Dietary intervention in prostate cancer patients: PSA response in a randomized double-blind placebo-controlled study
Article first published online: 21 OCT 2004
Copyright © 2004 Wiley-Liss, Inc.
International Journal of Cancer
Volume 113, Issue 5, pages 835–840, 20 February 2005
How to Cite
Kranse, R., Dagnelie, P. C., van Kemenade, M. C., de Jong, F. H., Blom, J. H.M., Tijburg, L. B.M., Weststrate, J. A. and Schröder, F. H. (2005), Dietary intervention in prostate cancer patients: PSA response in a randomized double-blind placebo-controlled study. Int. J. Cancer, 113: 835–840. doi: 10.1002/ijc.20653
- Issue published online: 8 DEC 2004
- Article first published online: 21 OCT 2004
- Manuscript Accepted: 30 JUN 2004
- Manuscript Received: 16 DEC 2003
- Unilever Research Laboratories, Vlaardingen, The Netherlands
- prostate cancer;
- prostate-specific antigen slope;
- doubling time;
The objective of this study was to show or to exclude an effect of dietary supplement on rising prostate-specific antigen (PSA) levels. We have studied the effect of a dietary supplement (verum, administered for 6 weeks) containing plant estrogens, antioxidants, including carotenoids, selenium and other putative prostate cancer inhibiting substances in a randomized placebo-controlled double-blind crossover study in 37 hormonally untreated men with prostate cancer and increasing PSA levels. Outcome measures were changes in the rates of change of serum concentrations of total and free PSA and changes in male sex hormone levels. Male sex hormone levels were significantly lower during the verum phase (DHT: −0.11 nmol/L, p = 0.005; testosterone: −1 nmol/L, p = 0.02). Total PSA doubling time was unaffected. Free PSA, which increased during the placebo phase (average doubling time of 68 weeks), decreased during the verum period (average half-life of 13 weeks; p = 0.02). In those men in whom the free androgen index decreased (21 out of 32), a significant decrease in the slopes of both total and free PSA was observed (p = 0.04). Overall total PSA doubling times did not increase significantly during verum. However, the study demonstrates that this dietary intervention reduces DHT and testosterone levels and increases free PSA doubling time (and total PSA doubling time in a relevant subgroup). If future studies confirm that these observations translate into a slowing of disease progression, a dietary intervention may become an attractive option for prostate cancer treatment and prevention. © 2004 Wiley-Liss, Inc.
Epidemiologic data strongly suggest that environmental factors play a role in prostate cancer progression. Prostate cancer incidence and mortality rates in Japan and other Asian countries are much lower than in Western countries. Around 1985, before the prostate-specific antigen (PSA) era, age-standardized prostate cancer incidence rates were roughly 60–90 per 100,000 per year in the United States versus 7 per 100,000 per year in Osaka, Japan.1 Interestingly, autopsy studies have demonstrated that latent prostate cancer of the focal noninvasive type is as common in Japan as it is in Western countries.2, 3 When Japanese men move to California, their risk for prostate cancer increases 4-fold.4 This suggests that environmental factors, most probably diet, play a central role in promoting progression of prostate cancer from the latent to the clinical stage. Diet may affect the tumori genesis directly or indirectly by affecting hormone levels (prostate cancer is a hormone-sensitive tumor). Soy, a major constituent of the Asian diet, is rich in phytoestrogens.5 Estrogens have been, and still are, in use as treatment for advanced prostate cancer.
