Sexual activity and prostate cancer risk in men diagnosed at a younger age


  • R.P. is the BAUS Section of Oncology Representative. R.E and K.R.M are joint senior authors. Lists available on request for The UK Genetic Prostate Cancer Study Collaborators, Surrey, and British Association of Urological Surgeons Section of Oncology, London, UK.

Polyxeni Dimitropoulou, University Department of Oncology, University of Cambridge, Box 279 (S4), Addenbrooke’s Hospital, Cambridge, Cambridgeshire CB2 0QQ, UK.
e-mail: or

Kenneth Muir, University of Nottingham Medical School, Queens Medical Centre, Nottingham NG7 2UH, UK.



To examine, in a case-control study, the association between the frequency of sexual activity (intercourse, masturbation, overall) and prostate cancer risk in younger men diagnosed at ≤60 years old.


In all, 431 prostate cancer cases and 409 controls participated and provided information on their sexual activity. In particular, the frequencies of intercourse and masturbation during the participants’ different age decades (20s, 30s, 40s, 50s) were collected.


Whereas frequent overall sexual activity in younger life (20s) increased the disease risk, it appeared to be protective against the disease when older (50s). Alone, frequent masturbation activity was a marker for increased risk in the 20s and 30s but appeared to be associated with a decreased risk in the 50s, while intercourse activity alone was not associated with the disease.


These findings could imply different mechanisms by which sexual activity is involved in the aetiology of prostate cancer at different ages. Alternatively, there is a possibility of reverse causation in explaining part of the protective effect seen for men in their 50s.


sexual activity


times per month


times per week


odds ratio


multivariate logistic regression


sexually transmitted disease/venereal disease.


In Western countries, prostate cancer is amongst the most common adult-onset cancers in men [1]. Evidence from recent studies supports the role of sexual activity (SA) as a factor in the risk of prostate cancer, but the exact mechanisms by which this takes place are unknown. Some studies have supported the hypothesis that the concentration of the male hormone androgen inside the prostate affects the risk of prostate cancer, and that, in general, prostate cancer is hormone-dependent [2]. Sexual drive is modulated by the male hormones, making men with a high SA prone to a higher risk of prostate cancer [3]. However, it remains inconclusive as to whether different levels of androgen in men determine the variation in risk [4].

Only age, family history of prostate cancer and ethnicity have so far been established as risk factors for the disease [5]. The probability of prostate cancer increases after 50 years of age and most men diagnosed are aged >65 years, but 25% are diagnosed when aged <65 years ( It has been shown that African-American men have a higher occurrence of aggressive prostate cancer than white men [6]. The risk is increased by two to three times for someone with an affected father or brother (especially for younger men) and increases proportionately with the number of affected relatives [7]. However, the genes identified to date account for a small fraction of cases.

We conducted a case-control study to assess the role of SA, including intercourse and masturbation, in prostate carcinogenesis in younger men. We defined a priori our variables of interest (intercourse, masturbation, overall SA) and the age periods (20s, 30s, 40s, 50s and lifetime) to be examined. A strength of this study is that not only intercourse was considered when defining activity, as is the case with most other studies. Other important points are that different decades of life, lifetime activity and early-onset disease were examined.


The Prostate Cancer Research Foundation Study was established to collect both epidemiological and genetic information on cases diagnosed at ≤60 years old; 431 symptomatic cases and 409 controls were used for the present analysis. Controls consisted of men in the community encouraged to enrol by their GP; GPs were asked to provide up to six controls for each case. Participants were sent a postal questionnaire, with questions about their lifestyle and their sexual habits. Frequencies for intercourse and masturbation during the participants’ different age decades (20s, 30s, 40s, 50s) were collected retrospectively. Other information that was requested for age decades spanning 20–60 years was the number of sexual partners, types of partner, age at first intercourse/masturbation, and presence of a sexually transmitted disease/venereal disease (STD/VD). Intercourse and masturbation were examined separately, and in a combined analysis. As the reported frequencies of intercourse and masturbation for each participant were ranges (e.g. two to three times per week, tpw), a range was produced additively for each reported pair of values and the mean was assigned as the overall frequency of SA. This variable was assessed for each age decade separately. The frequency categories were: none, less than once per month, 1–3 times per month (tpm), once a week, 2–3 tpw, 4–6 tpw and daily.

