Impact of Social Support Intensity on Walking in the Severely Obese: A Randomized Clinical Trial




Objective: There are few established methods for promoting physical activity (PA) in the severely obese. Because social support is a potential method for promoting PA, we compared mean steps/day during 18 weeks in severely obese outpatients receiving either standard support (SS) or added support (AS).

Methods and Procedures: Eighty severely obese outpatients from an obesity clinic were invited; 66 provided written consent, 55 were randomized, and 42 were included in final analyses (9 men, 33 women; age 44.4 ± 13.1 years; BMI 41.9 ± 5.5 kg/m2). All participants received a pedometer and a walking promotion booklet. In addition to SS, the AS group received ten 2-h group counseling sessions aimed at increasing weekly accumulated steps, every second week during the study. Each participant was asked to complete a 7-day walking diary every second week (10 observations).

Results: Baseline steps/day was 6,912 for the AS group and 5,311 for the SS group (P = 0.023). Data at 18 weeks showed that the AS group recorded 10,136 steps/day and the SS group 6,118 steps/day (P = 0.024). There was no allocation × time interaction (P = 0.46). During the follow-up period as a whole, the AS group recorded 1,794 more steps/day than the SS group (P = 0.0074).

Discussion: The AS group recorded more steps/day than the SS group, reaching a mean level of ∼10,000 steps/day. However, the nonsignificant interaction between allocation × time suggests that this difference was present already at baseline and did not increase during follow-up.


Caloric restriction and increased physical activity (PA) are vital in the treatment of obesity and its comorbidities (1,2). To gain the health benefits of PA fully, it is critical to build and establish a realistic routine that is maintained long term.

While there is ample evidence on the health effects of PA in the obese, there is little data on how to promote PA. Interventions that work in everyday life, so that long-standing PA routines become more feasible, are especially lacking (3). A review on clinical methods for improving adherence to PA (3) found some indirect support for the hypothesis that therapist and peer support can lead to either increased PA or improved weight control in the obese (4,5,6).

Moreover, in-depth interviews with patients from our outpatient clinic identified our program of therapist and peer support as the most salient theme for increased PA (Hemmingsson et al., data not shown). However, the role of support intensity for increasing PA is not well described by quantitative data.

Pedometers have been used frequently as a PA promotion tool in various settings with success (7,8,9,10,11,12,13,14), although there is limited data on its usefulness in severely obese individuals.

The aim of this study, therefore, was to provide proof of concept that social support intensity influences PA adoption (walking) in a population of severely obese outpatients. This was done by prospectively comparing PA levels measured during 18 weeks as steps/day from a pedometer, in two groups randomized to different degrees of social support. Compliance with the 10,000 steps/day recommendation was also compared between groups.

Methods and Procedures

Recruitment goal

The power calculation was based on the following criteria: power of 80%, and a 5% significance level, resulting in a power index of 2.8. A difference of ≥1,000 steps/day was considered to be a clinically relevant difference, with a predicted s.d. of 1,000. This gave the following sample size formula (n/group): 2 × (2.8 × 1,000/1,000)2 = 16 (rounded upward). This means that complete data from 32 participants would be needed to reject H0. However, because we estimated a drop-out rate of 20%, based on previous experiences, the recruitment goal was set to 40 participants.


Participants were recruited from the obesity treatment program at the Obesity Unit, Centre for Metabolism and Endocrinology, Karolinska University Hospital Huddinge, Stockholm, Sweden. The clinic's 2-year obesity self-help treatment program relies exclusively on behavior change, as opposed to using drugs or surgery, to induce weight loss and weight loss maintenance.

Eighty outpatients, who were about to start the normal care program, were informed about the study (August 2002), and asked whether they would be interested in participating (Figure 1). Sixty-six of these provided written informed consent (August 2002). Between consent and time of randomization (December 2002), 11 individuals withdrew from the study, leaving 55 eligible for randomization. Of these, seven dropped out before the end of the study at 18 weeks (reasons for dropout were not further investigated), and six did not produce walking logs that met the criteria for inclusion (see below). Forty-two participants were therefore included in the final analysis. There were no significant differences in age, sex, or BMI between those who dropped out between randomization and follow-up (n = 13) compared with those who completed the study (n = 42).

