Background

The World Health Organization considers people with class I obesity if they have a body mass index (BMI) ≥ 30 kg/m2, class II obesity if they have a BMI ≥ 35 kg/m2, and class III obesity with a BMI ≥ 40 kg/m2 [1]. Obesity is considered a risk factor for diseases such as hypertension, heart failure, coronary heart disease [2], diabetes mellitus [2, 3], sleep apnea, and osteoarthritis [3]. Compared to normal weight, class II and III obesity are related to higher mortality rates from all causes [4]; the average survival rate is reduced by 2–4 years for people with class II obesity and by 8–10 years for people with class III obesity [5]. In economic terms, obese people have worse results in the world of work, lower wages, and higher health costs [6].

Weight and fat loss are related to improvements in obese people’s health, a reduction in the inflammatory markers associated with diabetes [7], a decrease in blood pressure [8], and improvements in terms of cardiovascular diseases [7]. Bariatric surgery is an effective treatment option for reducing weight in people with morbid obesity [9], on average losing around 12% of total body weight in the 6 months after surgery, and up to 45% over 3 years [10]. Physical activity (PA) is also one of the main approaches that influences and improves people’s health [11], decreasing cardiovascular risks [12] and coronary heart diseases [13]. Exercise training diminishes comorbidities related to obesity, such as asthma and sleep problems [14] as well as reducing insulin resistance, hypertension, and blood lipids [15]. Furthermore, PA plays an important role in the amount of weight recovered after weight loss and helps reduce weight progressively [16].

Given the increased number of people with morbid obesity [17], the proven short- and long-term effectiveness of bariatric surgery [18] and the possibilities presented by PA in relation to maintenance and improvement of risk factors suggest that PA could help those patients who suffer weight regain after bariatric surgery [19]. The present review and meta-analysis arise from the need to accurately assess whether PA following bariatric surgery has a positive effect on weight loss and to try to determine what type of exercise is most effective for that purpose. Previous systematic reviews [20, 21] and one meta-analysis [22] have studied weight loss caused by exercise following bariatric surgery, although the authors failed to take into account the specific characteristics of each training program, which can undoubtedly influence their effect.

The aims of this systematic review and meta-analysis were (a) to analyze the effects of exercise training after bariatric surgery in relation to weight loss; and, in case of being effective for a higher weight loss, (b) to determine what type of training is the most appropriate for weight loss in people undergoing bariatric surgery.

Material and Methods

Protocol and Registration

This study has been carried out according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses, the PRISMA Statement [23], using the PRISMA checklist as a reference (Table A supplementary files), and the Cochrane Handbook for Systematic Reviews of Interventions [24]. This systematic review and meta-analysis has been registered in PROSPERO, the International Prospective Register of Systematic Reviews, with the ID CRD42018097444.

Searches

The literature search was performed systematically using the MEDLINE, EMBASE, Scopus, Cochrane, and Web of Science databases, with the deadline of May 23, 2019.

The following equation was used for the search (“Bariatric Surgery” OR “stapling stomach” OR “weight loss surgery” OR “obesity surgery” OR “weight reduction surgery” OR “Biliopancreatic Diversion” OR “Duodenal switch” OR “laparoscopic band” OR “lap band” OR “gastric band” OR “gastric banding” OR “Gastric Bypass” OR “Gastroplasty” OR “gastric sleeve” OR “sleeve gastrectomy” OR “gastric bypass surgery” OR “gastric bypass” OR “Roux-en-Y Gastric Bypass” OR “Maestro Rechargeable System” OR “gastric balloon” OR “gastric bubble” OR “ballobes balloon” OR “Greenville gastric bypass”) AND (“physical exercise” OR “Physical Therapy” OR “physical activity” OR “physical education” OR “physical training” OR exercise OR fitness OR sport OR “Exercise Movement” OR “exercise program” OR “Complementary Therapies” OR “physiotherapy” OR “physio therapy” OR “therapeutic exercise” OR “Occupational Therapy” OR “Exercise therapy”) AND (“body mass index change” OR “weight maintenance” OR “weight loss” OR “weight regain” OR obesity OR overweight).

Eligibility Criteria

Articles showing either the effect of physical activity on patients who had undergone bariatric surgery or which carried out an experimental intervention were included as eligible for further review.

