Risk of testicular cancer in cohort of boys with cryptorchidism

Introduction

Cryptorchidism is the most common congenital genitourinary abnormality in males and is important because of its associations with infertility and malignancy. Patients usually receive orchidopexy or hormonal treatment to bring the testis into the scrotum in the belief that this reduces these risks. As evidence has mounted that substantial risks of infertility and malignancy remain in boys who have had orchidopexy late in childhood,1 there has been a move towards early operation because adverse histological changes begin in the second year of life if the testis remains undescended.2 The effect of age of orchidopexy on the risk of malignancy remains unclear, however,3 4 5 6 7 as do the risks of malignancy in different types of cryptorchidism–for instance, unilateral and bilateral.

Most studies of risk of testicular cancer in relation to cryptorchidism have been of case-control design,3 5 6 8 with considerable potential for misclassification and bias6 in ascertaining maldescent. It is impossible to assess the effect of type of operation or of operative complications on risk in studies dependent on patients' or parents' recall. Cohort studies following up patients with undescended testis can overcome these problems, but those conducted have not been large and have not analysed risks in relation to type and number of operations and whether biopsy was performed.7 9 10 11

The Hospital for Sick Children, Great Ormond Street, London, has retained the case notes of its patients for the past 45 years. We followed up cancer incidence and mortality in patients who had orchidopexy at the hospital.

Subjects and methods

We identified all boys with a discharge diagnosis of cryptorchidism during 1951-64 by searching discharge lists. The case notes of each of these patients were examined, and patients were included in the study if they had had orchidopexy or hormonal treatment during the study period and did not have any major congenital malformations or any syndromes of which cryptorchidism formed a minor part. We included patients whose first orchidopexy was before 1951 but who had a subsequent orchidopexy during the study period; their follow up was counted only from the orchidopexy within the study period. For each subject we used a standardised abstraction schedule to obtain data from hospital notes on demographic details, type of maldescent and its treatment, other diseases, and biopsies and other investigations.

To obtain follow up information we sent identifying data for the cohort members to the NHS Central Register. The register includes records of virtually all British residents of England and Wales, and the register staff extracted information on mortality, cancer registrations since 1971, and emigrations. We also used the register to find, where possible, the patients' current general practitioners. The general practitioners were sent a questionnaire asking if the patient had had further operations since he was last seen at the Hospital for Sick Children and whether testicular cancer had occurred. We sent two repeat mailings if there was no response.

We used the EPICURE computer program12 to compare observed numbers of deaths and cancers in the cohort with expected numbers. These expected numbers were calculated by applying the person years at risk in the cohort according to age and calendar year to the corresponding mortality and cancer registration rates for males in England and Wales. Confidence intervals and significance tests for these relative risks were obtained by likelihood based inference. All P values cited are two sided.

To enable analysis of risk in relation to variables that can have different values for the two testes in an individual–for instance, the age at which the testis was brought into the scrotum or whether biopsy had been done on the testis–we conducted most analyses of risk of testicular cancer per testis rather than per person. We stratified these analyses by side of the testis (that is, we calculated observed and expected numbers separately for left and right testes) to prevent negative confounding by side: there is a tendency for testicular cancer at the study ages to be right sided13 14 and probably for cryptorchidism to be left sided,10 15 which would reduce the apparent risk of ipsilateral malignancy. Since national data on testicular cancer have not been collected by side we estimated the expected incidence rates for each side by multiplying the age specific incidence rates of testicular cancer in England and Wales by the age specific proportions of tumours that were left and right sided in data from the South Thames region, 1958-77.14

Follow up for death was to 30 September 1994 or to emigration if earlier. For cancer incidence, follow up was to 30 June 1990 or to death or emigration if earlier. National mortality data were not available for 1991 onwards and therefore 1990 data were used as the comparison for 1990-94. Similarly, cancer registration data for 1988-90 were not available and 1987 data were used as the comparison for these years.

Results

A total of 1182 boys without major malformations had treatment for cryptorchidism at the hospital during 1951-64. Twenty eight (2.4%) could not be traced at the NHS Central Register, and a further 30 (2.5%), although traced, had no record that they had ever been registered with a general practitioner. The remaining 1124 patients who were traced and had follow up information from the register formed the study cohort. We received questionnaire replies from general practitioners for 792 (70%) of these subjects.