Within Europe, low prostate cancer incidence and mortality rates are observed in Mediterranean countries.6, 7 Low southern European rates have been suggested to be related to the consumption of tomatoes, which contain high levels of antioxidants, notably lycopene.8, 9 It is noteworthy that many isoflavonoids (present in the Eastern diet) and lignans (present in the fruits and vegetables of the Mediterranean diet, but also in cereals such as rye) are phytoestrogens. Many of these phytoestrogens exhibit antioxidant properties in addition to their estrogenic/antiestrogenic effects.8, 9, 10
The working hypothesis of the present study was that a cocktail of phytochemicals might affect male sex hormone levels and, simultaneously, the rates of increase of PSA and free PSA levels, which are markers for progression of prostatic cancer.11 This hypothesis and the choice of the components of the dietary supplement were based on evidence from epidemiologic studies. Increased use of some of the components such as selenium12, 13 and vitamin E14 have been shown to correlate with a decreased incidence of prostate cancer in prospective studies. Evidence has also been presented that (iso)flavonoids (green tea) and phytosterols may have a preventive effect.15, 16, 17 In this study of a dietary supplement, PSA levels may relate to tumor mass and, as far as the slope of PSA is concerned, with proliferative activity. However, since various substances have been shown to interfere with PSA per se, therapeutic efficacy cannot be derived from this study.18 Total PSA in serum analytically consists of free PSA (fPSA) and PSA bound to α-1-antichymotrypsin (ACT).19, 20 Free PSA was recently shown to relate to several molecular forms of PSA, of which one I-PSA may be more specific for cancer than for benign prostatic hyperplasia (BPH) or benign prostatic tissue.21
Material and methods
This is a double-blind randomized placebo-controlled 2-arm crossover intervention study of a dietary supplement in prostate cancer patients with rising PSA and without systemic treatment.
Planned study population and inclusion and exclusion criteria
Based on the sample size calculation, 15–20 patients were to be included in each treatment arm of this study. Participants were patients of the Departments of Urology of the Erasmus Medical Center and the St. Franciscus Gasthuis and were recruited during the fall of 1998.
Eligible for inclusion were men of any age with confirmed rising PSA levels >0.1 ng/ml and no clinical evidence of (recurrent) prostate cancer (PC) after radical prostatectomy, radiotherapy, or under watchful waiting. Prior to entry, PSA levels of >0.1 ng/ml had to be measured on >2 occasions at a >3-month interval. In patients studied after radiotherapy or during watchful waiting, rising PSA levels were established by visual inspection of PSA levels over time. Exclusions include any hormone or other systemic or radiotherapy after definitive treatment.
Intervention and timing
Participants were given a dietary supplement (verum) in the form of a water-dissolvable powder to be taken 3 times a day and 20 g of margarine per day. The experimental supplement contained a number of experimental ingredients (Table I), while the control supplements (placebo) margarine and drink contained none of these. Compliance was asserted by monitored dietary records and compliance markers. Source details of the compounds are available on request.
|Margarine (20 g)|
|Vitamin E (50 mg α-tocopherol)|
|Phytosterols (1.5 g)1|
|Selenium (0.2 mg organic selenium in 0.5 g bakers yeast)|
|Placebo beverage (3 servings of 200 ml/day)|
|Caffeine similar to supplement|
|Beverage (3 servings of 200 ml/day)|
|100 mg phytoestrogens|
|60 mg genestein|
|40 mg daidzein|
|10 mg luteine|
|10 mg lycopene|
|10 mg palm carotenoids (including some α-carotinoids)|
The time schedule and the details of interventions and diagnostic steps are indicated in Figure 1. Participants were requested not to change diet during the study (except from taking the supplements).
The primary endpoint of the study was the slope of the rise in PSA (dPSA) compared between placebo and verum. dPSA was defined as the slope of the regression line through all weekly 2log-transformed PSA measurements during the concerned period. Total PSA (tPSA) and fPSA were analyzed separately and in a similar fashion. Secondary endpoints were plasma levels of luteinizing hormone (LH), testosterone (T), 5α-dihydrotestosterone (DHT) and sex hormone binding globulin (SHBG) compared between intervention and placebo. The free androgen index (FAI) was defined as the ratio of testosterone and SHBG levels. Compliance with the use of the supplement was evaluated by measuring plasma concentrations of the phytosterols daidzein and genestein for the beverage and β-carotene for the use of margarine.
Statistical considerations and power
The sample size calculation was based on variations observed in the slope of PSA after radical prostatectomy. Details relating to this sample size calculation were reported earlier.11 It was estimated that a 70% reduction in PSA slope could be detected with a power of 90% with 20 subjects per treatment arm. It was therefore decided to include 30–40 patients in the study.
Participants taking <90% of all supplements were to be excluded from the analysis. Participants experiencing clinical progression or an unexpected rise of PSA were immediately informed. If, upon confirmation, treatment was to be changed, the patient was allowed to complete the dietary intervention but was excluded from the statistical analysis. This applied to all interventions that are known to interfere with PSA levels.
