https://orcid.org/0000-0002-7908-4213
Abstract
People with type 2 diabetes (T2D) have higher risks of cancer incidence and death.
We aimed to evaluate the relationship between dietary and physical activity–based lifestyle intervention and cancer outcomes among prediabetes and T2D populations.
We searched for randomized controlled trials with at least 24 months of lifestyle interventions in prediabetes or T2D populations. Data were extracted by pairs of reviewers and discrepancies were resolved by consensus. Descriptive syntheses were performed, and the risk of bias was assessed. Relative risks (RRs) and 95% CIs were estimated using a pairwise meta-analysis with both a random-effects model and a general linear mixed model (GLMM). Certainty of evidence was evaluated using the Grading of Recommendations Assessment, Development, and Evaluation framework, and trial sequential analysis (TSA) was conducted to assess if current information is enough for definitive conclusions. Subgroup analysis was performed by glycemic status.
Six clinical trials were included. Among 12 841 participants, the combined RR for cancer mortality comparing lifestyle interventions with usual care was 0.94 (95% CI, 0.81-1.10 using GLMM and 0.82-1.09 using random-effects model). Most studies had a low risk of bias, and the certainty of evidence was moderate. TSA showed that the cumulative Z curve reached futility boundary while total number did not reach detection boundary.
Based on the limited data available, dietary and physical activity–based lifestyle interventions had no superiority to usual care on reducing cancer risk in populations with prediabetes and T2D. Lifestyle interventions focused on cancer outcomes should be tested to better explore their effects.
Type 2 diabetes (T2D), prediabetes, and cancer are major global public health issues (1, 2). According to the IDF Diabetes Atlas, it is estimated that about 10.5% of the world's adult population were living with diabetes in 2021, and this number is expected to increase to 11.3% in 2030 and 12.2% in 2045 (3). Prediabetes is a critical risk factor for the development of diabetes and its related complications, and the prevalence of prediabetes is increasing in many parts of the world (4, 5). Besides that, cancer is becoming one of the leading causes of death worldwide due to the rapid increase in population and aging (6, 7). According to GLOBOCAN, the estimated number of new cancer cases and cancer deaths in 2018 was 17 and 9.6 million, respectively, and the number is still increasing.
Epidemiological evidence suggests that people with T2D are at increased risk of developing and dying from several types of cancers, including but not limited to liver, pancreatic, ovarian, colorectal, lung, breast, hepatocellular, gallbladder, endometrial, and biliary cancers (8-12). Potential hypothesized mechanisms include the effect of hyperglycemia, insulin resistance, hyperinsulinemia, insulin-like growth factor-1 levels, dyslipidemia, increased leptin, decreased adiponectin, and chronic inflammation on cancer risk (8, 10, 11).
Previous studies found that lifestyle interventions may be an effective method of reducing the incidence of T2D among high-risk populations (13, 14), and prospective data showed that adherence to a healthy lifestyle was associated with a lower risk of mortality in those with T2D (15). In addition to the effects on diabetes, lifestyle interventions with dietary changes and increases in physical activity frequency and intensity may also be effective in preventing cancer development and progression (16-18).
To our knowledge, no medications or interventions have been found to be effective in reducing cancer risk in participants with diabetes, and the effects of lifestyle interventions on cancer outcomes are poorly understood. Therefore, this study aimed to evaluate the relationship between dietary and physical activity–based lifestyle interventions and cancer outcomes among diabetes and prediabetes populations. We hypothesized that lifestyle interventions could achieve significant reductions in cancer incidence and mortality directly and indirectly by 1) modifying environmental risk factors for cancer, 2) reducing the incidence of diabetes in the prediabetes populations, and 3) improving clinical outcomes in participants with diabetes. If our hypothesis is true, the effectiveness of lifestyle interventions to reduce adverse health outcomes will be reinforced, and further support the importance and need for public health actions for lifestyle intervention implementation.
Materials and Methods
Search Strategy and Study Selection
The protocol of this systematic review was prospectively registered in the International Prospective Register of Systematic Reviews—PROSPERO (CRD42021207431), and this manuscript is reported in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist.
Based on the PICOTS acronym, available in Box 1, one investigator (K.P.Z.) searched MEDLINE, Cochrane Central Register of Controlled Trials (CENTRAL), Embase, and Web of Science, until August 23, 2022. Database searches were supplemented by review of trial protocol records (eg, Clinical Trials and ISRCTN Registry) and unpublished trials, as well as reference lists of relevant publications. There were no language or date publication restrictions. A list of search terms for each database is available in the supplementary material (19) (Supplementary Table S1). Studies had to be randomized clinical trials (RCTs) with at least 24 months of active intervention, including adults with prediabetes or T2D and comparing intensive lifestyle interventions to active controls. Lifestyle interventions were supposed to be based on dietary and physical activity recommendations, whereas control groups comprised usual care, standard advice, or placebo in studies with a pharmacological treatment arm.
