Soy and isoflavone consumption and subsequent risk of prostate cancer mortality: the Japan Public Health Center-based Prospective Study

Abstract

Background

Although many epidemiological studies have reported the preventive effects of soy products and isoflavones on prostate cancer, our previous studies reported that the association between soy and isoflavones and prostate cancer incidence differed according to stage. It is more important to identify modifiable risk factors related to lethal prostate cancer. Here, we investigated the association between soy, soy products and isoflavones intake and prostate cancer mortality, in a prospective study in Japan.

Methods

We conducted a population-based prospective study in 43 580 Japanese men with no history of cancer or cardiovascular disease (aged 45-74 years). Participants completed a validated questionnaire which included 138 food items. We followed participants from 1995 to 2016. Hazard ratios (HRs) and 95% confidence intervals (CIs) of prostate cancer mortality were calculated according to quintiles of soy products and isoflavones intake, using Cox hazard proportional hazards regression.

Results

During 16.9 years follow-up, we registered 221 deaths from prostate cancer. Isoflavones and soy products intake was associated with an increased risk of prostate cancer death, with multivariate HR_{Q5 vs. Q1}=1.39, 95% CI = 0.87-2.20, p for trend = 0.04 for isoflavones and multivariate HR_{Q5 vs. Q1}=1.76, 95% CI = 1.10-2.82, p for trend = 0.04 for soy food.

Conclusions

Our study suggested that high intake of soy and isoflavones might increase the risk of prostate cancer mortality.

Soy, isoflavone, prostate cancer death, prospective, JPHC Study

Key Messages

• The present data from the Japan Public Health Center-based Prospective Study show that isoflavones and soy products intake was associated with an increased risk of prostate cancer death in Japanese men.

• Regarding the type of soy food, miso intake tended to show a positive association with prostate cancer mortality, but natto and tofu intake showed a null association.

• This result is plausible under the following explanation. First, some epidemiological studies have shown a positive association between isoflavones and soy products and advanced prostate cancer. Second, the preventive effects of soy and isoflavones might be weaker against advanced and lethal prostate cancer due to their lack of estrogen receptor β. Third, biphasic effects of isoflavones may lead to the progression of prostate cancer in patients receiving hormone therapy, such as anti-androgenic agents.

Introduction

Although prostate cancer was the second most common cancer in men worldwide in 2018, the estimated age-standardized rates per 100 000 person-years (ASR, world standard) of prostate cancer incidence and mortality substantially diverge between Asia and Western countries, at 13.9 and 4.7 in Eastern Asia, 73.7 and 7.7 in Northern America and 85.7 and 13.5 in Northern Europe.¹ Given that prostate cancer incidence is much higher in Japanese immigrants in the USA or Brazil compared with the Japanese in Japan,^2,3 the wide range of prostate cancer incidence can be partly explained by the effects of environmental factors, including diet. In particular, the high consumption of soybean, which contains the phytoestrogens known as isoflavones, is one possible reason for low prostate cancer incidence in Asia.⁴

Although the World Cancer Research Fund stated in 2007 that soy (pulses) was a ‘limited suggestive factor’ for decreasing prostate cancer risk,⁵ with continuous updating of the project it is now currently listed as ‘limited no conclusion’.⁶ According to two meta-analyses, consumption of soy food⁷ and phytoestrogen⁸ is associated with a decreased risk of prostate cancer, especially in Asian populations. We also identified an inverse association between soy products, isoflavone and prostate cancer in the Japan Public Health Center-based Prospective Study (JPHC Study).^9,10 Although we showed that soy products and isoflavones had preventive effects only on localized cancer, studies using information on intake¹⁰ and plasma level⁹ have shown that higher isoflavone levels were associated with an increased risk of advanced prostate cancer, which is more lethal. Given that autopsy studies show that the incidence of latent or clinically insignificant prostate cancer in Asian and the USA is similar,^11,12 this inconsistency in the effect of isoflavones by stage may suggest that isoflavones only delay the discovery of prostate cancer. In terms of cancer prevention, 5-year survival rates for localized prostate cancer are 100% in Japan,¹³ rendering the identification of modifiable risk factors for death due to prostate cancer the more significant goal.