Loss to follow-up and adverse events
In case of suspected supplement-related adverse events, the code was to be broken. Patients lost to follow-up and dropouts were to be included in the analysis only for completed study periods.
Ethical approval was obtained from the institutional medical ethical committee. All participants gave written informed consent.
All participants were seen weekly by a dietician (M.C.v.K., Ineke Kersten). The frequency of dietary records and other follow-up procedures is indicated in Figure 1. Adverse events were systemically recorded.
Serum was separated immediately after blood collection and stored at −80°C; processing of the samples was conducted blinded to the treatment group. Total PSA and free PSA in serum were measured by use of the Immulite assay of Diagnostic Products Corporation (DPC, Los Angeles, CA). Since free PSA levels are typically quite low, an extremely sensitive assay (detection limit 0.01 ng/ml) was chosen for the present study.
Testosterone (coated tube radioimmunoassay; DPC), 5α-dihydrotestosterone (enzyme-immune assays provided by Diagnostics Biochem Canada, London, Ontario, Canada), LH and SHBG (Immulite kits; DPC) were determined in samples collected every odd week. Interassay variations for these assays were 4.8% for LH, 9.0% for SHBG, 8.1% for testosterone and 8.4% for DHT. Since only free testosterone is biologically active, the FAI (defined as the molar ratio of T and SHBG) was determined and used as a secondary endpoint together with the DHT index (DHTI, obtained by dividing DHT levels by SHBG levels). All blood samples were drawn between 08:00 and 13:00 o'clock. Quality control was performed by assaying samples from the same serum pool repeatedly within every assay run.
Antioxidants including vitamin E in plasma and the phytoestrogens daidzein, genestein, equol (a metabolite of daidzein) and enterolactone (a lignan not present in the verum as a negative control) were measured by mass spectrometry. These determinations were carried out as previously described.22 Vitamin E was analyzed according to the techniques previously described.23
All measured laboratory data were first plotted as a function of time for visual inspection. Simple descriptive statistics (mean, median, standard deviations) and nonparametric statistical tests (Wilcoxon's signed-rank test) were used to study the serum levels of the antioxidants, phytoestrogens and metabolites of phytoestrogens.
All PSA values were 2log-transformed to account for the exponential increase of PSA over time. Slopes for free and total PSA and hormone levels were determined for each period. The PSA and free PSA slopes were estimated by fitting a straight line through their 2log-transformed values. The doubling time can be calculated as 1/PSA slope if the slope is positive (i.e., if PSA increases over time). The half-life can be calculated identically if the slope is negative (i.e., if PSA decreases over time). The term “response” is used below both for hormone levels and for PSA/free PSA slopes.
The changes in tPSA and fPSA doubling times and in hormone levels were studied by means of analysis of variance. During each period, the observed change is assumed equal to the sum of average levels, a treatment effect, a period effect, an interaction between the period and the treatment effect and some residual variance. The treatment effect estimates the response that can be attributed to the effect of taking the supplement only. The period effect estimates changes attributed to the study period. The interaction term (or carryover effect) was evaluated in the model and subsequently removed because no significant effect was found. The 3 described variables, treatment effect, period effect, interaction and residual variation, were estimated by means of a linear regression procedure. In addition, the method proposed by Senn24 for analyzing crossover studies was applied. The resulting data are not given in detail for space considerations.
Spearman's rank correlation coefficients were calculated to explore the relation between differences in PSA or free PSA slopes from the verum compared to the placebo phase of the study and simultaneous differences in male sex hormone levels (T, DHT, SHBG, LH, FAI and DHTI). Two-tailed statistics and 95% confidence intervals are reported throughout this article; p-values < 0.05 were considered statistically significant.
