Adherence to the Mediterranean diet and grade group progression in localized prostate cancer: an active surveillance cohort

Abstract

BACKGROUND:

Mediterranean diet (MD) may be beneficial for men with localized prostate cancer on active surveillance (AS) due to its anti-inflammatory, anti-lipidemic, and chemopreventive properties. We prospectively investigated adherence to the MD with Gleason score progression; and explored associations by diabetes status, statin use and other factors.

METHODS:

Men (n=410) with newly diagnosed prostate cancer on an AS protocol completed a baseline food frequency questionnaire; and MD score was calculated across 9 energy-adjusted food groups. Cox proportional hazards models were fit to evaluate multivariable-adjusted associations of the MD score with progression-free survival (PFS), defined as an increase in Gleason grade group (GG) score over a biennial monitoring regimen.

RESULTS:

In our cohort, 15% of men were diabetic, 44% used statins, and 76 men progressed (median follow-up: 36 months). Following adjustment for clinical factors, higher adherence to the MD was associated with lower risk of GG progression among all men [HR and 95% CI per 1-unit increase in MD score: 0.88 (0.77–1.01)], non-white men [0.64 (0.45–0.92); P-interaction=0.07] and men without diabetes [0.82 (0.71–0.96); P-interaction=0.03]. When examining joint effects of the MD score and statin use, we observed similar risk reduction among men with high MD score who did not use statins, as compared to men with low/moderate MD scores with no statin use.

CONCLUSIONS:

The MD was associated with lower risk of GG progression in men on AS and is consistent with prior reports of the MD and reduced cancer morbidity and mortality.

INTRODUCTION

The Mediterranean diet (MD) is a well-established dietary pattern linked to anti-inflammatory and anti-lipidemic properties.^1,2 It is characterized by a foundation of vegetables, fruits, grains, legumes and fish; limited intake of meat and dairy; moderate intake of alcohol; and a healthy balance of monounsaturated fat (e.g., from olive oil, nuts, and seeds) relative to saturated fat (e.g., from red meat and butter). A multitude of studies and large international prospective cohorts have demonstrated that greater adherence to the MD is associated with a lower incidence of cardiovascular disease (CVD), multiple cancers^3,4, and overall mortality.⁵ The randomized PREDIMED study has also shown that a number of CVD and cancer risk factors, including insulin sensitivity, lipid profiles, lipoprotein particles and inflammatory markers, are improved by adherence to the principles of the MD.⁶ Furthermore, observational studies indicate that men who follow a MD pattern are less likely to develop overall³ and aggressive⁷ prostate cancer (PCa). Limited evidence also suggests that overall survival may be improved in men who adhere to the MD following PCa diagnosis.⁸

PCa is unique in that over 90% of men diagnosed following prostate biopsy have disease that is localized to the prostate⁹, and nearly half may have a low grade tumor that is eligible for active monitoring and potential delayed intervention, a management strategy termed “active surveillance” (AS).¹⁰ Radical PCa treatment can cause changes in health-related quality of life, such as declines in urinary and sexual function^11,12; therefore there is great interest in identifying modifiable risk factors for disease progression in men whose PCa is managed with AS. While a number of dietary factors or nutrients, such as calcium and Vitamin D, have been assessed in the setting of PCa risk, including advanced or aggressive disease,^13–15 none have been shown to limit disease or clinical factor progression. As such, laboratory values and biopsy results remain the foundation of risk assessment on AS.¹⁶

Therefore, we aimed to investigate the association between adherence to the MD and clinical disease progression in a group of localized PCa patients carefully followed on a prospective clinical protocol. We hypothesized that closer adherence to the principles of the MD or higher MD scores at baseline enrollment would be associated with improved grade group-progression free survival (PFS). Considering that other pharmacologic interventions in these patients may either mask or synergize with the effects of the MD score, we further explored the joint or modifying effect of statin use and diabetes status on the relationship between MD score and risk of progression.