Here, we investigated the association between soy, soy products and isoflavone intake and prostate cancer mortality, in a prospective study in Japan.

Methods

Study subjects

The JPHC Study is an ongoing prospective study which was initiated in 1990 for Cohort I and in 1993 for Cohort II (68 722men). The study design has been described in detail elsewhere.¹⁴ Subjects in Cohort I are on the whole middle-aged residents (40-59 years) who had registered their addresses in five public health centre (PHC) areas (Iwate, Akita, Nagano, Okinawa-Chubu and Tokyo), and those in Cohort II were all residents (40-69 years) who had registered in six PHC areas (Ibaraki, Niigata, Kochi, Nagasaki, Okinawa-Miyako and Osaka). After initiation of the study, we excluded ineligible subjects (n = 158), namely those with: non-Japanese nationality (n = 31); migration before the start of the follow-up period (n = 121); incorrect birth data (n = 6); or death before the starting point or loss to follow-up (n = 2984). A total of 65 422 men were eligible for participation in this study.

We asked participants to complete the questionnaire, including items related to diet, three times, once every 5 years. The food frequency questionnaire (FFQ) in the baseline survey had fewer items (44 food items for Cohort I and 52 food items for Cohort II). In contrast, the 5-year follow-up survey used a more comprehensive self-administered FFQ which included 147 food and beverage items with standard portions/units and nine frequency categories. Due to this greater detail, the starting point in the present study was therefore the 5-year follow-up survey, and the subjects were followed from 1995 for Cohort I and from 1998-99 for Cohort II until the end of follow-up. Among eligible subjects, 48 306 men (74%) returned valid responses to the 5-year follow-up questionnaire. To minimize the effect of changing diet after a diagnosis of severe disease, subjects who had a self-reported past history of cancer or cardiovascular disease (n = 1864) were excluded, leaving 46 422 men.

This study was approved by the institutional review board of the National Cancer Center, Tokyo, Japan (Figure1).

Figure 1

Flow chart of study participants in this JPHC Study. All resident men aged 40-59 years in five public health centre (PHC) areas and all resident men aged 40-69 years in six PHC areas at baseline, excluding persons who were ineligible (non-Japanese nationality, late report of migration occurring before the start of the study, incorrect birth data), had died, or were lost to follow-up. All resident men in 11 public health centre (PHC) areas at 5-year follow-up. Participants who answered the questionnaire (74%), excluding persons with a self-reported past history of cancer or cardiovascular disease andpersons who reported an extreme total energy intake

Follow-up

Subjects were followed from the starting point (the 5-year follow-up survey) until the date of death, emigration from Japan or the end of follow-up (31 December 2009 in the Tokyo area, 2012 in the Osaka area and 2016 in other areas), whichever came first. Changes in residence status, including survival, were obtained annually through the residential registry in the municipality of each study area. For subjects who moved to other municipalities, these were obtained through the municipal office of the area to which they had moved and were followed until they died. In general, mortality data for residents included in the residential registry are sent to the Ministry of Health, Labour and Welfare and coded in the national Vital Statistics. Death from prostate cancer was classified using the ICD-10 classifications (C61). Residency and death registration are required by the Basic Residential Register Law and Family Registry Law, respectively, and the registries are believed to be complete. Among questionnaire responders at the starting point, 14 094 men (20.3%) died and 126 (0.3%) were lost to follow-up during the study period.