Assignment and participants
Thirty-seven eligible patients consented and were assigned by individual randomization to either of the 2 treatment arms (Fig. 2). Color coding of verum and placebo (orange and blue) was used to achieve double blinding. Twenty-six participants had previously undergone a radical prostatectomy, 6 had been treated with radiotherapy, whereas the remaining 5 patients were managed by watchful waiting with their primary tumor in situ. At baseline, the following descriptive statistics were recorded (range, median value): age, 54–81, 70 years; weight, 58–107, 81 kg; PSA, 0.13–87.3, 3.24 ng/ml; testosterone, 10.1–32.1, 14.4 nmol/L; DHT, 0.86–5.92, 1.68 nmol/L. Eighty-three percent of the patients had clinical stage T1 or T2; in 60% of the patients who underwent a radical prostatectomy grade 1 or grade 2 tumors were found. All participants were recruited within a 2-week period.
Figure 2 gives a detailed account of recruitment, randomization, dropouts and time periods for those who completed the study. Other reasons for refusals (n = 14) included distance, inconvenience of frequent visits, vacation plans and unavailability for primary contacts. Six men who were eligible, registered and randomized declined for various reasons. There were no serious adverse events; one patient reported abdominal pains and decided to discontinue. He stopped during the washout period and was excluded because of lack of blood samples. No participants experienced clinical progression during the trial period; the code was broken for all 32 eligible participants at the time of completion of the study. Since the supplement was not available on the market, men who wished to continue on a similar regimen were given dietary advice. Severe violations in dietary intake were not seen.
Table II lists the serum levels of the food compounds for the different phases in the study. As expected on the basis of the chemical composition of the verum supplement, the serum levels of the 6 compounds listed in Table II were significantly higher during the intervention period as compared to the placebo period. For the carotenoids a 2-fold and for the phytoestrogens daidzein and genestein approximately 100-fold increases in serum levels were observed. The levels of enterolactone (p = 0.04) and equol (p = 0.09) also increased. Neither fat nor energy intake as measured by the 7-day food records changed over time; a slight (400 g in median weight) but statistically significant increase in weight was observed.
|Food compound (unit of measurement)||Placebo, mean, median, standard deviation||Verum, mean, median, standard deviation||p-value1|
|α-β-carotene (nmol/l)||608, 515, 410||1,369, 1,248, 768||< 0.0001|
|α-tocopherol (vitamin E) (μmol/l)||29, 29, 7||33, 32, 8||< 0.0001|
|lutein (nmol/l)||122, 115, 47||260, 239, 101||< 0.0001|
|Lycopene (nmol/l)||305, 296, 140||650, 649, 238||< 0.0001|
|Daidzein (ng/ml)||8.0, 7.0, 5.7||534, 520, 233||< 0.0001|
|Genestein (ng/ml)||17, 13, 15||1,589, 1,470, 739||< 0.0001|
|Enterolactone (ng/ml)||9, 5.8, 10.5||11, 9.4, 9.8||0.04|
|Equol (ng/ml)||0.34, 0.10, 0.79||52.1, 0.15, 103.1||0.09|
Table III shows the changes in total and free PSA kinetics between verum and placebo periods. Table IV summarizes changes in PSA doubling times. The doubling time of total PSA increased from an average of 41 weeks during the placebo period to 44 weeks during the verum period. This treatment effect was not statistically significant. If the analysis of variance was repeated for only those participants (21 men out of 32) in whom a decrease in free androgen index levels was observed, a significant treatment effect for total PSA slopes was found. In these men, the average PSA doubling time was 36 weeks during the placebo phase and 115 weeks during the verum phase of the study (p = 0.04; Table IV).
