A systematic review of dietary, nutritional, and physical activity interventions for the prevention of prostate cancer progression and mortality

Given the long-term, although potentially fatal, nature of prostate cancer, there is increasing observational evidence for the reduction in disease progression and mortality through changes in lifestyle factors. We systematically reviewed dietary, nutritional, and physical activity randomized interventions aimed at modifying prostate cancer progression and disease-specific mortality, including a detailed assessment of risk of bias and methodological quality. Forty-four randomized controlled trials of lifestyle interventions, with prostate cancer progression or mortality outcomes, were identified. Substantial heterogeneity of the data prevented a meta-analysis. The included trials involved 3,418 prostate cancer patients, median 64 men per trial, from 13 countries. A trial of a nutritional supplement of pomegranate seed, green tea, broccoli, and turmeric; a trial comparing flaxseed, low-fat diet, flaxseed, and low-fat diet versus usual diet; and a trial supplementing soy, lycopene, selenium, and coenzyme Q10, all demonstrated beneficial effects. These trials were also assessed as having low risk of bias and high methodological quality (as were seven other trials with no evidence of benefit). The remaining trials were either underpowered, at high or unclear risk of bias, inadequately reported, of short duration or measured surrogate outcomes of unproven relationship to mortality or disease progression, which precluded any benefits reported being reliable. Large, well-designed randomized trials with clinical endpoints are recommended for lifestyle modification interventions.


Introduction
Prostate cancer is the most common cancer in men in the Western world [1]. In the UK, for example, it accounts for a quarter of newly diagnosed cancers [2] and one in eight men will receive a prostate cancer diagnosis [3]. Prostate cancer is often localized and grows slowly, so men may live for many years with the disease. However, prostate cancer may behave more aggressively and is an important cause of morbidity and mortality [4]. Given the long-term chronic, but potentially fatal, nature of the disease, there is growing interest in low-toxicity interventions in the tertiary prevention of morbidity and mortality due to prostate cancer. This is of particular importance as noninvasive active surveillance, as treatment for localized disease, becomes more widely implemented and increases in popularity as a strategy for reducing potential overtreatment [5]. As the number of cancer survivors in the USA increases beyond 13 million [6], the American Society of Clinical Oncology highlights the need for clinician and survivor to understand secondary prevention and lifestyle modifications that could benefit their prostate, as well as overall, health [7]. Observationally, poor diet, low levels of physical activity, and obesity are thought to play an important role in cancer, including the progression of prostate cancer [8][9][10][11][12][13]. Higher levels of physical activity have been associated with reduced rates of overall, and prostate cancer-specific, mortality [14]. World Cancer Research Fund International guidelines for cancer prevention include being physically active for at least 30 min every day, limiting consumption of energy-dense foods, eating a variety of vegetables, fruits, wholegrains, and pulses, and limiting consumption of red and processed meats [9]. Published systematic reviews in the field have tended to examine only one specific nutritional element, such as soy isoflavones [15], or have not always focused specifically on prostate cancer [16,17]. Those that explored the implications of diet and nutrition more broadly often looked at risk of disease development, not progression and mortality [18], or did not include physical activity interventions [18,19]. Systematic reviews with a focus on physical activity failed to include diet and nutrition interventions or restrict the population to prostate cancer patients [20]. Where diet, nutrition, and physical activity interventions have been reviewed, primary outcomes were not progression or mortality, but measures such as body weight [21], or all cancers and pre-invasive lesions were included [22]. Additionally, some reviews have not focused purely on randomized controlled trials (RCTs), which introduces further potential for bias [18], and study methodology and risk of bias were not always assessed [16,19].
We therefore conducted a systematic review of dietary, nutritional, and physical activity interventions aimed at modifying prostate cancer progression and mortality in men with prostate cancer. We update and broaden the scope of previous systematic reviews [15][16][17][18][19][20][21][22] and undertake detailed assessment of risk of bias and methodological quality.

