Intended for healthcare professionals


Aetiology of testicular cancer: association with congenital abnormalities, age at puberty, infertility, and exercise

BMJ 1994; 308 doi: (Published 28 May 1994) Cite this as: BMJ 1994;308:1393
  1. D Forman,
  2. M C Pike,
  3. G Davey,
  4. S Dawson,
  5. K Baker,
  6. C E D Chilvers,
  7. R T D Oliver,
  8. C A C Coupland
  1. Imperial Cancer Research Fund Cancer Epidemiology Unit, Gibson Building, Radcliffe Infirmary, Oxford OX2 6HE
  2. Section of Epidemiology, Institute of Cancer Research, Sutton, Surrey.
  3. Department of Medical Oncology, Royal London Hospital, London.
  4. Department of Public Health Medicine and Epidemiology, University of Nottingham Medical School, Nottingham.
  1. Correcpondence to: Dr Forman.
  • Accepted 10 March 1994


Objective : To determine the risk of testicular cancer associated with undescended testis, inguinal hernia, age at puberty, marital status, infertility, vasectomy, and amount of exercise.

Design : A population based case-control study with a questionnaire administered by an interviewer and with relevant supplementary data extracted from general practioners' notes.

Setting : Nine health regions within England and Wales.

Subjects : 794 men, aged 15-49 years, with a testicular germ cell tumour diagnosed between 1 January 1984 and 1 January 1987; each had an age matched (within one year) control selected from the list of their general practioner.

Results : There was a significant association of testicular cancer with undescended testis (odds ratio 3.82; 95% confidence interval 2.24 to 6.52) and inguinal hernia (1.91; 1.12 to 3.23). The excess risk associated with undescended testis was eliminated in men who had had an orchidopexy before the age of 10 years. There were positive associations with earl age at voice breaking, early age at starting to shave, and infertility. There was a significant association with a sedentary lifestyle and a moderate protective effect of exercise. There was no association with vasectomy.

Conclusion : This study confirms previous reports that developmental urogenital abnormalities result in an increased risk of testicular cancer. The trend to perform orchidopexy at younger ages may reduce the risk associated with undescended testis. The increased risks associated with early age at puberty and low amounts of exercise may be related to effects of exposure to endogenous hormones. Changes in both of these factors may partly contribute to the increasing rates of testicular cancer observed in the past few decades.

Clinical implications

  • Clinical implications

  • Testicular cancer is increasing in incidence and is the most common form of cancer in men aged 15 to 44

  • This study confirms previously established risks associated with undescended testis and infantile inguinal hernia, which in this population were fourfold and twofold, respectively

  • The excess risk associated with undescended testis was eliminated in men who had had an orchidopexy before the age of 10 years

  • An early age at puberty, as assessed by age at starting to shave, age at voice breaking, and age at first nocturnal emissions, was a risk factor for testicular cancer

  • There were associations between testicular cancer and a sedentary lifestyle and lack of exercise


In the past few decades there has been a continuous rise in the incidence of testicular cancer in the United Kingdom,1,2 United States,3 and in most other white populations. 4,5 In England and Wales testicular cancer is now the most common form of cancer in men aged 15 to 44 years, 6 and apart from a history of undescended testis and infantile inguinal hernia no risk factors have been identified with certainty.7

There have been 11 published case-control studies of testicular cancer based on interview,*RF 8-21* the largest of which included 333 cases and 729 controls.19,20 All but one of these studies16 17 were conducted in North America. There are also six record based case-control studies.*RF 22-27 In general, positive results from these studies have not been found consistently, and the relative risks have been small.

In an attempt to increase understanding of the aetiology of testicular cancer we carried out a large population based case-control study within England and Wales. We report the first detailed analysis of our results and examine the associations of risk with urogenital abnormalities, age at puberty, marital status, infertility, land amount of exercise.

Subjects and methods

The study was carried out in nine health regions: South West, Wessex, South East Thames, South West Thames, North Thames, Oxford, South Wales, North West, and Yorkshire, Within each region a geographical area was defined and cases and controls resident only in this study area, covering a total population of 19.5 million, were eligible for inclusion. The study was approved by the central ethics committee of the BMA and by local ethics committees in each region.

