Toward the optimal strategy for sustained weight loss in overweight cancer survivors: a systematic review of the literature

A systematic review of the literature was conducted. Six databases (PubMed, Embase, Psychinfo, Web of Science, Cinahl and Central) were searched for relevant papers in January 2016. The following search terms were used: ((‘nutritional status’ OR (‘nutritional’ AND ‘status’) OR ‘nutrition’ OR ‘nutritional sciences’ OR (‘nutritional’ AND ‘sciences’) OR ‘diet’ OR ‘dietary’ OR ‘dietary supplements’)) AND (‘neoplasms’ OR ‘cancer’ OR ‘oncology’) AND (‘cancer patients’ OR ‘cancer survivors’) AND ((‘Intervention Studies’ OR ‘intervention’ OR ‘counselling’ OR ‘counseling’ OR ‘nutritional support’ OR (‘nutritional’ AND ‘support’) OR (‘nutrition’ AND ‘support’) OR ‘nutrition support’ OR ‘health promotion’)).

References that were retrieved from the database searches were exported to Endnote X5 and combined into one database with all retrieved references. Of duplicate references one was deleted. Two researchers (STAJ and MH) simultaneously and independently screened and labelled the titles, abstracts and the full-texts of all retrieved papers. First, the titles of the retrieved papers were screened and labelled in Endnote. If at least one of the researchers indicated that an abstract should have been read based on screening the title, both researchers subsequently read the abstract. Second, abstracts were screened and labelled in Endnote. If at least one of the researchers indicated that a full-text should have been read based on screening the abstract, both researchers subsequently read the full-text of the article. Finally, full-texts were read and labelled independently by both researchers.

We included original research articles describing the results of a lifestyle intervention (including a diet component) in adult (≥18 years) overweight (BMI ≥ 25) cancer survivors. Since it is expected that readiness to adopt long-term health behaviour changes is enhanced after completion of initial treatment when patients are primarily coping with the treatment and its side effects, only interventions that have been applied after completion of initial treatment (i.e. surgery, chemotherapy and radiotherapy) were included. Hormonal therapy was not considered to be initial treatment.

Because lifestyle interventions without a diet component are not likely to be able to achieve long-term weight loss maintenance, lifestyle interventions aiming to promote exercise or physical activity alone were excluded. Furthermore, a paper was excluded when it described non-human research, when a paper was not written in the English language, when it did not involve a lifestyle intervention, when the study population did not consist of overweight cancer survivors only, when the paper did not involve original research, when the intervention was not delivered after completion of initial treatment, when the lifestyle intervention did not include a diet component, when no results of the intervention were described, when weight was not included as an outcome and when the study population was younger than 18 years of age. A paper was also excluded when no abstract or full-text was available (e.g. in case of a congress abstract).

Inconsistencies between the researchers with regard to whether or not a paper should have been included in the review were discussed until consensus on inclusion or exclusion of the paper was achieved.

The following data were extracted from the included articles: first author, year of publication, country, study design, type of cancer, sample size (total sample size, and if applicable sample size of the intervention and the control group), sex (percentage of female participants), mean age with standard deviation (SD), time after diagnosis or treatment, dropout rate, duration and type of intervention (physical activity plus diet vs. diet only), follow-up after the end of the intervention, mean baseline BMI and body weight in kilogrammes (kg) with SD, mean weight change in kilogramme with SD and percent weight change from baseline (Table 1). Table 1 provides an overview of the effect of the included lifestyle interventions on weight loss and weight loss maintenance. An intervention was considered to be effective in inducing weight loss when mean weight loss from preintervention to postintervention was significantly (p < 0.05) higher in the intervention group compared with mean weight loss in the control group in a randomized controlled trial (RCT). In case an RCT compared two or more interventions (e.g. two different diets) or in single arm pretest-posttest studies, an intervention was considered to be effective when a significant difference (p < 0.05) in weight between preintervention and postintervention was found. An intervention was considered to be effective in inducing long-term weight loss maintenance when mean weight loss from preintervention to 1 year postintervention was significantly (p < 0.05) higher in the intervention group compared with mean weight loss in the control group or when significant (p < 0.05) weight loss was found from baseline to 1 year follow-up after the end of the intervention for single arm studies. Long-term weight loss maintenance was defined as weight loss of at least 10% of body weight maintained for at least 1 year [21]. Study results were interpreted in the context of study design. An RCT with a usual care control group, an attention control or a less intensive intervention control group was considered to be the preferred study design with regard to interpretation of the effectiveness of the study and is referred to as a robust study design.

The following characteristics of the included lifestyle interventions were extracted and described in Table 2: the aims of the intervention, the theoretical framework on which the intervention was based, a description of the control condition and details on intervention components [26], such as by whom the intervention was delivered, the frequency and length of intervention contacts, the format of intervention contacts, the context in which the intervention was delivered and the content of the intervention. BCTs were used to describe the content of the intervention.