P — Adults with criteria for prediabetes or type 2 diabetes
I — Intensive lifestyle interventions including dietary and physical activity recommendations
C — Less-intensive lifestyle interventions, usual or standard care, general advice, or general advice + placebo
O — Incidence of cancer and cancer death
T — Minimum of 24 mo of intervention
S — Outpatients
After the removal of all duplicates, 5 pairs of trained reviewers independently screened titles and abstracts and evaluated the full text of the selected studies using EndNote reference manager (version X7.17, Thomas Reuters, 2011). The same reviewers extracted data using a prepiloted form. Information retrieved included first author, year of publication, country, baseline characteristics of participants, sample size, total follow-up period, description of the intervention and control groups, incidence of cancer, and cancer mortality. Disagreements in studies’ selection and discrepancies in the data extraction were resolved by consensus or by a third reviewer (F.G.). If one of the study outcomes was missing, study authors were contacted to provide missing data and to resolve any uncertainty during the data extraction process. If the authors did not respond or could not provide the missing information, the study was removed from that portion of the analysis.
In cases where there were multiple treatment groups, we combined all the lifestyle intervention groups as the intervention group and used the group with minimum intervention as the control for the meta-analysis. If there was more than one publication of the same study, we selected the most recent article that reported the outcome within the time point of interest.
Outcomes
This is a prespecified analysis of the previously published protocol (PROSPEROCRD42021207431) with main outcomes of cancer cumulative incidence (cases of cancer diagnosed during the study period divided by the total number of participants) and the number of cancer deaths among participants.
Risk of Bias and Certainty of Evidence
The risk of bias of the included studies was assessed using the Cochrane tool to evaluate the risk of bias (RoB 2.0) (20), according to potential bias from the randomization process, deviations from the intended intervention, missing outcomes data, measurement of outcome, and selection of the reported result. The analysis was conducted by 2 independent reviewers (K.P.Z. and M.Z.) and discrepancies were settled by consensus or a third reviewer (F.G.). Overall study ratings were classified as “high risk of bias,” “some concerns,” or “low risk of bias.”
The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework was used to assess the overall certainty of evidence and was reported as a summary of findings table. The evidence was graded as “high,” “moderate,” “low,” or “very low,” depending on study designs and risk of bias, inconsistency in findings, indirectness, imprecision, publication bias, magnitude of effect, dose-response gradient, and other plausible residual bias or confounding (21).
Data Analysis
We performed a descriptive synthesis of all included studies. A pairwise meta-analysis was conducted using random-effects models with the Der Simonian and Laird method as a variance estimator, and the results were confirmed by the (binomial-normal) general linear mixed model (GLMM). The effects of randomized treatment on the cumulative number of events were estimated as relative risks (RRs) during the overall follow-up periods in accordance with the intention-to-treat principle. The RRs and 95% CIs were estimated from a log-binomial model. Heterogeneity was assessed using the I² statistic. Meta-analysis was not performed when the number of studies reporting a given outcome was less than or equal to 2. The potential for publication bias was not statistically assessed using a funnel plot because there were fewer than 10 studies included in the quantitative synthesis and the power of the test cannot distinguish chance from real asymmetry (22). A P value lower than .05 was considered statistically significant, and meta-analyses were performed in R statistical software version 1.4.17 with package meta v.4.18-2.
For cancer mortality, trial sequential analysis (TSA) was performed to assess if the current information is sufficient for definitive conclusions (23). We used a random-effects model with conservative trial monitoring boundaries set by Lan-DeMets-O’Brien-Fleming as the α-spending function. The information size was calculated based on a type I error of 5%, a β of 20%, a cancer mortality rate of 5.67% in the control group, and an expected RR reduction of 15% in the intervention group, based on the results of previous observational studies on adherence to dietary and physical activity cancer-prevention guidelines and cancer mortality outcomes (24-26). A conservative value of 15% was chosen considering that the RR reported in these studies compared the groups with the highest level of adherence with the groups with the lowest level of adherence and that some of the control groups of the studies included in this meta-analysis had some level of lifestyle intervention. Trial sequential analysis was performed using Trial Sequential Analysis Viewer Version 0.9.5.10 Beta (27).
Subgroup analysis was performed by glycemic status (prediabetes or T2D). Two sensitivity analyses were conducted. One excluded trials reporting cancer deaths as a loss to follow-up or with follow-up period not exceeding 2 years. The other was performed under the assumption that all unclassified deaths were cancer deaths. A statistically significant association was considered if the P value was less than .05.
Results
Characteristics of Included Studies
Of the 23 998 records after the removal of duplicates, we evaluated 492 records for full-text analysis and included a total of 6 studies in this review (as shown in Fig. 1). The reasons for exclusion in each phase are also shown in Fig. 1.
Flowchart of searches and selection of studies.
