Long-term weight trajectories

Table 6 shows that men in the FFIT-FU-I group gained 2.59 kg (95% CI 1.61 to 3.58 kg; p < 0.001) between the 12-month and the 3.5-year measures (i.e. 2.44%, 95% CI 1.61% to 3.27%, of their baseline weight). This equates to a weight gain of 1.04 kg per year. Nevertheless, 3.5 years after baseline measurement, men in the FFIT-FU-I group still weighed 2.90 kg less on average, demonstrating a sustained weight benefit from taking part in the FFIT programme. Meanwhile, men in the FFIT-FU-C group (who had the opportunity to take part in the FFIT programme under ‘routine delivery’ conditions immediately after the 12-month measures) lost 2.03 kg (95% CI 1.08 to 2.98 kg; p < 0.001) or 1.79% (95% CI 0.92% to 2.65%) of their baseline weight during the same time period. The mean between-group difference in weight trajectories was 4.62 kg (95% CI 3.26 to 5.99 kg; p < 0.001) or 4.23% (95% CI 3.02% to 5.43%; p < 0.001).

TABLE 6

Change in weight in the FFIT-FU-I and FFIT-FU-C groups between 12 months and 3.5 years (mixed-effects linear regression models)

Figure 4 illustrates the data shown in Table 4, and clearly demonstrates that both groups weighed less at 3.5 years than at baseline.

FIGURE 4

Mean weight (kg, 95% CI) of participants in the FFIT-FU-I and FFIT-FU-C groups at 12 months and 3.5 years after the FFIT RCT baseline measurements. Note that the y-axis [weight (kg)] does not start at zero.

Sensitivity analyses

Sensitivity analyses on the primary outcome (change in weight) were conducted to assess the sensitivity of the main analyses.

• Loss to follow-up sensitivity analyses used a variety of assumptions about the long-term weight outcomes of men who had not taken part in follow-up measures at 3.5 years.
• Baseline time points sensitivity analyses assessed the fact that both intervention and comparison groups had the opportunity to take part in the FFIT programme, but at different times.

In the loss to follow-up sensitivity analyses, the return-to-baseline and last-value-carried-forward methods were used to impute missing weight data at 12 months and 3.5 years for men who were not followed up. As men who were not followed up at 3.5 years (the not followed-up cohort) were, on average, heavier than the total followed-up cohort at baseline [112.8 kg (SD 17.2) vs. 107.8 kg (SD 17.1)] and 12 months [109.8 kg (SD 18.3) vs. 104.8 kg (SD 17.3)], the return-to-baseline sensitivity analysis is the most conservative and is reported here. The results of the last-value-carried-forward sensitivity analyses are shown in tables i and ii in Report Supplementary Material 7.

Table 7 shows that, in the return-to-baseline sensitivity analysis, the mean weight loss at 3.5 years was 1.81 kg (95% CI 1.09 to 2.52 kg) in the RCT intervention group (including imputed values) and 1.85 kg (1.12 to 2.58 kg) in the RCT comparison group (including imputed values). As in the main analyses, both figures were still significantly different from baseline, but there was no between-group difference.

TABLE 7

Return-to-baseline sensitivity analysis: change in weight at 3.5 years in all men in the RCT intervention and comparison groups

Table 8 shows that, between 12 months and 3.5 years, men in the RCT intervention group (including imputed values) gained 3.15 kg (95% CI 2.37 to 3.92 kg; p < 0.001) and those in the RCT comparison group (including imputed values) lost 1.30 kg (95% CI 0.59 to 2.01 kg; p < 0.001). Although, as expected, there were slight differences in these values from the main analyses, the change in weight over time in each group remained significant, as did the between-group difference in weight trajectories (4.45 kg, 95% CI 3.40 to 5.50 kg; p < 0.001).

TABLE 8

Return-to-baseline sensitivity analysis: change in weight in the RCT intervention and comparison groups between 12 months and 3.5 years

Full details of the baseline time points sensitivity analyses are provided in table iii in Report Supplementary Material 7 and confirm sustained weight loss in the FFIT-FU-I group at 3.5 years post intervention.