During follow up to 30 September 1994, 28 of these subjects died, 62 emigrated, 14 entered the armed forces, and 21 were lost to follow up because the NHS register had inactivated their records when they left their general practitioner and did not reregister elsewhere (they are likely to have emigrated). Total follow up was for 39 055 person years. The relative risk of death overall was 0.6 (95% confidence interval 0.4 to 0.8), and of non-cancer death 0.6 (0.4 to 0.9). Seven deaths from cancer occurred: one each from oesophageal, pancreatic, and lung cancers, Hodgkin's disease, and myeloid leukaemia, and two from testicular cancer. The relative risk of death from cancer was 0.8 (0.3 to 1.5), from testicular cancer was 5.0 (0.8 to 15.4), and from other cancers 0.6 (0.2 to 1.4).

Analyses of cancer incidence were conducted only from 1 January 1971 onward because the NHS Central Register does not contain data on cancers incident before then. Of the 1124 patients in the cohort, 49 were not included in the analysis of cancer incidence because they had died, emigrated, or otherwise been lost to follow up before 1971. Of the remaining 1075 men, 865 had had orchidopexy(s) only, 85 hormonal treatment only, and 125 both treatments.

Seventeen cancers were registered during 26 389 person years of follow up of the cohort. There were 12 testicular cancers (nine teratomas, two seminomas, and one mixed teratoma-seminoma), two non-melanoma skin cancers, one pancreatic cancer, one lung cancer, and one myeloid leukaemia. Two of the testicular cancers were incident on opposite sides, at different dates, in the same man. One registered case of testicular cancer led to death. The other death from testicular cancer occurred before 1971, in a man whose bilateral ectopic testes had been brought into the scrotum by orchidopexy at age 11 years without biopsy.

Compared with the general population, the relative risk of cancer in the cohort was 1.4 (0.8 to 2.2), of testicular cancer 7.5 (3.9 to 12.8), and of cancers other than testicular 0.5 (0.2 to 1.2). The analyses below are in terms of relative risks per testis not per person. They are based on 1405 undescended testes and 718 testes that were opposite an undescended testes and were themselves either descended (693) or not stated to be undescended (25) in the 1075 men. Twenty seven testes were not entered in the analysis because they had been excised before 1971.

The overall risk of malignancy in an undescended testis compared with a testis in a man in the general population was 11.3 (5.9 to 19.4) (table 1). One testicular cancer occurred in a descended testis opposite a maldescended testis (relative risk=2.1). All of the analyses below relate to malignancy in undescended testes. The risk of malignancy was greater, but not significantly so, for a maldescended testis opposite a maldescended testis (relative risk=14.3) than for a unilaterally maldescended testis (8.5). All testicular cancers occurred in testes with either inguinal (10.9 (5.0 to 20.4)) or ectopic (12.4 (3.1 to 32.1)) maldescent. The relative risk of malignancy was not significantly greater in testes described as small or atrophic at surgery (12.3 (2.0 to 38.0)) than in those not so described (11.1 (5.3 to 20.0)).

Ten of the cancers occurred after orchidopexy alone (relative risk=13.7) and one (a teratoma) after hormonal treatment alone (relative risk =9.5) (table 2). The relative risk of testicular cancer tended to be lower at older ages (P for trend=0.06) but was not related to age at descent of the testis (P=0.43). Risk was not related to whether herniorrhaphy had been performed on that side (relative risk=11.4 for herniorrhaphy, 11.2 for no herniorrhaphy). Most of the herniorrhaphies were done at the same time as an orchidopexy. There was no significant trend of risk with number of orchidopexies on the testis. Risk of malignancy after orchidopexy decreased significantly with duration since orchidopexy (table 2); effectively, however, this analysis was for categories of duration 10-14 years and greater, because the unavailability of cancer incidence data before 1971 meant that there were few person years (and no cancers) in the analysis for the period less than 10 years after orchidopexy. The most common methods of orchidopexy were the Dennis Browne, after which five testicular cancers occurred (relative risk=16.2 (5.8 to 34.9)), and the Torek, after which no malignancies occurred (0.12 expected). Thirty four of the orchidopexies were followed by haematoma and 19 by wound infection, but in none of these did malignancy follow.