Six RCTs were included in our study (28-33), contributing a total of 13 139 participants (Table 1). Three trials included populations with prediabetes (29-31), while the remaining trials included participants with diabetes (28, 32, 33). Two trials were conducted in Asian populations (30, 31) and more than half of the participants in the remaining 4 studies were of European ancestry (28, 29, 32, 33). The Da Qing study (31) had the youngest population at study entry, with a mean age of 45.2 years and mean body mass index (BMI) of 25.7. The ADDITION trial (32) had the oldest population, with a mean age of 60.2 years, and the Look AHEAD trial (28) had a population with the largest BMI, with a mean of 35.9.
Main characteristics of included studies
| Study, y | Population | Intervention group | Control group | Cancer incidence | Cancer death | Total follow-up, y | Risk of bias |
|---|---|---|---|---|---|---|---|
| Da Qing, 2019 | Prediabetes, 45.8% female, aged 45.2 ± 9.3 y, 25.7 ± 4.0, 41.3% current smoker | Dietary and/or physical activity prescription | Standard care + brochure about diet and physical activities but no specific advice | NI | INT (43/438, 9.8%) vs CON (14/138, 10.1%) | 30 | Low risk |
| ADDITION, 2019 | T2D, 42.1% female, aged 60.2 ± 6.9 y, 31.6 ± 5.6, 27.34% current smoker | Group or practice-based dietary and physical activity advice | Standard/Routine care according to each center | NI | INT (112/1678, 6.7%) vs CON (99/1378, 7.2%) | 10 | Low risk |
| DiRECT, 2019 | T2D, 40.9% female, aged 54.4 ± 7.6 y, 34.6 ± 4.4, 11.7% current smoker | Dietary prescription and physical activity advice | Standard/Routine care according to each center | INT (0/149, 0%) vs CON (5/149, 3.4%) | NI | 2 | Low risk |
| Thailand DPP, 2019 | Prediabetes, 79.7% female, aged 50.9 ± 6.4 y, 27 ± 4.6, current smoker NI | Group-based activities on healthy lifestyle behaviors | Standard care + one-time education | NI | INT (1/1030, 0.1%) vs CON (0/873, 0%) | 2 | Some concerns |
| DPPOS, 2021 | Prediabetes, 68.5% female, aged 50.4 ± 10.9 y, 34 ± 6.7, 7.1% current smoker | Dietary and physical activity prescription | Standard care + advice to increase exercise and improve diet | NI | INT (60/1079, 5.6%) vs CON (53/1082, 4.9%) | 21 | Low risk |
| Look AHEAD, 2022 | T2D, 59.4% female, aged 58.7 ± 6.8 y, 35.9 ± 6.0, 4.4% current smoker | Dietary and physical activity prescription | Standard care + group education about diet and physical activities | INT (332/2435, 13.6%) vs CON (352/2424, 14.5%) | INT (157/2570, 6.1%) vs CON (177/2575, 6.9%) | 16.7 | Low risk |
| Study, y | Population | Intervention group | Control group | Cancer incidence | Cancer death | Total follow-up, y | Risk of bias |
|---|---|---|---|---|---|---|---|
| Da Qing, 2019 | Prediabetes, 45.8% female, aged 45.2 ± 9.3 y, 25.7 ± 4.0, 41.3% current smoker | Dietary and/or physical activity prescription | Standard care + brochure about diet and physical activities but no specific advice | NI | INT (43/438, 9.8%) vs CON (14/138, 10.1%) | 30 | Low risk |
| ADDITION, 2019 | T2D, 42.1% female, aged 60.2 ± 6.9 y, 31.6 ± 5.6, 27.34% current smoker | Group or practice-based dietary and physical activity advice | Standard/Routine care according to each center | NI | INT (112/1678, 6.7%) vs CON (99/1378, 7.2%) | 10 | Low risk |
| DiRECT, 2019 | T2D, 40.9% female, aged 54.4 ± 7.6 y, 34.6 ± 4.4, 11.7% current smoker | Dietary prescription and physical activity advice | Standard/Routine care according to each center | INT (0/149, 0%) vs CON (5/149, 3.4%) | NI | 2 | Low risk |
| Thailand DPP, 2019 | Prediabetes, 79.7% female, aged 50.9 ± 6.4 y, 27 ± 4.6, current smoker NI | Group-based activities on healthy lifestyle behaviors | Standard care + one-time education | NI | INT (1/1030, 0.1%) vs CON (0/873, 0%) | 2 | Some concerns |
| DPPOS, 2021 | Prediabetes, 68.5% female, aged 50.4 ± 10.9 y, 34 ± 6.7, 7.1% current smoker | Dietary and physical activity prescription | Standard care + advice to increase exercise and improve diet | NI | INT (60/1079, 5.6%) vs CON (53/1082, 4.9%) | 21 | Low risk |
| Look AHEAD, 2022 | T2D, 59.4% female, aged 58.7 ± 6.8 y, 35.9 ± 6.0, 4.4% current smoker | Dietary and physical activity prescription | Standard care + group education about diet and physical activities | INT (332/2435, 13.6%) vs CON (352/2424, 14.5%) | INT (157/2570, 6.1%) vs CON (177/2575, 6.9%) | 16.7 | Low risk |
Abbreviations: CON, control group; INT, intervention group; NI, not informed.