The men were predominantly white (2.8% other races). Estimates of response rates for cases and controls are based on completed questionnaire returns from those participants who consented for their data to be used. The case response rate was 73% and the control response rate 74.1%.

In all, 486 (59%) of men reported a frequency of SA of ≥12 tpm in their 20s; the respective numbers for the 30s, 40s and 50s decreased to 391 (48%), 229 (28%) and 104 (13%).


Table 1 presents the demographic and other characteristics of the population used. Odds ratios (ORs) and CIs for disease association are presented after age-/race- adjusted analysis. The variables for combined SA (intercourse and masturbation) in the different age ranges were significantly positively correlated (all P < 0.001, Table 2).

Table 1.  Demographic and physiological characteristics for cases and controls
VariableCases, n (%)Controls, n (%)Adjusted* OR (95% CI)P
  • *

    adjusted for age at diagnosis and ethnicity. Significance level P < 0.05.

Age at diagnosis, years   0.016
 ≤52.9114 (27.2)126 (33.6)1.00 
 53–56.4125 (29.8)123 (32.8)1.15 (0.80, 1.64)0.446
 ≥56.5180 (43.0)126 (33.6)1.62 (1.15, 2.28)0.006
 non-white 19 (4.4)  4 (1.0)1.00 
 white408 (95.6)401 (99.0)0.22 (0.08, 0.67)0.007
Marital status   0.486
 never married 19 (4.5) 11 (2.7)1.00 
 widowed/separated/divorced 53 (12.5) 49 (12.2)0.61 (0.26, 1.42)0.249
 married353 (83.1)343 (85.1)0.63 (0.30, 1.36)0.243
Smoking   0.861
 non-smoker130 (30.4)133 (32.7)1.00 
 ex-smoker212 (49.6)195 (47.9)1.09 (0.79, 1.51)0.606
 smoker 85 (19.9) 79 (19.4)1.09 (0.72, 1.64)0.684
Age at first intercourse, years   0.066
 <20215 (51.4)184 (45.9)1.00 
 20–24142 (34.0)169 (42.1)0.71 (0.52, 0.97)0.030
 25–never 61 (14.6) 48 (12.0)1.03 (0.66, 1.6)0.902
Baldness (40s)    
 moderate/severe 96 (23.1) 59 (14.8)1.00 
 mild319 (76.9)341 (85.3)0.58 (0.40, 0.84)0.004
Body shape (20s)    
 thin/moderate362 (84.2)361 (89.1)1.00 
 full/obese 68 (15.8) 44 (10.9)1.56 (1.02, 2.37)0.038
Body shape (30s)    
 thin/moderate286 (66.7)297 (73.3)1.00 
 full/obese143 (33.3)108 (26.7)1.43 (1.05, 1.95)0.025
Body shape (40s)    
 not obese379 (88.8)378 (93.3)1.00 
 obese 48 (11.2) 27 (6.7)1.88 (1.12, 3.15)0.017
Having sore/ulcer    
 no380 (91.1)388 (96.3)1.00 
 yes 37 (8.9) 15 (3.7)2.92 (1.52, 5.61)0.001
 no363 (86.2)368 (91.1)1.00 
 yes 58 (13.8) 36 (8.9)1.68 (1.05, 2.7)0.032
No. of women partners    
 ≤5257 (61.5)277 (69.2)1.00 
 ≥6161 (38.5)123 (30.8)1.4 (1.03, 1.91)0.029
 yes (vs no) 20 (4.9) 10 (2.5)2.24 (0.91, 5.55)0.081
Non-specific urethritis    
 yes (vs no) 27 (6.5) 14 (3.5)1.71 (0.86, 3.38)0.125
Table 2.  Correlations (calculated from the averages of the frequency ranges) for intercourse, masturbation and combined SA, in different age decades