Food frequency questionnaire

In our FFQ, seven items of soy food, whose major ingredient is soybeans, referred to the consumption of tofu, natto [fermented soybean], miso [fermented soybean paste], yushidofu [pre-drained tofu], koyadofu [freeze-dried tofu], aburaage [deep-fried tofu] and soy milk. The questionnaire asked about the average consumption of 147 foods and beverages during the previous year. Total soy products are the sum of the seven soy food items. Miso was calculated from the consumption of ‘miso soup _*0.08’ because we regarded miso soup as the sum of approximately 92% water and 8% miso, based on the usual recipe. The frequency of miso soup consumption was divided into six categories: almost never, 1-3 days/month, 1-2 days/week, 3-4 days/week, 5-6 days/week and daily. Portion sizes were qualified, and the amounts were provided in three categories (less than 50% or equal to or more than 150% of a standard serving). Nine categories were used for soy food consumption: almost never, 1-3 times/month, 1-2 times/week, 3-4 times/week, 5-6 times/week, once/day, 2-3 times/day, 4-6 times/day and 7 or more times/day. Portion sizes were qualified, and the amounts were divided in three categories (less than 50% or equal to or more than 150% of a standard serving). Ten frequency categories were used for soy milk: almost never, 1-3 times /month, 1-2 times/week, 3-4 times/week, 5-6 times/week, 1 glass/day, 2-3 glasses/day, 4-6 glasses/day, 7-9 glasses/day and more than 9 glasses/day. The total consumption of soy food (g/day) and miso soup (ml/day) was calculated from these responses and of isoflavones (mg/day) was estimated using the values in a specially developed food composition table for isoflavones in Japanese foods.^15,16 In this study, isoflavones means daidzein and genistein in the sum of grams.

Our previous validation study¹⁷ using 14- or 28-day dietary records with 244 men and 254 women among subsamples for this study, found that the correlation coefficient (Spearman) between the energy-adjusted intake of soy products and miso soup consumption from the FFQ and from dietary records was 0.39 and 0.51 in men and 0.33 and 0.51 in women, respectively. With regard to reproducibility between the two FFQs administered 1 year apart, respective correlation coefficients of soy products and miso soup were 0.62 and 0.79 for men, and 0.56 and 0.81 for women.¹⁷

Among the 46 422 men who responded to the questionnaire and who did not have a history of cancer or cardiovascular disease, 2842 who reported an extreme total energy intake (upper 2.5% (>4190 kcal) or lower 2.5% (<993kcal)) were excluded, leaving 43 580 men. Finally, a total of 43 580 men, including 221 prostate cancer deaths, were used in the analysis.

Statistical analysis

Hazard ratios (HRs) and 95% confidence intervals (CIs) of prostate cancer mortality were calculated by quintile of energy-adjusted dietary intake of soy food and isoflavones, using a Cox proportional hazards regression model. The residual method was used for energy adjustment of intake of soy food and isoflavones.¹⁸ The model was adjusted for the following confounding factors: age at starting point (5-year groups),;study area (10 public health cenrers); smoking status (never, former, current); alcohol frequency (almost never, ≦4, ≧5 times/week); body mass index (<21, 21-22.9, 23-24.9, ≧25 kg/m²),;frequency of leisure time activity (almost never, <3, ≥3 times/week); history of diabetes mellitus (yes, no); history of screening in the starting year of the study (any of the following: blood pressure, electrocardiogram, chest X-ray, gastric photofluorography, and faecal occult blood test; yes, no); and consumption of green tea (never, ≤6 times/week, 1 cup/day, 2–3 cups/day, ≥4 cups/day), coffee (never, ≤6 times/week, 1 cup/day, 2–3 cups/day, ≥4 cups/day), vegetables (g/day) and fruits (g/day). Because we did not collect information for prostate-specific antigen (PSA) screening from all subjects, we used a history of screening in the starting year of the study as a surrogate marker. We created a separate code for missing data and treated missing data as dummy variables. We chose these as confounding factors because they were either known or suspected to be risk factors for mortality, or risk factors for prostate cancer, as previously identified in the JPHC Study.^19–24 In sensitivity analysis, we additionally estimated HRs by excluding deaths within 5 years after the starting point to diminish the possibility of reverse causality; p for trends were estimated by assigning the ordinal value for each category. Furthermore, we calculated p interaction values by a likelihood ratio test to compare Cox proportional hazards models with and without cross-product terms for smoking (never or ever smokers) and body mass index (< 25 or ≥25) and past history of diabetes mellitus (no or yes). All p-values were two-sided. The Cox proportional hazards regression model was conducted using PHREG command using SAS software (version 9.3; SAS Institute Inc., Cary, NC, USA). In addition, the Fine and Gray model was conducted to assess competing risk of death from all-cause mortality except for prostate cancer using the STCRREG command by STATA (version 15.0; StataCorp LP). Furthermore, we examined non-linear effects modelled with natural cubic splines, with four knots at the quartiles of distribution of the exposure variable using the MKSPLINE command by STATA.