METHODS

Study design and population

As described previously^17–19, the AS protocol was conducted by a multidisciplinary team of urologic surgeons, radiation oncologists and medical oncologists at MD Anderson Cancer Center to evaluate the safety of AS and is registered on clinical.trials.gov (NCT00490763). Men with localized PCa were eligible for the study. All patients underwent a confirmatory biopsy at study entry, with rare exception, and were evaluated every 6 months through clinical examination (digital rectal exam) and laboratory studies (serum PSA, testosterone). Prostate biopsies were repeated every 1–2 years; if a biopsy was negative, then the following year’s biopsy was omitted. All biopsies were performed using the trans-rectal ultrasound-guided technique with an 11-core multisite-directed biopsy scheme.¹ The option of treatment was offered to patients who had disease reclassification, defined as an increase in tumor volume (core number or percentage) or Gleason grade group (GG) increase, though patients who wished to remain on surveillance were allowed to do so if approved by their treating physician. A total of 560 patients who were diagnosed with GG 1 or 2 localized Pca were enrolled on to this prospective clinical protocol between February 2006 and February 2012. Among these, 501 provided a baseline food frequency questionnaire (FFQ). Following exclusion of patients with extreme total energy intake (defined as beyond twice the interquartile range of Box-Cox transformed intake, n=53), 410 had over 6 months of follow-up and were included in the MD score analysis.

Assessment of Mediterranean Diet Score and Patient Characteristics

Usual dietary intake at baseline and follow-up was assessed using a 170-item modified Block FFQ, as described previously.^17,20 FFQs were reviewed and coded by trained registered dietitians for completeness and acceptability. Total energy and nutrient intake were calculated by linking FFQ responses (frequency and portion size) to the US Department of Agriculture (USDA) Food and Nutrient Database for Dietary Studies.^17,21 Food groups were derived as described previously.^17,20

Given our relatively unique population of PCa patients, we constructed the MD score using population-specific cut points across a set of predefined dietary parameters, as described previously by Fung et al.²³ Seven food groups and two macronutrients define the MD score: vegetables, fruits, legumes, whole grains, fish, meat, dairy products, alcohol intake, and fat intake ratio (monounsaturated to saturated fatty acids). All components were standardized for total energy (per 1,000 kcal).²⁴ Individuals were assigned a value of 1 for each beneficial component (fruits, vegetables, legumes, whole grains and fish) whose consumption was at or above the median and for each detrimental component (meat and dairy products) whose consumption was at or below the median. In addition, a fat intake ratio at or above the median was assigned a value of 1. Alcohol intake was analyzed according to the National Institute on Alcohol Abuse and Alcoholism recommendations and the most recent evidence for PCa.²⁵ Moderate consumption was defined as any consumption up to 13 drinks per week and assigned a score of 1. Zero consumption or consumption defined as more than 13 drinks per week were assigned a score of zero. The MD score was then computed as the sum of scores across the 9 components with total score range of 0 to 9,with a higher score indicative of increased adherence to the MD score. Results were similar upon consideration of another version of the MD score that does not differentiate whole grains from total grains or cereals.^22,26

Baseline serum PSA, pathologic GG score, and summary tumor length (defined as the sum of diagnostic and confirmatory biopsy tumor length) were assessed at study enrollment.¹⁹ Other baseline patient characteristics were drawn from the medical record. BMI was calculated as weight (kg)/height (m)² and categorized based on the World Health Organization criteria.

Assessment of Outcomes

The primary outcome of interest was grade group progression, defined as any increase in GG following confirmatory biopsy. PSA kinetic changes, alone, were not used to reclassify patients. Patients were followed until grade progression, treatment, loss to follow-up, elective study removal, death or 12/31/2016 (the censor date of the study), whichever came first.