Controls were only frequency-matched to the cases (within 2 years of case diagnosis), therefore unconditional logistic regression analysis was used [8]. Our predefined analyses focused on intercourse activity, masturbation activity and their mean overall SA in each age decade and in the lifetime. The control distribution was used to determine percentiles, to group continuous variables. The lowest category for each age range was used as the reference group. P values are two-tailed and the 0.05 significance level was used for all analyses.

We used standard analysis approaches, i.e. starting with age- and ethnicity-adjusted analysis to identify potential confounders and then, adjusting for significant variables from this analysis in a multivariate logistic regression (MLR).


Initially, the analyses were adjusted for age and ethnicity (Table 3). Uncontrolled confounding might cause bias towards sicker or healthier men in one group or another, so several variables were then screened to detect associations with both prostate cancer and SA, i.e. to identify potential confounders. For this, we analysed each variable separately, adjusting for age/ethnicity (Table 1). The significance level used was 0.05. Variables having a confounding effect are those showing significant associations with both the disease and the SA variables of interest. Variables that were not disease-associated (age at first intercourse) or not associated with any of the SA variables (moderate exercise in the 40s) are not confounders, and were therefore not included in subsequent analyses (Table 1).

Table 3.  Associations between intercourse, masturbation activity, overall SA and lifetime SA, and prostate cancer risk
VariableCases, n (%)Controls, n (%)Adjusted* OR (95% CI)P
  1. *adjusted for age at diagnosis and ethnicity. Significance level P < 0.05 ; †calculated as the mean of overall SA across all age bands.

Frequency of intercourse    
20s   0.982
 ≤1 tpw126 (30.1)120 (30.5)1.00 
 2–3 tpw197 (47.1)187 (47.5)1.03 (0.74, 1.44)0.852
 4–7 tpw 95 (22.7) 87 (22.1)1.03 (0.69, 1.54)0.900
30s   0.697
 ≤1 tpw118 (28.6)124 (31.6)1.00 
 2–3 tpw221 (53.6)204 (51.9)1.15 (0.83, 1.60)0.410
 4–7 tpw 73 (17.7) 65 (16.5)1.05 (0.68, 1.63)0.831
40s   0.918
 ≤3 tpm 91 (22.2) 91 (23.3)1.00 
 1 tpw120 (29.3)124 (31.7)0.95 (0.63, 1.41)0.787
 2–7 tpw199 (48.5)176 (45.0)1.02 (0.70, 1.47)0.939
50s   0.487
 ≤3 tpm203 (51.5)190 (50.0)1.00 
 1 tpw106 (26.9)106 (27.9)0.86 (0.61, 1.22)0.406
 2–7 tpw 85 (21.6) 84 (22.1)0.81 (0.55, 1.18)0.275
Frequency of masturbation    
20s   0.002
 <1 tpm108 (27.2)124 (32.3)1.00 
 1–4 tpm153 (38.5)167 (43.5)1.05 (0.74, 1.50)0.773
 2–7 tpw136 (34.3) 93 (24.2)1.88 (1.27, 2.77)0.002
30s   0.004
 <1 tpm153 (40.8)178 (47.8)1.00 
 1–3 tpm 69 (18.4) 80 (21.5)1.04 (0.69, 1.57)0.865
 ≥1tpw153 (40.8)114 (30.6)1.72 (1.23, 2.41)0.002
40s   0.065
 never 81 (22.0)102 (27.6)1.00 
 1–3 tpm164 (44.4)164 (44.3)1.26 (0.87, 1.85)0.227
 1–7 tpw124 (33.6)104 (28.1)1.63 (1.08, 2.45)0.020
50s   0.831
 never108 (30.8)116 (32.2)1.00 
 ≤3 tpm156 (44.4)149 (41.4)1.09 (0.76, 1.56)0.641
 ≥1 tpw 87 (24.8) 95 (26.4)0.98 (0.65, 1.47)0.917
Overall SA, tpm    
20s   0.031
 0–10128 (30.3)125 (31.3)1.00 
 10.5–14125 (29.6)148 (37.1)0.82 (0.58, 1.18)0.284
 20–56169 (40.0)126 (31.6)1.31 (0.92, 1.86)0.137
30s   0.380
 0–8 95 (22.7)105 (26.4)1.00 
 10–14202 (48.2)197 (49.6)1.15 (0.81, 1.64)0.440
 20–48122 (29.1) 95 (23.9)1.33 (0.89, 1.98)0.164
40s   0.121
 0–4.5131 (31.4)136 (34.3)1.00 
 6–10112 (26.9)125 (31.6)0.86 (0.60, 1.24)0.428
 10.5–40174 (41.7)135 (34.1)1.24 (0.88, 1.75)0.212
50s   0.076
 0–2.5152 (37.9)125 (32.3)1.00 
 4–6125 (31.2)143 (37.0)0.67 (0.47, 0.95)0.026
 8–38124 (30.9)119 (30.7)0.77 (0.54, 1.10)0.152
Lifetime SA, tpm   0.486
 0–7.8131 (31.0)135 (33.8)1.00 
 8–11.6131 (31.0)132 (33.1)1.02 (0.72, 1.46)0.905
 12–39160 (37.9)132 (33.1)1.21 (0.86, 1.72)0.276