Results

During 785 411 person-years of follow-up (average follow-up, 16.9 years) for 43 580 men, a total of 221 deaths from prostate cancer were confirmed and included in the analyses.

Table 1 shows the characteristics of study subjects according to isoflavones intake. Men with higher isoflavones intake were older, smoked less, exercised more and reported a higher prevalence of diabetes mellitus and history of screening. Additionally, men with higher isoflavones intake drank less coffee and more green tea, and had a higher intake of fruits, vegetables and fibre.

Table 1

Characteristics of study subjects according to quintile of isoflavone intake

	Men (n = 43 580)
	Total isoflavone intake (median, mg)
	Lowest 13.5	Second 24.6	Third 35.0	Fourth 48.8	Highest 75.3
Number	8716	8716	8716	8716	8716
Age, years ± SD	55.7 ± 8.1	56.2 ± 7.9	56.2 ± 7.7	56.5 ± 7.6	57.6 ± 7.6
Body mass index, ± SD (kg/m2)	23.5 ± 3.0	23.5 ± 2.9	23.6 ± 2.8	23.6 ± 2.8	23.6 ± 2.8
Current smoker, %	51.4	48.7	46.9	45.5	41.3
Alcohol intake (≥5 days/week), %	49.7	51.9	50.4	50.7	47.0
Leisure time physical activity (≥3 times/week), %	20.2	21.3	22.1	22.2	24.8
Past history of diabetes mellitus (yes), %	5.0	5.6	6.2	6.5	8.7
Health checkup (yes), %	77.8	81.8	83.1	83.9	85.4
Green tea intake (daily), %	47.2	55.5	58.3	62.0	65.2
Coffee intake (daily), %	43.3	37.1	33.2	29.3	26.6
Vegetables intake ± SD (g/day)	163.3 ± 127.9	189.1 ± 121.0	201.7 ± 122.5	210.2 ± 123.1	231.0 ± 143.6
Fruits intake ± SD (g/day)	153.0 ± 157.7	170.4 ± 152.2	180.6 ± 147.1	190.3 ± 145.6	201.1 ± 156.9
Soy food intake ± SD (g/day)	30.5 ± 13.6	55.6 ± 16.5	77.3 ± 22.8	102.6 ± 32.7	178.3 ± 125.2
Miso intake ± SD (g/day)	9.1 ± 7.7	17.6 ± 10.8	22.2 ± 13.1	24.3 ± 14.4	25.6 ± 15.6
Tofu intake ± SD (g/day)	13.2 ± 8.6	21.8 ± 12.5	29.6 ± 17.1	39.3 ± 23.8	66.1 ± 59.8
Natto intake ± SD (g/day)	1.9 ± 2.9	5.0 ± 5.3	9.1 ± 7.8	17.0 ± 12.0	37.4 ± 32.7

SD, standard deviation.

Table 2 shows HRs of prostate cancer deaths according to intake of isoflavones, soy food, miso, tofu and natto. There was a positive association between intake of soy food and prostate cancer mortality (multivariate HR_{Q5 vs. Q1}=1.76, 95% CI = 1.10-2.82, p for trend = 0.04). A similar association was shown for isoflavones intake (multivariate HR_{Q5 vs. Q1}=1.39, 95% CI = 0.87-2.20, p for trend = 0.04). The association between miso and prostate cancer mortality was positive in the third and highest categories, with multivariate HRs (95% CI) _{Q3 vs. Q1} of 1.64 (1.02-2.62) and HRs (95% CI) _{Q5 vs. Q1} of 1.73 (1.08-2.79). The relation between miso and prostate cancer mortality was non-linear (p for trend = 0.09). Natto and tofu intake showed a null association with prostate cancer mortality (multivariate HRs (95% CI) _{Q5 vs. Q1} were 1.19 (0.75-1.90) for natto and 1.16 (0.77-1.75) for tofu).