Search strategy
Studies were identified through a systematic search of the following bibliographic databases from inception to July 2014: AMED, CINCH, the Cochrane library, Embase, MEDLINE, and Web of Science. The search strategy specified terms for RCTs, prostate cancer, dietary, nutritional, or physical activity interventions, and surrogate or clinical measures of prostate cancer progression or mortality (see Supplemental Data for Medline search strategy, Online Resource 1). Reference lists of all eligible full-text articles and all relevant systematic reviews that were identified were hand searched for additional studies.

Inclusion and exclusion criteria
To be eligible, studies had to be RCTs in men with prostate cancer who were randomized to dietary, nutritional, or physical activity interventions, which reported on surrogate or clinical measures of prostate cancer progression or mortality. Dietary or nutritional interventions were considered to be those that altered the intake of foods or dietary constituents either directly (e.g., by giving vitamin supplements) or indirectly (e.g., through nutrition education). Physical activity interventions were those involving any movement using skeletal muscles. RCTs that involved a combination of dietary, nutritional, and physical activity interventions were included. Outcomes were post-intervention effects on recognized surrogate measures of prostate cancer progression [Gleason score; prostate-specific antigen (PSA)] and clinical measures of prostate cancer progression (metastases, recurrence, disease-free survival, or prostate cancer mortality). An additional outcome was circulating insulin-like growth factor (IGF). We extracted data on any adverse events that were reported. There were no language restrictions. Commentaries and other related documents were excluded unless they provided additional data.

Data screening
All titles and abstracts were independently screened by two of three reviewers (LHM, RP, and VL) using pre-defined inclusion and exclusion criteria. Exact duplicates were removed. Any abstracts meeting the inclusion criteria were retrieved as a full article. These were then independently considered for inclusion by two of the three reviewers (LHM, RP, and VL). Any disagreements were resolved through discussion, and if necessary, the third reviewer was consulted. An additional 5 % of titles and abstracts were triple-screened for accuracy (LHM, VL, and SQ).

Data extraction
All data were extracted by one reviewer (VL) and doubleextracted by a second (LHM or RP) using a specifically designed data extraction form. Any disagreement was resolved by consulting the third reviewer. All extracted data were then checked for a final time by the third reviewer (LHM or RP). We extracted data on study characteristics, methodological quality (based on seven design and implementation questions), variables required for a Cochrane risk of bias assessment [23], and our pre-specified primary and secondary outcomes. The quality criteria assessed were: similarity of baseline characteristics and prognostic indicators between randomized arms; reporting of a power calculation and whether this sample size was achieved; reporting of withdrawal numbers and reasons by group; description of equal therapeutic time between groups. The risks of bias criteria assessed were: reporting of sequence generation; allocation concealment; blinding of participants, personnel, and outcome assessors; completeness of outcome data; and selective outcome reporting. Descriptions of what classifies as high and low risk can be found in Supplementary Table 1 (Online resource 2). Published protocols and trial registries were additionally searched, where available and when necessary, for further methodological detail related to methodological quality and risk of bias assessment. Authors were contacted if further data were required. Authors of non-peer-reviewed documents (such as conference abstracts) were contacted for published peer reviewed data; where none were provided, these were not included in the main analysis, but the description of the study included at the end of Tables 1 and  4.

Data analysis
Due to substantial heterogeneity across the studies in relation to intervention design, delivery mode and outcomes reported, formal pooling of the data by meta-analysis was not appropriate or possible. Therefore, a qualitative synthesis of all studies in a narrative format was undertaken.
The PRISMA statement was followed and adhered to [24]. The protocol was registered with PROSPERO International Prospective Register of systematic reviews, Ref: CRD42014008701.

Descriptions of studies
The search identified 12,037 titles and abstracts, of which 9,481 (79 %) papers that did not meet our inclusion criteria and 2,344 (19 %) exact duplicates were removed. The remaining full texts of 212 (2 %) papers were retrieved and read in full; 44 RCTs reported in 54 papers met the inclusion criteria (Fig. 1).

