1 Department of Public Health/Health Information Dynamics, Nagoya University Graduate School of Medicine, Nagoya, Japan.
2 Department of Preventive Medicine/Biostatistics and Medical Decision Making, Nagoya University Graduate School of Medicine, Nagoya, Japan.
3 Department of Public Health, Aichi Medical University School of Medicine, Nagakute, Japan.
4 Department of Public Health, Wakayama Medical University, Wakayama, Japan.
5 Department of Public Health, Showa University School of Medicine, Tokyo, Japan.
6 Department of Epidemiology, Research Institute for Radiation Biology and Medicine, Hiroshima University, Hiroshima, Japan.
7 Department of Clinical Epidemiology, Institute of Industrial Ecological Science, University of Occupational and Environmental Health, Kitakyushu, Japan.
8 Japan Collaborative Cohort Study for Evaluation of Cancer Risk.
Correspondence: Takaaki Kondo, Department of Public Health/Health Information Dynamics, Nagoya University Graduate School of Medicine, 65 Tsurumai-cho, Showa-ku, Nagoya 466-8550, Japan. E-mail: taka-ngy{at}umin.ac.jp
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Methods We used the baseline data from the Japan Collaborative Cohort Study for Evaluation of Cancer Risk (JACC Study), which was initiated during 19881990 in Japan. Association of the cancer history of the subjects parents with that of the subjects themselves and any of the subjects siblings was evaluated with odds ratios (OR) by the crude and generalized estimating equations (GEE) technique for four sites: stomach, colorectum, liver, and lung/bronchus.
Results The aggregation of a history of stomach cancer between parents and their offspring was evident with significant OR >2.5. The magnitude of the parentoffspring association of a disease history of the colorectum and liver was found to be greater than that for stomach cancer. Conversely, lung and bronchus cancer failed to demonstrate a significant aggregation.
Conclusions The hereditary and environmental influences shared by parents and offspring are likely to play a strong aetiological role in colorectal or liver cancer versus a weaker but still significant role in stomach cancer. In contrast, the aetiological role of familial predisposition to lung cancer was indeterminate, which suggests a predominant role of non-familial factors in the development of lung cancer.
Accepted 13 February 2003
The familial aggregation of stomach cancer has long been argued. As early as 1959, Macklin found that gastric cancer was unduly frequent in the families of gastric-cancer probands, whereas husbands and wives were not affected with gastric cancer more often than would be expected on the basis of random distribution.1 His conclusion stressed the importance of the genetic similarity, rather than the environmental one, between the members of the affected pairs as an explanation for the increased frequency. Studies suggesting the familial aggregation of stomach cancer, however, seem to have been of little interest in Western countries, probably because of its low morbidity rate relative to cancers at other sites in Western populations. In contrast, stomach cancer remains high in both mortality and morbidity in Japan, accounting for 17% of all deaths from malignancy. Recently, consistent results from a number of studies conducted in Japan have indicated that stomach cancer tended to aggregate among members of the patients family.2,3 Lately, we have found that a positive history of stomach cancer in one or more first-degree relatives was associated with a significantly increased risk of death from the disease in both men and women.4
In addition to stomach cancer, familial aggregation of cancer at other concordant sites has also been elucidated.5 Based on the family history information from a large-scale cancer registry database, elevated rates of site-specific cancer among family members of affected patients have been revealed for cancers of the breast, colon and rectum, and stomach. Moreover, a clear correspondence in cancer history between husbands and wives was demonstrated,6 which suggests the contribution of the long-term environmental influences shared by family members to familial aggregation of cancer.
Notably, the magnitude of the familial aggregation is known to vary from site to afflicted site. According to the study of Ogawa et al.,2 stomach cancer is less likely to aggregate among family members than cancers of the colon and rectum, or breast. In order to verify the presence of a site-specific aggregation of cancer history between parents and their offspring, with special reference to the magnitude of the association between stomach and other cancers, we attempted to analyse the dataset from a cohort study to explore the possible relationship of the cancer history of parents with that of their offspring. Our approach can be compared with that of reported studies in that we used the baseline data from a large-scale, multi-centre, population-based study which had been developed to evaluate factors related to cancer mortality.7
![]() |
Subjects and Methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
A standardized questionnaire was used to obtain basic demographic characteristics in addition to details on smoking, drinking and dietary habits, history of previous illness, and history of disease in parents and siblings. If any history of cancer was identified, the affected site was coded according to the Ninth Revision of the International Classification of Diseases (ICD-9). Cancer history in the parents was defined as present if the questionnaire stated that either the subjects father or mother (blood-related) had been affected, whereas a history of cancer in the offspring was confirmed when the subject, or any of subjects brothers or sisters (blood-related), had a history of the disease. Since our study was community-based, multiple counting of the same parents was likely to occur when any one of a subjects brothers or sisters was also a subject of this study, and could eventually jeopardize the independence of measurement. To avoid this potential problem, we began by grouping the subjects into four subpopulations according to their birth order; eldest, second eldest, third eldest, and the remainder. By conducting the subsequent statistical analysis on a subpopulation basis, every history of a subjects parents was given equal weight as a single count.
