Intended for healthcare professionals


Perinatal characteristics in relation to incidence of and mortality from prostate cancer

BMJ 1996; 313 doi: (Published 10 August 1996) Cite this as: BMJ 1996;313:337
  1. Anders Ekbom, associate professora,
  2. Chung-cheng Hsieh, professorb,
  3. Loren Lipworth, research fellowc,
  4. Alicja Wolk, associate professora,
  5. Jan Ponten, professora,
  6. Hans-Olov Adami, professora,
  7. Dimitrios Trichopoulos, professorc
  1. a Departments of Cancer Epidemiology and Pathology, Uppsala University, S-751 85 Uppsala, Sweden
  2. b Cancer Center, University of Massachusetts Medical Center, Worcester, MA, USA
  3. c Department of Epidemiology and Center for Cancer Prevention, Harvard School of Public Health, Boston, MA, USA
  1. Correspondence to: Dr Ekbom.
  • Accepted 1 June 1996


Objective: To test the hypothesis that factors causing morbidity and mortality from prostate cancer may operate in utero.

Design: Matched case-control study of singleton men born between 1874 and 1946 at one hospital.

Setting: Uppsala University Hospital.

Subjects: 250 patients with prostate cancer and 691 controls, including 80 patients who died from prostate cancer and their 196 matched controls.

Main outcome measures: Mother's age at menarche, parity, pre-eclampsia or eclampsia before delivery, age at delivery and socioeconomic status; case or control's birth length and weight, placental weight, prematurity derived from gestational age, and presence of jaundice.

Results: Both pre-eclampsia (odds ratio 0, 95% confidence interval 0 to 0.71) and prematurity (0.31, 0.09 to 1.04) were inversely associated with incidence of prostate cancer. Among subjects born full term, placental weight, birth weight, and ponderal index (weight/height*RF 3*) showed non-significant positive associations with prostate cancer incidence, and stronger associations with mortality.

Conclusion: Prenatal exposures that are likely correlates of pregnancy hormones and other growth factors are important in prostate carcinogenesis and influence the natural course as well as the occurrence of this cancer.

Key messages

Key messages Recent evidence suggests that events in utero may affect the risk of prostate cancer later in life

In this study 250 men born since 1874 at one hospital and who developed prostate cancer were identified and compared with 691 controls

Pre-eclampsia and prematurity were associated with a reduced risk of prostate cancer and ponderal index was associated with an increased risk of mortality from prostate cancer

Prenatal exposures that are likely correlates of pregnancy hormones and other growth factors seem to be of importance in the prostate carcinogenesis


Prostate cancer comprises about 8% of all new cancers among men globally,1 and the incidence seems to be increasing in many countries and ethnic groups.2 Although this increase may be partly due to increased detection rates,3 a genuine increase in the incidence of clinically aggressive cancers is likely since mortality from the disease is also increasing.2 The highest incidences have been reported in North America and northern Europe and the lowest in south east Asia.1 4 In the United States, incidence and mortality are substantially higher among African-Americans at all ages.4

Evidence exists that early life events have a role in prostate cancer, although it is not as strong as for breast cancer.5 The incidence of prostate cancer increases in the first generation after migration to areas where it is more common6 and recent epidemiological studies also suggest a link with early life.7 8 Intrauterine factors have already been associated with testicular cancer9 and breast cancer.10 11 Some recent reports have addressed the possible role of the intrauterine environment in prostate cancer. Ross and Henderson suggested that the intrauterine period may be aetiologically relevant to prostate cancer12; a small study of 21 prostate cancer cases indicated a positive relation between high birth weight and the risk of prostate cancer13; and molecular researchers have postulated that early life genomic events may be essential for the subsequent occurrence and progression of adult cancer,14 15 leading to the inference that early life events could have a more powerful influence on mortality than incidence.16

As part of a programme to evaluate intrauterine and perinatal influences on the occurrence of hormone related cancers we have used databases in Sweden that allow linkage of pregnancy and birth records since the last century with national cancer and death registry records on prostate cancer.

Subjects and methods

Because there is no private inpatient treatment in Sweden, hospital medical services are population based and linked to the county in which the patient lives. We identified for study all men who were born at the University Hospital in Uppsala, resident in Uppsala county since 1 January 1947, and were alive in 1958 using the unique 10 digit national registration numbers introduced for all Swedish citizens in 1947.17 The first six digits provide birth date (year, month, and day). The seventh and eighth digits provide information on county of residence.