Excluded studies
Of the 212 texts read in full, 160 (75 %) were excluded. Thirty-eight did not involve a diet, nutrition, or physical activity intervention, and these included Ernst et al. [81], Peng et al. [82], and Sternberg et al. [83]. Twenty-two did not include prostate cancer progression or mortality as an outcome, for example James et al. [84], Zhang et al. [85], and Lee et al. [86]. Trials that only included a small proportion of prostate cancer patients within their total sample and had analyzed the data as a whole were excluded, for      [88] (''patients with a biopsy negative for prostate cancer comprised the principal study sample'' p520). Figure 1 highlights all reasons for exclusion.

Risk of bias
Overall, most of the included papers demonstrated high risk of bias on the majority of criteria or failed to adequately report how they had conducted the study on these essential criteria (    [64], and Bourke et al. [66] were assessed to have relatively low overall risk of bias.

Methodological quality
The methodological quality of the trials was variable; although it was generally acceptable in the majority of the RCTs, some scored very low (Table 3). In particular, only 15 RCTs reported that they had reached an adequately powered sample size and reasons for withdrawals were described for only 20 RCTs.

Single-factor dietary interventions
Calcitriol (vitamin D3) The effect of calcitriol supplementation up to 2 months prior to radical prostatectomy was reported in three RCTs [27,28,49]. Men were randomized in the three trials, respectively, to doses of 10 lg vitamin D daily versus no supplement; 400 versus 10,000 versus 40,000 IU vitamin D3 daily; and 0.5 lg/kg calcitriol daily versus placebo. In all three trials, there was little evidence of an effect of vitamin D3 on change in total PSA, IGF-I, cell apoptosis, or proliferation.
Lycopene Lycopene supplementation up to 6 weeks prior to radical prostatectomy was investigated in two RCTs reported in five publications [41,[53][54][55][56]. Men were randomized in the two trials, respectively, to doses of 15, 30, or 45 mg lycopene daily versus no supplementation and 15 mg lycopene versus usual care. No between-group differences in PSA change [41,[53][54][55][56], IGF-I change [53][54][55][56], or cellular response [41,[53][54][55][56] were observed. In a trial assessed as having high or unclear risk of bias on six of the seven criteria, where low risk of bias was only attributed to selective outcome reporting, Ansari and Gupta [71][72][73] randomized men, with advanced or metastatic disease, to orchiectomy alone versus orchiectomy plus 2 mg lycopene supplementation twice daily. A difference in change in PSA between the groups at 24 months was observed (p \ 0.001); fewer intervention men had a clinically raised PSA indicating progression than the control arm (p \ 0.05); there were fewer bone metastasis in the intervention group (p \ 0.02), and prostate cancer mortality was lower in the intervention group (p \ 0.001).