![]() |
Statistical analyses |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
In the second approach, we employed a multivariate logistic model in which the disease history of the mother and father was treated as a function of the status of the disease history of the parents subjects and siblings. Let Y1, Y2, and Y3 denote the history status of the subject, subjects brother (if present), and subjects sister (if present), respectively, with Yj = 1 if case and Yj = 0 if control. Besides, let Covar denote the vector of covariates for the subject. The multivariate model we adopted can be expressed by two ordinal logistic regression equations with shared coefficients:
![]() |
![]() |
In our analysis, some study areas were not included because of the total absence of data on the past or family history in each area, and individuals with unknown birth order were excluded in the present study as well, which reduced the number of eligible subjects to 79 540. To assure sufficient sample size for each subpopulation, we restricted the application of statistical procedures to three major categories of the birth order; eldest, second eldest, and third eldest. As a statistical software package, we ran the Statistical Analysis System (SAS) release 6.12 installed at the Nagoya University Computation Centre. The model fitting was performed using the freq procedure for the estimation of univariate OR and the genmod procedure for GEE to adjust for the correlation of observations within families.10
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
A previous history of lung cancer was unlikely to aggregate between parents and offspring as the OR failed to indicate statistical significance, although the number of offspring with a previous disease history was too small to definitively establish that conclusion.
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
In our first statistical approach for the estimation of crude OR, the usual case-control analysis was done to examine the association of disorder status of the parents with the presence of the disorder in their offspring. Overestimations of the familial aggregation possibly attend this measure11 because information was reduced by pooling both parents into a single observation unit. The subject, the subjects brother, and subjects sister were also clustered into a single unit. To overcome the bias toward overestimation, we fitted a multivariate logistic model in which the disorder status of each of the parents (father or mother) was predicted by the status of each of the offspring (subject, subjects brother, or subjects sister) with intrafamilial correlations accounted for by the GEE technique. Besides, the effect of the subjects age was considered in the model as a covariate. We found that these two methods yielded explicitly similar results in estimation of OR. This finding suggests that the overestimation of familial aggregation resulting from the clustering of the relatives information on the disease history is not as substantial as anticipated.
The major finding in our study is that a familial aggregation between parents and offspring was observed for three concordant sites, i.e. stomach, colorectum, and liver, in addition to all-site cancer. The fact that this finding was consistently observed across all three categories of subjects birth order serves to underscore our conjecture that a family history of cancer among parents significantly increases the risk of cancer morbidity at some concordant sites among their offspring.
The OR observed in the range 2.52.6 for the aggregation of a history of stomach cancer somewhat approximate the figure in our previous paper,3 where we found the familial aggregation of stomach cancer to be less prominent than that of stroke, hypertension, tuberculosis, or diabetes mellitus. While the OR for familial aggregation varied from site to site of each cancer, some similarities were evident between our result and the earlier one reported by Ogawa et al.; i.e. stomach cancer is lower than cancers of the colorectum or liver in the magnitude of OR for the aggregation of disease history. This consistency in the relative magnitude of site-specific aggregations could reflect a difference in the aetiological contribution of shared hereditary and environmental influence. These shared factors may be less determinant for the familial aggregation of stomach cancer than for colorectal and liver cancers.
Lung and bronchus cancer failed to demonstrate any significant evidence of a parentoffspring aggregation, though the limited number of observed patients is likely to compromise the validity of such a conclusion. The lack of clear evidence for familial aggregation of lung cancer is consistent with the previous literature,2 while a case-control study indicated an increase in the risk of lung cancer in young adults but not in an older group,12 suggesting differences in the risk factors between younger and older people. Although ages at which the history of illness occurred were not available in our data, taking into consideration the mean age of the subjects, it is likely that a great proportion of the parents with lung cancer were affected at advanced ages. Presumably, acquired susceptibility due to the effects of some risk factors not shared by family members may outweigh any shared effects in the development of lung cancer in later life. In particular, smoking is a predominantly non-familial risk factor for lung cancer, and we have previously shown that the effect of smoking on lung cancer mortality continues into later life even after cessation of the smoking habit.13
A mechanism to explain the familial aggregations we observed remains to be elucidated. Mutations of some genes have been proven to confer a great deal of lifetime risk in familial settings.14 These include Rb in retinoblastoma, WT1 in Wilms tumour, and p53 in Li-Fraumeni syndrome. However, this type of mutation is estimated to be quite uncommon, e.g. inherited mutations in p53 accounting for less than 1% of breast cancer patients, even at young ages.15 More frequent types of gene variants classified as a genetic polymorphism are thought to carry high population risks.16 By influencing metabolic activation, detoxification, or the elimination of carcinogens, the interaction between low-penetrant gene susceptibility and environmental factors shared both by the parents and offspring is likely to contribute to the occurrence of a familial aggregation of cancer. Some examples of the genecarcinogen interaction in relation to gastrointestinalcancer are heterocyclic amines and CYP1A1, alcohol and the ADH phenotype, and so on.