The exclusion criteria for this systematic review were (a) papers not written in English or Spanish; (b) studies that do not report the outcome weight; (c) studies in which the intervention is performed before bariatric surgery; (d) studies in which the population investigated are non-humans; (e) studies in which participants are under 18 years old; (f) papers that combine physical activity with other types of intervention, medications, and nutrition among others; (g) retracted studies; (h) duplicate studies; and (i) non-selectable publications, as in the case of reviews, guidelines, interviews, comments, or case studies.

The literature review was independently and simultaneously performed by two reviewers (AC and IC-R). Disagreements were sorted out either through consensus or with the participation of a third party (EG).

Data Selection

The data gathered from the selected studies were as follows: (a) the year of the study; (b) the study design; (c) the main features and the type of physical activity intervention; (d) the population’s characteristics, number, sex, and age; and (e) the pre- and post-surgery weight.

Assessment of the Risk of Bias

For randomized controlled trials (RCTs), the Cochrane Collaboration’s tool was used for assessing risk of bias in randomized trials [25]. This tool measures the risk of bias based on six domains: selection bias, performance bias, detection bias, attrition bias, reporting bias, and other biases. For non-randomized control trials (non-RCTs), the quality assessment tool for quantitative studies [26] was used. This tool considers seven domains: selection bias, study design, confounders, blinding, the data collection method, and withdrawals and dropouts.

To combine both tools for evaluating the risk of bias for the reviewed articles, each of the sections to be evaluated was designated as strong, moderate, or weak, and the articles were classified as having a low risk of bias (without weak ratings), a moderate risk of bias (one weak rating), or a high risk of bias (two or more weak ratings) [27, 28].

Quality assessment and data extraction were carried out independently and simultaneously by two reviewers (AC and IC-R). Any differences were sorted out either through consensus or with the participation of a third researcher (EG).

Statistical Analyses

The software used to perform the statistical analyses of the meta-analysis was StataSE V.14.0 (StataCorp LP., College Station, TX, USA). The analysis variable (the result) was the weight loss at the end of the treatment in each of the groups; the quantification of the effect was calculated through the standardized mean difference (SMD) and through the unbiased Hedges estimator [29]. Weightings and standard deviations (SDs) were extracted for each study and group before and at the end of the treatment. In the case of not having the SD values, these were imputed as the average value of each group [30]. On the other hand, the differences of the weight means and the standard deviation of the difference for each of the articles were calculated; for the latter, a correlation of 0.59 was considered between the values before and after starting the treatment [31]. A positive SMD value indicated greater weight loss in the intervention group compared to the control group. The DerSimonian-Laird random effects method was used and the 95% confidence intervals (CI) were calculated [32].

As a test to evaluate heterogeneity, we estimated the I2 statistic [25]—values of 0% indicated non-heterogeneity, whereas values of 25%, 50%, and 75% were interpreted as having a low, moderate, and high level of heterogeneity, respectively. In addition, the Q statistic and its P value were calculated.

The publication bias was evaluated by the funnel plot and the Egger test was performed.

To complete the statistical analysis, certain details were contemplated: (I) when two or more studies obtained the data from the same database, only the main study was taken into account; (II) those papers with two intervention groups were analyzed as two individual studies; and (III) pre-operative body weight data were collected using the baseline and post-intervention data from the evaluation performed immediately after the surgery.

Subgroups analyses were carried out considering the following characteristics of the training intervention protocols: (I) the type of physical activity intervention, (II) the start of the intervention after the surgery, (III) the duration of the intervention, (IV) the type of exercise, and (V) the total exercise time per week.

Meta-regressions with random effects were employed using the aggregate level data to know the effect of the intervention and the heterogeneity in relation to (I) the average age of the participants, (II) the time per session, and (III) the length of the intervention.

Finally, a sensitivity analysis was conducted by performing the calculations again, without each of the studies, to know the robustness of that particular study.

Results

Systematic Review

Of the 10543 studies obtained in the search, 26 documents were finally selected (Fig. 1). These were carried out in different countries—10 from the United States, 6 from Brazil, 2 from Iran, 4 from Denmark, and 1 from Belgium, Sweden, Italy, and the UK, respectively. All the studies included were experimental, most of them with an RCT design (21) while 5 were non-RCTs. All 26 studies were published between 2011 and 2018 (Table 1).