Pomegranate Three trials [25,26,34,78] randomized men to pomegranate extract supplements, one in men for four weeks prior to radical prostatectomy [25]; one for up to 18 months following radiotherapy, prostatectomy, hormone therapy or ADT, or cryotherapy [26,34]; and one following radiotherapy, prostatectomy, hormone therapy or ADT, chemotherapy, or watchful waiting for 28 days [78]. In men randomized to 2 g of pomegranate extract daily (including 1.2 g of polyphenol), there was no difference in measures of cell proliferation, progression, or change in PSA [25]. In the studies that randomized men at a variety of TNM classification of malignant tumors stages to pomegranate extract following definitive treatment, no differences were observed in median PSA doubling time or PSA change between experimental versus control groups [26,34]. The third trial, which was assessed to have low risk of bias, found no between-group differences in PSA change [78]. Due to the nature of the interventions, we modified the Cochrane guidelines to separate blinding of participants and blinding of personnel; this addressed the difficulty in blinding participants in some of the interventions considered. Each trial was given a low (?), high (-), or unclear (?) risk of bias score for each dimension. Where a full paper was not available risk of bias was not assessed, as it was felt this would be a biased assessment without full data available   [50,75,76]. In men randomized to 30 mg genistein daily for three to six weeks prior to prostatectomy, differences in favor of the experimental, versus control, group were reported for percentage change in PSA (p = 0.051), cellular response (p = 0.033), and cell proliferation (p \ 0.001). However, the trial was assessed as having high or unclear risk of bias on four of the seven criteria, and it was assessed to have low risk of bias for sequence generation and blinding of participants, as well as selective outcome reporting [75,76]. Comparably, in a trial with relatively low risk of bias, in men undergoing watchful waiting, randomization to 60 mg genistein daily versus an isocaloric placebo for 12 weeks had no impact on mean change in PSA [50].
Selenium Two RCTs investigated the effects of selenium [37,52] or selenium and silymarin [70] supplementation: one in men with localized disease on active monitoring, active surveillance, or watchful waiting supplemented with 200 or 800 lg selenium versus placebo for up to 60 months; and the other in men following prostatectomy who were supplemented with selenomethionine (240 lg selenium and 570 mg silymarin) or placebo for 6 months. There were no between trial group differences in measures of PSA in either trial.
Other nutritional interventions There was no evidence of any effect of any of the following interventions: Qilan capsules (consisting of astragalus, fenugreek, gynostremma, pentaphyllan, and smilaz glabra) given for 4 weeks versus placebo in men who had undergone hormone therapy or ADT and orchiectomy on PSA outcomes [69]; 50 mg of beta carotene on alternate days (for an unreported duration) versus placebo in men undergoing radiotherapy on prostate cancer mortality [31]; 800 mg of polyphenol E versus placebo daily for 3-6 weeks prior to undergoing prostatectomy on changes in PSA, IGF-I, or Gleason score or on tissue measures of cell proliferation, cell apoptosis, angiogenesis [32]; or 2.4 g eicosapentaenoic acid daily versus no intervention for 24 months in men who had undergone radical prostatectomy on PSA failure [74]. However, in men not undergoing active treatment, 295 g of rye bread versus a wheat bread control for 11 weeks resulted in increased cell apoptosis (p \ 0.05) in the intervention group, although no effect on change in PSA or IGF-I were reported. This trial was assessed to have high or unclear risk of bias on four of seven criteria, and low risk of bias was found for blinding of participants and outcome assessors, as well as selective outcome reporting [77].