Figure 1.

Flowchart of study procedures and participants.

Participant inclusion criteria were referral to the obesity unit, BMI > 30 kg/m2, and no contraindication for PA identified by a physician.

Experimental design

Two programs with different support intensity were compared: standard support (SS) and added support (AS), see Table 1. SS consisted of three components: (i) standard care at the Obesity unit (see below for further details), (ii) a validated pedometer (Yamax Digiwalker, SW-701; Yamax, Tokyo, Japan), and (iii) a PA behavior change booklet. The booklet was written specifically for the study, where the content was based on the previously mentioned literature review on PA behavior change for the obese (3).

Table 1.  Outline of intervention content for the added support (AS) group and the standard support group (SS)
inline image

During the second semester (18 weeks) of standard care, the patients met for 2-h group sessions every month, where the overall focus was on maintaining achieved weight loss. This included support for increased PA, but also dietary changes, body weight diaries, and home assignments (such as low-fat/low-sugar cooking, stress management, rewards, and relapse prevention).

The group allocated to AS received, in addition to the three components of standard care (described above), ten 2-h sessions in groups of eight participants, every second week. The 10 additional group sessions were specifically designed for walking promotion. The main difference between the groups was therefore in quantity:

•    Five 2-h, monthly standard care group meetings for the SS group

•    Fifteen 2-h biweekly group meetings for the AS group (5 standard care meetings + 10 added walking promotion meetings)

However, it could also be argued that there was a difference in intervention quality between groups because the AS group attended theory-based meetings specifically focused on increased walking.

The theoretical framework of behavior change, for the additional meetings in the AS group, was the Transtheoretical model (15). This model outlines 10 processes of change: 5 cognitive and 5 behavioral. Three processes of change were selected to comprise the core of the behavior change principles: raising awareness (increased understanding of the need for intervention, e.g., by comparing current PA levels with public health recommendations, or by identifying day-to-day variations in PA), countering (swapping an unhealthy behavior for a healthy behavior, e.g., taking a walk after dinner instead of watching TV, or cycling instead of driving to work), and helping relationships (support from family and friends, fellow outpatients and therapists, e.g., walking groups at work, or encouragement from a spouse).

The activity booklet contained separate sections on the selected three processes of change, along with sections on general activity information regarding behavior change, such as solving activity-related problems, avoiding injuries, and establishing a walking routine.

The focus of the PA prescriptions was on increasing walking, building up to the recommended level of 10,000 steps/day, if possible (16). If baseline levels were low (e.g., 3,000 steps/day), the participant was encouraged to increase walking gradually in increments of 2,000 steps/day, and if feasible reach 10,000 steps/day toward the end of the follow-up.

All participants, including those in the SS group, were instructed to set their own PA goals, where they made gradual and realistic increases in their everyday walking, by making small but numerous changes, such as taking the stairs instead of the lift, using public transport as opposed to the car, and walking instead of taking the bus.

Everyone was encouraged to interact with the pedometer as much as possible and to study the booklet. The importance of coming to group sessions, both during the SS program and the additional group sessions for the AS group, was especially emphasized. During all group sessions, the staff mainly worked on reinforcing positive aspects, boosting self-efficacy, and increasing autonomy.

The study was approved by the Karolinska Institute ethical review board at Karolinska University Hospital, Huddinge.

Data collection

The main outcome variable was mean steps/day, which was measured using the Yamax Digiwalker SW-701 pedometer (Yamax), a validated criterion pedometer (17). Participants were instructed to record their daily step tally (from the time they put on their clothes until bed time) in a log every day, for 7 consecutive days. The mean steps/day tallies from each log were calculated thereafter. Logs were filled in every second week during the study, meaning a maximum of 10 mean steps/day observations (comprising 70 days) for each participant.