Fig. 1
figure 1

Literature search preferred reporting items for systematic reviews and meta-analyses consort diagram

Full size image
Table 1 Studies included in the systematic review
Full size table

The internationally accepted criteria to undergo bariatric surgery are a BMI ≥ 40 kg/m2 or a BMI ≥ 35 kg/m2 if comorbidities are present that put the patient’s health at risk. Most of the selected studies included any type of surgical technique, apart from nine of them which included only laparoscopic surgery with RYGB, and one study on the sleeve gastrectomy technique. The sample size of the studies varied between 12 and 120 subjects, most of them including participants of both sexes, except for 5 studies in which only women participated.

The physical activity interventions carried out were supervised in 12 of the studies, programmed in 4, while 8 were mixed interventions, combining programmed and supervised. Only 1 study used a counseling intervention type. Aerobic exercise alone was used in 11 studies, resistance exercise alone in 2 of them, whereas a combination of both was used in 10 studies, and other alternative types of exercise in 3 studies. The reviewed studies used exercise durations between 20 and 85 min per session. The maximum length of the intervention was 40 weeks and the shortest was 1 week. The start of the intervention varied between 5 days and 24 months after surgery.

Risk of Bias

Regarding the risk of bias, as can be seen in Tables 2 and 3, from the 26 studies analyzed, 10 of them (38.5%) show a high risk of bias, 6 (23.1%) show a moderate risk of bias, and the other 10 (38.5%) show a low risk of bias. Analyzing the sections considered in both assessment tools individually, we can observe that all non-RCTs (n = 5, 100%) present deficiencies (i.e., scored as weak) in the blinding domain study, while in the case of RCTs, 38.1% (n = 8) and 14.3% (n = 3) show detection bias and performance bias, respectively. Among the non-RCT studies, deficiencies in the selection bias and the design of the study were present in 40% of the studies (n = 2) compared to 5.3% (n = 1) showing selection bias among the RCTs.

Table 2 The Cochrane collaboration’s tool for assessing the risk of bias in randomized trials
Full size table
Table 3 The quality assessment tool for quantitative studies
Full size table

Meta-Analyses

The pooled SMD estimate did not show a greater significant weight loss in favor of the intervention (exercise) group, with a small effect size and no significant differences (SMD = 0.15; 95% CI = − 0.02, 0.32; p = 0.094), as well as low heterogeneity (I2 = 0%; p = 0.999) (Fig. 2).

Fig. 2
figure 2

Forest plot of the weight loss standardized mean difference between the control group and the intervention group. SMD standardized mean difference, CI confidence interval

Full size image

Subgroup Analyses

When considering the specific characteristics of the physical activity intervention (i.e., type of intervention, start of the intervention after surgery, duration, type of exercise, and total time per week), none of the subgroups analyses showed a significant difference in favor of the control group or the intervention group, with heterogeneity being very low in all cases (I2 = 0.0%) (Table 4).

Table 4 Stratified analysis according to the characteristics of the exercise program
Full size table

Meta-Regressions

The meta-regression analyses showed no heterogeneity based on the participants’ mean age (p = 0.902), nor on the length of the intervention (p = 0.377) or the time devoted to each exercise session (p = 0.807) (Table 5).

Table 5 Meta-regression of weight loss with mean age, length of intervention, and time per session
Full size table

Sensitivity Analysis and Publication Bias

Once the impact of each study was verified in the final result, eliminating each study individually, no changes were observed in the overall results. As seen in the funnel plot (Fig. 3) and once the Egger test was performed, there was no evidence of significant publication bias risk (p = 0.208).

Fig. 3
figure 3

Funnel plot with Egger test

Full size image

Discussion

Our systematic review and meta-analysis provide insight into how a physical activity program performed after bariatric surgery can affect weight loss. The data obtained from the analysis of a total of 749 pooled bariatric surgeries did not show significant positive results in favor of exercise for people who underwent bariatric surgery. No significant results were found in terms of weight loss in favor of physical exercise after bariatric surgery when compared to the usual postoperative care.

Three previous systematic reviews [20, 21, 33] concluded that exercise following bariatric surgery is positive in increasing weight loss and improving other factors such as muscle loss or cardiovascular risk. Similarly, a recent meta-analysis [22] obtained significant results in favor of physical exercise, in contrast to our results, which are slightly positive for the intervention group, although not significant. This might be because some of the studies included in these previous reviews carried out multi-component interventions; meaning that physical exercise was accompanied by something else. Also, discrepancy between our results and those from previous systematic reviews and meta-analysis may be due to the fact that other authors directly calculated the difference in means between the groups, instead of calculating the difference in the standardized means. In addition, the number of articles included in our meta-analysis is higher than in previous weight-loss reviews, and the studies included in this meta-analysis only performed exercise, without any other type of intervention, such as diet or psychological support.