Multiple factor dietary interventions
Isoflavones Three studies explored the effect of combinations of isoflavones within individual supplements [30,38,39]. In a study assessed as being of low risk of bias, Bosland et al. [30] found no effect on recurrence-free survival of powdered soy protein (combining genistein, daidzein, and glycitein) compared with calcium caseinate given for 24 months in a population of men who had undergone radical prostatectomy. Others found no effect of combinations of isoflavones on PSA measures in men due to undergo radical prostatectomy [39] or men undergoing active surveillance for a period of 6 months [38]. However, a trial assessed to have high to unclear risk of bias on four of seven criteria, where only blinding of participants, completeness of outcome data and selective outcome reporting were assessed as low risk of bias, randomising men awaiting radical prostatectomy to either 50 g of heattreated soy, or 50 g of heat-treated soy plus 20 g of linseed, or placebo, found a difference in change in PSA between the soy-only and the placebo group (p = 0.02) and a difference in change in free-total PSA ratio between the two intervention groups (p = 0.007) and the soy-only and placebo group (p = 0.01) [60].

Other complex nutritional supplement interventions
In an RCT of men undergoing active surveillance or watchful waiting, a capsule containing pomegranate seed, green tea, broccoli and turmeric in a capsule versus placebo given for 6 months was associated with a reduced rise in PSA (p = 0.0008) and an increase in the percentage of participants with stable PSA at 6 months (p = 0.00001). This trial was reported to have relatively low risk of bias [64]. Schroder et al. [63] randomized men undergoing radiotherapy or radical prostatectomy to a supplement consisting of soy, lycopene, selenium and coenzyme Q10. The intervention was associated with improved measures of PSA during follow-up, and was assessed to have low risk of bias on five of seven criteria, however, unclear risk of bias for sequence generation and blinding of outcome assessors. Further to this, several trials of various combinations of nutrients in a variety of populations of men with prostate cancer observed no evidence for any differences [45,47,51,61,62].
Low-fat diet combined with other nutritional elements Three studies combined low-fat diet with another nutritional element [35,42,44]. Aronson et al. [35] randomized men due to undergo radical prostatectomy to 4 weeks of a daily low-fat and fish oil diet versus a Western diet, the trial was assessed to have high or unclear risk of bias for five of seven criteria, and low risk of bias was only awarded for sequence generation and selective outcome reporting. No differences were noted in change in mean PSA or change in mean IGF-I, but there was a reduction in cell proliferation (p = 0.026). Li et al. [44] compared a daily low-fat, high-fiber and soy protein (40 g) diet with a standard recommended control diet given for 48 months in men who had undergone radical prostatectomy. A difference was observed in IGF-I change between the groups (p = 0.04); however, none was seen for change in PSA. This RCT was reported to have unclear or high risk of bias on four of seven criteria, and low risk of bias was only awarded for sequence generation, blinding of outcome assessors and selective outcome reporting. Demark-Wahnefried et al. [42] randomized men due to undergo radical prostatectomy to flaxseed, low-fat diet, or flaxseed and low-fat diet versus usual diet, over an average of 31 days. The trial was assessed to have low risk of bias for five of seven criteria, and blinding of participants and of personnel were assessed to show high risk of bias; there was a change in proliferation rate between the flaxseed only and control groups (p = 0.0013) but no difference between apoptotic rate, median change in PSA, or median change in IGF-I. Aronson et al. [36] randomized men undergoing active monitoring to a low-fat diet, which included 35 mg of soy protein per day for 1 month, versus a Western diet but observed no differences in mean change in PSA or mean change in IGF-I between experimental and control groups. Three RCTs included an educational element within their complex nutritional intervention [29,40,43] but found no consistent effects in men awaiting radical prostatectomy or undergoing radiotherapy, active monitoring, or active surveillance on PSA or IGF-I outcomes.

Physical activity
Four RCTs reported a physical activity intervention, resistance and/or aerobic training [58,59,67,68], but found no consistent effects in men who had undergone hormone therapy, androgen deprivation therapy, or radiotherapy on PSA-based measures of progression. One of these four trials was reported to have relatively low overall risk of bias [68].

Combination interventions
Four RCTS combined both a nutritional and a physical activity element in their intervention [33,46,48,57,65,66]. All of these implemented an aerobic or aerobic and resistance training program in combination with a nutritional element. There was no consistent effect in men who had undergone radiotherapy, radical prostatectomy, active surveillance, or were on ADT. Only one of these trials had relatively low overall risk of bias [66]. Further information about all studies can be found in Table 4.

Adverse events
A variety of adverse events was reported in the included RCTs, and these most often included gastrointestinal events, such as mild abdominal pain, constipation, diarrhea and nausea; also reported were myalgia, including aches and pains and fever like symptoms, such as chills.