Objectively measured physical outcomes

Table 9 shows that there were sustained improvements in waist circumference, BMI, percentage body fat, and systolic and diastolic BP at 3.5 years and 2.5 years after taking part in the programme (for the FFIT-FU-I group and FFIT-FU-C group, respectively). There were no between-group differences.

TABLE 9

Change from baseline in objectively measured physical outcomes in the FFIT-FU-I and FFIT-FU-C groups at 3.5 years

Self-reported physical activity

Table 10 shows that the total PA was significantly higher at 3.5 years than at baseline in both the FFIT-FU-I group [800.0 MET-minutes per week, interquartile range (IQR) –120 to 2514 MET-minutes per week] and the FFIT-FU-C group (919.0 MET-minutes per week, IQR –186 to 2909 MET-minutes per week), and that there were no significant between-group differences. A similar pattern of results was observed for vigorous and moderate PA, and for walking. Table 10 also demonstrates a sustained reduction in time spent sitting in both groups.

TABLE 10

Change from baseline in self-reported PA in the FFIT-FU-I and FFIT-FU-C groups at 3.5 years

Self-reported eating and alcohol intake

Table 11 shows that the consumption of fatty food and sugary food scores at 3.5 years were significantly reduced from baseline in both groups, and that there were no between-group differences. Fruit and vegetables consumption was significantly higher at 3.5 years in both groups, and, again, there were no between-group differences. Similar sustained improvements were also evident for portion sizes of cheese, meat, pasta and chips, and for alcohol consumption; all showed sustained reductions from baseline, and no between-group differences.

TABLE 11

Change from baseline in self-reported food and alcohol consumption in the FFIT-FU-I and FFIT-FU-C groups at 3.5 years

Self-reported psychological health and quality-of-life outcomes

Table 12 shows that improvements in self-esteem, positive and negative affect, and physical and mental HRQoL were sustained to 3.5 years in both groups, and that there were no between-group differences.

TABLE 12

Change from baseline in self-reported psychological health outcomes in the FFIT-FU-I and FFIT-FU-C groups at 3.5 years

Randomised controlled trial secondary outcome trajectories

Table 13 shows significant differences in the trajectories of many RCT secondary outcomes from 12 months to 3.5 years between the FFIT-FU-I and the FFIT-FU-C groups. These are likely to reflect the fact that the FFIT-FU-C group had the opportunity to take part in the FFIT programme during this time (from month 13). The results for the FFIT-FU-C group, therefore, reflect both the changes men made as a result of taking part in the FFIT programme and the maintenance, or lack of maintenance, of these changes over the 2.5 years following the completion of the programme.

TABLE 13

Change in RCT secondary outcomes between 12 months and 3.5 years (mixed-effects linear regression models)

Taking each group separately, the FFIT-FU-I group showed some regain in waist circumference and BMI, and increases in systolic and diastolic BP between 12 months and 3.5 years (although these were all significantly improved from baseline), whereas the FFIT-FU-C group showed tendencies towards improvement in these outcomes over this time period.

In relation to self-reported PA, although the FFIT-FU-I group showed some reduction in total and vigorous PA between 12 months and 3.5 years, levels of moderate PA and walking appeared to be sustained. The FFIT-FU-C group showed improvements in total and moderate PA, and in walking. Between-group differences in moderate PA and in walking were not significant.

The FFIT-FU-I group also demonstrated sustained reductions in intake of fatty and sugary foods, portion sizes of cheese and meat, and total alcohol consumption from 12 months to 3.5 years, but consumption of fruit and vegetables decreased. The FFIT-FU-C group showed improvements in all dietary outcomes (apart from fruit and vegetables consumption) between 12 months and 3.5 years.

Finally, improvements in all psychological outcomes, apart from mental HRQoL, were sustained between 12 months and 3.5 years in the FFIT-FU-I group. The FFIT-FU-C group demonstrated improvements in self-esteem, positive affect and mental HRQoL during this period.

Baseline predictors of long-term weight loss

None of the prespecified baseline predictors (age, BMI, education level, socioeconomic status, marital status, orientation to masculine norms or number of long-standing illnesses) was significantly associated with weight loss at 3.5 years. For specific estimates and p-values, refer to tables 7.1–7.7 in Report Supplementary Material 8.