Table 2

Adjusted hazard ratios (HRs) of mortality by prostate cancer according to isoflavone and soy food intake in men

	Intake by quintile
	Lowest	Second	Third	Fourth	Highest	p for trend
Isoflavone intake
Median (mg)	13.5	24.6	35.0	48.8	75.3
Person-years	143 571	147 846	148 971	150 029	148 017
No. of prostate cancer deaths	37	33	44	51	56
HR1 (95% CI)	1.00	0.82 (0.51-1.32)	1.10 (0.71-1.73)	1.28 (0.82-2.00)	1.30 (0.83-2.03)	0.06
HR2 (95% CI)	1.00	0.85 (0.53-1.37)	1.16 (0.73-1.82)	1.35 (0.86-2.13)	1.39 (0.87-2.20)	0.04
Excluding first 5 years
No. of prostate cancer deaths	34	32	41	47	51
HR2 (95% CI)	1.00	0.90 (0.55-1.47)	1.17 (0.73-1.87)	1.37 (0.85-2.20)	1.37 (0.85-2.22)	0.06
Soy food
Median (g)	29.8	52.2	72.6	98.7	156.0
Person-years	143 298	147 607	149 327	150 426	147 776
No. of prostate cancer deaths	29	38	49	39	66
HR1 (95% CI)	1.00	1.18 (0.73-1.93)	1.42 (0.89-2.26)	1.08 (0.66-1.78)	1.62 (1.03-2.57)	0.07
HR2 (95% CI)	1.00	1.25 (0.76-2.04)	1.49 (0.93-2.40)	1.16 (0.70-1.93)	1.76 (1.10-2.82)	0.04
Excluding first 5 years
No. of prostate cancer deaths	27	34	47	35	62
HR2 (95% CI)	1.00	1.21 (0.72-2.03)	1.54 (0.94-2.51)	1.13 (0.67-1.91)	1.78 (1.09-2.90)	0.04
Miso
Median (g)	2.9	10.5	18.2	26.1	38.0
Person-years	139 090	144 833	150 085	152 801	151 625
No. of prostate cancer deaths	29	40	53	35	64
HR1 (95% CI)	1.00	1.25 (0.77-2.02)	1.61 (1.01-2.58)	1.01 (0.60-1.69)	1.73 (1.08-2.78)	0.08
HR2 (95% CI)	1.00	1.26 (0.78-2.05)	1.64 (1.02-2.62)	1.03 (0.61-1.73)	1.73 (1.08-2.79)	0.09
Excluding first 5 years
No. of prostate cancer deaths	25	39	48	32	61
HR2 (95% CI)	1.00	1.40 (0.84-2.24)	1.68 (1.02-2.77)	1.07 (0.62-1.84)	1.87 (1.13-3.08)	0.08
Natto
Median (g)	0	3.5	9.4	18.4	39.1
Person-years	210 073	131 442	133 250	133 007	130 662
No. of prostate cancer deaths	78	31	25	43	44
HR1 (95% CI)	1.00	0.90 (0.58-1.41)	0.77 (0.47-1.28)	1.27 (0.82-1.98)	1.16 (0.74-1.82)	0.28
HR2 (95% CI)	1.00	0.91 (0.58-1.42)	0.78 (0.47-1.30)	1.32 (0.84-2.08)	1.19 (0.75-1.90)	0.22
Excluding first 5 years
No. of prostate cancer deaths	74	31	24	37	39
HR2 (95% CI)	1.00	0.94 (0.60-1.49)	0.76 (0.45-1.28)	1.15 (0.71-1.86)	1.06 (0.65-1.73)	0.62
Tofu
Median (g)	7.4	15.9	24.8	37.8	69.8
Person-years	147 727	149 368	148 363	146 805	146 171
No. of prostate cancer deaths	45	36	43	42	55
HR1 (95% CI)	1.00	0.82 (0.53-1.27)	0.95 (0.62-1.44)	0.89 (0.58-1.36)	1.09 (0.73-1.63)	0.55
HR2 (95% CI)	1.00	0.84 (0.54-1.31)	0.99 (0.64-1.52)	0.93 (0.60-1.44)	1.16 (0.77-1.75)	0.38
Excluding first 5 years
No. of prostate cancer deaths	43	33	41	39	49
HR2 (95% CI)	1.00	0.81 (0.51-1.28)	0.99 (0.64-1.53)	0.92 (0.58-1.44)	1.09 (0.71-1.68)	0.53

