MC Perrin Department of Psychiatry, School of Medicine, New York University, 550 1st Avenue, New York, NY 10017, USA and Department of Epidemiology, Mailman School of Public Health, Columbia University, 722 West 168th Street, New York, NY 10032, USA , MB Terry Department of Epidemiology, Mailman School of Public Health, Columbia University, 722 West 168th Street, New York, NY 10032, USA , K Kleinhaus New York State Psychiatric Institute, 1051 Riverside Avenue, New York, NY 10032, USA , L Deutsch Unit of Epidemiology, The Hebrew University-Hadassah School of Public Health, Ein Kerem, Jerusalem, 91120, Israel , R Yanetz Unit of Epidemiology, The Hebrew University-Hadassah School of Public Health, Ein Kerem, Jerusalem, 91120, Israel , E Tiram Unit of Epidemiology, The Hebrew University-Hadassah School of Public Health, Ein Kerem, Jerusalem, 91120, Israel , R Calderon Unit of Epidemiology, The Hebrew University-Hadassah School of Public Health, Ein Kerem, Jerusalem, 91120, Israel , Y Friedlander Unit of Epidemiology, The Hebrew University-Hadassah School of Public Health, Ein Kerem, Jerusalem, 91120, Israel , O Paltiel Unit of Epidemiology, The Hebrew University-Hadassah School of Public Health, Ein Kerem, Jerusalem, 91120, Israel , S Harlap Department of Psychiatry, School of Medicine, New York University, 550 1st Avenue, New York, NY 10017, USA and Department of Epidemiology, Mailman School of Public Health, Columbia University, 722 West 168th Street, New York, NY 10032, USA

Abstract

Background Diabetes is known to be associated with cancer of the pancreas, though there is some debate as to whether it is a cause or a consequence of the disease. We investigated the incidence of pancreatic cancer in a cohort of 37926 Israeli women followed for 28–40 years for whom information on diabetes had been collected at the time they gave birth, in 1964–1976, in Jerusalem. There were 54 cases of pancreatic cancer ascertained from the Israel Cancer Registry during follow-up.

Methods We used Cox proportional hazards models to adjust for age at baseline and explore effects of other risk factors, including ethnic groups, preeclampsia, birth order and birth weight of offspring.

Results We observed no cases of pancreatic cancer in the women with insulin dependent diabetes; however, there were five cases in the women with gestational diabetes. The interval between the record of diabetes in pregnancy and the diagnosis of pancreatic cancer ranged from 14–35 years. Women with a history of gestational diabetes showed a relative risk of pancreatic cancer of 7.1 (95% confidence interval, 2.8–18.0).

Conclusion We conclude that gestational diabetes is strongly related to the risk of cancer of the pancreas in women in this population, and that gestational diabetes can precede cancer diagnosis by many years.

Background

Cancer of the pancreas is the fourth highest cause of death from cancer among women in the US [1]. It is generally diagnosed at an advanced stage. Only a small proportion of tumors can be surgically resected [2], and many are resistant to chemotherapy or radiation [2,3]. Thus, the lethality of pancreatic cancer is high, with the mortality rate (9.2/100000) approximating the incidence rate (9.5/100000) among women [4]. Diabetes is well known to be associated with pancreatic cancer [5,6]. There has been a long-standing debate, however, as to whether this is a consequence or an antecedent of the pancreatic tumor; evidence exists supporting both views [3]. On the one hand, patients with newly diagnosed pancreatic cancer frequently have type 2 diabetes mellitus of recent onset; removal of the tumor often ameliorates its manifestations [3]. On the other hand, individuals with long-standing diabetes have also been shown to be at increased risk of pancreatic cancer [3].

During pregnancy, women become progressively more insulin resistant as a result of weight gain and release of placental hormones [7]. While most women are able to compensate with increased secretion of insulin and experience only minor changes in plasma glucose levels, those with gestational diabetes mellitus are unable to compensate for the increased resistance and become hyperglycemic [7]. Risk factors for gestational diabetes mellitus include older age, family history of diabetes and high body mass index (BMI) and ethnicity [8]. In the US, the prevalence of gestational diabetes mellitus is approximately 7% [9], though it varies by ethnic group. It is more common among African-Americans, Hispanics, Asians and Native Americans than among non-Hispanic Whites [10-14]. Short-term consequences include fetal macrosomia and other neonatal morbidities; long term sequelae place the mother and offspring at increased risk of type 2 diabetes mellitus [9]. A few investigators have studied gestational diabetes mellitus and gestational glucose intolerance as risk factors for breast cancer and other cancers [15,16] but none, to our knowledge, have investigated gestational diabetes mellitus in relation to pancreatic cancer.