Selection of cases and controls

All men aged 15 to 49 years diagnosed as having a testicular germ cell tumour between January 1984 and January 1987 (dates subject to some regional variation) and who were resident in one of the defined study areas were included. The date of diagnosis was taken to be the date of the first biopsy specimen that yielded positive results. The main sources of cases were major treatment centres for oncology or radiotherapy, or both, and regional cancer registries. For every case two controls were chosen from the list of the general practitioner with whom the case was registered. The controls' dates of birth were matched to within one year, and controls had to have been registered with the general practitioner at the time of diagnosis of the case. Only one control was interviewed, the second name being kept as a reserve if the first could not be interviewed. In some instances it was necessary to choose further controls.

The study was restricted to white men with no previous malignancy. Men with a severe mental handicap or psychiatric condition (as determined by a consultant or general practitioner) which would make an interview impossible were also excluded, as were cases diagnosed abroad and controls whose names were still on the general practitioner's list but who had moved out of the study area before control selection. Cases and controls were first contacted by letter from the consultant or general practitioner, respectively, and informed of the purpose of the study. The letter was followed by a telephone call or a visit from interviewing staff within three weeks if no reply had been received. Strenuous efforts were made to maximise response rates. Failure was accepted only if a man was untraceable, had moved outside the study area, or returned the reply form stating that he did not wish to take part.


Interviewing took place between June 1984 and April 1988. The median time between case diagnosis and interview was 10 months (range 0-50 months). Each case-control pair was seen by the same female interviewer, and interviews took place at general practitioners' surgeries, at the man's home, at his place of work, or in a hospital


The interview included questions on personal history, sexual development and behaviour, medical history, marital history and children, lifestyle including sport and exercise, occupational history and exposures, and family history. Every control was given a pseudodiagnosis date, the date on which he was exactly the same age as his matched case was at diagnosis. A reference age was defined as the age of the case and control one year before the diagnosis or pseudodiagnosis date. Most data collected referred to events happening before the reference age. As controls were always interviewed after their matched case they had a slightly longer time interval between their reference and interview dates (median three months longer).

After interview, and with the interviewees' consent, data on medical history, particularly with respect to undescended testis, inguinal hernia, and testicular trauma, were abstracted from general practitioners' notes by the interviewers with a structured form. For cases details of their testicular tumours were abstracted from their hospital notes and a copy of their pathology reports obtained. Hospital notes were used only to confirm the diagnosis. Abstracts from general practitioners' notes and mother's questionnaires were used to confirm the history of undescended testis, hernia, and trauma.

A diagnosis of undescended testis was regarded as definite only if there was evidence of successful orchidopexy or of surgical investigation at which a testis was found to be missing or impossible to correct. Otherwise, the testis had to be still undescended at reference age. Testes that were retractile in either childhood or adulthood or that descended spontaneously were not regarded as undescended. There was also a third category of “possible undescended testis” in which there seemed to be a genuine undescended testis but there was insufficient evidence for a definite diagnosis. Included in this category were any apparent undescended testes which were successfully treated with hormonal treatment. All the possible undescended testes were included in the category of no undescended testis for the analyses.

Statistical methods

The EGRET package was used for statistical analysis using multivariate conditional logistic regression methods for individually matched case-control studies. 28 Relative risks were estimated by odds ratios with 95% confidence intervals. A case-control pair was excluded from the odds ratio estimation if the information for either the case or the control was not known for the variable in question.The numbers of pairs on which each analysis was based are given in the tables.

Significance tests (P values) quoted are two sided. Tests for trend were calculated either across categories or by using recorded levels for continuous variables as appropriate (see footnotes on tables). Odds ratios are presented both unadjusted and adjusted for a history of undescended testis and inguinal hernia (the latter diagnosed at ages less than 15 years).


Within the predefined geographical areas of the study 863 eligible cases were identified. Interviews were completed with 794 (92.0%) of these men. Fourteen of the 863 eligible patients (1.6%) refused to participate, 12 (1.4%) were too ill to participate, 27 (3.1%) died before an interview could be arranged, and 16 (1.9%) had moved outside the study area. Of the 794 first selected individually matched controls, 609 (76.7%) were interviewed. Eighty four (10.6%) refused to participate, 26 (3.3%) could not be contacted, and 14 (1.8%) were not approached on the advice of the general practitioner. Sixty one controls (7.7%) had moved from their registered addresses and should not response rate was therefore 83.1% (609/(794-61)).To replace the 185 controls and 43 third or subsequent controls were interviewed.

