Since more than 50 years, there is an ongoing, yet unexplained epidemic of germ-cell testicular cancer in many populations. Previous studies suggest that environmental factors operating during fetal life could disturb testicular development and result in neoplastic transformation of the primordial germ cells.1 In a hypothesis article, Clemmesen suggested, based on the observation of parallel trends of testicular cancer and female bladder cancer, that the rising prevalence of smoking during pregnancy may explain the rising trend in testicular cancer.2 Besides containing numerous potent carcinogens, tobacco smoke has several physiological effects, such as reducing placental blood flow and interfering with pregnancy estrogens, that could disturb normal testicular cell differentiation.3 In accordance with Clemmesen's hypothesis, we have recently reported a strong geographical and temporal correlation between female smoking prevalence and testicular cancer incidence in the Nordic countries.4 Several other data lend indirect support to this hypothesis, including reports of elevated risks of testicular cancer among sons of women with lung cancer.5, 6 Nevertheless, previous analytic studies on testicular cancer report no association with maternal smoking.7, 8, 9, 10, 11, 12 These studies, though, have used questionnaire information, and the inherent validity problems of selection and recall bias in assessing smoking during pregnancy retrospectively hinder conclusive inference. Therefore, we have used the unique infrastructure of high-quality registries in Sweden to conduct a nested case–control study on testicular cancer, where data on maternal smoking were collected during pregnancy, and where selection and recall bias should not be a problem.
For no apparent reason, the incidence of testicular cancer has increased to epidemic proportions in many countries. Pregnancy smoking has been suggested to be a cause. Previous analytical studies have been negative, but the inherent difficulties in retrospective assessment of this exposure have led to no definite conclusion. We have conducted a population-based case–control study on 192 cases of testicular germ-cell cancer—born in Sweden in 1973 onwards and aged ≥15 at cancer diagnosis—and 494 matched controls, where data on maternal smoking were collected during pregnancy. We found no association with testicular cancer for maternal smoking during pregnancy (OR, 0.91; 95% CI, 0.64–1.30), and there was no evidence of a dose–response effect. We conclude that the epidemic rise in testicular cancer in many populations is not due to the surge in smoking among women. © 2007 Wiley-Liss, Inc.
Material and methods
The Swedish Medical Birth Register includes prospectively collected information on the antenatal and perinatal period for 98% of all deliveries in Sweden.13 Since 1973, antenatal care centers and delivery units have used comprehensive charts to record information in a standardized manner. At registration for antenatal care, which occurs before the 15th week of gestation in more than 95% of the pregnancies,14 information about smoking habits is either reported as number of cigarettes smoked per day or categorized into 3 groups: nonsmoker, less than 10 cigarettes per day, and at least 10 cigarettes per day. Perinatal information, including birth weight and gestational duration, is recorded at the delivery unit. The charts are stored together with the birth records in the archives of the delivery units. The major part of the information in the charts (including smoking status since 1983) is forwarded to and computerized by the Birth Registry.
The Swedish National Cancer Registry was established in 1958 and obtains mandatory reporting from both clinicians and pathologist on all newly diagnosed malignant neoplasms. Information on site and histopathology of the tumors are recorded. For testicular cancer (code 178, International Classification of Diseases, 7th revision) the completeness of registration is estimated to be more than 95% complete.15
The study base in the present study is defined by the person time experienced by the cohort of singleton males in the Medical Birth Registry born in June 1973 onwards. There were 266 cases of testicular germ-cell cancer (age ≥15 at cancer diagnosis) in the study base before 2002, ascertained through the nationwide Cancer Registry. Information from the Cancer Registry allowed us to assign cases to 1 of the 2 major histological groups of testicular germ-cell cancer: seminomas and nonseminomas. The latter included cancers of mixed histological pattern. Potential controls were the first 3 men born at the same hospital after a case, and who were alive and without testicular cancer at the time of diagnosis of the corresponding case. Information on the study variables for cases and controls was retrieved from the Birth Registry. For those born before 1983, data on smoking were collected manually from the birth records in the delivery archives. We excluded 32 cases (12%) and 79 controls (10%), for whom the chart was untraceable, and 41 (15%) cases and 117 (15%) controls, for whom there was no information on smoking (mainly because smoking was not registered in 1 Swedish county before 1983). After exclusion of another 2 controls with missing values in covariates, and 1 case and 108 controls in risk sets with no case or no controls, 192 cases and 494 control subjects remained for analysis.
We estimated odds ratios (ORs) with corresponding 95% confidence intervals (CIs) through conditional logistic regression, using SAS statistical software (version 9.1, SAS Institute, Cary, NC) procedure PHREG.