Mediators of long-term weight loss

Unlike the predictors, which are measured at baseline only, mediators of long-term weight loss were measured at multiple time points (12 weeks, 12 months and 3.5 years). Therefore, using these results to interpret the effect of each mediator on weight change, especially when the groups took part in the FFIT programme at different times, is not straightforward. Consequently, we performed multiple analyses at single time points. Mediators that were measured at 12 weeks and 12 months were analysed only in the FFIT-FU-I group, as the FFIT-FU-C group had yet to receive the intervention. The groups were combined for mediators measured at 3.5 years. The full results of the mediator analyses are presented in Report Supplementary Material 9. No 12-week variables (apart from objectively measured weight in the FFIT-FU-I group) showed any associations with long-term weight outcomes.

Table 14 shows that increases in self-reported total and vigorous PA, positive affect and physical HRQoL at 12 months all predict a lower weight at 3.5 years. In addition, increased total, vigorous and moderate PA, reduced intake of fatty and sugary foods, increased intake of fruit and vegetables, smaller portions of cheese, meat, pasta and chips, and increased positive affect, self-esteem and physical HRQoL at 3.5 years are associated with a lower weight at 3.5 years. Finally, increased time spent sitting at 3.5 years is associated with a higher weight at 3.5 years.

TABLE 14

Significant behavioural and psychological mediators of weight loss at 3.5 years

Table 15 shows that higher scores in relation to the self-determination theory constructs of autonomous regulation for diet and PA, perceived autonomy for diet and PA, perceived competence for diet and PA, and relatedness with peers (intimacy) and family members (acceptance), as well as higher satisfaction with current diet and PA, were associated with a lower weight at 3.5 years. In contrast, greater amotivation in relation to diet and PA was associated with a higher weight at 3.5 years.

TABLE 15

Significant mediators of long-term weight loss in relation to self-determination theory constructs and satisfaction with health behaviours assessed at 3.5 years

Table 16 shows that self-reported routinisation of PA (walking, gym attendance and taking other forms of exercise), and healthy eating behaviours (eating regular meals, reducing portion sizes and reducing intake of fatty and sugary foods, sugary drinks and calories), self-monitoring of weight, reading the labels on food packaging, and ongoing contact with other participants and coaches from the FFIT programme were associated with a lower weight at 3.5 years.

TABLE 16

Significant mediators of long-term weight loss in relation to routinisation of health behaviours, use of BCTs, contact with people from the FFIT programme, and major life events assessed at 3.5 years

Table 17 shows that greater weight loss at the end of the initial 12-week active phase of the FFIT programme (from objectively measured weight at 12-week RCT measurements in the FFIT-FU-I group, and self-reported retrospectively at 3.5 years by the FFIT-FU-C group) was associated with a lower weight at 3.5 years. Weight gain in the 12 months before starting the FFIT programme (from the RCT 12-month measurements in the FFIT-FU-C group only) was associated with a higher weight at 3.5 years.

TABLE 17

Significant mediators of long-term weight loss in relation to attendance at FFIT sessions and weight change during the RCT

Table 18 demonstrates that self-reported injuries that limit walking at 12 months predicted higher weight at 3.5 years, and that self-reported lower limb joint pain frequency, upper body and lower limb joint pain that limit activities, injuries that limit walking and using stairs, and total types of limitations due to injury were all positively associated with weight at 3.5 years.

TABLE 18

Significant mediators of long-term weight loss in relation to self-reported injury and joint pain, assessed at 12 months or 3.5 years

Retention to the Football Fans in Training follow-up study

The use of the strategies that had been successful in achieving a retention rate of 92% at the 12-month RCT measurements⁴⁴ resulted in the retention of 65% of the original RCT cohort at the 3.5-year follow-up study. There were some differences between men who took part in the 3.5-year measurements and those who did not. Men who were followed up were slightly older, more likely to be in paid work and more likely to own their home than those who were not followed up, perhaps reflecting a more stable lifestyle among those who were followed up. Men who took part in the 3.5-year measurements were also less overweight and had lower BP at baseline than those who did not take part in the follow-up study. Nevertheless, there were no differences in the RCT 12-month weight outcomes between men who were followed up and men who were not followed up, suggesting that success at maintaining weight loss to 12 months did not influence participation in the 3.5-year measures.