HR1, adjusted for age, area; HR2, adjusted for age, area, smoking, alcohol frequency, body mass index, leisure time activity, history of diabetes mellitus, screening and intake of green tea, coffee, vegetables and fruit.

These associations were not attenuated after the exclusion of subjects who died in the first 5 years. Additionally, in the competing risk approach, the subhazard ratios for death from all-cause mortality except for prostate cancer were not substantially different (data not shown). Furthermore, no interactions were seen after stratification by smoking (p_interaction=0.13), body mass index (p_interaction=0.18) and past history of diabetes mellitus (p_interaction=0.71). Additionally, we further adjusted for family history of prostate cancer up to 5 years before the starting point, because we did not collect information of family history at the starting point. However, the results were not changed (data not shown).

Discussion

In this study, we found that high intake of soy and isoflavones might be associated with increased risk of prostate cancer mortality. To our knowledge, this is the first prospective study to report that isoflavones and soy intake are associated with an increased risk of prostate cancer mortality. Our results support previous reports of an ecological study of prostate cancer mortality. Nagata reported a positive association between soy intake and prostate cancer mortality, albeit this was reduced after adjustment for mean age, proportion of smokers and alcohol intake.²⁵

Many previous epidemiological studies have reported the protective effects of isoflavones intake^7,10,26–31 and blood or urine isoflavones level^9,32 on prostate cancer incidence. However, a recent prospective study showed that intake of isoflavones, genistein and daidzein, was associated with an increased risk of advanced prostate cancer, with a relative risk (95% CI) for the highest versus lowest intake of 1.91 (1.25-2.92) for isoflavones, 1.51 (1.02-2.22) for genistein and 1.80 (1.18-2.75) for daidzein.³³ Additionally our previous paper from the JPHC Study,^9,10 a study in Japanese Americans,³⁴ and the European Prospective Investigation into Cancer and Nutrition (EPIC) study³⁵ reported a tendency of a positive association between isoflavones or soy food and advanced or high-grade prostate cancer: relative risk (95% CI) for daidzein intake,¹⁰ miso intake¹⁰ and plasma daidzein⁹ in the JPHC Study was 1.43 (0.63-3.28), 2.79 (1.19-6.55) and 1.64 (0.34-7.87), respectively; relative risk for tofu intake in the study among Japanese Americans was 1.26 (0.67-2.39)³⁴ and relative risk for plasma genistein in EPIC was 1.57 (0.72-3.40).³⁵ Nevertheless, all showed preventive associations against localized or low-grade cancer in these studies.^9,10,35 These results suggest that soy and isoflavones might prevent only localized prostate cancer, and conversely have an adverse effect on lethal cancer.

Soy isoflavones are structurally similar to 17β-estradiol and can bind to estrogen receptors beta (ER-β).³⁶ This is considered to be one of the mechanisms of their preventive effects on prostate cancer. However, our present findings do not support the hypothesis that isoflavone intake has preventive effects on advanced prostate cancer. Although isoflavones have much higher affinity for ER-β than estrogen receptors alpha (ER-α), most studies have reported that ER-β expression levels decrease with increasing Gleason grade score or recurrent tumours.^36–38 On this basis, the preventive effects of soy and isoflavones via ER-β might be weak on advanced and lethal prostate cancer, which has a lower ER-β content than localized prostate cancer.

