the contemplation or maintenance stages (24). However, the University of Rhode Island Change Assessment Scale instruments were not completed at baseline; thus, the short follow-up interval was indeterminate; in addition, weight loss was an indicator of a negative energy balance, not a target behavior. Greene and Rossi (25) reported, “Between 12 and 18 months, participants progressing at least 1 stage reduced their fat intake to a greater extent that participants who did not progress”; however, the theory suggested that all stage progressions should not be equally predictive of decreased dietary fat consumption. Jeffery et al. found no relationship between baseline SOC and weight loss over a 3-year period (26). However, they used a staging algorithm that defined the action stage in terms of dieting only; the authors did not consider portion control along with other important target behaviors such as planned physical exercise or usual activity in their definition of the action stage(s). Macqueen et al. found no association between baseline SOC for weight loss and the percentage of weight loss over a 4- to 6-week period (27). Marcus et al. found a significant relationship between baseline SOC for physical activity and physical activity 6 months later in a sample of employees (28). Plotnikoff et al. reported that only 45% of their longitudinal predictions based on the Transtheoretical Model were supported (29); however, the authors did not have an independent measure of exercise, which was the target behavior.


This study is one of the first to document that the elapsed time in action or maintenance for multiple weight lossrelated target behaviors is longitudinally related to weight loss (or weight gain) over a 2-year period. Future studies could focus on identifying those environmental, psychosocial, and programmatic factors that increase (or decrease) the amount of time in action or maintenance (for relevant target behaviors) over prolonged follow-up periods. However, no investigator has yet tested whether an intervention based on the consistent application of stage-matched processes of change will produce better weight management outcomes than an intervention based on the consistent application of stage-mismatched processes of change (47). Plotnikoff et al.’s work challenges the standard application of the processes of change to exercise behavior by suggesting that behavioral processes of change are useful in the preaction stages for exercise, which is contrary to SOC theory (29).

Acknowledgments We thank Yen-Pin Chiang, Agency for Healthcare Research and Quality Project Officer; Keding Hua, statistical programming; Margaret Reitenbach, administrative support; and physicians and patients from the study practices for their time and efforts. This study was supported by Grants 1 R01 HS08803-01A2 through 5 R01 HS08803-04 and 3 R01 HS08803-02S1 through 3 R01 HS08803-04S1 from the Agency for Healthcare Research and Quality and the National Institute of Diabetes, Digestive, and Kidney Diseases and by consecutive Nutrition and Exercise Studies grants (1998 to 2002) from the Summa Health System Foundation. Author contributions were as follows: study concept and design, E.E.L., D.G.J., W.D.S., and K.S.S.; acquisition of data, K.R.B., W.D.S., C.F.C., and K.F.S.; analysis and interpretation of data, E.E.L. and D.G.J.; drafting of the manuscript, E.E.L., D.G.J., W.D.S., K.S.S., K.R.B., and C.F.C.; critical revision of the manuscript for important intellectual content, E.E.L., D.G.J., W.D.S., and K.S.S.; statistical expertise, D.G.J. and E.E.L.; obtained funding, E.E.L., W.D.S., and D.G.J.; administrative, technical, or material support, K.R.B. and E.E.L.; and study supervision, K.R.B. and E.E.L. REACH trial results were presented at the American Academy of Family Physicians, National Research Network’s Convocation of Practices, Improving Patient Care through Healthy Behaviors Research (Arlington, VA, March 20 to 23, 2003). References 1. French SA, Story M, Jeffery RW. Environmental influences on eating and physical activity. Annu Rev Public Health. 2001;22:309 –35. 2. Blundell JE, Cooling J. Routes to obesity: phenotypes, food choices and activity. Br J Nutr 2000;Mar 83(Suppl 1):S33– 8.