Logs were reviewed and clarified during group sessions for the AS group if something was unclear. Logs from the SS group were collected when they were attending group sessions as part of standard care.

At least 4 days of recording per week (including 1 weekend day) were needed for the log to be included. Each participant was also required to hand in at least three approved logs (three participants from each group failed to do so and were excluded from analyses). Anthropometric data (height, weight, and waist circumference) were collected in a fasting condition (no shoes, wearing only underwear) at baseline and after 18 weeks. Waist circumference was measured with a measuring tape, at the point midway between the iliac crest and the lower rib.


In the main analysis of steps/day, we used baseline steps/day as a covariate. In total, 174 of a maximum 220 logs (22 participants × 10 logs) were approved for the AS group (73.6%) and 121 of 220 for the SS group (65.3%), with an approximately consistent pattern of missing data both within and between groups. Owing to the missing data and to avoid loss of power, we used multiple imputations through the Solas predictive model method (Solas software, version 3.2; Statistical Solutions, Saugus, MA). Unlike other procedures for handling missing data such as “baseline value carried forward” or “last observed value carried forward,” the Solas predictive model of data imputation reflects the variance of the existing data, also taking into account the variance of imputed values (18). Five datasets were created, taking into account the within-subject effect and the between-subject factor. Each dataset was then analyzed using a linear mixed effects model (procedure MIXED, SAS version 9.1.3; SAS Institute, Cary, NC).

The SAS procedure MIANALYZE was used to combine the estimated parameters into one summarizing model. Cook's distance and various residual plots revealed the influence of individual observations on the fitted regression lines.

Participants were stratified into two groups (≥10,000 steps/day and <10,000 steps/day) according to published recommendations (16). The odds ratio for complying with the recommendation was analyzed using logistic regression (adjusted for baseline steps/day). This analysis used the same five datasets with imputations from the analysis of steps/day as a continuous variable. A P value of <0.05 was considered statistically significant.


Baseline characteristics are shown in Table 2. The sample mean BMI was above the cutoff point for grade III obesity (>40 kg/m2). The sample consisted of 9 men and 33 women, reflecting the general men/women ratio of our standard care program. Attendance at the additional group sessions for the AS group was 73.2%.

Table 2.  Sample characteristics, mean (1 s.d.)
inline image

Baseline steps/day was 6,912 (1 s.d.: 2,757; 95% confidence intervals (CIs): 5,320–8,504) and 5,311 (s.d. 2,657; 95% CI: 2,850–6,775) for the AS and SS groups, respectively (Pbetween = 0.023). At final follow-up 18 weeks from baseline, the AS group recorded 10,136 steps/day (s.d.: 3,928; 95% CI: 8,183–12,089) and the SS group 6,118 steps/day (s.d.: 3,121; 95% CI: 3,231–9,004), Pbetween = 0.024. During the follow-up period as a whole, the AS group recorded 1,794 more steps/day than the SS group (Pbetween = 0.0074), see Table 3 and Figure 2. There was a linear increase by 136 steps/day (P < 0.001) with each passing week. The allocation × time interaction was not significant (P = 0.46), indicating that the difference in walking did not increase between groups over time.

Table 3.  Regression analyses of walking (steps/day) at follow-up between the standard support group (SS, coded 1) and the added support group (AS, coded 2)
inline image
Figure 2.

Steps/day in the two groups. Solid line indicates added support (AS) group; dotted line indicates standard support (SS) group. Error bars show 95% confidence intervals.

A −8.15 steps per day2 quadric trend was estimated and, though not statistically significant (P = 0.10), indicating that the process of walking adoption may be quadratic, and that the process reached a plateau at ∼10,000 steps/day toward the end of follow-up.

The AS group also showed higher compliance with the 10,000 steps/day recommendation than the SS group (14% compared to 0% at baseline, and 56% compared to 0% at 18 weeks). In comparison with the SS group, the overall odds ratio for walking 10,000 steps/day (adjusted for baseline differences) was 6.0 (95% CI: 1.1–31.6, P = 0.04) favoring the AS group (Figure 3).