MethodsThis study relies on an ongoing population-based cohort study derived from births, with follow-up till the present day of all offspring and their parents. The Jerusalem Perinatal Study recorded all 92408 births in 1964–1976 to residents of a defined geographic area. Subsets of mothers were interviewed in 1965–1968 (N = 11467 births) in antenatal clinics and in 1974–1976 (N = 16912 births) 1–3 days after birth [17]. The present analysis focuses on the mothers of the 84781 offspring born in the three largest obstetric units, where the study included active surveillance of maternal and obstetric conditions; data on maternal and obstetric information was copied from the labor ward log at the time of birth using separate rubrics in the Perinatal Study's pre-coded forms [17] that allowed for a record of maternal "diabetes" (presumed to be insulin-dependent juvenile diabetes or type 1) and "pre-diabetes", corresponding, approximately, to gestational diabetes mellitus. In that era, all pregnant women were screened for glycosuria at each antenatal visit; if found positive, they would be referred for an oral glucose tolerance test.

In 2004–2005, using the national identity numbers that are assigned to citizens of Israel, we traced and ascertained the vital status of 40898 mothers in this cohort through linkage with Israel's National Population Registry. Then, we linked the cohort to the Israel Cancer Registry. The Israel Cancer Registry, which was started in 1961, is 95.7% complete for pancreatic cancer [18]. Names, identity numbers and other identifying information were removed from the file that was analyzed collaboratively in New York and Israel. The study was approved by the Institutional Review Boards of both participating institutions.

Statistical analyses

We included all cases of first primary malignant pancreatic cancers, as defined by the International Classification of Diseases for Oncology, 3rd edition (site code: C25 and fifth digit morphology code of 3) that were diagnosed between the first observed birth and 31 December 2004. We used Cox proportional hazards models to estimate the relative risk (hazard ratio) of pancreatic cancer in women with a diagnosis of gestational diabetes mellitus in any pregnancy compared to women without a diagnosis of gestational diabetes mellitus during the entire study period (1964–1976) using the PHREG procedure available in SAS 9.0 (SAS Institute Inc, Cary, NC, USA). Women were followed from the "baseline birth" (i.e., first time they gave birth in 1964–1976) until death, date of diagnosis of any cancer, or end of follow-up period (31 December 2004).

Variables included in the model were those that altered the age-adjusted estimate of the relative risk by more than 10%. We also tested whether a covariate for time period (year of first observed birth) affected the estimate, or was predictive of pancreatic cancer risk. As it did not effect the estimate and did not predict pancreatic cancer risk, we did not include the term in the final model. Age at the first observed birth was treated as a continuous variable, missing values (N = 50) being assigned to the mean (26.2 years). Other variables were treated as dichotomies, or sets with a dummy, coded 1 (if present) or 0 (if absent). Variables tested included categories of birth order at the last observed pregnancy, ethnic ancestry based on the woman's father's place of birth (Israel, other Western Asia, North Africa and Europe etc. – the latter including the Americas, sub-Saharan Africa and Australasia; no information was available for the origin of the woman's mother), social class (based on husband's occupation at last observed birth), categories of education, presence of other specific complications of pregnancy in any observed birth, and birth defects, low (< 2.5 kg) or high (4.0 kg or more) birth weight in one or more offspring. Unless otherwise stated, categories of missing data in other variables (most affected less than 0.1% of women) were included in the reference groups. The results are presented as relative risks, adjusted for age or for other variables, with 95% confidence intervals.

Numbers and exclusions Of the 40898 women traced, 37980 (92.9%) delivered at least once in one of the three largest hospitals in Jerusalem, where complete diabetes information was collected for the cohort. Untraced women were similar in age at first observed birth, more often unmarried and more likely to be of European ancestry. Untraced women had a lower prevalence of gestational diabetes mellitus (0.4%) compared to a prevalence of 1.0% among women who were traced. Among women who were successfully traced, the incidence of pancreatic cancer was not significantly related to hospital of birth. We also excluded 41 women who were diagnosed with various malignancies prior to their first observed birth in the study, and an additional 13 women (none with pancreas cancer) who were diagnosed with gestational diabetes mellitus in one pregnancy and type 1 diabetes mellitus in another pregnancy during the study period.