|Parameters||Mean response, 2-sided p-value||Treatment effect, 2-sided p-value||Period effect, 2-sided p-value|
|Total PSA slope2 log (ng/ml)/(week)|
|n = 32||0.024 (p < 0.001)||−0.0018 (p = 0.84)||0.0014 (p = 0.87)|
|n = 21 (FAI ≤ 0)||0.0184325 (p = 0.01)||−0.0194612 (p = 0.04)||−0.0121827 (p = 0.19)|
|Free PSA slope2 log (ng/ml)/(week)|
|n = 32||−0.032 (NS)||−0.093 (p = 0.02)||0.14 (p < 0.001)|
|n = 21 (FAI ≤ 0)||−0.0291954 (p = 0.29)||−0.1023862 (p = 0.05)||0.1068415 (p = 0.04)|
|Testosterone (ng/ml)||15.5 (p < 0.001)||−0.98 (p = 0.018)||−0.13 (p = 0.75)|
|Free androgen index (no dimension)||0.20 (p < 0.001)||−0.0095 (p = 0.07)||−0.0061 (p = 0.23)|
|Dihydrotestosterone (ng/ml)||1.90 (p < 0.001)||−0.11 (p = 0.005)||−0.059 (p = 0.12)|
|Dihydrotestosterone index (no dimension)||0.026 (p < 0.001)||−0.0014 (p = 0.13)||−0.0017 (p = 0.07)|
|SHBG (ng/ml)||83.4 (p < 0.001)||−2.25 (p = 0.23)||2.39 (p = 0.21)|
|LH (ng/ml)||7.01 (p < 0.001)||−0.11 (p = 0.73)||−0.34 (p = 0.30)|
|Marker||Subgroup||Doubling time during verum (95% CI)||Doubling time during placebo (95% CI)||p-value of the difference in doubling times|
|Total PSA||All men (n = 32)||44 (32 to 71)||41 (30 to 63)||0.84|
|Men with decreased FAI (n = 21)||115 (53 to −678)||36 (26 to 56)||0.042|
|Free PSA||All men (n = 32)||−13 (−25 to −9)||68 (19 to −43)||0.02|
|Men with decreased FAI (n = 21)||−12 (−20 to −8)||45 (14 to −40)||0.049|
The results of the analysis of variance described above are not given in detail for reasons of space. The 2 statistical methods used to analyze the data yielded identical results (analysis of variance and Senn's method24). The analysis showed a significant treatment effect for free PSA (slope decreased from a doubling time of 68 weeks to a half-life of 13 weeks; p = 0.02), testosterone (decreased 0.98 nmol/L; p = 0.02), DHT (decreased 0.11 nmol/L; p = 0.005) and FAI (decreased 0.01; p = 0.07; Table III).
Differences in the slopes of total PSA between the verum and placebo phase were significantly correlated with simultaneous differences in the free androgen index (Spearman's rank correlation = 0.47; p = 0.01). Differences in the free PSA slopes were significantly correlated with differences in the total PSA slopes and with differences in LH levels and DHT (details are on file).
A significant period effect was observed in one arm of the study for free PSA only. Free PSA slopes also decreased significantly from placebo to verum in the 21 men with a decrease in FAI and plasma testosterone levels. In these men, the doubling time was 45 weeks during placebo and changed to a half-life of 12 weeks during verum (p = 0.049). During the verum period, DHT levels and testosterone levels decreased significantly; the FAI decreased at borderline significance.
Prompted by the epidemiologic evidence referred to above, we have studied the effect of a dietary supplement on PSA and free PSA in a placebo-controlled crossover study. The study utilized innovative methodology to establish the potential efficacy of new preventive measures or agents for treatment of prostate cancer. We show that by using PSA slopes for comparison, a small sample size and a short time period can be utilized provided PSA progression has been established and quantified.
No significant treatment effect was seen in relation to total PSA. This may be in part due to the heterogeneity of the patient material, which is reflected in a great variability of the PSA levels seen during the run-in period25 and to the relatively short intervention period. At the same time, a significant treatment effect on free PSA slopes/doubling time was seen. Free PSA levels, which increased during the placebo phase with a doubling time of 68 weeks, decreased during verum treatment with an average half-life of 13 weeks. The significant period effect, which was observed in the analysis of free PSA only, may be explained by the presence of spontaneously (i.e., nonintervention-related) decreasing free PSA levels in a number of men (despite increasing total PSA levels) combined with the lower limit of the ultrasensitive free PSA assay used (lower limit of detection and resolution 0.01 ng/ml). In such men, the lower limit of detection is more likely reached in the second period of the study than in the first (the longer the time interval of spontaneous decrease, the lower the levels of free PSA will become). If free PSA levels are below the detection limit, the assay measures a steady 0.01 ng/ml corresponding to a free PSA slope of zero. Thus, in men with spontaneously decreasing free PSA levels and a low free PSA level at the baseline (many of the postradical prostatectomy patients), a period effect is likely to occur. If, in addition, verum affects free PSA slopes, a mix of a period and a treatment effect results.