The results obtained in our systematic review and meta-analysis slightly disagree with the conclusions of previous reviews, showing no significant differences in favor of physical exercise once the bariatric surgery has been completed successfully. The effect (standardized mean difference, SMD) does not achieve statistical significance, which could indicate that (i) longer/more intense/better designed physical exercise programs are needed to elicit greater weight loss after bariatric surgery, or (ii) physical exercise after bariatric surgery must be accompanied by long-term changes in eating habits [16].

In a previous randomized controlled trial by Creasy et al. [34], obese/overweight participants lost on average 3.8 ± 3.0 kg when performing supervised physical activity, compared to an average weight loss of 5.1 ± 3.3 kg among those who were in the programmed (unsupervised) physical activity group. In our meta-analysis, the data reveal a non-significant trend toward greater weight reduction in those studies carrying out programmed (unsupervised) physical activity. This may occur due to an improvement in the psychological processes, self-efficacy, and autonomous motivation produced by performing physical activity after bariatric surgery [35].

Regarding the type of exercise, aerobic training shows the highest effect size, although this is also not significant. In contrast, previous studies on obese and overweight people (not bariatric patients) showed that a combination of resistance and aerobic training is the most effective way to lose weight [36]. Furthermore, a minimum of 150 min/week of moderate intensity physical activity has been proposed for developing and maintaining fitness [37]. In our results, however, the greatest weight loss was observed (although again not significant) in those interventions lasting less than 150 min/week. These two results might be related to the scarcity of studies containing sufficient samples combining aerobic and resistance exercise that entailed more than 150 min/week of exercise in patients who had undergone bariatric surgery. In addition, the studies did not control other factors that may affect weight loss, such as sleep, physical activity outside the program, or nutritional habits, which may condition the study results in relation to weight loss.

Physical activity interventions in obese people lasting less than 16 weeks were associated with increased energy expenditure but not with a reduction in body weight [38, 39]. According to our analysis, interventions lasting more than 16 weeks seem to produce greater weight loss, although this effect was not statistically significant compared to the control groups (no exercise). This might occur due to the increase in energy expenditure and the induction of lipolysis [40]. Some evidence suggests that weight loss could be higher if physical activity were combined with dietetic education, caloric reduction, or changes in eating habits [38, 40].

Limitations

Certain limitations that may have affected the results should be considered. First of all, some studies (five of them) did not report the standard deviation for body weight after the intervention. Secondly, not having the correlation between pre- and post-measures to estimate the standard deviation of the difference—it was for this reason that 0.59 was used. Thirdly, the samples, the inclusion/exclusion criteria, and the follow-up time varied greatly across studies. Moreover, there was a moderate risk of bias for the studies included. Finally, we could not analyze the exercise intensity performed in the reviewed programs because it was not possible to unify the criteria.

Conclusions

Based on the results of our meta-analysis, one can conclude that exercise training in patients who have undergone bariatric surgery does not seem to be effective in achieving greater weight loss compared to the usual postoperative care, and no particular training is more effective than another for losing weight in patients who have undergone bariatric surgery. There is, however, a consensus that large enough RCTs of exercise have yet to be done—so doing meta-analyses of early, small cohorts can be misleading. The reviewed studies used different surgery types, exercise types, exercise durations, and interventions were in general poorly reported, making them difficult to be properly combined. It may be too early to make any conclusions, and reporting null findings could possibly dissuade people from adding exercise to their post-operative lifestyle changes, which could take them away from the myriad of positive effects of exercise. Even if the lack of effects on weight loss could be rigorously confirmed, exercise after bariatric surgery may help maintain lean body mass, improve cardiovascular health, psychological well-being, and increase adherence to training, among other benefits [41].

Future Directions

This line of research, which combines bariatric surgery and exercise training, should continue in order to elucidate the most appropriate type of exercise, as well as to determine the design and implementation of training programs with greater frequency, volume, intensity, and/or with different types of exercises, in such a way that is most appropriate for this population. Studies are also necessary in which patients begin to train straight after surgery to take advantage of the window of opportunity for behavioral change as soon as possible, not only to achieve greater weight loss but also to improve other parameters that affect these patients’ health status such as muscle mass loss, cardiovascular parameters, biochemical markers, and respiratory parameters, among others. As demonstrated in our weight loss meta-analysis, all these research areas require better designed (and better reported) studies with sufficient sample sizes.