Main characteristics of included studies
| Study, y | Population | Intervention group | Control group | Cancer incidence | Cancer death | Total follow-up, y | Risk of bias |
|---|---|---|---|---|---|---|---|
| Da Qing, 2019 | Prediabetes, 45.8% female, aged 45.2 ± 9.3 y, 25.7 ± 4.0, 41.3% current smoker | Dietary and/or physical activity prescription | Standard care + brochure about diet and physical activities but no specific advice | NI | INT (43/438, 9.8%) vs CON (14/138, 10.1%) | 30 | Low risk |
| ADDITION, 2019 | T2D, 42.1% female, aged 60.2 ± 6.9 y, 31.6 ± 5.6, 27.34% current smoker | Group or practice-based dietary and physical activity advice | Standard/Routine care according to each center | NI | INT (112/1678, 6.7%) vs CON (99/1378, 7.2%) | 10 | Low risk |
| DiRECT, 2019 | T2D, 40.9% female, aged 54.4 ± 7.6 y, 34.6 ± 4.4, 11.7% current smoker | Dietary prescription and physical activity advice | Standard/Routine care according to each center | INT (0/149, 0%) vs CON (5/149, 3.4%) | NI | 2 | Low risk |
| Thailand DPP, 2019 | Prediabetes, 79.7% female, aged 50.9 ± 6.4 y, 27 ± 4.6, current smoker NI | Group-based activities on healthy lifestyle behaviors | Standard care + one-time education | NI | INT (1/1030, 0.1%) vs CON (0/873, 0%) | 2 | Some concerns |
| DPPOS, 2021 | Prediabetes, 68.5% female, aged 50.4 ± 10.9 y, 34 ± 6.7, 7.1% current smoker | Dietary and physical activity prescription | Standard care + advice to increase exercise and improve diet | NI | INT (60/1079, 5.6%) vs CON (53/1082, 4.9%) | 21 | Low risk |
| Look AHEAD, 2022 | T2D, 59.4% female, aged 58.7 ± 6.8 y, 35.9 ± 6.0, 4.4% current smoker | Dietary and physical activity prescription | Standard care + group education about diet and physical activities | INT (332/2435, 13.6%) vs CON (352/2424, 14.5%) | INT (157/2570, 6.1%) vs CON (177/2575, 6.9%) | 16.7 | Low risk |
| Study, y | Population | Intervention group | Control group | Cancer incidence | Cancer death | Total follow-up, y | Risk of bias |
|---|---|---|---|---|---|---|---|
| Da Qing, 2019 | Prediabetes, 45.8% female, aged 45.2 ± 9.3 y, 25.7 ± 4.0, 41.3% current smoker | Dietary and/or physical activity prescription | Standard care + brochure about diet and physical activities but no specific advice | NI | INT (43/438, 9.8%) vs CON (14/138, 10.1%) | 30 | Low risk |
| ADDITION, 2019 | T2D, 42.1% female, aged 60.2 ± 6.9 y, 31.6 ± 5.6, 27.34% current smoker | Group or practice-based dietary and physical activity advice | Standard/Routine care according to each center | NI | INT (112/1678, 6.7%) vs CON (99/1378, 7.2%) | 10 | Low risk |
| DiRECT, 2019 | T2D, 40.9% female, aged 54.4 ± 7.6 y, 34.6 ± 4.4, 11.7% current smoker | Dietary prescription and physical activity advice | Standard/Routine care according to each center | INT (0/149, 0%) vs CON (5/149, 3.4%) | NI | 2 | Low risk |
| Thailand DPP, 2019 | Prediabetes, 79.7% female, aged 50.9 ± 6.4 y, 27 ± 4.6, current smoker NI | Group-based activities on healthy lifestyle behaviors | Standard care + one-time education | NI | INT (1/1030, 0.1%) vs CON (0/873, 0%) | 2 | Some concerns |
| DPPOS, 2021 | Prediabetes, 68.5% female, aged 50.4 ± 10.9 y, 34 ± 6.7, 7.1% current smoker | Dietary and physical activity prescription | Standard care + advice to increase exercise and improve diet | NI | INT (60/1079, 5.6%) vs CON (53/1082, 4.9%) | 21 | Low risk |
| Look AHEAD, 2022 | T2D, 59.4% female, aged 58.7 ± 6.8 y, 35.9 ± 6.0, 4.4% current smoker | Dietary and physical activity prescription | Standard care + group education about diet and physical activities | INT (332/2435, 13.6%) vs CON (352/2424, 14.5%) | INT (157/2570, 6.1%) vs CON (177/2575, 6.9%) | 16.7 | Low risk |
Abbreviations: CON, control group; INT, intervention group; NI, not informed.