The Behaviour Change Technique Taxonomy version 1 (BCTTv1) was used to extract the BCTs that were used in the included interventions [28]. The BCTTv1 provides detailed definitions of 93 BCTs and includes examples of each BCT. This taxonomy has shown to be a reliable method for extracting information about intervention content and identifying potentially active ingredients associated with effectiveness [29, 30].

Two researchers (MH, MvS) independently coded intervention and control group content of all included interventions using the BCTTv1. Both coders were trained in applying the BCTTv1. When both coders independently coded the same BCT, the BCT was considered to be present. When only one of the coders coded a particular BCT, that BCT was discussed and only considered to be present if consensus was reached. When discrepancies could not be resolved through discussion, a third experienced coder was consulted (SM), and the BCT was considered to be present when two out of three coders deemed the BCT to be present.

When the authors of an included study referred to another publication for further details on intervention content, this other publication was also used to code BCTs. When authors of an included study mentioned that the content of their intervention was (partly) described elsewhere, but no reference was provided, only the description of the intervention content as mentioned in the included publication was used to code BCTs.

BCTs were identified and synthesized across interventions that were found to be effective after evaluation in an RCT with a usual care control, an attention control or a less intensive intervention control group. The frequency of identified BCTs was quantified across these effective interventions (see Table 3).

Most of the included interventions were conducted in the USA (n = 22) [42‐44, 46‐59, 61‐63, 65, 66, 69‐72]. The other interventions were conducted in Canada [45], Spain [67], Italy [60], Australia [64] and the Netherlands [49] (see Table 1). The majority of the included interventions were offered to overweight breast cancer survivors (n = 17) [44‐66, 68, 69, 79], of which five were offered to postmenopausal overweight breast cancer survivors [44‐49]. The other interventions were offered to overweight endometrial cancer survivors (n = 2) [71, 72], overweight colorectal cancer survivors (n = 1) [41], a mixed study sample of overweight breast and endometrial cancer survivors (n = 1) [70] and a mixed study sample of overweight colorectal, breast and prostate cancer survivors (n = 1) [42, 43]. The number of participants varied from 10 [57] to 1510 [62] but was relatively low (n < 50) in the majority of the interventions (n = 16) [41, 44, 45, 48, 51‐53, 55‐57, 63‐66, 68, 69, 71, 79]. Fifteen papers described 13 feasibility or pilot studies [41, 44, 45, 49‐51, 55‐57, 63, 64, 68‐70, 79]. Intervention effects were assessed using an RCT (n = 16) [42, 43, 48‐56, 58, 59, 61‐63, 66, 71, 72], a single arm pretest-posttest design (n = 10) [41, 44, 45, 57, 60, 64, 65, 68‐70, 79] or a three-arm non-RCT (n = 1) [46, 47]. Two of the 16 RCTs had a wait-list control design [42, 43, 53], two compared the effect of two different diets [48, 52], one assessed the additional effect of spirituality counselling in addition to a combined dietary and physical activity intervention [51], 10 had a usual care control, an attention control or a less intensive intervention control group (i.e. a robust study design) [50, 54‐56, 58, 59, 61‐63, 66, 71, 72] and in one RCT, no details about the control group was mentioned [49]. The three-arm non-RCT had a usual care control group [46, 47]. Effects were mostly assessed directly after the end of the intervention only (n = 19). However, in 8 out of the 27 intervention studies, a follow-up assessment after the end of the intervention was conducted, at 12 weeks [45], 6 months [53, 54, 64, 71, 72], 12 months [43] and 2 years after the end of the intervention [49].

In 22 out of the 27 interventions, a statistically significant decrease in body weight was found directly after intervention completion [42‐50, 53‐61, 64‐66, 68‐72, 79], with a weight loss of <5% of baseline body weight after half of the effective interventions, a weight loss of 5–10% after 7 interventions [48, 54‐56, 58‐60, 64, 69] and a weight loss of ≥10% after 3 interventions [44, 46, 47, 68, 79]. Percent weight loss from baseline could not be calculated for one effective intervention, since mean baseline body weight was not reported [49]. Seven out of the eight intervention studies with a follow-up assessment after intervention completion showed a significant decrease in body weight at follow-up of 2.07 to 5.84% of baseline weight [43, 45, 49, 53, 64, 71, 72] (Table 1).