Discussion
Among 54 papers reporting the results of 44 RCTs that explored dietary, nutritional, and physical activity interventions in men with prostate cancer, there was a large degree of heterogeneity with regard to intervention aims, methods of implementation and outcomes, with the quality of the research often being poor. Only three of ten studies with the lowest risk of bias and highest methodological rigor found a possible beneficial effect; a study in men undergoing watchful waiting or primary active surveillance suggested that a capsule containing pomegranate seed, green tea, broccoli and turmeric improved PSA kinetics in the intervention group compared to the control arm [64]. A study randomising men due to undergo radical prostatectomy to flaxseed, low-fat diet, or flaxseed and low-fat diet versus usual diet, over an average of 31 days, demonstrated a change in proliferation rate between the flaxseed only and  Polyphenon E (800 mg daily) versus P Due to undergo surgery: 100 % PSA: taken at 3-6 weeks; Absolute change in PSA mean (SD), ng/ml; I:    Kucuk et al. [53][54][55][56] Lycopene ( [42]. Finally, in a trial that randomized men undergoing radiotherapy or radical prostatectomy to a supplement consisting of soy, lycopene, selenium and coenzyme Q10, the intervention was associated with improved measures of PSA during follow-up. It should be noted that despite PSA being the most widely available, and cited, biomarker for prostate cancer, taken alone it may not be an appropriate surrogate marker of long-term therapeutic benefit in prostate cancer trials, which has not been proven to be a suitable replacement for a final survival endpoint [89].
Of the remaining studies with low risk of bias; an RCT randomising men, who were due to undergo radical prostatectomy, to calcitriol or control reported no difference in cell apoptosis or rise in PSA [49]. A trial where patients were randomized to 15 mg lycopene versus usual care reported no between-group differences in PSA change, IGF-I change, or cellular response. A study which scored low for all risk of bias, and also had excellent methodological quality, randomized participants to take pomegranate extract supplements following radiotherapy, prostatectomy, hormone therapy or ADT, chemotherapy, or watchful waiting for 28 days, and found no between-group differences in PSA change [78]. A complex isoflavone intervention, with strong methodological vigor and low risk of bias, resulted in no change in recurrence-free survival between groups [30], similarly, a trial, with relatively low risk of bias, assessing the effect of genistein supplementation on men undergoing watchful waiting were randomized to 60 mg genistein daily versus an isocaloric placebo for 12 weeks, no impact on mean change in PSA was reported [50]. It should be noted that these previous studies support World Cancer Research Fund International guidelines which state ''Don't use supplements to protect against cancer'' [9]. One study, identified as being relatively low risk of bias, which combined aerobic and resistance training with diet advice [66], reported no difference in PSA or IGF-I outcomes. Finally, a resistance training intervention in men due to undergo ADT concluded no change in PSA [68], and this RCT was reported to have relatively low risk of bias. Most of the other studies reviewed were assessed as having high or unclear risk of bias, often with poor methodological vigor, so it is not possible to draw any conclusions from those studies. This is the first systematic review, to our knowledge, which has combined interventions of modifiable lifestyle risk factors, with a primary outcome of prostate cancer progression or mortality. This is clinically relevant as it is unlikely that patients would make changes to a singular lifestyle behavior [90]; instead, for example, a clinician may recommend changes to diet alongside an increase in physical activity. As the number of cancer survivors living for longer increases [7], particularly for those with prostate cancer who are turning to active surveillance [5], further understanding of diet, nutritional, and physical activity interventions is of great importance.
The review was systematic and comprehensive and had no language restrictions. All papers were at least doubleextracted. Risk of bias and methodological quality were assessed by at least two independent reviewers. This review does have some limitations. The primary outcomes of the review were not always reported as primary outcomes in the papers. Thus, we relied on reported secondary outcomes and the RCTs may not have been powered to detect differences in these outcomes. Meta-analysis was not possible due to heterogeneity of trial design, outcomes and statistical presentation; however, a qualitative synthesis was conducted. The limited quality of most of the RCTs, and the possibility of publication bias (which we were unable to formally assess in the absence of a meta-analysis), restricted the definitive conclusions that could be drawn.

Conclusion
The complex nature of dietary, nutritional, and physical activity interventions, along with the slow-growing nature of prostate cancer that causes difficulties in measuring longterm clinically relevant change, makes research in this area difficult. Poor quality, variability in methodology, inconsistency of results, and a variety of proxies for prostate cancer progression make firm conclusions hard to draw. The RCTs identified in our review were generally likely to be underpowered, appeared to be at high or unclear risk of bias and were often inadequately reported, intervened for only short durations and followed-up men for surrogate outcomes of questionable relationship to clinical outcomes. Such trials are unlikely to have any clinical impact and should be abandoned in favor of large, well-designed trials with endpoints that will impact on clinical practice. These findings are in line with previous systematic reviews which concluded that the impact of interventions could not be reliably estimated due to limited, and low-quality, RCTs [18,22].