Testicular cancer: Marked birth cohort effects on incidence and a decline in mortality in southern Netherlands since 1970

Authors

  • Rob Verhoeven,

    1. Eindhoven Cancer Registry, Comprehensive Cancer Centre South, Eindhoven, The Netherlands
    2. Department of Epidemiology and Biostatistics, Radboud University Nijmegen Medical Centre, The Netherlands
    Search for more papers by this author
  • Saskia Houterman,

    1. Eindhoven Cancer Registry, Comprehensive Cancer Centre South, Eindhoven, The Netherlands
    2. Regional Study Group for Urological Oncology IKZ, Eindhoven, The Netherlands
    Search for more papers by this author
  • Bart Kiemeney,

    1. Department of Epidemiology and Biostatistics, Radboud University Nijmegen Medical Centre, The Netherlands
    2. Department of Urology, Radboud University Nijmegen Medical Centre, The Netherlands
    Search for more papers by this author
  • Evert Koldewijn,

    1. Regional Study Group for Urological Oncology IKZ, Eindhoven, The Netherlands
    2. Department of Urology, Catharina Hospital, Eindhoven, The Netherlands
    Search for more papers by this author
  • Jan Willem Coebergh

    Corresponding author
    1. Eindhoven Cancer Registry, Comprehensive Cancer Centre South, Eindhoven, The Netherlands
    2. Department of Public Health, Erasmus University Medical Centre Rotterdam, The Netherlands
    • Eindhoven Cancer Registry, Comprehensive Cancer Centre South (IKZ), PO Box 231, 5600 AE Eindhoven, The Netherlands
    Search for more papers by this author
    • Fax: +31-40-2971610.


Abstract

The aim of our study was to interpret the changing incidence, and to describe the mortality of patients with testicular cancer in the south of the Netherlands between 1970 and 2004. On the basis of data from the Eindhoven Cancer Registry and Statistics Netherlands, 5-year moving average standardised incidence and mortality rates were calculated. An age-period-cohort (APC) Poisson regression analysis was performed to disentangle time and birth cohort effects on incidence. The incidence rate remained stable for all ages at about 3 per 100,000 person-years until 1989 but increased annually thereafter by 4% to 6 in 2004. This increase can almost completely be attributed to an increase in localised tumours. The largest increase was found for seminoma testicular cancer (TC) patients aged 35–39 and non-seminoma TC patients aged 20–24 years. Relatively more localised and tumours with lymph node metastases were detected in the later periods. APC analysis showed the best fit with an age-cohort model. An increase in incidence of TC was found for birth cohorts since 1950. The mortality rate dropped from 1.0 per 100,000 person-years in 1970 to 0.3 in 2005, with a steep annual decline of 12% in the period 1979–1986. In conclusion, the increase in incidence of TC was strongly correlated with birth cohorts since 1945. The increase in incidence is possibly caused by in utero or early life exposure to a yet unknown risk factor. There was a steep decline in mortality in the period 1979–1986. © 2007 Wiley-Liss, Inc.

Testicular cancer (TC) is a rare neoplasm, accounting for 0.8% of all male cancers worldwide,1 but it is the most common malignancy among men aged 15–44 in developed countries. Of all TCs, 95% are germ cell tumours which are grouped histologically into seminomas and nonseminomas.2 The latter have an age-specific incidence peak 10 years earlier (20–35 years) than seminomas (30–45 years).2 The non-seminoma TC group has multiple histologies, including tumours with mixed histology containing seminoma tissue.

Incidence rates across Europe ranged from 2 per 100,000 person-years in Spain to 10 in Denmark in the period 1993–1997.3 Trends in incidence in almost all European countries are characterized by steep increases in the past few decades, particularly among adolescents and young adults. The smallest increase over a period of 20 years was found for Switzerland (28%, from 5.1 to 6.5 per 100,000), and the largest for France (126%, from 3.5 to 7.9 per 100,000).3 Incidence in the Netherlands has also increased, but it is not clear when the incidence started to rise.4 In contrast to incidence, TC mortality has dropped by about 70% in the USA and Western Europe since the 1970s.5 The pace of the decrease was different among countries, depending on the ability of the health care system to give specialised care to TC patients. This decrease in mortality is attributed to the introduction of cisplatin chemotherapy, which has proven to be the most effective treatment for non-seminoma TC.6

It has been reported consistently in northern and middle European countries that the period of birth affects the incidence of TC.7, 8, 9, 10, 11 Birth cohorts from around 1930 and later exhibited a higher incidence of TC, some countries had an attenuation in the increasing trend TC incidence in the birth cohorts from 1930 until 1945. The trend in incidence has been described by both an age-cohort (AC) model and an age-period-cohort (APC) model in many different western populations.7, 8, 10, 11, 12, 13 An AC model suggests that age and time of birth affect the incidence of TC, while an APC suggests that the incidence is also influenced by the time period of TC diagnosis.