Figure 3.

Compliance with the 10,000 steps/day recommendation in the two groups.


The results from this study show that the AS group increased the amount of accumulated steps by 47% and the SS group by 15%, between baseline and follow-up. However, the statistical test of interaction between allocation and time was not significant, suggesting that the difference between groups in accumulated number of steps were similar despite exposure to different intensities of social support.

The overall difference in walking between groups was also reflected in the data on compliance with the 10,000 steps/day recommendation. The AS group was six times more likely (odds ratio, adjusted for baseline differences) to comply with the recommendation. At 18 weeks, 56% of participants in the AS group accumulated ≥10,000 steps/day (14% at baseline).

Although our data do not support a positive effect of added social support, it was clear that increases in walking, even up to 10,000 steps/day, were feasible for the severely obese. In the AS group, mean walking levels toward the end of the follow-up were relatively stable at ∼10,000 steps/day, i.e., the level of steps/day that was generally recommended. Programs to promote walking in the severely obese can therefore be effective by using the relatively simple combination of social support, a pedometer, and walking diaries. Studies from other settings confirm the usefulness of this combination (7,8,9,13,14).

The current recommendation for the prevention of weight regain suggests accumulation of 60–90 min/day of moderate intensity PA (19). It is unclear, however, if this constitutes an appropriate general recommendation for severely obese individuals. PA recommendations of lower quantity, such as walking 10,000 steps/day, may be better received by obese individuals. Indeed, there is evidence suggesting that the recommendation to accumulate 10,000 steps/day may increase PA levels to a greater extent compared with the recommendation of walking briskly for 30 min/day in previously sedentary women (11). The main focus of this study was to provide prescriptions adapted to each individual, based on their baseline values. Regular behavioral counseling was used to facilitate incremental increases in walking. This more realistic PA prescription could then be part of an overall obesity treatment package, including dietary changes, meal replacements, drugs, or surgery.

A related point of confusion in the discussion about PA recommendations is how to translate steps/day into minutes of moderate intensity PA. Research suggest that it takes ∼30 min to walk 3,000 steps (20), and that 100 steps/min may be close to the lower cutoff for moderate intensity PA (i.e., three metabolic equivalents) in men and women (20). Therefore, it is sometimes assumed that 10,000 steps/day equals 90 min of moderate intensity PA.

However, as a relatively sedentary individual accumulates between 5,000 and 7,500 steps/day (8,11,16,21,22) at an intensity usually below moderate (i.e., <100 steps/min), it could be argued that the accumulation of ∼10,000 steps/day (a normal sedentary day with two 15-min walks) is approximately equal to 30 min of moderate intensity PA per day.

Accumulating 10,000 steps/day is likely insufficient to prevent weight regain (23), although it is likely to confer other health benefits. Studies have suggested that PA of light-to-moderate intensity is suitable for activity promotion (24), and over time may also prevent weight gain (25,26). Furthermore, regular PA of light-to-moderate intensity such as walking may also improve the metabolic profile and reduce the risk of the metabolic syndrome and type 2 diabetes (27,28,29). Increases in PA intensity, once an individual has managed to stabilize their activity level above the 10,000 steps/day level, will lead to even further increases in energy expenditure (30), thereby enhancing overall weight control (19).

We deliberately elected against analyzing anthropometry between groups, for two reasons. First, to avoid problems related to multiple significance, i.e., type 1 error. Second, we did not power the study to detect changes in body weight or waist circumference. An adequately powered trial for analyzing changes in body composition would require a larger sample size and longer follow-up. Furthermore, the chances of rejecting H0 for any health parameter would be reduced even more by the concurrent weight loss program.

The results from this study should be interpreted considering the following limitations. First, our design did not allow a blinded study. Participant blinding was deemed unrealistic owing to continued participation on the standard care program. How this may have has an effect on the outcome is unknown. Moreover, the fact that the participants were attending another treatment program—albeit of similar content—may also have biased our findings.