Our study is subject to some limitations associated with information bias, including among others, Japans medical culture. This discourages physicians from informing cancer patients of their true diagnosis because of fear of the incurable nature of the disease, with the result that measurement errors related to under-reporting are likely to ensue. By classifying the subjects and their siblings into the single category of offspring, we attempted to reduce the consequence of this bias, though the eventual effects of such a countermeasure for the bias could not be tested.
Various recall biases in relation to the previous history of the subjects siblings or parents may well be involved. Since individuals with a parental cancer history became more conscious of the disease than those without one, they might be prone, at least subconsciously, to evoke the recollection of such a family history, which can result in the differential misclassification of the existence of the history. Another problem concerns the ICD-9 coding of the disease. As we could not ascertain a diagnosis in any cancer history self-reported by the subjects, the misclassification or omission of the ICD-9 codes could not be ruled out, which also renders our results less convincing. In this respect, a study to confirm the reliability of the particulars of a given previous and family history is underway.17
In summary, our analysis revealed that a history of stomach cancer among parents is associated with an increased risk for the same cancer among their offspring. When OR were calculated to quantify the risk association for some affected sites, stomach cancer had a lower OR than cancers of the colorectum and liver, in agreement with the previous literature. Our result suggests the importance of family-shared factors whose aetiological influence is likely to differ substantially from site to site of cancer history.
KEY MESSAGES
|
![]() |
Acknowledgments |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
2 Ogawa H, Kato I, Tominaga S. Family history of cancer among cancer patients. Jpn J Cancer Res 1985;76:11318.[ISI][Medline]
3 Toyoshima H, Hayashi S, Hashimoto S et al. Familial aggregation and covariation of diseases in a Japanese Rural Community: comparison of stomach cancer with other disease. Ann Epidmiol 1997;7:44651.[CrossRef]
4 Yatsuya H, Toyoshima H, Mizoue T et al. Family history and the risk of stomach cancer death in Japan: differences by age and gender. Int J Cancer 2002;97:68894.[CrossRef][ISI][Medline]
5 Hemminki K, Vaittinen P. Effect of paternal and maternal cancer on cancer in the offspring: a population-based study. Cancer Epidemiol Biomarkers Prev 1997;6:99397.[Abstract]
6 Kato I, Tominaga S, Suzuki T. Correspondence in cancer history between husbands and wives. Jpn J Cancer Res 1990;81:58489.[ISI][Medline]
7 Aoki K. Research activities of epidemiology in Japan. Report by the Research Committee of the Ministry of Education, Science, Sports and Culture on Evaluation of Risk Factors for Cancer. J Epidemiol 1994;6:107S13S.
8 Ohno Y, Tamakoshi A. Japan Collaborative Cohort Study for Evaluation of Cancer Risk sponsored by Monbusho (JACC Study). J Epidemiol 2001;11:14450.[Medline]
9 Hudson JI, Laird NM, Betensky RA. Multivariate logistic regression for familial aggregation of two disorders. I. Development of models and methods. Am J Epidemiol 2001;153:50005.
10 Allison PD. Logistic Regression Using the Sas SystemTheory and Application. Cary, NC: SAS Institute Inc, 1999.
11 Khoury MJ, Flanders WD. Bias in using family history as a risk factor in case-control studies of disease. Epidemiology 1995;6:51119.[ISI][Medline]
12 Kreuzer M, Kreienbrock L, Gerken M et al. Risk factors for lung cancer in young adults. Am J Epidemiol 1998;147:102837.[Abstract]
13 Wakai K, Seki N, Tamakoshi A et al. Decrease in risk of lung cancer death in males after smoking cessation by age at quitting: findings from the JACC study. Jpn J Cancer Res 2001;92:82128.[ISI][Medline]
14 Sinha R, Caporaso N. Diet, genetic susceptibility and human cancer etiology. J Nutr 1999;129:556S59S.[ISI][Medline]
15 Sellers TA. Genetic factors in the pathogenesis of breast cancer: their role and relative importance. J Nutr 1997;127:929S32S.[Medline]
16 Perera FP. Environment and cancer: who are susceptible? Science 1997;278:106873.
17 Zhu SK, Toyoshima H, Kondo T et al. Short- and long-term reliability of information on previous illness and family history as compared with that on smoking and drinking habits in questionnaire surveys. J Epidemiol 2002;12:12025.[Medline]