Ratio of free and total PSA
The quotient of free and total PSA, FT ratio, has been introduced as a tool to differentiate BPH from prostate cancer (i.e., to differentiate between and not within individuals; low FT ratios correspond to an increased prostate cancer risk).26, 27 The use of the FT ratio within patients, e.g., to assess preoperatively the pathologic stage or monitor disease progression, is less well established.28, 29 In our study, no FT ratio treatment effect was found (data not shown).
We found that T and DHT levels decreased significantly by 0.98 nmol/L (6%) and by 0.11 nmol/L (6%) during the verum period when compared to the placebo phase. Simultaneously, a significant increase of tPSA doubling time from 36 to 115 weeks (p = 0.04) was seen in those 21 men whose free androgen index decreased during verum. In parallel, the free PSA levels, which increased during the placebo phase with a doubling time of 68 weeks, decreased during verum supplementation with an average half-life of 13 weeks (p = 0.049). When percent free testosterone calculated on the basis of the law of mass action30 was used instead of the free androgen index, our conclusions did not change qualitatively (the correlation coefficient for the relationship between FAI and percent free testosterone was 0.82).
Differences in plasma testosterone levels and in the free androgen index between the placebo and the verum period were positively correlated with simultaneous differences in the total PSA slopes. Also, the increase in free PSA showed a marginally significant negative correlation with differences in LH and free DHT index. It has been observed by others that, within certain PSA ranges, the levels of free PSA correlate with LH.31 Similar differences in androgen levels were seen in a comparison of Dutch and Japanese men32 and in a study of healthy Finnish men put temporarily on a vegetarian diet.33 Obviously, these small changes may only be indicative of other mechanisms that impact on PSA levels.
This study in part confirms our working hypothesis that during the verum period a significant decrease in fPSA is seen. During verum, testosterone and DHT levels decreased significantly, FAI levels decreased at borderline significance levels, while LH levels were unaffected. A significant decrease in the slope of fPSA and tPSA is seen in men in whom the free androgen index decreased during verum. At this point, we may only speculate about which substance in the verum cocktail may have caused the effects. As we have observed a hormone response, we think it is reasonable to assume that the phytoestrogens are at least in part involved. Treatment effects in fPSA and tPSA slopes correlated positively with each other and with treatment effects in the free androgen index. Based on the relatively small testosterone treatment effect, the fact that a consistent free and total PSA response is only observed in men with a decreasing FAI during verum and the absence of an LH effect, we hypothesize that the observed treatment effect may be related to effects on peripheral testosterone synthesis from androstenedione. Phytoestrogens block the activity of the enzyme 17β-hydroxysteroid dehydrogenase type 534 that converts androstenedione into testosterone.35 The conversion of androstenedione into estrone regulated by aromatase and the conversion of estrone to the biologically active estradiol by 17β-hydroxysteroid dehydrogenase type 1, however, might remain unaffected by the intervention. Intervention thus decreases testosterone levels and increases estrogen levels (since more androstenedione is available for conversion into estradiol). The “decreased testosterone level induced LH increase” may therefore be compensated by the “estradiol increase induced LH decrease,” leaving overall LH levels during testosterone unaffected as observed. Considering the setup of this study, judgment concerning a possible effect of the antioxidants is impossible.
The question whether the impact on fPSA and tPSA slopes seen correlates with a stabilization or decrease in tumor mass remains unresolved. This issue can be dealt with by using human prostate cancer lines transplantable in nude mice. In this setting, PSA and tumor mass can be studied simultaneously as previously suggested.11, 18
This study would not have been possible without the work and dedication of Bertrand van Zelst, Ellen van den Berg, Ineke Kersten, Mathilde Dijk, Dr. Bert Blijenberg, Monique Roobol, Dr. Wim Hop and Dr. Theo Stijnen. Karin van Alphen provided patient and effective secretarial support.
- 6Nutrition and cancer. Oxford: ISIS Medical Media, 1996., , , , , .
- 24The AB/BA cross-over: how to perform the two-stage analysis if you can't be persuaded that you shouldn't. Rotterdam: Erasmus University, 1996. 93–100., .