Among the 1110 periods of unprotected intercourse studied, there was no clear evidence that male gonadal dose of ionizing radiation lowered the probability of pregnancy during the first thirteen months of unprotected intercourse.
The relationship between exposure to ionizing radiation and fecundability has little been studied,[11] with our study being the first on the relationship between medical X-ray examinations and fecundability. We attributed a mean gonadal dose for each type of X-ray examination and quantified gonadal dose for each man rather than dose to the whole body, less relevant biologically. Our detailed questionnaire allowed to adjust for female behavioral factors likely to influence fecundity. We included periods of involuntary infertility lasting 12 months or more; such periods are most often excluded in retrospective studies of time to pregnancy, inducing bias [25] However, the power of our study is probably limited because assessment of male gonadal dose relied on a retrospective questionnaire, which can induce classification errors.
When we split in two the period of unprotected intercourse, the point estimates of the HR of pregnancy associated with doses of 5.00 mGy or more were greater than one for the first four months of the period of unprotected intercourse (meaning an increased probability of pregnancy with male gonadal dose) and lower than one for months five to thirteen. Couples conceiving during the first four months are the most fecund ones, so that the HR of pregnancy associated with doses for this time-period reflects the effect of exposure among the most fecund couples. Symmetrically, the HR of pregnancy for the months 5 to 13 reflect the effect of exposure among the least fecund couples. Although the sample sizes were too small to draw firm conclusions, there was some evidence of gonadal doses having different effects among the least fecund couples (those not conceiving in the first four months of attempt) and the most fecund couples. If real, such a difference would have sense from a biological point of view. Variations in sperm concentrations are known to influence the probability of pregnancy only in the range between 0 and 50 millions spermatozoa/ml[34, 35] Supposing schematically that men from the more fecund couples had sperm concentrations above 100 millions/ml and that the effect of ionizing radiation is to halve sperm concentration, male gonadal dose would not induce any effect on the probability of pregnancy in this group. If, on the other hand, the proportion of men with a sperm concentration lower than 50 millions/ml were high among the least fecund couples, then an effect of gonadal exposure to ionizing radiation could be expected in this group. Although not clear statistically in our study, a possibly different effect of gonadal dose between the first months of the period of unprotected intercourse and the following time period thus has some biological plausibility. Further studies are needed to confirm or rule out such a different sensitivity to doses of ionizing radiation between more fecund and less fecund couples.
One cohort, describing the influence of male exposure to ionizing radiation on the probability of pregnancy, compared subjects treated for cancer before adulthood and controls never treated for cancer[11] In this cohort, the probability of pregnancy was estimated from time between marriage and a first pregnancy. Radiotherapy and chemotherapy with alkylating agents were associated with a decreased probability of pregnancy. For patients who did not receive chemotherapy, the probability of pregnancy tended to be more strongly decreased for radiotherapy carried out below compared to above the diaphragm. The mechanisms underlying these associations are not obvious because radiotherapy below the diaphragm may more often be carried out to treat testes or other cancers associated with a decreased fecundity independently of any treatment[36] Moreover, fecundity was also reduced for survivors who received no treatment at all, compared to controls[11] Another study assessed primary infertility, defined as involuntary infertility with a medical consultation and not followed by a live birth until the age of 40 years[9] No deleterious effect of male dose of ionizing radiation was highlighted. The frequency of primary infertility was 2.5% among male workers not monitored for exposure to ionizing radiation; among monitored workers, the frequency of primary infertility varied from 2.2% for external doses between 0 and 2.49 milliSievert to 3.2% for doses above 100 milliSievert (9 cases). Occupational exposure to ionizing radiation was prospectively assessed, but primary infertility corresponds to a very severe and rare endpoint. Also, couples having involuntary infertility who did not consult a doctor for infertility were not considered as cases. These couples may represent a large proportion of couples with involuntary infertility; in a Danish study, about half of the couples with 12-month involuntary infertility had chosen to consult a medical doctor [12].
Study subjects
The participants were not informed of the endpoints of the study until the start of the interview. The women who refused to participate (about 30% of the eligible women [20]) had similar numbers of children to those who agreed to participate (not shown). Therefore, it is unlikely that fecundity was related to the refusal of the contacted women.