3. Mokdad AH, Bowman BA, Ford ES, Vinicor F, Marks JS, Koplan JP. The continuing epidemics of obesity and diabetes in the United States. JAMA. 2001;286:1195–200. 4. Hill JO, Wyatt HR, Reed GW, Peters JC. Obesity and the environment: where do we go from here? Science. 2003;299: 853–5. 5. Auld GW, Nitzke SA, McNulty J, et al. A stage-of-change classification system based on actions and beliefs regarding dietary fat and fiber. Am J Health Promot. 1998;12:192–201. 6. Boyle RG, O’Connor PJ, Pronk NP, Tan A. Stages of change for physical activity, diet, and smoking among HMO members with chronic conditions. Am J Health Promot. 1998; 12:170 –5. 7. Brug J, Glanz K, Kok G. The relationship between selfefficacy, attitudes, intake compared to others, consumption, and stages of change related to fruit and vegetables. Am J Health Promot. 1997;12:25–30. 8. Bull FC, Eyler AA, King AC, Brownson RC. Stage of readiness to exercise in ethnically diverse women: a U.S. survey. Med Sci Sports Exerc. 2001;33:1147–56. 9. Campbell MK, Reynolds KD, Havas S, et al. Stages of change for increasing fruit and vegetable consumption among adults and young adults participating in the national 5-Day for Better Health community studies. Health Educ Behav. 1999; 26:513–34. 10. Curry SJ, Kristal AR, Bowen DJ. An application of the stage model of behavior change to dietary fat reduction. Health Educ Res. 1992;7:97–105. 11. Glanz K, Kristal AR, Tilley BC, Hirst K. Psychosocial correlates of healthful diets among male auto workers. Cancer Epidemiol Biomarkers Prev. 1998;7:119 –26. 12. Greene GW, Rossi SR, Reed GR, Willey C, Prochaska JO. Stages of change for reducing dietary fat to 30% of energy or less. J Am Diet Assoc. 1994;94:1105–10. 13. Greene GW, Rossi SR, Rossi JS, Velicer WF, Fava JL, Prochaska JO. Dietary applications of the stages of change model. J Am Diet Assoc. 1999;99:673– 8. 14. Hargreaves MK, Schlundt DG, Buchowski MS, Hardy RE, Rossi SR, Rossi JS. Stages of change and the intake of dietary fat in African-American women: improving stage assignment using the Eating Styles Questionnaire. J Am Diet Assoc. 1999; 99:1392–9. 15. Herrick AB, Stone WJ, Mettler MM. Stages of change, decisional balance, and self-efficacy across four health behaviors in a worksite environment. Am J Health Promot. 1997; 12:49 –56. 16. Laforge RG, Velicer WF, Richmond RL, Owen N. Stage distributions for five health behaviors in the United States and Australia. Prev Med. 1999;28:61–74. 17. Ling AM, Horwath C. Defining and measuring stages of change for dietary behaviors: readiness to meet fruit, vegetable, and grain guidelines among Chinese Singaporeans. J Am Diet Assoc. 2000;100:898 –904. 18. Logue E, Sutton K, Jarjoura D, Smucker W. Obesity management in primary care: assessment of readiness to change among 284 family practice patients. J Am Board Fam Pract. 2000;13:164 –71. 19. Marcus BH, Simkin LR. The stages of exercise behavior. J Sports Med Phys Fitness. 1993;33:83– 8. OBESITY RESEARCH Vol. 12 No. 9 September 2004

1507


20. Nigg CR, Burbank PM, Padula C, et al. Stages of change across ten health risk behaviors for older adults. Gerontologist. 1999;39:473– 82. 21. Reed GR, Velicer WF, Prochaska JO, Rossi JS, Marcus BH. What makes a good staging algorithm: examples from regular exercise. Am J Health Promot. 1997;12:57– 66. 22. Sarkin JA, Johnson SS, Prochaska JO, Prochaska JM. Applying the transtheoretical model to regular moderate exercise in an overweight population: validation of a stages of change measure. Prev Med. 2001;33:462–9. 23. Sutton K, Logue E, Jarjoura D, Baughman K, Smucker W, Capers C. Assessing dietary and exercise stage of change to optimize weight loss interventions. Obes Res. 2003;11:641– 52. 24. Anderson WA, Apovian CM. Using the stages of change model to target effective interventions for obese patients. Am J Clin Nutr. 2002;75:362–3S. 25. Greene GW, Rossi SR. Stages of change for reducing dietary fat intake over 18 months. J Am Diet Assoc. 1998;98:529 –34. 26. Jeffery RW, French SA, Rothman AJ. Stage of change as a predictor of success in weight control in adult women. Health Psychol. 1999;18:543– 6. 27. Macqueen C, Brynes A, Frost G. Treating Obesity: can stages of change model help predict outcome measures? J Hum Nutr Diet. 1999;12:229 –36. 28. Marcus BH, Eaton CA, Rossi JS, Harlow LL. Self-efficacy, decision-making and stages of change: an integrative model of physical exercise. J Appl Soc Psychol. 1994;24:489 –508. 29. Plotnikoff RC, Hotz SB, Birkett NJ, Courneya KS. Exercise and the transtheoretical model: a longitudinal test of a population sample. Prev Med. 2001;33:441–52. 30. Prochaska JO, Velicier WF, Rossi JS, et al. Stages of change and decisional balance for 12 problem behaviors. Health Psychol. 1994;13:39 – 46. 31. Logue E, Smucker W, Sutton K, Jarjoura D, Baughman K, Capers C. REACH: using the transtheoretical model to change eating and exercise behavior. Presented at the conference of the American Academy of Family Physicians, National Research Network’s Convocation of Practices, Improving Patient Care Through Healthy Behaviors Research. Arlington, Va, March 2003. 32. Logue E, Sutton K, Jarjoura D, Smucker W, Baughman K, Capers C. A transtheoretical model-chronic disease care intervention for obesity in primary care: a randomized trial. Final Report for grants 1 R01 HS08803-01A2 through 5 R01 HS08803-04. 2003. Agency for Healthcare Research and Quality. 33. Baughman K, Logue E, Sutton K, Capers C, Jarjoura D, Smucker W. Biopsychosocial characteristics of overweight