The largest risk of testicular cancer was in relation to biopsy (table 3). Five testicular cancers occurred in testes that had had biopsy samples removed, giving a relative risk of malignancy in such testes of 66.7 compared with 6.7 for testes that had not had biopsy samples taken (difference in risk between groups, P<0.001). Two testes in which biopsy was followed by malignancy were in the same man, but the risk of malignancy after biopsy was still highly significant if only one testis from this man was included in the analysis. The large risk relating to biopsy was present both for a maldescended testis in a patient with bilateral maldescent and for a unilaterally maldescended testis (table 3), and in each instance the risk was significantly greater than for a corresponding testis that had not had a biopsy sample removed.

The operation notes for the five cases in which testicular cancer had occurred after biopsy revealed no features in common other than the biopsy. Two of the testes were ectopic and three inguinal. In two an inguinal hernia was also present and in one hypospadias. In three orchidopexy had been done once; two had required repeat operations, in one instance with repeat biopsy. In no case were any operative or postoperative complications mentioned. Three of the testes were described as of normal size, one as smallish, and one as two thirds the size of the contralateral testis. No reason was stated for biopsy in any case.

On histological examination biopsy material from the four cases where biopsy was done at ages 10 years or younger showed tubules containing immature Sertoli cells and occasional spermatogonia but no Leydig cells (M C Parkinson, personal communication). This was the pattern seen in most biopsy specimens taken at these ages from cryptorchid testes in which cancer had not subsequently occurred.16 In the fifth case, where the biopsy was at age 14, and in one case in which biopsy was repeated at age 15 there were mature Sertoli cells, occasional spermatogonia, and Leydig cells–again a usual pattern in maldescent at these ages (M C Parkinson, personal communication). One of the biopsy specimens, taken at age 7, showed features of dysgenesis. Carcinoma in situ was found in a repeat biopsy specimen from a testis taken at age 15 but was not seen in the first biopsy at age 9. Further details of this case have been reported.16 None of the biopsy specimens showed atrophy. Judging from the shapes of the pathological specimens and the dates of the biopsies, all had been obtained by open, not needle, biopsy.

Discussion

The raised risk of testicular cancer in men with cryptorchidism has been known for 140 years ormore.17 The overall relative risk we found is similar to that reported in most case-control studies (relative risks 5-10),3 5 6 8 18 and in previous cohort studies, which with the exception of a study with only two cases of malignancy,10 have found relative risks of 5 to 7.7 9 11 The relative risks in cohort studies (including ours) that used general population rates to generate expectations are about 10% underestimates of the results if comparison could have been made with rates in men without cryptorchidism because the general population rates include malignancies in men with undescended testes in the general population. Nevertheless, adding 10% to our result would still leave it compatible with the ranges above.

Our analyses of cancer incidence omitted the years before 1971 because national cancer incidence follow up was not then available. The person years lost were mainly under age 20; this reduced the power of our study at these ages, when testicular cancer is uncommon, but did not in itself bias it since both person years and cancers in the period were omitted. A bias could have occurred if the effect of cryptorchidism was to create a small pool of susceptible testes, in many of which testicular cancer occurred before 1971 leading to withdrawal from follow up. Only one death from testicular cancer occurred before 1971, however, despite a relatively high fatality from the malignancy at that time, and no cases before 1971 were reported in the questionnaire responses from general practitioners.

Compared with expectations from the general population the relative risks of developing and dying from cancers other than testicular were reduced but not significantly lower in the cohort, and the risk of non-cancer deaths was significantly decreased. This raises the possibility of incomplete follow up. Flagging of deaths by the NHS Central Register is highly complete,19 so the mortality deficit is unlikely to be due to underascertainment. Furthermore, if there had been appreciable underascertainment of deaths we might have expected to have received questionnaire responses from general practitioners explaining that their patients had died. No such responses were received (we did not mail general practitioners of patients known to have died). Notification of cancer incidence is less complete than for mortality,19 which is why we also contacted general practitioners for cancer information. This revealed one testicular cancer (which since it had been registered with a cancer registry we included in the analysis) and one brain tumour, not notified by the NHS Central Register. The low death rate might reflect socioeconomic selection of patients treated for cryptorchidism at the hospital or might have been due to chance. Whatever the reason, it could not plausibly explain the findings for testicular cancer.