Table I shows the distribution of cases by age and by region of residence at diagnosis. Most diagnoses occurred between the ages of 20 and 39 years. There were 446 right sided and 345 left sided cancers. Two men had bilateral disease at diagnosis and for one man the side was unknown. There were 400 tumours described as pure seminoma and 394 tumours of other (including combined) germ cell histological types. Differences in risk by histological type will be presented elsewhere.


Numbers of cases of testicular cancer by age group and by region of residence at diagnosis

View this table:

Table II shows the risk of testicular cancer associated with undescended testis and inguinal hernia. Both were significant risk factors for testicular cancer. The overall odds ratio for a history of undescended testis was 3.82 (95% confidence interval 2.24 to 6.52). Nineteen cases and no control men had bilateral undescended testis (odds ratio (infinity) (5.86 to (infinity))), whereas 46 cases and 17 controls had unilateral undescended testis (2.71 (1.55 to 4.72)). Twelve of the 46 cases with unilateral undescended testis had cancer in the contralateral testis to that which was maldescended (odds ratio 1.42 compared with 4.02 for the risk to the ipsilateral testis). Confirmation of undescended testis from the general practitioners' notes was present for 57 of the 65 case (87.7%) and 16 of the 17 controls (94.1%). The remaining eight cases and one control provided sufficient information themselves to allow a classification of the undescended testis as definite. Four cases and five controls had a possible undescended testis, of whom two cases and one control had hormonal treatment.


Numbers (percentages)* of cases and controls and odds ratios (95% confidence intervals) for diagnosis of undescended testis and inguinal hernia

View this table:

Age at correction in men with unilateral undescended testis had a strong effect on the risk of cancer. Men who were successfully operated on before the age of 10 years did not have an increased risk (0.60 (0.22 to 1.65)) whereas those corrected at or after the age of 10 years of who had an uncorrected testis had a significantly increased risk (7.67 (2.30 to 25.53) for correction at 10-14 years; (infinity) (1.22 to (infinity)) for correction at ages older han 14 years, and 3.00 (0.97 to 9.30) if uncorrected). The odds ratio for men who had an undescended testis corrected at or after the age of 10 was significantly increased in comparison with men who had an undescended testis corrected before 10 years (6.75 (1.55 to 29.48)). Among men who had a successful correction the trend of association between age at orchidopexy (as a continuous variable) and risk of cancer (X2=5.53;P=0.019) was significant. Apart from one case with unilateral undescended testis who had a correction at the age of 3 years no orchidopexies were carried out below the age of 5 years. Fifteen of the 19 cases with bilateral undescended testes had successful corrections on both sides, four before the age of 10 years, 10 at 10-14 years, and one over 14 years. Because there were no controls with bilateral undescended testes the odds ratios for these categories could not be calculated

There were 16 cases (12 unilateral, four bilateral) and four controls (all unilateral ) with uncorrected undescended testis. We confirmed that of these, nine cases and one control had never had an orchidopexy whereas seven cases and three controls had had an operation but the testes either could not be found or could not be put into the scrotum.

Although a history of inguinal hernia in the absence of undescended testis was significantly associated with testicular cancer (1.91 (1.12 to 3.23)), this risk was confined to men who had a hernia diagnosed before the age of 15 years (2.64 (1.32 to 5.28)).

Table III shows results from questions relating to age at onset of puberty. There was significant trends of decreased risk of cancer with increasing age at voice breaking (P=0.010), increasing age at which shaving started (P=0.007), and age at first recalled nocturnal emissions (P=0.031). There was no significant trend with age at first masturbation (P=0.099).


Numbers (percentages)* of cases and controls and odds ratios (95% confidence intervals) by variables relating to age at puberty

View this table:

Table IV shows that there was no association between risk of cancer and marital status (0.98 (0.74 to 1.30) for ever versus never married), although there was a significant relation among men who had been married of decreased risk with later age at first reported having experienced homosexual intercourse (1.35 (0.56 to 3.28)); 48 cases and 42 controls refused to answer this question.