Long-term weight loss

Long-term weight trajectories

Between 12 months and 3.5 years, as expected, there were marked differences in weight trajectories between the groups. Men in the FFIT-FU-I group regained almost half of the weight they had lost (2.59 kg, 95% CI 1.61 to 3.58 kg) compared with men in the FFIT-FU-C group who lost 2.03 kg (95% CI 1.08 kg to 2.98 kg) over this period. This difference probably reflects the fact that men in the comparison group had the opportunity to take part in the FFIT programme at the start of this period, whereas those in the intervention group had had this opportunity 12 months earlier. As a result, the 12-month to 3.5-year weight trajectories of men in the FFIT-FU-C group reflect both their initial weight loss during the FFIT programme and its subsequent maintenance or lack of maintenance, whereas the weight trajectories of the FFIT-FU-I group simply reflect maintenance or lack of maintenance post programme.

Long-term secondary outcomes in the intervention group

Men in the FFIT-FU-I group showed sustained improvements in physical, behavioural and psychological outcomes at 3.5 years. Improvements in physical outcomes (in addition to weight) included sustained reductions in BMI, waist circumference, percentage body fat, and systolic and diastolic BP. Self-reported PA was higher at 3.5 years than at baseline and, on average, men sat less, continued to eat fewer fatty and sugary foods, continued to eat more fruit and vegetables, had smaller portion sizes and showed a sustained reduction in the amount of alcohol that they reported drinking. This group also showed sustained improvements in self-esteem, positive and negative affect, and physical and mental HRQoL at 3.5 years.

Long-term secondary outcomes in the comparison group

Men in the FFIT-FU-C group showed improvements from baseline in all secondary outcomes, and to levels similar to those seen in the FFIT-FU-I group. Thus, at 3.5 years, BMI, waist circumference, and systolic and diastolic BP were all reduced in the FFIT-FU-C group. Men reported increased PA and decreased sitting time, and also showed improvements in consumption of fatty and sugary foods, fruit and vegetables, alcohol intake and portion sizes. Self-esteem, positive and negative affect, and physical and mental HRQoL were also improved at 3.5 years.

Long-term secondary outcome trajectories

As with weight, men in the FFIT-FU-I group showed some attenuation of the impact that the FFIT programme had across other physical, behavioural and psychological outcomes over the long term, although all were still significantly better at 3.5 years than they had been at baseline. There was some regain of waist circumference and BMI, and increases in systolic and diastolic BP, as well as reductions in self-reported total and vigorous PA, fruit and vegetable consumption and mental HRQoL between 12 months and 3.5 years. Nevertheless, there was no significant reduction during this period in levels of moderate PA and walking. In relation to diet, men were still managing to limit their intake of sugary foods and alcohol and their portion sizes of cheese and meat. Finally, improvements in all psychological outcomes, apart from mental HRQoL, were sustained from 12 months to 3.5 years.

Again, as with weight, the long-term secondary outcome trajectories of the FFIT-FU-C group reflect both the initial impact of the FFIT programme and the subsequent maintenance, or lack of maintenance, of the changes made. This group showed significant improvements from 12 months to 3.5 years in BMI but not in waist circumference or BP (the lack of improvement in BP may reflect our protocol at the RCT baseline measures, whereby any man with an elevated BP recording was advised to consult his GP). Men in this group also demonstrated increases in self-reported total and moderate PA, and in walking. Dietary improvements between 12 months and 3.5 years included reductions in the consumption of fatty and sugary foods (but no significant changes in fruit and vegetables intake), in portion sizes of all key foods and in alcohol intake. Finally, men also showed improvements in self-esteem, positive affect and mental HRQoL between 12 months and 3.5 years.

Significant between-group differences in the trajectories of most of these secondary outcomes reflect the fact that many FFIT-FU-C men took part in the FFIT programme immediately after the 12-month measures (i.e. in autumn 2012).

Baseline predictors of successful long-term weight loss

None of the prespecified baseline predictors showed a significant relationship with weight loss at 3.5 years in either group.

Mediators of successful long-term weight loss