After screening each variable separately, factors found to have a confounding effect were body shape in the 20s, 30s and 40s, baldness in the 40s, number of women partners in the lifetime, presence of STD/VD and having a genital sore/ulcer (Table 1). Therefore, these were controlled for in a MLR analysis to examine the association of SA variables (intercourse, masturbation, overall SA) with disease. Other variables screened were vasectomy and condom use; these had no confounding effect (results not shown).

Multivariate analyses

Factors remaining significant in the MLR were retained in the model. Non-significant terms were excluded from the model, provided that excluding them did not inflate the se of the examined variables considerably, while changing very little the corresponding effect size. Results from the MLR are shown in Table 4. A detailed explanation of the impact of the confounders adjusted for in each analysis is given in Table 4. In most analyses, confounders remaining significant were age at diagnosis, ethnicity, ulcer/sore, body shape in the 40s (for the 40s and 50s analyses). However, inherently, different factors are included in different analyses, e.g. baldness in the 40s can only be included in the 40s and 50s analyses, but not in the 30s and 20s analyses. In all multivariate analyses, body shape in the 20s and 30s, total number of women partners in the lifetime, and presence of STD/VD were not significant and removed, taking account of the effect and se of the examined variable/s.

Table 4.  MLR analysis of sexual determinants of prostate cancer in study participants
Variable*Adjusted OR (95% CI)P
  • *

    Significance level P < 0.05; non-significant terms were removed from the model. All non-significant terms removed had no considerable influence on the effect and se of the variable/s under examination. Body shape in the 20s, total number of women partners in the lifetime, and presence of STD/VD were non-significant in all SA analyses described, and were therefore always removed from the regression models. Similarly for body shape in the 30s in all analyses for the 30s, 40s and 50s.

  • 1

    For intercourse and masturbation SA in the 20s: terms contributing to the model are age at diagnosis, ethnicity, ulcer/sore. The three terms described above were removed from the model. Masturbation (20s) was also adjusted for intercourse SA in the 20s.

  • 2

    For intercourse and masturbation SA in the 30s: factors remaining significant in the model are age at diagnosis, ethnicity, ulcer/sore. Non-significant variables intercourse and masturbation in the 20s were removed from the model. Masturbation SA (30s) was also adjusted for intercourse SA in the 30s.