Numbers (percentages)* of cases and controls and odds ratios(95% confidence intervals) by reported marital status, sexual preference, low fertility, and vasectomy

View this table:

Table IV also shows the relation between infertility and risk for testicular cancer. After all cases with bilateral undescended testes were excluded 13 cases and five controls reported that they had been diagnosed with either low fertility or sterility (2.66 (0.94 to 7.54)). Confirmation of this diagnosis was available from the general practitioners' notes for nine cases and three controls (2.95 (0.79 to 10.99)).

There was no overall association between testicular cancer and having had a vasectomy (1.09 (0.77 to 1.52))

The risk of testicular cancer decreased with increased amount of exercise and increased with increased sedentary time (table V). The trends at age 20 years and at reference age were significant for exercise (P=0.010 at 20 years; P=0.018 at reference age) and for time spent seated (P=0.024 at 20 years; P=0.006 at reference age). Those who engaged in 15 or more hours a week of exercise had a substantially reduced risk (0.62 (0.42 to 0.91) at 20 years; 0.54 (0.32 to 0.90) at reference age). At reference age there was a 71% increase in risk in men who spent 10 or more hours


Numbers (percentage) of cases and controls and odds ratios (95% confidence intervals) by hours of exercise a week and hours spent seated a day at age 20 years and at reference age.

View this table:


This is the largest interview based case-control study of testicular cancer carried out to date. Other strengths of the study derive from our high response rate, use of population based controls, and supplementation of information obtained at inteview with that from general practitioners' notes and from the mothers of cases and controls. We believe therefore that our results are particularly reliable, although they will not be free of the reporting bias between cases and controls which may be present in retrospective studies. Some of the questions relating to sexual behaviour might be particularly sensitive to this bias. There is also a potential problem of control selection bias for two reasons. firstly, although a high proportion of the population is registered with a general practitioner, men aged 15-49 are particularly likely to be unregistered, especially if single, healthy, and mobile, and would not be included in our sampling frame. Secondly, even though the response rate of controls of 83.1% can be regarded as good for a study of young and middle aged men, the non-responders may be systematically different in terms of their history and behaviour, It will therefore be important to consider whether out results may be due to reporting or selection biases. Because cases and controls were selected from the same general practitioners' lists, there is also a possibility of overmatching - that is, both groups having similar geographical and socio-economic backgrounds. We do not believe that this effect is likely to be strong and, if present, would bias any risks towards unity.

Congenital abnormalities

Risks associated with undescended testis and inguinal hernia were as expected from previous studies29,30 - that is, bilateral undescended testes carry a substantial risk of cancer whereas unilateral undescended testis or an inguinal hernia early in life carry more moderate risks (both about threefold in this study). Our overall risk estimate of 3.82 associated with undescended testis is somewhat lower than that reported previously. Chilvers and Pike reported a summary estimate of 5.8 from an overview of nine studies published since 1979.30 There are two explanations for this reduction. Previous estimates have been based largely on recall of the diagnosis by cases and controls. Men may recall retractile or late descending testes as undescended testis and cases, who may know of the association between undescended testis and cancer, may be more likely to misclassify their condition in this way. This could result in an over-estimate of the relative risk. Misclassification of undescended testis has been reduced in our study by seeking verification of the diagnosis in the general practitioners' notes. We report that 8.2% of cases and 2.1% of controls had a definite undescended testis; if we had relied on the responses to the questionnaire alone, the proportions would have been 11.6% and 3.0%, respectively. There was therefore a tendency for cases to overreport a history of undescended testis relative to controls.

A second reason for a reduced risk estimate may arise from the increasing incidence of undescended testis over time.31,32 If the population prevalence of undescended testis has increased since earlier studies took place while the proportion of cases with undescended testis has remained constant, this will produce a reduced relative risk*RF 32 as we have shown by splitting our study group at the median age.