Among the 6 trials, the duration of total follow-up ranged from 2 to 30 years, with a mean and median of 13.61 and 13.35 years, respectively. The Thailand DPP (30) and DiRECT (33) were RCTs of a 2-year intervention, whereas the other studies had posttrial phases (28, 29, 31, 32). The Da Qing study (31) had a 6-year active intervention and a 24-year posttrial follow-up. Both the ADDITION (32) and DPPOS (29) trials had a 5-year intervention phase, and an additional duration of posttrial observations. The Look AHEAD study (28) conducted an active intervention for 4 years, after which there was a more flexible and less-intense intervention period of follow-up.
Smoking status of participants was reported in 5 of the 6 studies (28, 29, 31-33), with only the Thailand DPP having no information (30). The highest number of baseline smokers was 41.3% in the Da Qing study (31) followed by 27.3% and 11.7% in the ADDITION (32) and DiRECT (33) trials, respectively. Less than 10% of the participants in the DPPOS (29) and Look AHEAD (28) trials were current smokers. Alcohol consumptions were reported differently across studies. In the Look AHEAD study (28), 59.7% of the participants reported consuming alcohol in the past 2 years. Of all those who responded to the question about alcohol consumption in the DPPOS study (29), 75.3% reported less than 1 drink per week, while 4.4% reported more than 7 drinks per week. The DiRECT trial (33) reported an average of 5.5 units per week of alcohol drinking, and the median weekly alcohol intake in the ADDITION trial (32) was 4 units. Baseline alcohol consumption was not reported in the Thailand DPP and Da Qing trials (30, 31). Participants in the intervention group of the Da Qing study were advised to reduce alcohol intake (31), while all participants in the DPPOS study were encouraged to avoid excessive alcohol intake and to quit smoking (29).
In terms of lifestyle interventions, the Thailand DPP conducted only group activities on diet, exercise, and weight management (30), whereas the remaining 5 studies included some individual sessions (28, 29, 31-33). Of the 5 studies, 4 had dietary prescriptions (28, 29, 31, 33), and 1 gave advice on dietary choice (32). The DiRECT (33) trail underwent a total diet replacement in the first phase of the study, which provided a micronutrient-replete 825 to 853 kcal/day liquid formula diet and a soluble fiber supplement, followed by a food-based diet with an individually tailored energy prescription. In the first phase of the Look AHEAD study (28), participants were prescribed a restricted energy intake diet (1200-1500 kcal/day for individuals < 114 kg and 1500-1800 kcal/day for individuals ≥ 114 kg) with less than 30% of calories from fat and less than 10% from saturated fat. Participants were also encouraged to replace 2 meals with a liquid shake and 1 snack with a meal bar, and to eat more fruits and vegetables. After phase one, calorie targets were personalized based on participants’ weight loss goals. In the Da Qing study (31), participants with a BMI less than 25 were prescribed a diet of 25 to 30 kcal/kg body weight, 55% to 65% carbohydrate, 10% to 15% protein, 25% to 30% fat, as well as a recommended intake of more vegetables and less sugar, while individual goals for total calorie consumptions were set for other participants to lose weight at a rate of 0.5 to 1.0 kg per month until they achieved a BMI of 23. In contrast to the goal of reducing total caloric intake in the Look AHEAD (28) and Da Qing (31) trials, the DPPOS (29) aimed at reducing fat intake so that 25% of calories came from fat, and the calorie goal was calculated by taking the daily calories needed to maintain the participant's starting weight and subtracting 500 to 1000 calories/day to achieve a weight loss of 1 to 2 pounds per week.
Three trials provided recommendations for increasing physical activity (30, 32, 33), whereas the other 3 specified exercise prescriptions (28, 29, 31). The Thailand DPP (30) and ADDITION (32) trials gave suggestions for lifestyle changes, including physical activity. The DiRECT (33) advised physically capable participants to increase daily physical activity and provided recommendations for step-counters to reach an individual sustainable maximum. Participants in the Look AHEAD trial (28) were initially instructed to walk at least 50 minutes per week, then increased to 125 minutes/week or more and 175 minutes/week or more of physical activity by week 16 and 26, respectively. In the Da Qing study (31), participants were taught to increase the amount of leisure physical exercise by at least 1 U/day or 2 U/day, depending on their individual circumstances. The exercise goal of the DPPOS (29) was to achieve and maintain a level of at least 150 minutes/week of moderate-intensity physical activity.
In addition, pharmacological treatment advice was given in the ADDITION trial (32), and participants in the DiRECT (33) were instructed to withdraw antihypertensive and diuretic medications. No weight-loss medication was used in the DPPOS (29), and no additional information on medication was found in the other trials.