  • 3

    For intercourse and masturbation SA in the 40 s: terms remaining significant are age at diagnosis, ethnicity, ulcer/sore, body shape in the 40s, masturbation in the 20s. Baldness in the 40s, intercourse in the 20s and 30s, masturbation in the 30s were non-significant and removed. Intercourse (40s) was adjusted for masturbation SA in the 40s (and vice versa).

  • 4

    For intercourse and masturbation SA in the 50s: factors significant in the model are age at diagnosis, ethnicity, ulcer/sore, body shape in the 40s, masturbation in the 20s and 40s. Intercourse in the 20s, 30s and 40s, masturbation in the 30s, baldness in the 40s were non-significant, and were removed from the model, in addition to the three terms described above. Intercourse (50s) was adjusted for masturbation SA in the 50s (and vice versa).

  • 5

    For overall SA in the 20s: factors contributing to the model are age at diagnosis, ethnicity, ulcer/sore. Non-significant terms were named above.

  • 6

    For overall SA in the 30s: significant terms in the regression model are age at diagnosis, ethnicity, ulcer/sore. Overall SA in the 20s was non-significant and was removed from the model.

  • 7

    For overall SA in the 40s: age at diagnosis, ethnicity, having ulcer/sore, body shape in the 40s, baldness in the 40s remained significant in the model. Overall SA in the 20s and in the 30s were non-significant and removed.

  • 8

    For overall SA in the 50s: factors remaining in the model as significant are age at diagnosis, ethnicity, ulcer/sore, baldness in the 40s and overall SA in the 40s. Non-significant terms are body shape in the 40s, and overall SA in the 30s and 20s.

  • 9

    For lifetime SA: significant terms are age at diagnosis, ethnicity, baldness and body shape in the 40s, having ulcer/sore.

Frequency of intercourse  
20s1 0.952
 ≤1 tpw1.00 
 2–3 tpw1.05 (0.74, 1.49)0.803
 4–7 tpw0.99 (0.65, 1.51)0.972
30s2 0.403
 ≤1 tpw1.00 
 2–3 tpw1.26 (0.88, 1.79)0.203
 4–7 tpw1.05 (0.66, 1.68)0.823
40s3 0.611
 ≤1–3 tpm1.00 
 1 tpw1.08 (0.70, 1.68)0.718
 ≥2–3 tpw1.22 (0.81, 1.83)0.343
50s4 0.912
 ≤3 tpm1.00 
 1 tpw1.08 (0.73, 1.58)0.706
 2–7 tpw0.99 (0.65, 1.49)0.948
Frequency of masturbation  
20s1 0.005
 <1 tpm1.00 
 1–4 tpm1.04 (0.73, 1.50)0.818
 2–7 tpw1.79 (1.21, 2.67)0.004
30s2 0.007
 <1 tpm1.00 
 1–3 tpm1.04 (0.68, 1.58)0.861
 ≥1 tpw1.69 (1.20, 2.40)0.003
40s3 0.571
 1–3 tpm1.26 (0.82, 1.91)0.290
 1–7 tpw1.19 (0.71, 1.99)0.512
50s4 0.008
 never active1.00 
 ≤3 tpm0.64 (0.34, 1.20)0.162
 ≥1 tpw0.30 (0.14, 0.66)0.003
Overall SA, tpm  
20s5 0.040
 10.5–140.82 (0.57, 1.18)0.291
 20–561.29 (0.90, 1.85)0.160
30s6 0. 421
 10–141.20 (0.83, 1.71)0.334
 20–481.31 (0.87, 1.96)0.197
40s7 0.083
 6–100.81 (0.55, 1.18)0.272
 10.5–401.23 (0.86, 1.75)0.256
50s8 0.003
 4–60.50 (0.33, 0.77)0.001
 8–380.49 (0.30, 0.79)0.003
Lifetime SA, tpm9 0.735
 8–11.60.99 (0.69, 1.44)0.975
 12–391.13 (0.79, 1.62)0.509

Stepwise logistic regression was also used, but this technique might model ‘noise’ and cause over-fitting of the data, and thus is considered useful only in an exploratory fashion. Use of established knowledge and variable screening separately in model selection before performing MLR is preferred [9].