Another limitation was the short duration. This means that we do not have information on adherence to the more active lifestyle. The issue of adherence is critical, because the health benefits of increased PA, for example, an increase in accumulated steps by 20–30%, are possibly accumulated over a long period of time.

Because the AS-program was added onto an already existing treatment scheme, this makes it unsuitable for direct implementation. However, our aim was not to develop and evaluate a ready-to-be-used program, which could then be implemented widely; instead, our aim was to provide proof of concept that a pedometer-driven program of social support can be sufficient to achieve meaningful increases in PA even in the severely obese. Future studies that will more directly compare feasible and effective programs are needed, so that best practice can be established and implemented.

Some of the baseline measurements for the AS group were not performed until after the first group meeting. It is likely that this contributed to the observed differences at baseline, i.e., some of the behavior change had already occurred. To be able to plan participation, the AS group was made aware of their allocation ∼6 weeks before the start of the intervention. This may also have contributed to the observed differences in walking at baseline. Randomization with stratification for baseline steps/day, so that balance would be achieved, was not deemed feasible, because participants needed to plan their participation in the added group sessions with >1 week's notice. Therefore, the difference in steps/day at baseline limited our possibility to fully reject H0 for time × allocation interaction.

Some of the baseline activity diaries from both groups were rejected because of faulty pedometer readings (i.e., <500 steps/day), probably due to attachment difficulties with loose clothing. As the study progressed, the participants gradually learned how to attach the pedometer properly, i.e., to minimize underreporting. Lack of “true” steps at baseline could have led to an overestimation of observed increases in steps/day. It is, however, unlikely that this biased our results regarding between-group comparisons, because any potential measurement error was likely to be random between groups. To avoid problems of this nature, it may be wise to train participants in pedometer use before collecting baseline data.

Another important issue when measuring walking in obese individuals is the minimization of pedometer tilt away from the vertical plane, which will lead to underestimation of accumulated steps (31). All participants were instructed to avoid tilting, if possible, by attaching the pedometer to the lower back. Since both groups lost similar amounts of weight, it is unlikely that between-group comparisons of steps/day were biased due to tilted pedometers.

Missing data was expected because we wanted the participants to fill in their logs every second week, in total 10 observations (mean steps/day during the 7-day period) per participant. Overall, 73.6 and 65.3% of activity diaries were included in the final analyses for the AS and SS groups, respectively. Although unlikely, we cannot exclude that this difference between groups may have biased our findings.

Our study also has some positive features. The statistical power was sufficient to reject H0 for steps/day. Problems with type 2 error are otherwise common in intervention studies of PA. The power was in part boosted by frequent data collection, in this case 10 observations per participant during 18 weeks. The frequent data collection periods also added insight into the behavior change process, i.e., the initial increase and the subsequent plateau.

Strengths of this study also include the use of a criterion pedometer for data collection (17). Provided that the participants correctly recorded their daily steps/day tallies (we can only assume that they did), it can be argued that our PA data were objective, as opposed to self-report, which tends to be overestimated (32).

Furthermore, we choose to include severely obese individuals which may mean that the observed results from this study may be generalizable to overweight and other obese groups. Any walking challenges specifically related to excessive fatness (skin chafing, social physique anxiety, low fitness, mechanical problems, pains, and aches) is likely to be exposed in severely obese people. If we accept this point, increased accumulated number of steps should be equally possible in overweight and obese class 1 individuals.

In summary, we found that mean steps/day reached ∼10,000 steps/day in the AS group, which was significantly higher than the SS group. However, because the allocation × time interaction term was not significant, it was not possible to conclude that the difference was due to social support. However, given the increased walking between baseline and follow-up our data suggest that PA promotion can be feasible for many of the severely obese.


This study was supported by the Swedish National Centre of Research in Sports and the Stockholm Centre of Public Health. We are grateful to the Department of Exercise and Health Sciences, University of Bristol, UK, for intellectual support, and to Maria Klingvall, Registered Nurse, and Dr Andersson from the outpatient team at the Obesity unit. We sincerely thank all our participants.


The authors declared no conflict of interest.