Analytical studies of possible risk factors (mainly those occurring early in life) have shown that cryptorchidism, a hypotrophic (<12 ml) or atrophic testis, Klinefelter's syndrome and a family history of testicular tumours among first-degree relatives are associated with TC among Caucasians.14 Subfertility, maternal in utero exposure to cigarette smoke and hormones, early age at puberty, decreased levels of androgen and high intake of dairy products have been associated with an increased risk of TC, but the relevance of these risk factors remains unclear.14

The aim of our study was to detect trends in incidence and mortality of TC in the South of the Netherlands from 1970 to 2004. In addition, we aimed to clarify whether the trend in TC incidence follows an AC pattern or an APC pattern.

Abbreviations:

AD, age-drift; AC, age-cohort; AP, age-period; APC, age-period-cohort; EAPC, estimated annual percentage changes; ECR, Eindhoven Cancer Registry; ICD-O, International Classification of Diseases for Oncology; TC, testicular cancer.

Material and methods

Patients

The Eindhoven Cancer Registry (ECR) has collected data on all patients with newly diagnosed cancer in the southern part of the Netherlands since 1955.15 Until 1988, only patients diagnosed in the eastern part of the area were registered, this was a population-based coverage of 1.0 million inhabitants. Thereafter, also patients living in the middle and western parts were registered. Nowadays, the registry covers a population of 2.4 million inhabitants. The area offers good access to specialised medical care in 9 general hospitals and 2 large radiotherapy institutes. Information on diagnosis, staging and treatment was extracted from the medical records by trained registrars after notification by pathology laboratories and the medical records departments of the hospitals.

All patients diagnosed with testicular cancer between 1970 and 2004 were included in the study. The tumours are grouped according to histological origin, as described in the third revision of the International Classification of Diseases for Oncology (ICD-O):16 seminomas (ICD-O codes: 9060-9064), nonseminomas (ICD-O codes: 9065-9085, 9100-1902, 9105) or other. The stage grouping of the TNM-classifications of TC changed over time in such a way that it became impossible to compare the different stage groups over time. We have therefore chosen to categorise the extent of the disease as localised (any T, N = 0 and M = 0), lymph node metastases (any T, N > 0 and M = 0), distant metastases (any T, N and M > 0) and stage unknown. Patients with stage unknown (n = 23) were excluded from analyses that were stratified by stage.

Incidence

Five-year moving average age-standardised incidence rates were calculated per 100,000 person-years for the total group of TC and for seminomas and nonseminomas separately. The total incidence of TC was compared to the total incidence of TC in the Netherlands, which was available for the period 1989–2003 (available at http://www.ikcnet.nl/page.php?id=41, accessed Jan 20, 2007). Standardisation was performed according to the European standard population. Moreover, stage and age-specific incidence rates were computed. Because stage was recorded reliably from 1980 onwards, only patients diagnosed in 1980 and later were included in the stage distribution analysis.

Evaluation of the trend in incidence was performed by calculating the estimated annual percentage changes (EAPC) for different time periods.

Age-period-cohort models

For the age-period-cohort (APC) models, patients aged <15 (1.0% of all cases) and ≥60 (7.2% of all cases) were excluded because of the small number of cases in these age groups. The population was divided into 5-year age groups (15–19, …, 54–59), 5-year calendar periods (1970–1974, …, 2000–2004) and matching 10-year birth cohorts (1915–1924, …, 1980–1989), which means that each cohort overlaps the next cohort by exactly 5 years. Birth cohorts will be denoted by the central birth year, for example the 1915–1924 birth cohort is denoted by 1920. The GENMOD procedure of the SAS package was used to fit a series of Poisson regression models, to estimate the separate effects of age, time of diagnosis and birth cohort on the trend in incidence, according to the methods described by Clayton and Schifflers.17, 18 To test the goodness-of-fit of the models with the observed incidence rates and to test the models against one another, deviances and differences between the deviances with appropriate degrees of freedom were used.17, 18 The terms preceding the word model indicate which variables are used to describe the data. So an AC model means that the incidence data is described by both the age and cohort variables. The age-drift (AD) model means that the incidence data are described by the age variable and a drift parameter.

Birth cohorts with data for less than 3 diagnostic periods will not be presented.

Mortality

Mortality data were available from Statistics Netherlands for the period 1970–2005. Five-year moving average European standardised mortality rates per 100,000 person-years were calculated and were compared to the Dutch mortality rates (available at http://www.ikcnet.nl/page.php?id=217, accessed Jan 20, 2007). In addition, trend EAPC analysis was performed for different time periods. For the period 1970–1988, the Dutch mortality rates were only available as crude mortality rates.