In both age-ethnicity adjusted analyses and multivariate analyses, data on intercourse and masturbation were analysed separately and then a combined analysis was conducted. Terms corresponding to activity in each decade were entered simultaneously in the regression model, as the disease-activity association in any specific age band might be confounded by activity in preceding decades (Table 2).

Numbers reported in the tables are crude/unadjusted. Results are those in adjusted analyses for age-ethnicity, and additional factors.

The SA results are presented first separated into intercourse and masturbation, and then as a combined analysis of overall activity.

Frequency of intercourse or masturbation

There was no significant association for frequency of intercourse in any age band (Table 3). The frequency of masturbation category of 2–7 tpw in the 20s (compared with < once per month) produced an OR of 1.88 (adjusted P = 0.002) with an overall regression P = 0.002 (Table 3). The respective result in the MLR was OR 1.79 (adjusted P = 0.004, overall P = 0.005). The frequency of masturbation of once a week or more in the 30s was associated with the risk of prostate cancer (OR 1.72, adjusted P = 0.002, reference category < once per month) with an overall regression P = 0.004. The respective result in the MLR was OR 1.69 (adjusted P=  0.003, overall P = 0.007). In the 40s the 1–7 tpw category produced an OR of 1.63 (adjusted P = 0.020) compared with no activity, but there was no significance overall and this was also the case for the MLR result. In the MLR for the 50s, the frequency of masturbation of once/week or more yielded an OR of 0.30 (adjusted P = 0.003, overall P= 0.008) compared with no activity (Tables 3,4).

Combined analysis

When intercourse and masturbation data were analysed together, the mean of minimum and maximum frequency pairs were obtained and a grouped variable was created to reflect the frequency of SA. The mean of the grouped variable reflects the mean frequency of overall SA in the different age groups. For activity during the 20s to 50s, the mean (sd), respectively, was 15.2 (9.7), 12.9 (7.8), 9.75 (7.0) and 6.0 (5.6). For the 20s, comparing the 20–56 tpm category with 0–10 tpm gave an OR of 1.31 (adjusted P = 0.031). For the 50s, when the 4–6 tpm category was compared with the lowest category (0–2.5 tpm) the OR was 0.67 (adjusted P = 0.026) and showed a protective effect of moderate SA against low SA (there was no significance overall; Table 3). The respective result of the MLR for the 20s was an OR of 1.29 (overall P = 0.040). For the 50s, the MLR results were significant (overall P = 0.003) with the 8–38 tpm category reaching an OR of 0.49 (adjusted P = 0.003) when compared with the 0–2.5 tpm reference category (Table 4).

Lifetime analysis

We also examined overall SA over the lifetime; the mean of the variables corresponding to SA in the four age bands reflects overall SA from 20 to 60 years old. We grouped the lifetime SA variable, based on the thresholds of the control distribution. We repeated the analyses described above, also including the family history of prostate cancer. However, the respective results are not presented, as the sample was reduced considerably and the risk estimates are unreliable with wide CIs.


The main effects on the risk of prostate cancer (that remain consistent after including potential confounders) are those for overall SA in the 20s, and masturbation in the 20s and 30s, while masturbation and overall SA in the 50s were significant. The overall pattern arising from these results is that frequent SA is positively associated with an increased risk of prostate cancer in younger men, but for men in their 50s there is evidence of an inverse association/protective effect of masturbation and overall SA. Possible reasons for this are presented below. Specifically for men in their 50s the protective effect was apparent even for low SA. Masturbation activity in the 20s and 30s was consistently associated with high risk, but for the 40s there was no significant association with any kind of SA. The protective effect found by Giles et al.[10] applied only to ejaculations experienced in early adult life. In the present results, a slightly protective effect in early adults applied only to moderate overall SA. When we assessed overall SA across the lifetime, there was no significant pattern (Tables 3 and 4).

