METHODS

The present study included offspring to licensed pesticide applicators who were followed up for incident cancer in the Cancer Registry. Information was linked from several registers: pesticide applicators licensed 1965–76, Multigeneration Register, Cancer Registry, and Cause of Death Register. The linkage was possible since each register included the personal identification number. This number is unique to all citizens in Sweden and consists of 10 digits: two each for year, month, and day of birth; a three digit serial number; and a check digit.

The cohort of offspring of male pesticide applicators was created by linking the computerised cohort of 20 245 pesticide applicators licensed 1965–76 with the Multigeneration Register. In total 27 329 offspring born between 1958 and 1994 were identified.

Date of death was derived from the Cause of Death Register when an offspring was deceased. The offspring cohort was thereafter followed up in the Cancer Registry from date of birth until death or 31 December 1994. Notification to the Cancer Registry of all malignant and some benign tumours has been compulsory for all physicians and pathologists since the establishment in 1958. Each case is thus in general reported from two different sources.

The expected number of incident cases in the offspring cohort was calculated on the basis of annual cancer incidence in fiveyear age groups for the whole country. The standardised incidence ratio (SIR) was computed as the ratio between the observed and expected number of cases. The observed number of cases was assumed to be Poisson distributed, and a 95% confidence interval for SIR was derived by means of Poisson distribution tables.³³

A priori hypotheses based on previous studies were that this cohort had increased risk of tumours of the nervous system, kidney cancer, leukaemia, lymphoma, soft tissue sarcoma, and testicular cancer.

Since the number of statistical significant findings is directly related to the number of significance tests performed, the possibility of some findings being due to chance is likely to be enhanced when testing several hypotheses. However, no correction for multiplicity has been performed in the calculations of confidence intervals.

RESULTS

Among the offspring, 51.5% were males and 48.5% were females. Table 1 presents the distributions of year of birth among the children and father’s year of license for pesticide application. The mean age at diagnosis was 9.3 (SD 6.4). Table 2</tblref> shows the year of diagnosis distribution.

Table 3 presents observed and expected numbers of cases and SIR for selected cancer sites. In total 51 cases of cancer were observed in the follow up of 472 461 person-years, significantly lower than the expected 73.0 (SIR 0.70, 95% CI 0.52 to 0.92).

Tumours of the nervous system

This site was the most common, amounting to 39% of all cases of cancer in the cohort. The observed number of tumours was close to the number expected, 20 cases; SIR was 1.01 (95% CI 0.62 to 1.56). The majority of the tumours were localised in the brain and the meninges.

Leukaemia

A statistically significant reduced risk for leukaemia was found (SIR 0.43, 95% CI 0.19 to 0.86). Seven of the eight observed cases were acute lymphatic leukaemia; the other was a blast and stem cell leukaemia.

Lymphoma

With regard to non-Hodgkin’s lymphoma, three cases were observed as against 4.8 expected (SIR 0.63, 95% CI 0.13 to 1.83). For Hodgkin’s disease, there were five cases observed versus 3.7 expected (SIR 1.36, 95% CI 0.44 to 3.17).

Testicular cancer

With regard to testicular cancer, two cases were observed and 1.7 expected (SIR 1.19, 95% CI 0.13 to 4.28).

Kidney cancer

An SIR of 0.27 (95% CI 0.00 to 1.52) was found, based on one observed versus 3.7 expected cases.

Soft tissue sarcoma

Only one case was found versus 2.7 expected (SIR 0.38, 95% CI 0.00 to 2.09).

Cancer sites not included in the a priori hypotheses

Noteworthy are two non-significant results; an increased SIR for endocrine glands of 1.62 (95% CI 0.33 to 4.73) and a decreased SIR for bone tumours of 0.54 (95% CI 0.06 to 1.93).

DISCUSSION

Offspring of male pesticide applicators, who have used the most acutely toxic pesticides in agriculture, were in this study found to have a statistically significant decreased overall risk of cancer compared with the whole population. None of the a priori hypotheses could be confirmed. For leukaemia a significantly decreased risk was found.