Results

Incidence

In total, 1,165 patients were diagnosed with TC between 1970 and 2004 (53% seminoma, 45% nonseminoma and 2% other). The age-standardised 5-year moving average incidence rate increased from 2.9 per 100,000 males in 1970 to 6.1 in 2004 (Fig. 1). Incidence rates of 3.3 and 6.0 per 100,000 males in the ECR-region, for the years 1988 and 2003, were similar to such rates of 4.1 and 6.5 per 100,000 males in these same years in the whole Netherlands (data not shown). Because the largest increase seemed to take place in the period 1988–2004, an EAPC of 4.4% (95% confidence interval (CI) 3.0%–5.8%) was calculated, whereas it was −0.2% (95% CI −3.1% to 2.7%) for the period 1970–1987. This means that there was a significant annual increase of the incidence of 4.4% in the period 1988–2004 but no increase of incidence in the period 1970–1987. An incidence analysis of only the eastern part of the ECR region showed a similar TC incidence over time as that of the whole ECR region.

Figure 1.

Five-year moving average European standardised testicular cancer incidence and mortality rates per 100,000 person-years.

In the period 2000–2004, the highest age-specific incidence rate for seminomas was found for the 35–39 year age group (6.8 per 100,000 person-years) and for nonseminomas for the 20–24 year age group (6.7 per 100,000 person-years).

The 5-year moving average incidence rates according to stage are presented in Figure 2, which shows that almost the total increase in incidence can be attributed to the increase in local tumours. This increase was almost equally distributed over the seminoma and non-seminoma TCs. The annual increase over the period 1980–2004 in localised tumours of seminoma TC was 4.4% (95% CI 2.3% to 6.5%) and 6.2% (95% CI 4.4% to 8.0%) for non-seminoma TCs. For the tumours with lymph node metastases this was 3.7% (95% CI 1.0% to 6.4%) for the seminoma TCs and 4.3% (95% CI 1.8% to 6.8%) for the non-seminoma TCs.

Figure 2.

Five-year moving average European standardised testicular cancer incidence rates per 100,000 person-years according to stage.

APC models

The fits of the different models are presented in Table I. The AD model and the Age-Period (AP) model gave a poor fit, with a p-value for the goodness-of-fit test of <0.001 for both (a small p-value indicates that the model is significantly different from the observed incidence data). The AC model (p = 0.12) and APC model (p = 0.10) both showed a good fit.

Table I. The Goodness-of Fit for the Different Models
ModelDevianceDFp-value
  1. A small p-value indicates that the model is significantly different from the observed data.

Age179.754<0.001
Age-drift98.353<0.001
Age-period87.348<0.001
Age-cohort50.5400.12
Age-period-cohort45.8350.10

The AC model was significantly better than the AD model (p < 0.001), while the AP-model did not differ significantly from the AD model (p = 0.05). The APC model did not describe the data much better than the AC model (p = 0.46). We therefore concluded that the AC model provided the most parsimonious description of the data.

The birth cohort effect in the AC model is presented in Figure 3. Men who were born around 1975 had a 3 times (95% CI 2.2–4.2) higher risk of developing TC than those in the reference 1950 birth cohort.

Figure 3.

Relative risk of testicular cancer incidence per birth cohort with 95% Wald confidence limits, based on the age-cohort model (birth cohort 1950 is the reference cohort).

Mortality

The 5-year moving average mortality rate increased in the period 1970–1978 (Fig. 1) and declined steeply in the period 1979–1986 (from 1.0 per 100,000 to 0.4 per 100,000, with an EAPC value of −12.0% (p = 0.07)). In the period 1987–2005, the mortality rate fluctuated around a level of 0.4 per 100,000 and being similar to that of the entire Netherlands during the period of 1970 until 2005 (data not shown).

Discussion

A marked increase in incidence of TC has occurred since 1989 in the southern part of the Netherlands, which correlates with increasing risks for birth cohorts since 1945. There has been a marked decrease in mortality since 1979.

TNM testicular cancer stage classification changed several times between 1970 and 2004. To prevent misclassification of the tumours we have made our own stage classification, in which all tumours are classified in the same way. But through this stage classification it becomes more difficult to compare the study results to other studies.

The changes in histological classification over the years have no meaningful influences on our histological classification into seminoma and non-seminoma TC.

Unfortunately, the ECR does not have access to death certificates. The registry partly solve the problem of potential incompleteness by active registration based on a national computerised archive of pathological diagnosis (PALGA) and on clinical diagnosis derived from a computerised national hospital discharges system (LMR).