The main features in the present study, by contrast with some other studies of SA, is that the present sample comprised younger men (aged ≤60 years) from the UK population, and comprehensive SA of participants was examined, including intercourse and masturbation. The mean frequencies of overall SA in the different age ranges are similar to those of the population used in the study of Leitzmann et al.[11]. Our study findings corroborate the hypothesis proposed by the studies of Leitzmann et al.[11] and Giles et al.[10].

Participants were predominantly white and the low percentage of other races is unlikely to affect the results. Prostate cancer is age-associated [1] and usually appears in older men. The present study population of younger men is therefore one where the genetic component of the disease is stronger, but these cases are less likely to be screen-detected, as screening targets older men.

In some studies examining circulating testosterone levels there have been positive associations with the risk of prostate cancer [12]. There is some evidence showing a link between testosterone levels and sexual function [13,14], supporting the theory that androgenic factors which associate with high SA also increase prostate cancer risk. However, similar to the present findings for men in their 50s, some studies have suggested that frequent SA reduces toxicity and boosts immune function in the prostate region [15,16], and thus maintains a healthy organ. Increased risk can be conferred because of the accumulation of carcinogenic secretions in the prostate and prostatic fluid [16,17]. Although an increased risk of prostate cancer has been attributed to frequent general SA in some studies [3], and this would implicate hormone-related factors, there are several studies suggesting negative associations with risk, an example being that for increasing age at first intercourse/masturbation [18,19]. Our findings support the association of decreased risk with increasing age at first intercourse (Table 1).

Previous results from studies of reported ejaculation frequencies or sexual intercourse and prostate cancer are mixed, and include studies reporting inverse [10], positive [20] or no associations [19]. Where the present study differs from many others of the same type is in the early-onset type of prostate cancer, in the examination of frequency of SA rather than the frequency of intercourse only, and in the assessment of associations in different decades. Some studies [10,11] have suggested that a high frequency of ejaculation is protective against prostate cancer. A general view that seems to emerge from the present results for the 50s group corroborates the theory of Isaacs [17], supported by the studies of Leitzmann et al.[11] and Giles et al.[10]. It might be that toxin accumulation, e.g. through restraining SA, increases the probability of disease [21,22]. One of the means of eliminating chemical carcinogens and toxins from the prostate area or altering the composition of prostatic fluid is through SA [23], and this might explain the protective effect seen in the 50s. Mishina et al.[24] and Hayes et al.[18] examined the frequency of sexual intercourse and found positive associations with disease for younger men, but inverse associations for those aged >60 years; this is in agreement with the present results. However, there is a possibility that disease is affecting SA and/or its reporting by participants, thus the protective effect might be due to reverse causation or bias.

As in previous studies, early age at first intercourse [18,19], many sexual partners and STDs [18,20] were shown to increase prostate cancer risk. One hypothesis is that certain sexual habits might expose a man to infection, leading to the development of prostate cancer [20], but this theory has not been substantiated in other studies [25], and in the present study we found no significant associations with intercourse activity either. Infection through sexually transmitted agents does not appear to be a causal factor here; it might have a small role in increasing the risk of disease, or it might be interacting with a causal factor.

As the present data were collected retrospectively, the possibility cannot be totally excluded that differential measurement error of reported SA between cases and controls might play a role. We tried to minimize bias due to any potential uncontrolled confounding by identifying variables indicating differences between cases and controls, and including them simultaneously in a multivariate analysis (Table 4). Furthermore, the controls used here are similar to control groups in other studies and are therefore representative of the general population. Finally, the cases and controls were age-matched and selected from the same catchment area.