Statistical Analysis

The MD score was categorized into tertiles (low, medium, and high) based on the population distribution and for consistency with previous studies examining MD score and PCa risk.⁸ Differences across tertiles for continuous and categorical variables were tested using Analysis of Variance and Chi-square tests, respectively. We evaluated the association between MD score and PFS by using Cox proportional hazards (PH) regression models with person-years as the underlying time metric. Hazard ratios (HR), 95% confidence intervals (CI), and P values for linear trend (using the median value within tertiles) are reported across medium and high MD score, with the lowest tertile representing the referent group. We confirmed the PH assumption was met through assessment of interaction terms for the exposures with follow-up time. We also evaluated the association between MD score and progression in the subgroup of men with GG1 disease at baseline. Finally, we evaluated an alternate outcome in the entire cohort: progression to grade group 3 or higher disease.

We examined three models in the analyses: 1) a base model adjusted for age; 2) a clinical model additionally adjusted for baseline clinicopathologic factors including PSA and summary tumor length (defined as the sum of total tumor, measured in mm, found on biopsy cores from the diagnostic and confirmatory prostate biopsies)²⁷; and 3) a comprehensive model additionally adjusted for factors associated with MD score (testosterone, BMI and added sugar intake). We further evaluated other clinical, lifestyle, and demographic factors potentially related to MDS and grade progression, including tumor stage, smoking status and race. However, none of these factors appreciably modified the crude hazard ratio or the final models; and thus, were not included. We moreover assessed whether the association between MDS and grade progression varied by baseline clinicopathologic, lifestyle, and demographic factors, including age group, race, smoking status, testosterone level, presence of diabetes and median BMI, total energy and added sugar intake. Due to interest in inflammation- and lipid-related effects of the MD, we performed a joint effects analysis evaluating the effect of statin use and low/intermediate vs. high MD score, using patients with a low/intermediate MD score and statin use as the referent. All statistical tests were 2-sided and were considered statistically significant at P < 0.05. Statistical analyses were conducted using STATA version 13.1 (StataCorp, College Station, TX) and SAS (version 9.4, Cary, NC, USA).

RESULTS

Baseline characteristics by low (1–3 points), medium (4–5 points), and high (6–9 points), MD score (broken into tertiles) are displayed in Table 1. Overall, the median MD score was 4 (IQR 3–5) median age was 64.4 years (SD 8.4 years), 358/410 (87.3%) had GG1 disease based on their diagnostic prostate biopsy, and 297/410 (72.4%) had a single core of PCa. Low BMI (P<0.01), testosterone level (P<0.01), and added sugar intake (P<0.01)) were associated with higher MD score on univariate analysis. Over a median follow-up of 36 months (range 6–126 months), 18.5% of patients (76/410) experienced grade progression; and 12 patients died of other causes without documented progression.

Table 1:

Baseline characteristics of men with localized prostate cancer on active surveillance (n=410) by tertiles of Mediterranean Diet score

Characteristic		Low (0–3) N=141	Med (4–5) N=171	High (6–9) N=98	P value *
Age, y	Mean (SD)	62.8 (8.5)	64.9 (8.0)	65.8 (8.6)	0.02
Race					0.61
	White	113 (80.1)	147 (86.0)	80 (81.6)
	Black	14 (9.9)	12 (7.0)	7 (7.1)
	Other/unknown	14 (9.9)	12 (7.0)	11 (11.2)
PSA, ng/dL	Mean (SD)	4.2 (2.7)	4.0 (2.4)	4.2 (2.8)	0.82
Summation tumor length, mm **	Mean (SD)	2.2 (2.5)	2.1 (2.5)	2.3 (2.9)	0.84
First biopsy core positivity					0.66
	Single	106 (75.2)	122 (71.4)	69 (70.4)
	Multiple	35 (24.8)	49 (28.7)	29 (29.6)
First biopsy Gleason Grade Group Score					0.12
	Gleason GG1	129 (91.5)	148 (86.6)	81 (82.7)
	Gleason GG2	12 (8.5)	23 (13.5)	17 (17.4)
BMI, kg/m2	Mean (SD)	29.4 (4.6)	28.4 (4.0)	27.5 (4.1)	<0.01
Testosterone, ng/dl	Mean (SD)	379 (136)	408 (155)	444 (166)	<0.01
Statin use					0.71
	Yes	58 (41.1)	77 (45.0)	45 (45.9)
	No	83 (58.9)	94 (55.0)	53 (54.1)
Smoking status					0.43
	Ever	81 (57.5)	101 (59.1)	50 (51.0)
	Never	60 (42.6)	70 (40.9)	48 (49.0)
Alcohol intake, drinks/day	Mean (SD)	0.6 (1.2)	0.7 (1.0)	0.9 (1.0)	0.11
Hypertension					0.52
	Yes	65 (46.1)	89 (52.1)	51 (52.0)
	No	76 (53.9)	82 (47.9)	47 (48.0)
Diabetes Mellitus					0.89
	Yes	19 (13.5)	26 (15.2)	15 (15.3)
	No	122 (86.5)	145 (84.8)	83 (84.7)
Usual dietary intake (per day)
Added sugar, tsp/1,000 kcal	Mean (SD)	5.8 (3.5)	4.8 (2.9)	4.2 (2.1)	<0.01
Total energy, kcal	Mean (SD)	2558 (1081)	2302 (1020)	2216 (711)	0.02