Incidence and APC models

The incidence of TC increased by 110% in the period 1970–2004: seminoma TC started to increase in 1990 and non-seminoma TC some years earlier. This increase can be attributed almost entirely to the increase in localised TC tumours. As a result of the increasing incidence and the decreasing mortality, the prevalence of TC in the Netherlands increased from 37 per 100,000 person-years in 1990 to 64 per 100,000 in 2002 and the prognosis is that the prevalence in 2010 will be 132 per 100,000.19 As a consequence the claim for medical care in the Netherlands will rise markedly, since active follow-up for localised TC seems to be efficacious (unlike the situation in many other tumour sites20, 21).

The fact that the rate of the incidence increase in the eastern part of the ECR region was comparable to the incidence increase of the whole ECR region indicates that the expansion of the ECR region in 1988 did not introduce a bias in the TC incidence.

The incidence of nongerm cell TC was low and relatively constant during the whole study period and did not influence the overall TC incidence.

The fact that the incidence of both seminoma and non-seminoma TC is increasing suggests that one or more mutual risk factors, probably introduced by changes in environment and lifestyle, might be responsible for the increase in incidence. Because the AC model gave the most efficient fit of all APC models, and the increased risks of incidence for birth cohorts since 1950 also suggest that the risk factor exerts its effect in utero and/or early in life. When the risk factor would exert its effect later in life, it would probably affect boys and men of different ages. An increased incidence would then be more attributable to period effects then to birth cohort effects as is found in our study.

The result of our AC model is comparable to that of 2 other studies, which analysed multiple European populations and found the best fit to be an AC model for most populations.7, 11 A study in the United States found the best fit with an APC model, but the birth cohort effect was dominant.12 All 3 studies found an increase in risk with successive birth cohorts, which is comparable to the results of our study.

There are several hypotheses explaining the increase of TC incidence. One of them suggests that there is an increase in incidence of several testicular diseases (for example cryptorchidism and TC) through changes in genetic and/or environmental factors, including endocrine disrupters.22 Another hypothesis suggests that lifestyle changes such as an increase in maternal age and an increase in the number of first-born children causes the increase of TC incidence.23 In the Netherlands, the average age of the mother at birth of her first-born child increased from 24.9 years in the 1960s to 26.4 years in the 1980s and the average number of children per mother decreased from 3.0 to 1.5. This resulted in an increase in the percentage of first-born children from 35% in the 1960s to 45% in the 1980s. Although these changes are in the same period as the increase in relative risk of birth cohorts on TC, we cannot verify this hypothesis in our data because these possible risk factors of TC were not registered in the ECR.

The risk factors that are responsible for the increase in incidence may initiate or promote the development of both histologies, in utero and/or early in life, but apparently have a shorter latency time for nonseminomas than for seminomas, explaining the difference in the age-peaks of seminoma and non-seminoma TC.

The incidence of TC has been increasing for many decades in most European countries, but with a varying start.3, 7 A study of TC in Scotland found that the increase in incidence was more pronounced in the age-group <40 than in the age-group ≥40.24 The rise in incidence seems to have started later in the ECR-region, compared to most other European countries, but the absolute incidence and the age-distribution of the incidence have become similar to those of other European countries.

We also found a small nonsignificant decrease in the relative risk for birth cohorts 1935 and 1940, as was found in Denmark, France, Norway, Italy, Slovenia, Spain and Sweden.7, 10, 11

Through the relative small population in our study, it was not possible to perform a more detailed APC-analysis for example smaller period and cohort groups or separate analysis for seminoma and non-seminoma groups.

Mortality

The mortality rate dropped from around 1 per 100,000 person-years in the mid 1970s to 0.4 in 1986 and fluctuated thereafter between 0.2 and 0.5. However, in this last period the numbers became very small. This pattern of decrease is comparable to the mortality rates found in the populations of the United States and the European Union. The steep decrease is probably related to improved TC survival in that time.24, 25, 26, 27 There was a small and nonsignificant annual decrease of −1.2% (p = 0.61) in mortality in the area of the ECR in the period 1987–2005. And in almost the same period (1987–2004) the incidence increased by 4.2% (p = <0.001). Thus although the incidence was rising, mortality did not rise. This can be attributed to the higher percentage of less-aggressive tumours, which have higher survival rate than more aggressive tumours, and increased survival rates for TC in the ECR region.27 These increased survival rates can be attributed to the introduction of cisplatin-containing chemotherapy in the 1970s.6

Conclusion

A marked increase in incidence of TC was observed in the south of the Netherlands, predominantly for tumours with lower aggressiveness and for both the seminoma and non-seminoma TCs. This occurred predominantly in birth cohorts since 1945; in utero and/or early life influences seem likely. The marked decrease in mortality since the 1970s was most likely caused by improved treatment.

Future investigations should focus on factors that influence the development of testicular tumours.

Acknowledgements

We thank Huub Straatman for his support in the analyses of the APC-models. The Comprehensive Cancer Centre South enabled this study.

Ancillary