Like all questionnaire data collected in case-control studies, there is some potential for differential recall of information between cases and controls. Most participants were diagnosed before returning the questionnaire, and others had a diagnosis afterwards or around the same time as the questionnaire return, introducing a small potential for recall bias. However, we think that this is unlikely to be significant, as the questionnaire was carefully worded to ask for unbiased responses, and was relatively long, so differential recall between any one type of exposure is less likely. Furthermore, the study by Fransson [26] indicated that patients tended to remember their baseline sexual function as having been better than it had actually been. In the worst-case scenario, where data are returned after a diagnosis of prostate cancer, Fransson’s study means that worse rather than better sexual function is associated with disease. This is in line with our findings, with increased sexual function during the 50s being protective. Therefore, our results are unlikely to have been affected by recall bias. In addition, Karakiewicz et al.[27] showed that responses to erectile function domain questions appear to be most reliable. There might be some bias introduced from those responding after the diagnosis, so that the disease would have decreased their activity, although the overall protective effect is strong in both the overall SA and for masturbation. This shows that results from the rest of the participants, not introducing such bias, are in the same direction. Thus, the protective effect observed cannot be attributed entirely to bias. It is most likely that some of the effect detected is due to a true protective influence of SA in the 50s on the prostate, possibly explained by the reduced carcinogenicity/toxicity theory presented above.

A reduced SA in older men due to prostate problems/disease [28], i.e. reverse causation, could be an explanation for the inverse associations in men in the 50s group, and the contrast with the high activity-increased risk association seen in younger ages. We assessed the likelihood of reverse causation, i.e. disease affecting SA levels, by comparing the age at diagnosis against the age band at which the lowest level of SA was reported by each participant. Most men in the study (691) were in their 50s and their mean age at diagnosis was 54.7 years. There was little difference in the percentages of low reported SA between cases and controls in each decade. The numbers of participants with a diagnosis before/after the lowest SA period were calculated. There were 239 cases and 205 controls with the diagnosis preceding the period of lowest SA while there were 150 participants in each disease status group with their reported lowest SA being earlier than their diagnosis. The data show that there is reverse causation, but the difference measured between the above groups was not significant. There are also many instances in men, in their sixth decade mainly, where the age at diagnosis was within the decade and the lowest SA was also reported within the decade, but because a period rather than the exact age is available for reported lowest SA, we cannot be certain whether the diagnosis preceded or followed the lowest SA point. To allow for the possibility of reverse causation, we presumed that the diagnosis preceded the lowest SA point in the above situation.

However, the data also show that controls reported similar patterns for frequency of SA to those of cases. If the disease was responsible for low SA reported from cases, reporting from the healthy population would be expected to be in the opposite direction. Thus it cannot be concluded with certainty that it is prostate cancer that interferes or causes modification in SA in men in their 50s. Finally, participants identified as being treated for prostate cancer or different types of sexual dysfunction (through androgen/oestrogen drugs) were not included in the analyses, to reduce bias due to treatment effects. We cannot conclusively provide a reason for the inverse association; SA might indeed be protective in men in their 50s. The possibility of reverse causation is less likely, but cannot be entirely excluded. As with all case-control studies, there remain some concerns as to the interpretation of causal relationships. As such, further studies to assess the relationship of SA in younger men with prostate cancer are merited, but our data suggest that there is a relationship.

This is the first study to focus on the role of total SA in prostate cancer risk with a young age at onset (≤60 years). Our findings suggest that the risk of prostate cancer diagnosed at ≤60 years is associated with high SA in earlier life (20s) and low activity in later life (50s), in agreement with the studies of Leitzmann et al.[11] and Giles et al.[10]. Intercourse was not associated with prostate cancer in the present study.


The authors thank Professor Tim Key for his helpful comments.


None declared. Source of funding: the study was funded by the Prostate Cancer Research Foundation and Cancer Research UK grant C5047/A3354.