^*ANOVA or chi square test, as appropriate

^**Summation tumor length is the sum of the tumor length of all positive biopsy cores from the first (diagnostic) biopsy and the second (confirmatory) biopsy that is completed at study enrollment

Multivariable-adjusted models evaluating the association between baseline MD score and PFS are shown in Table 2. Following adjustment for age and clinical characteristics, we observed a suggestive inverse association between high baseline MD score and lower risk of grade progression (high MD score vs. low MD score, HR=0.68, 95% CI=0.38–1.27, P-trend=0.06). Adjustment for additional factors did not noticeably change the observed relationship (HR 0.67, 95% CI 0.36–1.25, P-trend=0.05) Findings were consistent among men who had GG1 PCa at baseline (Supplementary Table 1) and also when evaluating the association between MD score and GG3 or higher disease in the broader cohort (Supplementary Table 2). Notably, for every one-unit increase in the MD score we observed a >10% lower risk of progression (HR 0.88, 95% CI 0.77–1.01; Table 2).

Table 2:

Multivariable-adjusted associations of Mediterranean Diet score with time to prostate cancer grade group progression

	Frequency		Base modela			Clinical modelb			Comprehensive modelc
	Number of events	N	HR	95% CI	P-value	HR	95% CI	P-value	HR	95% CI	P-value
MD score (tertiles)
0–3	24	119	Reference			Reference			Reference
4–5	27	139	0.78	0.47–1.30	0.95	0.78	0.47–1.30	0.34	0.77	0.46–1.28	0.32
6–9	12	77	0.69	0.38–1.27	0.14	0.68	0.36–1.25	0.21	0.67	0.36–1.25	0.21
P-value for trend			0.07			0.06			0.05
MD score (per 1 unit increase)			0.88	0.77–1.00	0.07	0.88	0.77–1.01	0.06	0.89	0.77–1.02	0.10

^aBase model: MD score + age

^bClinical model: MD score + age + tumor length + PSA

^cComprehensive model: MD score + age + tumor length + PSA + testosterone + BMI + added sugar

Figure 1 shows the multivariable-adjusted associations between baseline MD score (as a continuous variable) and PFS according to a priori selected potential modifying factors. Increased MD score was associated with improved PFS among men who did not have diabetes mellitus (HR 0.82, 95% CI 0.71–0.96, P-interaction=0.03), men who identified as non-white (HR 0.64, 95% CI 0.45–0.92, P-interaction=0.07), and men with high testosterone levels (HR 0.83, 95% CI 0.70–0.98; P-interaction=0.35).

Figure 1.

Association between baseline adherence to the Mediterranean diet score and disease progression according to selected characteristics. Risk of progression stratified by patient characteristics and evaluated in models of continuous MD score adjusted for age, tumor length, and PSA. Units: testosterone = ng/dL; daily kcal = kilocalories per day; added sugar = teaspoons per 1,000 kcal per day; age = years. * P<0.05.