For the control groups, 5 trials provided standard care (28, 30-33) and 1 used a placebo (29). Two studies continued with routine care according to each center, but did not provide additional information and advice on lifestyle (32, 33). The Da Qing study (31) provided brochures with general instructions on diet and increased physical activities, but no specific individual or group advice was given. The Thailand DPP (30) study provided general recommendations on diet, physical activity, or weight loss through a one-time education program. The control group in Look AHEAD study (28) had 3 group education sessions per year, focusing on diet and exercise without individual plans. The DPPOS (29) offered written information and 20- to 30-minute individual sessions on healthy lifestyle, including recommendations on following a diet guideline and increasing activity to 30 minutes per week to lose weight.
Regarding prevalence of cancer among participants at baseline, the DiRECT (33) excluded participants with known cancer and the DPPOS (29) excluded participants with cancer requiring treatment in the past 5 years. For cancer incidence of the Look AHEAD trial, data were obtained from participants with no cancer diagnosis at baseline and a median follow-up of 11 years (34). No information on exclusion criteria for cancer was found in other trials.
Risk of Bias
The results for the risk of bias analysis for cancer mortality and cancer incidence are shown in Fig. 2A and 2B. For cancer mortality, 4 of the 5 trials had low risk of bias in all domains (28, 29, 31, 32). One trial had some concerns in the randomization process due to divergences in baseline characteristics, and there was a risk of deviation from intended intervention because the protocol for that trial could not be found (30). For cancer incidence, the 2 trials had a low risk of bias in the overall classification (33, 34).
Risk of bias assessment of studies using the RoB 2.0 tool. A, Risk of bias assessment of studies with cancer mortality. B, Risk of bias assessment of studies with cancer incidence.
Cancer Incidence
Cancer incidence was reported in 2 studies (33, 34): 1 as an adverse event and the other as a tertiary outcome. Data were collected from general practitioner records and medical records. Meta-analysis was not performed because of the limited number of studies reporting cancer incidence.
Cancer Mortality
Five trials reported cancer mortality (28-32). Among them, 3 assessed all-cause mortality as a primary outcome and cancer death as 1 of the causes of mortality (28, 29, 31). One included cancer death as a secondary end point (32), and the other reported cancer death as one of the reasons for loss to follow-up (30). Data were collected through death certificates, medical records, interviews with relatives, and in nationwide death databases.
Among a total of 12 841 participants, the RR of dying of cancer in the intervention groups compared with control groups was 0.94 (95% CI, 0.82-1.09 using the random-effects model; 95% CI, 0.81-1.10 using GLMM). Meta-analysis results were similar for both methods. No statistically significant heterogeneity was found (Fig. 3). According to the GRADE system, the quality of evidence was moderate due to the imprecision of the result, as the CIs showed potential benefit and harm, crossing the null effect (Fig. 4).
Meta-analysis of the effect of dietary and physical activity–based lifestyle interventions compared to usual care on cancer death. *The square size of the risk ratio reflects the weight of the trial in the pooled analysis. Horizontal lines represent 95% CIs.
Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework to assess quality of evidence about the effect of dietary and physical activity–based lifestyle interventions compared to usual care on cancer death. aImprecision was considered as a reason for downgrading the certainty of evidence due to CIs showing potential harm and benefit.
Subgroup analysis showed a trend toward increased risk of cancer death in the intervention group among those with prediabetes and a trend toward decreased risk among those with diabetes; however, the results were imprecise because of the limited number of studies in each subgroup (Fig. 5).
Subgroup analysis of the effect of dietary and physical activity–based lifestyle interventions on cancer death by glycemic status of the study population. *The square size of the risk ratio reflects the weight of the trial in the pooled analysis. Horizontal lines represent 95% CIs.
In the sensitivity analysis excluding trials that reported cancer deaths as a loss to follow-up or with follow-up period not exceeding 2 years, only the Thailand DPP trial (30) met the exclusion criteria. A meta-analysis excluding the Thailand DPP trial (30) was conducted, and the result remained unchanged (Supplementary Fig. S1 (19)).
TSA revealed that the number of participants in our study did not reach the detection boundary or the required sample size to define the effects of lifestyle interventions in reducing cancer mortality. However, the cumulative Z curve reached the boundary for futility indicating that it will be unlikely to result in statistical significance, even if included new studies until the required information size of 21 597 randomly assigned patients is achieved (Fig. 6).
Trial sequential analysis (TSA) of the effect of dietary and physical activity–based lifestyle intervention compared to usual care on cancer death. *The uppermost and lowermost curves represent the beneficial and harmful trial sequential detection boundaries, respectively. The horizontal lines are the traditional statistically significant boundaries. Triangular lines on the right are the futility boundaries.