Among the partners of the participating women, 74.7% of the men eligible for inclusion accepted to answer the questionnaire and provided data on gonadal dose of ionizing radiation. The probability of pregnancy and the distribution of TTP were similar for participating and non-participating men. Therefore, it is unlikely that the participation of the man was associated with the fecundity of the couple.
Estimation of fecundity
Time to pregnancy is considered to be well recalled by women over a 15-year period in the case of pregnancy attempts leading to a live birth [21] The quality of recall of time of unprotected intercourse when the period of unprotected intercourse does not end with a pregnancy has not been studied to our knowledge. We had no information about the TTP of pregnancies that ended in miscarriages. Nonbirth outcomes like spontaneous abortions are commonly excluded from retrospective TTP studies, notably because there are doubts about the quality of recall of TTP for spontaneous abortions [21] This exclusion implies the hypothesis that X-ray radiations would have the same effect on TTP for periods of unprotected intercourse leading to a spontaneous abortion and for those leading to a live birth. The exclusion of unsuccessful attempts at pregnancy has been shown to limit the statistical power and bias TTP studies [23–25] We considered that this potential bias was of more concern than the potential bias induced by a poor quality of recall of such unsuccessful periods of unprotected intercourse, and therefore chose to include them in the analysis.
Since only a 9 to 12 months period extended from the end of the study period to the time of the interview, short TTP were over-represented among the periods of unprotected intercourse started at the end of the study period [21] This truncation might entail a bias in the case of time trends in exposure. As a check, we restricted the analysis of Table 2 to periods of unprotected intercourse started between 1985 and December 1997. The HR of pregnancy associated with a dose above 10 mGy was 1.38 (95%CI: 0.65–2.96, n = 27), very similar to the HR among the whole population, making any bias due to truncation unlikely.
Some covariates likely to influence TTP were not taken into account in the study. For example, the last contraceptive method used by the couple at the beginning of the unsuccessful attempts at pregnancy was not available.
Pregnancy planning
Pregnancies that happened while the couple was using birth control methods have no defined time to pregnancy. These couples might have a higher fecundity than couples whose pregnancies started after discontinuation of contraception [21] Differences in male doses between these groups would result in a pregnancy planning bias [21] We estimated the magnitude of possible pregnancy planning bias in our study by assuming that pregnancies occurring while contraception was being used corresponded to very fecund couples, and we assigned a TTP of one month to all of these pregnancies. The proportion of male gonadal doses above 5 mGy was 2.7% for pregnancies occurring while the couple was using a contraceptive method, and 5.3% for the periods of unprotected intercourse taken into account in our main analysis (p = 0.17). We estimated adjusted HR of pregnancy in this new dataset including pregnancies occurring while contraception was used; the adjusted HR of pregnancy for a dose of 10.00 mGy or more was 1.17 (n = 32), versus 1.44 (n = 31) when only the eligible couples of the study were taken into account. This suggests that pregnancy planning bias may have lead to an over-estimation of the HR of pregnancy associated with male gonadal dose, which might explain why most estimated HR of pregnancy were above 1.00.
Confounding by indication
A spurious statistical association between exposure to ionizing radiation and fecundity would be observed if X-ray examinations had been carried out as a consequence of diseases associated with decreased fecundity. This bias would correspond to confounding by indication [37] For example, percutaneous treatment with radioscopy can be used to treat varicocele, a condition that may be associated with a decreased fecundity independent of any treatment. Other examples include vas deferens opacification or a CT scanner of the abdomen or pelvis to estimate the spread of a testes cancer. None of the interviewed men reported these X-ray examinations, so confounding by indication is very unlikely in our study. Other diseases, such as vascular, neurological and endocrinal diseases, and drugs, such as neuroleptics and blood pressure drugs,[38] might indirectly influence fertility through erectile dysfunction or decreased libido. Such biases are unlikely because we excluded couples who declared a monthly sexual intercourse frequency of less than four.