Table 3 shows the results of a joint effects analysis assessing the relationship between statin use, MD score and PFS. With low/middle MD score and no statin use as the referent, high MD score and no statin use was associated with a potential improvement in PFS (HR 0.74, 95% CI 0.32–1.68), though results were not significant.

Table 3:

Joint effects of Mediterranean Diet score and statin use on time to prostate cancer grade group progression

	N	Number of events	Clinical Model*
			HR	95% CI	P-value
Group
MDS 0–5 and no statin	177	32	1.00	Ref	Ref
MDS 0–5 and uses statins	135	28	1.19	0.71–2.00	0.51
MDS 6–9 and uses statins	45	9	0.93	0.44–2.00	0.85
MDS 6–9 and no statin	53	7	0.74	0.32–1.68	0.47
P-value for trend					0.54

^*Clinical model adjusted for age, tumor length, and PSA

DISCUSSION

Baseline adherence to the MD appears to be associated with lower risk of GG progression in men with localized PCa managed on AS. This would suggest that consistently following a diet rich in plant foods, fish, and a healthy balance of monounsaturated fats may be beneficial for men diagnosed with early stage PCa. Although few of our findings reached pre-specified measures of statistical significance, likely due to a limited number of events, the magnitude and direction of this inverse relationship appeared to be consistent across a range of lifestyle and clinical factors, particularly in men who did not have diabetes or those who self-identify as non-white.

The MD score has previously been associated with significant improvements in CVD and related comorbidities^28,29 as well as a lower risk of developing and dying from cancer.³⁰ Several epidemiological studies have investigated the association between a priori or index-based dietary patterns and risk of PCa.^8,31,32 Lower risk of aggressive PCa was reported in men who adhered to the MD in the Multi-case-Control Study on Common Tumors in Spain cohort⁷ and the North Carolina-Louisiana PCa Project²⁶, though findings were mixed in recent meta-analyses.^3,33 Interestingly, an analysis of MD score in the Health Professional Follow-up (HPF) cohort did not show an association between increased MD score and prostate-related death,⁸ though an overall survival benefit was noted. This study included over 1000 patients with more advanced disease, representing a much more aggressive population than that managed on AS following PCa diagnosis.

The potential protective effect of high adherence to the MD and improved PFS in PCa patients on AS, is the main contribution of our work to the existing evidence in this area of research. Our group previously demonstrated that adherence to U.S. dietary recommendations or the healthy eating index (HEI-2015) was associated with improved grade group-PFS in this population.¹⁷ These and other trial findings targeting largely plant-based, low-fat dietary patterns³⁴ suggest that promoting overall healthy eating patterns, rather than individual diet components, may be the most relevant in terms of impacting cancer-related outcomes among men with localized, low-risk PCa. For example, in a phase III randomized trial (The Men’s Eating and Living or MEAL study) evaluating a validated phone-based counseling program to specifically increase vegetable consumption in men on AS, no difference in PCa PFS was reported (HR 0.96, 95% CI 0.75–1.24, P=0.76 for the treatment as compared to the control arm).³⁵ These data serve to reinforce that if successfully promoted/supported and adhered to, the MD (complete with its anti-inflammatory properties), could represent a promising and non-invasive intervention to reduce the risk of progression in PCa patients managed on AS.

A number of mechanisms support the potential protective effect of MD score in localized PCa patients commencing AS. The antioxidant and anti-inflammatory properties of the Mediterranean diet via a variety of plants foods and healthy balance of fats, may collectively support a systemic and tumor environment that inhibits progression.^36–40 Inflammation and CVD-related risk factors, such as C reactive protein, IL-6, lipid profiles and fasting glucose levels, lowered by MD adherence in the randomized PREDIMED and other studies may also be relevant for PCa patients.^{2,28,29,41,42} While measures of inflammation are complicated, particularly when accounting for genetic factors,⁴³ a twin study has also revealed that adherence to the MD is associated with higher fasting markers of circulating antioxidant capacity.⁴⁴ Statin use, while known for improving mortality in patients at risk for cardiovascular disease,⁴⁵ is also associated with decreased markers of acute and chronic inflammation in the prostate,⁴⁶ and may affect PSA and measures of proliferation in aggressive prostate tumors following prolonged use.⁴⁷ Using a randomized crossover design, Jula et al. demonstrated that the effect of Mediterranean diet adherence and statin use is independent and additive in terms of effects on both circulating lipids (such as LDL and HDL) and antioxidant measures such as serum alpha-tocopherol, B-carotene and oxidized LDL.⁴⁸ We similarly examined both the joint and modifying effects of statin use in our cohort and observed that higher adherence to the MD appeared to lower risk of GG progression among men who did or did not use statins. Intriguingly, MD score was associated with improved progression-free survival in men who did not have diabetes mellitus, while not providing apparent benefit for diabetics. These data suggest that elevated blood glucose or treatments for diabetes may interact with the systemic effects of MD adherence; however, we would caution that the small number of men with diabetes mellitus included in this population limit conclusions drawn from these findings.