After exclusion of non-randomized studies and feasibility or pilot studies, nine robust studies describing eight interventions reported a statistically significant higher reduction in body weight in the intervention group compared with the control group directly after intervention completion [42, 43, 53, 54, 58, 59, 61, 66, 71, 72]. Contrary, one intervention was not found to be effective in inducing weight loss directly after intervention completion [62]. The amount of weight loss in the intervention group varied across these eight effective interventions from 2.4% [42] to 6.8% loss of baseline weight [58, 59]. A loss of <5% of baseline weight was found after intervention completion in six interventions [42, 53, 61, 66, 71, 72, 80], whereas weight loss of 5–10% was found after two interventions [54, 58, 59]. Two robust studies reported a significant higher reduction in body weight in the intervention group compared with the control group at 6-month follow-up after the end of two 6-month interventions [71, 72]. Both studies reported a mean decrease of ~3% of baseline body weight at follow-up in the intervention group [71, 72].

Characteristics and intervention components are described for the eight interventions that were effective in inducing weight loss after evaluation in a robust study [42, 43, 53, 54, 58, 59, 61, 66, 71, 72] (Table 2).

All interventions promoted both diet and physical activity to induce weight loss. Weight loss goals of these effective interventions varied across studies: five studies aimed for a specific weight loss goal (5–10%) [42, 43, 61, 75, 78], one study more generally aimed to facilitate a modest sustained weight loss without prescribing a percentage of weight loss [58, 59] and in three studies, no specific weight loss goal was mentioned [53, 54, 66].

A theoretical framework on which the intervention was based was reported in six out of eight interventions. Most interventions (n = 5) [42, 43, 54, 61, 71, 72] were based on ‘social cognitive theory’ [73]. Other theoretical frameworks included the ‘transtheoretical model’ [42, 43, 74], behavioural determinants model [61, 76], motivational interviewing [61, 77], cognitive behavioural therapy [58, 59, 61, 75, 78] and standard behavioural treatment for obesity [58, 59].

In most interventions, registered dieticians were involved in applying the intervention [54, 61, 66, 71, 72]. In one study, the intervention was applied by a dietician trained in exercise physiology and behaviour modification counselling alone [54]. In other studies, the dietician applied the intervention together with a ‘primary investigator’ [71], counsellors with backgrounds psychology and exercise physiology [61], exercise physiologists trained in medical rehabilitation [66] and with a ‘physician, a psychologist and a physical therapist’ [72]. Other interventions were applied by a ‘health counsellor’ [42, 43], an ‘instructor’ and a ‘trainer’ [53], as well as ‘trained investigators and research staff’ [58, 59].

Intervention duration varied from 12 weeks [66] to 24 months [61]. Duration of most interventions was 6 months (n = 4) [53, 54, 71, 72]. Frequency of intervention contacts decreased over time in the majority of these effective interventions [42, 43, 54, 58, 59, 61, 71, 72], mostly from weekly contacts in the first weeks or months of the intervention to monthly contacts towards the end of the intervention.

Five out of the eight interventions consisted of both individual and group counselling [53, 58, 59, 61, 71, 72], and three consisted of individual counselling only [42, 43, 54, 66]. Most interventions consisted of a combination of face-to-face and telephone counselling (n = 5) [53, 58, 59, 61, 71, 72]. Three of these interventions also included contacts via e-mail and/or newsletters [61, 71, 72]. One intervention consisted of contacts through telephone and mailed print materials only [42, 43], and in one study, it was determined by group assignment whether participants received face-to-face counselling only or telephone counselling only [54].

In three interventions, it was not mentioned where the intervention took place [58, 59, 71, 72]. One intervention was fully home-based [42, 43]. Three interventions where partly home-based [54, 61, 66] and also involved face-to-face meetings (e.g. at an exercise facility) [66]. One intervention was clinic based and included exercise sessions at a commercial fitness centre [53].

In line with recent reviews on the effects of lifestyle interventions in cancer survivors with mixed diagnosis [34, 40], the vast majority of the included interventions were offered to female cancer survivors. Therefore, our results may not be generalisable to male cancer survivors. Since our review aimed to provide a broad overview of the scientific literature on lifestyle interventions in overweight cancer survivors after completion of initial treatment, we did not only include high-quality studies. Of the included studies in our review, a relatively small proportion had a robust study design. Moreover, in line with other reviews on the effects of lifestyle interventions in cancer survivors [35, 36, 38, 81], sample sizes of the included studies were generally small. We did not conduct a quality assessment of the included studies. We did, however, focus on the studies with a robust study design in the interpretation of our findings.

Due to heterogeneity across included studies in timing, duration, intensity and content of the intervention, we did not conduct a meta-analysis to estimate a mean overall intervention effect on weight. Moreover, since only one out of the nine studies with a robust study design did not report a significant intervention effect, we could not compare components of effective and ineffective interventions. Although it is possible to detect patterns and generate testable hypotheses about likely effective components, there is not the power to be able to draw conclusions about effective intervention components. Therefore, our findings with regard to the intervention components used in effective interventions should be interpreted with caution. Finally, due to inadequate or incomplete description of intervention content in the included publications, the number and the variety of BCTs used in each intervention may have been underestimated.