Our results suggest that men who had a unilateral undescended testis which was successfully corrected before the age of 10 years were no longer at increased risk of developing a cancer. This is consistent with the results of some 13,19 but not all33,34 studies which have addressed this issue, although there are few data on men with an age at correction below 10 years. If our findings are confirmed the recent trend to reduce the age at which correction is carried out should be encouraged and the effect on rates of testicular cancer closely monitored. As all but one of the corrections under 10 years were carried out above the age of 5, it would seem from these data that orchidopexy at particularly young ages (less than 5 years) is not necessary to remove the increased risk.

Inguinal hernia is often associated with undescended testis, and to examine the risk associated with hernia alone we carried out the analysis after excluding cases and controls with a definite undescended testis. In keeping with the results of most other studies29 we found a small increased risk associated with having had a hernia, and, consistent with results of several11,16,23 but not all13 studies, the risk increased for hernias diagnosed in childhood. Comparison between studies is complicated by the fact that not all analyses clearly distinguish between undescended testis and hernia and because there are misclassification problems for both conditions when self reporting is relied on.

Age at puberty

Our results give some support to the hypothesis that early age at puberty may be a risk factor for testicular cancer. Three variables associated with age at puberty - age at starting shaving, age at voice breaking, and age at first having nocturnal emissions - were all related to risk of cancer with significant trends suggesting that the later in life these events occurred the lower was the risk of cancer. The trend in adjusted odds ratios for reported age at first masturbation was in the same direction (table III), although the trend test was not significant. None of these trends were materially altered by excluding the “not yet” or “never” responders from the upper category. It is therefore plausible to speculate that some aspect of endocrine function may be associated with risk and that boys undergoing earlier onset of puberty, having a higher net exposure to postpubertal hormones, will be at increased risk.

Moss et al previously reported a twofold increased risk in men who underwent puberty (as assessed by reported age at appearance of public hair) at ages less than 14 years,15 whereas Depue et al reported no significant effect on risk of two of the variables used here, age at voice breaking and age at starting shaving.12 In a study in the United Kingdom Swerdlow et al asked directly about age at puberty and whether it was earlier, the same as, or later than that of classmates and found no effect.17 He also reported, as here, no association with frequency of shaving or age at first sexual intercourse (our data not shown). Assessment of age at puberty by recall is particularly difficult in men. There is, however, no reason for there to be systematic differences between cases and controls in recall, and the consequence of inaccuracy would be to reduce the observed magnitude of any real effect. This may partly account for the discrepancies between the studies. It should also be noted that the average age at puberty in boys is declining,35 an effect consistent with the increase in incidence of cancer. There was no association of risk of cancer with sexual preference or with marital status (table IV). There have been several studies showing an effect of marital status, although not always in the same direction25,26 36,37 but other recent case-control studies agree with our finding of no overall effect.14,17 Among men who have been married there was a significant trend of decreased risk with later age at first marriage.


When considering infertility it is necessary to exclude subjects with bilateral undescended testes. all of whom are likely to be subfertile or infertile.38 Men with infertility are known to have an increased prevalence of testicular carcinoma in situ,39,40 and it has been suggested that cancer and subfertility are both associated with a common underlying defect in spermatogenesis.41 if this is correct some association between infertility and increased risk may be anticipated. Although we observed such an association, of borderline significance, the small proportion of cases with reported low fertility (1.7%) suggests that this condition is not a major determinant of risk. Clinical data indicate, however, that a reduced sperm count is a common finding in men with testicular cancer. 42 This might be because tumour development directly leads to impaired spermatogenesis. Alternatively, impaired fertility might precede and be associated with testicular cancer but rarely by brought to clinical attention and thus would not be reported in retrospective studies such as this. One other case-control study reported risks of 6.0 and 10.0 associated with low sperm count and reported infertility, respectively,21 whereas two other studies have failed to find and association with fertility. 14 17 Two studies found no association with paternity,*RF 12,43 but this is a poor surrogate for fertility; of the 13 cases and five controls with confirmed low fertility in this study, six cases and four controls had fathered children.