We also observed that MD score may be associated with longer PFS in non-white men with PCa. Although this analysis was limited by sample size and is thus exploratory in nature; this subgroup was largely represented by African American men. Studies of tumors⁴⁹, circulating exosomes⁵⁰, and stromal fibroblasts⁵¹ from African American men with PCa have demonstrated a clear role for inflammation in the development of aggressive disease within this population. In the US, the rate of PCa diagnosis is more than 50% higher in African American men than non-Hispanic white men; and African American men are disproportionately burdened by PCa mortality, with age-adjusted rates 2–3 times higher than those of non-Hispanic white men.⁹ While additional research is certainly needed, MD-based dietary interventions targeting tumor aggression-related inflammation in this high-risk population could be a promising strategy to impact relevant oncologic outcomes such as PFS in men managed on AS..

Despite these findings, it is difficult to isolate aspects of the MD that may account for improvements in PFS among men with a high MD score. The MD includes a high consumption of foods such as fish, which has been related to a decreased incidence of PCa specific mortality.⁵² On the other hand, a high consumption of tomatoes and other fruits and vegetables rich in lycopene, also typical of the MD, may be associated with PSA levels in men with localized disease prior to surgery⁵³ and in those with metastatic disease following castration.⁵⁴ Mechanistically, lycopene consumption increases urinary and plasma levels of phytochemicals⁵⁵ and may affect androgen and estrogen metabolism along with a measure of oxidative stress response.⁵⁶ However, our study suggests that the sum of benefits gleaned from overall adherence to a Mediterranean-style diet may be greater than the individual parts for men at risk of PCa progression who are managed on AS.

Strengths of our study reside in the use of a prospective clinical AS protocol that includes an assessment of baseline diet, which, although self-reported, was sufficiently comprehensive (170 items) to derive the MD score. The pre-specified AS protocol yielded robust data regarding features of PCa diagnosis and clinical outcome assessment, allowingthe first investigation to directly examine associations between MD adherence and grade group progression in men on AS. The study is, however, limited by sample size and length of follow-up. Additionally, it was performed in a group of men with mostly low risk disease, potentially limiting its generalizability to a broader population of PCa patients. If confirmed in larger patient groups or validation cohorts, these findings may provide more definitive evidence for randomized interventions and subsequent recommendations for men who opt for AS to manage their PCa. Finally, residual or unmeasured confounding, particularly by physical activity, and bias and error in patient recall are potential limitations of these data, not allowing the drawing of strong or causal conclusions.

Among U.S. men with localized PCa on AS, adherence to the basic principles of the MD may lower risk of GG progression. Our suggestive findings warrant follow-up in larger studies of men enrolled on AS. Given that most men on AS are likely to have excellent prognosis, the known benefits of the MD on CVD morbidity and mortality, as well as the potential impact on PCa aggression, serves to bolster its promise to improve long-term quality of life in these patients.^12,57 At present, men with PCa who enroll on AS should continue to be encouraged to adhere to healthy lifestyles, such as those espoused by MD adherence, both before and after localized PCa diagnosis. Future studies are needed to verify patient- and cancer-specific effects of MD adherence in men with early stage PCa.