Missing information on specific cause of death was present in 4 of the 5 studies. In the Look AHEAD trial (28), there were 56 (10%) unknown deaths among 589 deaths in the control group and 66 (12%) unknown deaths among 549 deaths in the intervention group. In the DPPOS study (29), there were 14 (10%) unknown deaths among 143 deaths in the control group and 11 (7%) unknown deaths among 158 deaths in the intervention group. In the Da Qing study (31), there were 4 (5%) unknown deaths among 76 deaths in the control group and 4 (2%) unknown deaths among 185 deaths in the intervention group. In the ADDITION study (32), there were 20 (9%) unknown deaths among 219 deaths in the control group and 19 (8%) unknown deaths among 246 deaths in the intervention group. To assess whether the missingness affected the result, we performed a meta-analysis assuming that all unclassified deaths were cancer deaths. The direction and magnitude of the result remained unchanged (RR 0.95; 95% CI, 0.84-1.08 using the random-effects model and 95% CI, 0.84-1.09 using the GLMM).
Discussion
This systematic review and meta-analysis found that, based on the data we have, long-term RCTs and trials with posttrial follow-up did not suggest a superiority of lifestyle interventions to usual care on cancer outcomes in populations with prediabetes and T2D.
Potential Factors Influencing the Results
Cancer is a complex disease resulting both from genetic and environmental factors (35, 36), so there are many potential factors that could influence cancer outcomes of trials.
Compared to pharmacology trials, lifestyle intervention trials are inherently less standardized, as the intensity, frequency, duration, and format of the intervention vary from trial to trial. The active intervention phase durations varied considerably across these 6 trials, ranging from 2 years to 6 years (28-33). The Thailand DPP had the lowest intensity of intervention (30). Only group activities were conducted, and no prescriptions for diet and exercise were given. In contrast, the DPPOS (29) study employed an intense intervention, in which participants were given a prescription for diet to reduce fat intake and a prescription for physical activities to achieve 150 minutes of exercise per week. The care received by the control groups was also different. The DiRECT (33) and ADDITION (32) trials provided only standard care to control groups, while the others provided some level of lifestyle advice or education. Of the larger 3 trials, the ADDITION (32) and Look AHEAD (28) trials, which were performed in T2D populations, demonstrated a nonsignificant reduction in cancer mortality comparing intervention group to usual care, while the DPPOS (29) trial, which was performed in a prediabetes population, showed a nonsignificant increase in cancer mortality comparing the intervention group to usual care. Differences in the implementation of lifestyle interventions and differences in the effects of lifestyle interventions in diabetes and prediabetes populations may influence the cancer outcomes of trials.
A recent study suggests that bariatric surgery resulting in greater weight loss was associated with a 37% reduction in cancer incidence (37). However, we did not see an effect of similar magnitude among the 2 included trials conducted in individuals with T2D (33, 34). Since we did not find information on cancer screening administered to participants, cancer incidence of the 2 trials (33, 34) may have been underestimated. It is possible that studies employing appropriate age-specific cancer screening and intensive lifestyle intervention achieving comparable weight loss could demonstrate a similar effect to that found in the bariatric surgery study.
Age is the strongest risk factor for cancer incidence and mortality, and the probability of developing invasive cancer increases dramatically in older individuals (38, 39). Therefore, trials with younger populations or shorter follow-up periods may have a lower risk of cancer events. In the 2 studies with 2-year follow-up, the DiRECT trial (33) reported 6 cases of incident cancer and the Thailand DPP trial (30) reported 1 case of cancer death, demonstrating a much lower cancer risk compared to the other studies, and consequently making it difficult to show the effect of the interventions. Since many cancers develop after a prolonged induction period followed by a long latent period before diagnosis, studies to demonstrate a reduction in incidence or mortality may require a longer active intervention period, long-term posttrial follow-up, and larger sample size.
In addition, although there was no statistical heterogeneity, clinical heterogeneity remained across the 6 studies. Among the 6 studies we analyzed, 2 were of Asian populations (30, 31) and the rest consisted of mainly White populations (28, 29, 32, 33). Since cancer risk varies in populations of different ancestry (35), the effects of lifestyle interventions might be masked when combining populations with different ancestry. However, due to the limited number of studies of both outcomes, subgroup analyses were not performed. The proportion of women participants ranged from 40.9% to 79.7%. The most common types of cancer differ between men and women, and the risk of developing the same type of cancer varies (39). Two trials excluded participants with a cancer diagnosis at baseline (29, 33), whereas other trials did not (28, 30-32). Baseline smoking probabilities varied widely across the 6 studies (from 4.4% to 41.3%) (28-33). Alcohol consumption levels were not reported in a comparable manner, or not reported at all. As mentioned in the Da Qing study, the baseline prevalence of smoking was much higher in men than in women, and the intervention on the incidence of diabetes was less effective in smokers regardless of sex (31). Considering the aforementioned information, clinical heterogeneity between trials may influence the effectiveness of interventions and contribute to differences in cancer outcomes.