All newly diagnosed malignant tumours in Sweden must be reported by both the diagnosing physician and the pathologist or cytologist to one of six regional cancer registries, which pass data to the national cancer registry established in 1958.18 At the time of this study the national cancer registry was complete up to 31 December 1991, and the Uppsala regional cancer registry was complete up to 31 December 1994.

All men in the national and regional cancer registry who had prostate cancer (ICD-7 code 177) and the code for Uppsala county in their national registration number were included in the study. Cases could have been missed if a man born at the University Hospital moved out of the county before 1947 and subsequently developed prostate cancer.

From the parish in which each patient lived when prostate cancer was diagnosed, we were able to establish his place of birth and family name at birth. At this stage, those who were not born at the University Hospital were excluded. Two hundred and fifty cases of prostate cancer were identified among singletons in the study cohort. We used the Swedish death registry to identify patients who were still alive at the time of censoring the follow up information (December 1994), thus enabling us to evaluate both incidence of and mortality from prostate cancer.

For each case, singleton males born live to the first four or more mothers who were admitted after the case's mother were selected as potential controls. This insured that one individually matched control was alive at the time prostate cancer was diagnosed in the case. We used the cancer and death registries to check that the potential control man was alive and had not had prostate cancer diagnosed at the time of diagnosis or death of the case. A total of 691 controls were matched to the incident cases and 196 to the fatal cases.

For all cases and controls we manually abstracted from the standardised hospital maternity charts information on the mother's age at menarche, parity, pre-eclampsia or eclampsia before delivery, age at delivery, and socioeconomic status. We also recorded the following data on the child: singleton or twin, birth length and weight, placental weight, prematurity derived from gestational age, and presence of jaundice. The completeness of records exceeded 95% for every item used in this study.

We analysed the data using logistic regression conditional on the matching process.19 Different sets of models were generated for assessing incidence of and death from prostate cancer. For analysis of death rates we used only cases in whom prostate cancer was the cause of death and excluded controls who did not survive until the age of death of the corresponding case. In addition to the indicators of birth size (birth weight, birth length, placental weight) we used the ponderal index of body mass (kg/m3). This indicator is often used in neonatology to assess intrauterine growth,19 20 and it is believed to be particularly sensitive to growth during the last few weeks of gestation. For each of the birth size indicators that were alternatively introduced into the model to avoid problems of colinearity, the unit measure was set as the closest round number corresponding to one standard deviation increase. The level of significance was set at P<0.05 (two sided test).


Table 1 shows the distribution of cases of prostate cancer and controls according to birth cohort and age at diagnosis of all cases. The mean age at diagnosis was 70.5 years; 72% of the subjects were born before 1920.

Table 1

Age at diagnosis of all men who developed prostate cancer and those who died plus numbers of matched controls

View this table:

Table 2 gives the adjusted odds ratios for prostate cancer in relation to perinatal or maternal characteristics. Pre-eclampsia or eclampsia predicted a low risk of prostate cancer and was the only significant association. Prematurity and higher parity were non-significantly associated with low risk, and birth size and jaundice showed non-significant positive associations.

Table 2

Results of multiple logistic regression analysis for incidence of prostate cancer in relation to perinatal or maternal characteristics from a nested case-control study of 250 cases and 691 controls.* Values are numbers(percentages) unless stated otherwise

View this table:

Table 3 presents mutually adjusted odds ratios for death from prostate cancer in relation to perinatal or maternal characteristics. The effect of pre-eclampsia or eclampsia is no longer significant because of small numbers, but prematurity is significantly associated with reduced mortality from prostate cancer. All birth size indicators (with the exception of birth length, for which random measurement error is always proportionately high) were more strongly related to death from prostate cancer than to incidence, and the association with ponderal index was significant. The effects of both neonatal jaundice and parity were larger than for incidence of prostate cancer but still non-significant.