We found no association with testicular cancer for maternal smoking during pregnancy (OR, 0.91; 95% CI, 0.64–1.30), and there was no evidence of a dose–response effect (Table I). Older maternal age was weakly associated with risk of testicular cancer. No association was found with birth order, birth weight or gestational duration. Among controls, the mean birth weights were 3,638, 3,540 and 3,379 g among nonsmokers, light and heavy smokers, respectively. The corresponding mean birth weights among cases were 3,640, 3,476 and 3,504. Risk of seminoma was reduced among light smokers, but less so among heavy smokers (Table II).
|Characteristics||Cases (n = 192)||Controls (n = 494)||Crude OR1,2||Adjusted OR1,2|
|Subjects' age, mean ± SD1 (years)||22 ± 3||22 ± 3|
|Maternal smoking (cigarettes/day)|
|None||111 (58)||273 (55)||1.00||1.00|
|<10||34 (18)||102 (21)||0.82 [0.53–1.29]4||0.83 [0.52–1.30]|
|≥10||47 (25)||119 (24)||0.97 [0.64–1.47]||0.99 [0.65–1.52]|
|Maternal age (years)|
|<20||8 (4)||32 (6)||0.63 [0.28–1.43]||0.58 [0.24–1.37]|
|20–24||57 (30)||160 (32)||0.88 [0.59–1.31]||0.88 [0.58–1.33]|
|25–29||77 (40)||188 (38)||1.00||1.00|
|30–34||39 (20)||90 (18)||1.11 [0.70–1.75]||1.11 [0.70–1.76]|
|≥35||11 (6)||24 (5)||1.16 [0.52–2.58]||1.13 [0.50–2.54]|
|Each 5-year increase||1.14 [0.96–1.36]||1.14 [0.95–1.38]|
|Firstborn||83 (43)||216 (44)||0.98 [0.70–1.37]||1.14 [0.79–1.65]|
|Nonfirstborn||109 (57)||278 (56)||1.00||1.00|
|Gestational duration (tertiles, days)|
|<277||67 (35)||155 (31)||1.27 [0.84–1.90]||1.34 [0.87–2.04]|
|277–285||59 (31)||171 (35)||1.00||1.00|
|≥286||66 (34)||168 (34)||1.12 [0.74–1.70]||1.11 [0.73–1.70]|
|Birth weight (tertiles, g)|
|<3,330||58 (30)||164 (33)||0.77 [0.51–1.15]||0.73 [0.48–1.11]|
|3,330–3,770||76 (40)||165 (33)||1.00||1.00|
|≥3,780||58 (30)||165 (33)||0.77 [0.52–1.14]||0.78 [0.52–1.17]|
|Cases||Controls||Adjusted OR1,2||Cases||Controls||Adjusted OR1,2|
|Maternal smoking (cigarettes/day)|
|None||34 (77)3||69 (59)||1.00||77 (52)||204 (54)||1.00|
|<10||4 (9)||24 (21)||0.29 [0.09–0.99]4||30 (20)||78 (21)||1.00 [0.60–1.66]|
|≥10||6 (14)||23 (20)||0.49 [0.17–1.43]||41 (28)||96 (25)||1.18 [0.73–1.91]|
Our finding of no association between pregnancy smoking and testicular cancer is unlikely to be due to bias. First, information on smoking was collected at the time of exposure without knowledge of the outcome. Recall bias is thus precluded. Second, the study was based on registries with a high completeness of registration. This allowed us to ascertain virtually all cases originating from the study base.
In our data, smoking was self-reported, and thus possibly underreported.16 Such nondifferential misclassification could theoretically generate a false-negative result. The measurement of smoking in the Birth Registry is, however, known to be fairly accurate,13, 17 and we found an effect of smoking on birth weight consistent with other studies.16, 18
We lack information on smoking cessation after the first antenatal visit, and the potential effect of smoking cessation could not be assessed. It is, however, known that the majority of women who stop smoking because of pregnancy do so before antenatal registration.16 Moreover, the differences in mean birth weight between non- and heavy smokers were 136 and 259 g among cases and controls, respectively. This would, if anything, imply a higher cessation rate among mothers of cases compared with controls, and consequently an overestimation of a potential adverse effect of smoking on testicular cancer risk.
In subgroup analyses, we found a decrease in risk for seminomas associated with maternal smoking. This finding was based on 10 exposed cases and inverse to the prior hypothesis. We therefore consider chance to be the most likely explanation.
In conclusion, although pregnancy smoking cannot be entirely discarded as a minor risk factor for testicular cancer, we believe that this register-based study together with previous questionnaire-based studies provide strong evidence to rule out smoking as the cause of the epidemic rise in testicular cancer observed in many populations.
The work of Dr. Lorenzo Richiardi was conducted within the framework of projects supported by the project “Oncologia” San Paolo/FIRMS and the Italian Association for Research on Cancer (AIRC). We also thank Ms. Lena Brandt for her valuable contribution.