Comparison With Observational Studies
According to the 2018 cancer prevention recommendations from the World Cancer Research Fund and American Institute for Cancer Research (WCRF/AICR) (40) and the American Cancer Society guideline for diet and physical activity for cancer prevention (41), favorable lifestyle factors include maintaining a healthy body weight; being physically active; increasing consumption of whole grains, vegetables, and fruits; reducing intake of processed foods, red meat, sweet beverages, and refined grain products; limiting alcohol consumption; as well as breastfeeding infants. Previous studies suggested that adherence to WCRF/AICR or American Cancer Society guidelines are protective against cancer. A 2016 systematic review found a consistent association of high adherence to nutrition and physical activity cancer prevention guidelines and reductions in all cancer incidence (from 4% to 45%) and mortality (from 14% to 61%) (24). This association was further confirmed by another systematic review in 2020, which found that compliance with the 2007 WCRF/AICR recommendations was associated with a lower risk of breast, colorectal, and lung cancer incidence and all-cancer mortality (25). However, we did not find a statistically significant association between dietary and physical activity interventions and cancer outcomes among RCTs in our study. This inconsistency might be due to the following: 1) the nature of the study, as the high-adherence populations in the cohort study may have other protective environment factors and/or a higher economic status that allows for better access to care; 2) the guidelines in cohort studies include multiple lifestyle components (body weight, diet, physical activity, breastfeeding, and alcohol consumption), whereas RCTs mainly focused on diet and physical activity interventions and only 2 studies recommended reducing alcohol intake (29, 31); 3) the populations included in the observational studies are generally younger and healthier, whereas the RCTs included in our studies consisted of participants with prediabetes and T2D. Insulin resistance, a common condition in T2D, promotes hyperinsulinemia and further promotes the production of insulin-like growth factors. Overexpression of insulin receptors can stimulate normal cells involved in cancer progression and insulin-mediated mitogenesis. Therefore, high plasma glucose and a high endogenous insulin level may theoretically favor cancer development (8, 10, 12).
Future Perspective
Although this study did not find a statistically significant effect of lifestyle interventions on risk of cancer death in populations with prediabetes and diabetes, it identifies the directions for future research. Observational studies have shown that a healthy lifestyle, including a healthy diet and more physical activity, is protective against cancer; this finding should be further examined using RCTs to determine whether it is lifestyle or factors related to lifestyle that influence cancer risk. Although the TSA result demonstrated futility in testing diet and physical activity–based lifestyle interventions for preventing cancer outcomes in prediabetes and diabetes populations, subgroup analysis of our study suggested that the effect of diet and physical activity–based lifestyle interventions may differ between the prediabetes and T2D populations, which needs to be further explored in future trials. In addition, lifestyle intervention strategies that have been shown to be effective in reducing all-cause mortality, such as Mediterranean, of vegetarian diets, or the DASH diets (Dietary Approaches to Stop Hypertension) (42), should be tested with this aim to further evaluate the effectiveness of lifestyle interventions in prediabetes and diabetes populations in reducing cancer incidence and mortality. These new trials of lifestyle intervention should not only define cancer as a main aim and report cancer mortality and incidence as an adverse effect or end point, but also consider excluding participants with cancer or reporting baseline prevalence of cancer. Furthermore, adjudication of cancer outcomes is needed to ensure the quality of cancer outcome data. Smoking, alcohol consumption, and other environmental factors associated with cancer risk should be reported in the baseline characteristics.
Conclusion
Long-term RCTs and posttrial follow-up of diet and physical activity–based lifestyle interventions did not show a benefit in reducing cancer risk in people with prediabetes and T2D compared to usual care in this study. More RCTs with different lifestyle intervention strategies and long-term follow-up are needed. Future studies should include cancer incidence and mortality as study end points or adverse events to assess the effects of lifestyle interventions more comprehensively.
Funding
This work was supported in part by the Research and Events Incentive Fund of Hospital de Clínicas de Porto Alegre (FIPE-HCPA 2020-0507); K.P.Z., P.P.T., and B.F.S. received a scholarship from the Fundação Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES—Brazil); and P.E.C. received a scholarship from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq—Brazil). The research group received a grant from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq—Brazil) (CNPq/MCTI/FNDCT No. 18/2021—Process: 420065/2021-0). M.Z. received a scholarship from the Johns Hopkins Bloomberg School of Public Health.
Author Contributions
K.P.Z., P.P.T., and F.G. were responsible for conceptualization and protocol design. All authors were responsible for data interpretation, manuscript writing, and critical review for important intellectual content. M.Z., K.P.Z., P.P.T., and V.C. performed data acquisition and analysis. M.Z., K.P.Z., P.P.T., P.E.C., and F.G. obtained the financing. V.C., M.S., and F.G. were the supervisors and the guarantors of this manuscript.
Disclosures
All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf and declare no support from any organization for the submitted work. The authors declare no relationships or activities that could appear to have influenced the submitted work. The authors declare no conflict of interest.
Data Availability
Original data generated and analyzed during this study are included in this published article or in the data repositories listed in “References.”
References
Abbreviations
- BMI
body mass index
- GLMM
general linear mixed model
- RCT
randomized clinical trial
- RR
relative risk
- T2D
type 2 diabetes
- TSA
trial sequential analysis
- WCRF/AICR
World Cancer Research Fund and American Institute for Cancer Research