Table 3

Results of multiple logistic regression analysis for death from prostate cancer in relation to perinatal or maternal characteristics from a nested case-control study of 80 cases and 196 controls *

View this table:


This case-control study nested within an explicitly defined cohort, and as such its validity is identical to that of the underlying cohort. Thus, information bias could occur only by non-differential misclassification, which weakens the associations found. Moreover, selection cannot introduce bias because it is unrelated to ascertainment of outcome. Some controls may have had subclinical prostate cancer but the study base (population at risk) for clinical prostate cancer also includes people in this category. Whereas implications of chance variation can be and were measured, confounding requires the existence of risk factors and demonstration of associations of these risk factors with the exposures under consideration. The only established risk factors for prostate cancer are age and race, and both of them were accounted for through matching and conditional analysis for age and because there are few non-white Swedish citizens. There is no agreement about the quantitative importance of other suggested risk factors for prostate cancer, including fat intake,22 23 24 vasectomy,24 25 26 and physical activity,7 27 and no evidence that any of these is associated with perinatal characteristics.


Sex steroids may affect prostate cancer given their role in normal growth and function of the prostate.28 29 However, epidemiological studies of some steroid hormones (oestradiol, testosterone, androstenedione) and prostate cancer in humans have proved inconclusive.30 31 32 33 The activity of 5(alpha)-reductase (involved in the conversion of circulating testosterone to the biologically active dihydrotestosterone) is higher among white Americans than among low risk native Chinese,34 but its role has not been evaluated in an epidemiological study of prostate cancer.

Although we were primarily examining the effect of intrauterine influences on the incidence of prostate cancer, mortality may be more strongly related to perinatal variables than incidence.16 Our choice of exposure variables was constrained by availability, but several of them are thought to reflect intrauterine growth, which is partly modulated by steroid hormones.35 36 37 38 Ponderal index is an indicator of late stage intrauterine growth.20 Moreover, concentrations of oestrogens and other hormones are generally lower in women with preeclampsia or eclampsia,35 8 40 and there is some evidence that primiparous women have higher oestrogen concentrations in pregnancy than multiparous women.41 42 We found that prematurity and pre-eclampsia or eclampsia reduced the risk of prostate cancer and that birth size was more strongly associated with mortality than incidence. These findings agree with our hypothesis and support a role for sex hormones in prostate cancer.


The existence of associations between perinatal factors and prostate cancer indicate that early life is important in the long carcinogenesis process. It seems unlikely that the associations were confounded by postnatal exposures, and there is no evidence to suggest confounding. The substantially stronger associations of most of the perinatal characteristics with death from prostate cancer than with incidence further suggest that not only the future occurrence but also the biological features of the tumour are programmed in utero. This may explain why the strong geographical variations in incidence of prostate cancer are not mirrored by a similar pattern in the prevalence of subclinical disease detected at necropsy.43

Loeb and other investigators have postulated that an early phenomenon, perhaps affecting factors controlling genetic stability, is needed to explain the high mutation rate required for the accumulation of the genetic changes leading to overt cancer.14 15 44 According to this hypothesis, if clonal cells remain genetically unstable even after the appearance of cancer, perinatal determinants of the early event should be expected to forecast metastasis of, and death from, a particular cancer even better than its incidence. This theory provides a possible explanation for our epidemiological findings. It is unclear how high concentrations of growth hormones in utero would influence subsequent mutation rates, but it could be through enlargement of a pool of potential target cells for malignant transformation or cell proliferation.

Our working hypothesis and results are compatible with two of the most striking descriptive characteristics of the disease. Clinical prostate cancer and mortality from prostate cancer are substantially higher among Americans and Europeans than among Asians,1 4 and within the United States incidence of and mortality from the disease are substantially higher among black than among white Americans.4 9 Birth weight is lower among Japanese and Chinese than Americans,45 and pregnancy steroids are much higher among black women.9 Moreover, long term increases in the incidence of prostate cancer and mortality have been preceded by long term increases in weight and obesity of the newborn.

It may be stretching biologic plausibility that a cancer which appears clinically in elderly people would have an origin in utero. However, it has been shown recently that prostatic intraepithelial neoplasia is already quite common among men in their 20s.46 47

We thank Ann Almqvist, Ulrika Lund, and Monica Rundgren for abstracting the data.


  • Conflict of interest None.

  • Funding This study was supported by grants from the Swedish Cancer Society. LL received a training award in cancer epidemiology from the US National Institutes of Health/National Cancer Institute (T32CA09001-19A2).


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
View Abstract