An explanation of the decreased cancer risk could be that the offspring were brought up on farms and have had a healthier environment with regard to air pollution, exhausts, smoking, diet, drinking water, less stress, more physical activity, etc. There could, however, be a number of potentially hazardous agents on a farm, such as pesticides, solvents, motor vehicle fuels, lubricants, dusts, manure gases, infectious agents, mould, antibiotics, fungi, etc., some of which are known or suspected carcinogens. The fact that a blood relationship exists does not guarantee that a child has shared the environment with the father. However, farmers have the lowest rate of divorces in Sweden; for example, in 1960 the rate was 0.5% among farmers compared to 2.5% in the general population (personal communication with Gunilla Ringbäck, The National Board of Health and Welfare). Parental cancer has been shown to modify the risk of cancer in the offspring.³⁴ Decreased risks in both pesticide applicators and their offspring suggest environmental and/or hereditary factors.

Children born 1958–60 and who have died before 1961 are not included in the offspring cohort. This will have the effect of underestimating the risk of cancer. However, a very conservative estimate is that 0.3 offspring would have been diagnosed with any type of cancer and died during these years.

Applicators smoke less than the general population and this has been used as an explanation for their lower incidence of cancer compared to the general population. Less passive smoking could contribute to the lower risk of cancer in their offspring.

Another question is whether the children are exposed enough to increase the cancer risk and whether it is feasible to use “license for pesticide application in agriculture” as a proxy for pesticide exposure, since neither the applicators nor their children have an increased cancer risk. There is no individual information of pesticide use in the pesticide applicator cohort. A random sample of 268 applicators was studied with reference to pesticide trade name, application days per year and years, application method, wearing protective clothing, tobacco habits, and occupational history during the 1950s, 1960s, and 1970s.²⁸ That study showed that few of the applicators worked full time applying pesticides. The most common occupations were in agriculture and/or forestry (70%), and in horticulture (10%). The remainder worked in a variety of occupations. The majority had used pesticides, some already during the 1950s. Wearing protective clothing has, despite the special education, been only slightly more frequent than among farmers in general, and as late as in the 1970s about 50% never or seldom wore mask, glasses, or dress when applying pesticides.

In Sweden, the sales of phenoxy acid herbicides and other pesticides increased steeply from the introduction in the late 1940s until the middle of the 1970s.^35,36 Licensed pesticide applicators and farmers are the occupational groups most exposed to pesticides. Since 1965 it has been mandatory to have a licence for handling the most acutely toxic pesticides in Sweden. A cohort of 20 245 pesticide applicators with licences issued between 1965 and 1976 was established earlier. Cancer risks in the cohort have been studied and reported in several papers.^{26–28,37,38} Around 70% of the pesticide applicators were occupied in agriculture and the majority had used pesticides, some as early as in the 1950s.

There are some advantageous features specific for this study. The large cohort of pesticide applicators is unique and consists of the occupational group most exposed to pesticides in Sweden. Another advantage is that the Swedish Cancer Registry is population based, with high quality and coverage.^1,39,40

Another advantage of this cohort study is that it is free from recall bias. Many studies of childhood cancer and pesticide use suffer from methodological limitations. Most of the previous epidemiological investigations have been case-control studies with few exposed cases and the possibility of recall bias.

A disadvantage of the study is that individual confounding factors and detailed pesticide exposure are not available, either for the pesticide applicators or their offspring.

“License for pesticide application in agriculture” was used as surrogate for pesticide exposure. However, the cancer risks associated with pesticide exposure in the general population would not be higher than among pesticide applicators and their offspring.

To further investigate a possible relation between paternal exposure to pesticide and cancer risk in their offspring, we plan to extend the cohort by adding pesticide applicators licensed in year 1977 and onwards by using information from The Swedish Board of Agriculture.