Estimation of male gonadal dose of ionizing radiation
Male gonadal exposure was defined as the sum of the estimated gonadal doses of all X-ray examination received by the man between the age of 18 and the beginning of the period of unprotected intercourse. A window of exposure starting at puberty might have been more relevant, but we believed it unrealistic to obtain information on the type and localization of X-ray examinations during teenage years in a questionnaire to the man. Moreover, X-ray radiations received in adulthood seem able to have an effect at various stages of spermatogenesis, including on germinal cells [14] We could not study the short-term effect of exposure to ionizing radiation (e.g. in the year before the beginning of the period of unprotected intercourse) because of a too small number of exposed men. X-ray examinations performed during the period of unprotected intercourse considered were excluded to guarantee that exposure precedes the health outcome studied; indeed, the longer the TTP, the higher the likelihood of an X-ray examination being performed during the period of unprotected intercourse.
Estimations of testicular dose of ionizing radiation for a given type of X-ray examination can vary by up to a factor of 10. The patient dose depends on equipment type (applied potential), radiographic technique (number of X-ray films, fluoroscopic screening time, X-ray beam projection and centering point position) and anthropomorphic characteristics[31, 39] When such information is not available, classification errors in exposure occurs. In a study exploring the association between preconception paternal exposure to medical ionizing radiation and offspring birth weight, [7] exposure was classified by a dichotomous variable indicating whether the testes were likely to be in the field of the examination or not, and not by a variable with 6 categories as in our study. The question of knowing which approach leaves more room to classification bias deserves further investigation.
The type, anatomical localization and date of the X-ray examination may not always be well recalled, especially for examinations carried out a long time ago. Questions on X-ray examinations in three different sections of the questionnaire and questions on the medical history of men probably limited recall bias. We found no source of data allowing to estimate the frequency of X-ray examinations according to sex and age in France. Therefore, the magnitude of recall bias on exposure could not be directly evaluated. When we restricted the analysis to men aged 35 years or less at interview to limit the length of the recall period, adjusted HR of pregnancy associated with doses of 5.01 to 10.0 mGy and doses above 10.01 mGy were respectively 0.31 (3 observations, 95% CI, 0.04–2.50) and 0.50 (8 observations, 95% CI, 0.10–2.57), indicating that recall bias might have biased towards a lack of association the estimated effect of gonadal exposure to X-rays. Data on medical examinations in private and public health care institutes are more exhaustive since 2004–2005. This may allow, in future studies, a more precise quantification of X-ray examinations performed. A study on the quality of recall of X-ray examinations among thyroid cancer cases and controls [40] showed that, compared to medical records, some X-ray examinations were overdeclared in retrospective questionnaires whereas others were underdeclared; the associations between exposure and cancer risk were similar for both ways of assessment of exposure. Recall bias may also exist for adjustment factors, thus leaving residual confounding; the quality of recall of the frequency of sexual intercourse over a 15-year period, for instance, may be poor, as illustrated by the fact that there were 20% of missing data for this question.
Some of the men recruited in the Beaumont-Hague ward had been potentially exposed to ionizing radiation occupationally. We collected information on whether the man had ever worked in the vicinity of radioactive material or X-ray devices, or had ever carried a gamma-radiation dosimeter. The adjusted HR of pregnancy for doses of X-rays of 10.00 mGy or more was 1.32 (95%CI: 0.56–3.08, n = 19) for the 563 men potentially exposed to ionizing radiation at work, compared to 1.65 (95%CI: 0.54–5.07, n = 12) for the 460 unexposed men, giving no evidence that our estimate was strongly biased by potential occupational exposure.
Treatment of missing data on X-ray examinations
For 209 periods of unprotected intercourse (19% of the observations), either the date or the anatomical localization was missing for one of the X-ray examinations. The missing information for these examinations were imputed by using a single imputation approach[32] Excluding these missing values would have meant excluding the corresponding periods of unprotected intercourse. This could have resulted in a bias. When these 209 periods of unprotected intercourse were excluded, the adjusted HR of pregnancy for male doses of 10.00 mGy or more was 1.71 (95%CI: 0.78–3.74, n = 25), compared to 1.44 (95%CI: 0.73–2.86, n = 31) when the missing data was imputed.
Assuming no bias, the probability to detect a decrease in fecundability by 50% for men exposed to a gonadal dose above 10 mGy was about 70% for a significance level of 0.05 [22].