Sharpe and Skakkebaek have proposed that a common factor, exposure to environmental oestrogens in utero, may be responsible for both a reported decreased sperm count and increased risk of testicular cancer in developed countries.*RF 41 Tarone et al have reported an excess of both fertility problems and testicular cancer in American servicemen who were in Vietnam, and they suggest that this might be due to exposure to toxic pesticides or defoliants.44

The idea that vasectomy should be an important risk factor for testicular cancer has always been unlikely given that this cancer often occurs before the age at which men usually undergo the operation. Nevertheless, there have been concerns about such an association 45,46 and our results add to several other negative reports on this issue.47


This is the first report that a lack of exercise and sedentary lifestyle may increase the risk of testicular cancer, although several studies have found higher rates in men with professional, non-manual occupations48,49,50 Our observed effects were not large, with no more than a doubling of risk comparing the groups at the highest and lowest extremes for the relevant variables (table V). For active exercise there was no major effect except a significant protection in those who reported very high amounts of exercise (15 or more hours each week). For hours spent seated each day there was a graded response with highly significant trends; these trends were virtually unchanged after adjustment for social class, indicating that this was not a social class effect. Clearly these results need to be confirmed in other studies and as yet should not be regarded as indicative of a causal relation. Changing patterns of manual work and exercise, however, together with the earlier onset of puberty could, in part, offer an explanation for the secular increase in incidence rates of testicular cancer seen in most Western countries.

General model

Our results are in accord with a model of development of testicular cancer based on increased exposure to pituitary gonadotrophin. These hormones may increase rates of germ cell mitosis and there by influence tumour development. When testes become atrophic, a state that occurs with both undescended testis and infertility, reduced synthesis of gonadal hormone may result in feedback inhibition of pituitary function and thus an excess production of gonadotrophins.51

Patients with Down's syndrome, which is associated with both testicular cancer52 and undescended testis,53 have increased concentrations of gonadotrophin54,55 whereas patients with Kallmann's syndrome, who have insufficient gonadotrophin secretion, never develop testicular cancer.56 Cumulative exposure of the testis to gonadotrophin will also be associated with age at puberty and may be associated with the amount of physical exercise.57 In women it is known that regular exercise can induce luteal phase dysfunction58 and, at extremes, may delay menarche and induce amenorrhoea.59 There is also evidence that circulating concentrations of one specific gonadotrophin, follicle stimulating hormone, are higher in men with stage I testicular cancer,42 especially in those who later develop second primary cancers.60

This model is clearly speculative and the relevant epidemiological observations in our study are for the most part weak and unconfirmed. The hypothesis does, however, suggest several important priorities for future research - notably, on the hormonal profiles of men with and without testicular cancer, the role of factors such as exercise which may influence such profiles, and the control that hormones exert over rates of mitosis of germ cells. Our data on the protective effect of early correction of unilateral undescended testis would also indicated that the deleterious effects of the position of an ectopic testis can be overcome. In contrast, the presence of bilateral undescended testes seems to have an irreversible harmful effect. This would suggest the benefit of examining and comparing hormonal concentrations in men with different types of undescended testis and with different ages of correction.

The regional collaborators in the study group were R A Cartwright (Leukaemia Research Fund Centre, University of Leeds); P C Elwood (MRC Epidemiology Unit, Cardiff); J Birch (Department of Epidemiology and Social Oncology, University of Manchester, Manchester); and C Tyrell (Plymouth Oncology unit, Freedom Fields Hospital, Plymouth). The interviewing staff were R Brett (Oxford); T Bush, V Isbell (Wessex;) A Cornwell, R Steer, S Thistlethwaite (Yorkshire); H Gellman (North West), J Hughes, M Llewellyn (Wales); A Ardern-Jones, A Allen, E Hilton, B Lloyd, S McVeigh, M Thorne, P Trowbridge (Thames); and S Reid (South West).

This study was supported by the Imperial Cancer Research Fund, the Cancer Research Campaign, and the Medical Research Council. We thank M Slattery (Wessex Regional Health Authority); R G Skeet (Thames Cancer Registry), F Landells, C Jenkinson (ICRF Genetic Epidemiology Laboratory, Leeds), and F Ledbetter (Welsh Office) for help in case finding; the family practitioner committees (now family health services authorities) that helped with control selection; all the consultants and general practitioners who allowed us to interview their patients; and, most importantly, all the patients and controls who willingly helped us with the study. We thank Drs T Key and H Moller for helpful comments. The manuscript was prepared by Dawn Pearson and Julia Bennett.