1508


34.

35.

36.

37.

38.

39.

40.

41.

42. 43. 44.

45. 46.

47.

and obese primary care patients: do psychosocial and behavior factors mediate sociodemographic effects? Prev Med. 2003; 37:129 –37. Spitzer RL, Kroenke K, Williams JB. Validation and utility of a self-report version of PRIME-MD: the PHQ primary care study. Primary Care Evaluation of Mental Disorders. Patient Health Questionnaire. JAMA. 1999;282:1737– 44. Gregg EW, Williamson DF. The relationship of intentional weight loss to disease incidence and mortality. In: Wadden TA, Stunkard AJ, eds. Handbook of Obesity Treatment. New York: Guilford Press; 2002, pp. 125– 43. Ware JE, Kosinski M, Keller SD. SF-12: How to Score the SF-12 Physical and Mental Health Summary Scales. 2nd ed. Boston, MA: The Health Institute, New England Medical Center; 1995. Perri MG, Corsica JA. Improving the maintenance of weight loss in behavioral treatment of obesity. In: Wadden TA, Stunkard AJ, eds. Handbook of Obesity Treatment. New York: Guilford Press; 2002, pp. 357–77. Wing R. Behavioral weight control. In: Wadden TA, Stunkard AJ, eds. Handbook of Obesity Treatment. New York: Guilford Press; 2002, pp. 301–16. Blair S, Leermakers EA. Exercise and weight management. In: Wadden TA, Stunkard AJ, eds. Handbook of Obesity Treatment. New York: Guilford Press; 2002, pp. 283–300. Kleinbaum DG, Kupper LL, Morgenstern H. Validity: general considerations. In: Kleinbaum DG, Kupper LL, Morgenstern H, eds. Epidemiologic Research, Principles and Quantitative Methods. New York: Van Nostrand Reinhold; 1982, pp. 183–93. Kleinbaum DG, Kupper LL, Morgenstern H. Selection bias. In: Kleinbaum DG, Kupper LL, Morgenstern H, eds. Epidemiologic Methods, Principles and Quantitative Methods. New York: Van Nostrand Reinhold; 1982, pp. 194 –219. Ware JH. Interpreting incomplete data in studies of diet and weight loss. N Engl J Med. 2003;348:2136 –7. Wing RR, Hill JO. Successful weight loss maintenance. Annu Rev Nutr. 2001;21:323– 41. Stunkard AJ. Binge eating disorder and the night-eating syndrome. In: Wadden TA, Stunkard AJ, eds. Handbook of Obesity Treatment. New York: Guilford Press; 2002, pp. 107–21. Bandura A. The anatomy of stages of change. Am J Health Promot. 1997;12:8 –10. Povey R, Conner M, Sparks P, James R, Shepherd R. A critical examination of the application of the Transtheoretical Model’s stages of change to dietary behaviours. Health Educ Res. 1999;14:641–51. Quinlan KB, McCaul KD. Matched and mismatched interventions with young adult smokers: testing a stage theory. Health Psychol. 2000;19:165–71.