Furthermore, an additional mechanism may be biphasic effects of isoflavones. Sex hormone levels might determine the function of isoflavones.³⁹ An animal experiment reported that genistein was anti-androgenic in the testis in intact male mice, but showed androgen agonistic activity by reporter gene activity in castrated male mice.⁴⁰ On this basis, isoflavones might lead to the progression of prostate cancer in patients receiving hormone therapy, such as anti-androgenic agents. Recent research reported that pre-diagnostic diet has an effect on treatment failure,⁴¹ although our study unfortunately did not have information on diet after cancer diagnosis and we did not confirm whether pre-diagnostic isoflavone intake affects patients treated with hormone therapy, due to a lack of treatment information. However, a relatively high proportion of patients appear to be treated with hormonal therapy. Among 1125 patients treated with radiotherapy between 1995 and 2013 in Japan, 867 patients with stage T1–T4, N0, M0 prostate cancer received androgen deprivation therapy (ADT) (77%) with radiotherapy.⁴² In Swedish data also, use of ADT with radiotherapy for locally advanced prostate cancer is high, at 90% in 2006 and 2012.⁴³ Further study on prostate cancer survivors is needed.

Our study showed that soy food and miso increased the risk of prostate cancer mortality, whereas natto showed no association. Although the reason for this discrepancy is not clear, content of fibre and protein per gram is higher in natto (6.7 g and 16.5 g per 100 g natto, respectively) than in miso (4.1 g and 13.1 g per 100 g miso, respectively). Kim et al. recently reported that a pre-diagnostic diet characterized as low-sugar/high-fibre plus low-fat/high-protein was associated with a decreased risk of treatment failure following radical prostatectomy in clinically early-stage prostate cancer.⁴¹ Fibre and protein intake may overcome the adverse effects of isoflavone intake on the progression of prostate cancer. Furthermore, we cannot rule out the possibility of residual confounding, which we were unable to evaluate.

Our study has several limitations. First, we did not collect information on treatment for prostate cancer or PSA screening, and we cannot rule out the possibility that treatments and PSA screening had different effects on prostate cancer death. Second, soy food intake was evaluated only a single time, at the study starting point, and the possibility of misclassification due to habitual change cannot be excluded. However, with regard to the reproducibility of estimations between two questionnaires administered 1 year apart, Spearman’s correlation coefficient for the soy products was around 0.80, which means it was relatively high.¹⁷ Third, we did not collect family history of prostate cancer at the starting point and were therefore unable to adjust for it. However, when we adjusted for family history of prostate cancer using data collected 5 years before the starting point, our results did not change. Fourth, unknown residual confounding might have affected our results, although we adjusted for several confounding factors in statistical analyses. Notwithstanding these limitations, this study represents a novel assessment of the associations between soy and isoflavones intake and prostate cancer mortality in a large prospective cohort study among Japanese men, who consume more soy food than Western populations. Other advantages were a high response rate (around 80%) and negligible loss to follow-up, which minimized selection bias.

In conclusion, our study suggested that high intake of soy and isoflavones might be associated with an increased risk of prostate cancer mortality. Further studies to confirm our results are needed. Data are available at [https://epi.ncc.go.jp/en/jphc/805/index.html].

Funding

This study was supported by the National Cancer Center Research and Development Fund (since 2011) [23-A-31(toku), 26-A-2 and 29-A-4], Grant-in-Aid for Cancer Research from the Ministry of Health, Labour and Welfare of Japan (from 1989 to 2010) (19shi-2) and the Ministry of Agriculture, Fishery and Forestry, Japan (JPJ005336).

Acknowledgements

JPHC members are listed at the following site (as of April 2018): [https://epi.ncc.go.jp/en/jphc/781/8233.html].

Author contributions

N.S.: responsible for management of the study, data collection, statistical analysis, data interpretation, and manuscript drafting; M.I., T.Y., T.S., S.S., M.I.: reviewed and edited the manuscript, data collection, and contributed to the discussion; S.T. (principal investigator): obtained funding and designed, initiated and organized the study. N.S. will act as guarantor for the paper.

Conflict of interest

None declared.