Study Population and Measurements

The study population comes from the Metabolic syndrome and Cancer project (Me-Can) which includes cohorts with 578,700 participants from Norway, Austria and Sweden. In these cohorts, health examinations data have been collected on height, weight, blood pressure, blood levels of glucose, total cholesterol, triglycerides, and smoking status. Time period of data collection spanned from 1972 to 2006. A detailed description of Me-Can and inclusion criteria for participants in this study has been previously described [18].

Follow-up and Endpoints

Linkages have been performed with cause of death and vital status registries of the respective countries in order to identify those cases with incident gallbladder cancer (ICD-7): 155.1). Endpoints for the study were set at the date of the first cancer diagnosis, emigration, death, or December 31, 2003 (Austria), 2005 (Norway) and 2006 (Sweden).

Statistical Analysis

The statistical analysis of this study is similar to a previously published study by the same study group [19]. In brief, Cox proportional hazards regression models, with age as the time variable, were fitted to obtain hazard ratios, denoted as relative risks (RRs), of primary GBC incidence with 95% confidence intervals (95% CI). We did our main analyses with both sexes combined as there was no significant interaction between sex and each of the MetS factors. As in the previous publications of Me-Can studies, analyses were undertaken with exposures as quintiles, standardized z-score continuous variables as well as bi-categorical values using the WHO defined cut-off points of the determinant variables.

Quintile Analysis

Quintile cut-off points for the exposure variables were calculated within each cohort and sex. For glucose, cholesterol and triglycerides, cut-offs were additionally stratified by fasting time before blood sampling (>8 hours, fasting or ≤8 hours, non-fasting). The models were further stratified for the seven cohorts, sex and year of birth (five categories: ≤1929, 1930–39, 1940–49, 1950–59, and ≥1960), and adjusted for age, smoking status (three categories: never, former and current smokers) and for BMI where appropriate. The lowest quintile was used as a reference. Mean levels within the quintiles of exposure variables were used to test for linear trend.

Standardized z-score Analysis

In addition to the quintile analysis, we also performed statistical tests with the exposures on a continuous scale. Standardized scores let each determinant to be investigated in the same scale making a uniform comparison possible. We transformed the existing values to standardised variables (z-scores), with zero as mean and one as standard deviation (z = (x - μ)/σ). As in the quintile analysis, the transformation is stratified by cohort, sex, and fasting time. Skewed variables (glucose and triglycerides) were logarithmically transformed prior to standardisation. Metabolic syndrome (MetS) score was constructed by adding the individual z-scores, and further standardization of the resulting sum. The adjustments and the stratifications in the z-score analysis are the same as in the quintile analysis.

Analysis by WHO Standards

We also estimated risks in two categories according to cut-offs defined by WHO as follows: overweight (BMI 25–<30 kg/m²), obesity (BMI ≥30 kg/m²), hypertension (systolic blood pressure ≥140 mmHg and/or diastolic blood pressure ≥90 mmHg), impaired glucose tolerance (fasting glucose 6.0–6.9 mmol/l), diabetes (fasting glucose ≥7.0 mmol/l), hypertriglyceridemia (fasting triglycerides ≥1.7 mmol/l), and hypercholesterolemia (fasting total cholesterol ≥6.2 mmol/l). For blood glucose and lipids only those individuals who had fasted >8 hours prior to blood draw were included [20]–[22]. The same adjustment and stratification scheme was used as in the models with quintile and continuous exposure variables.

Random Errors

All risk estimates were adjusted for random error in exposure measurements, based on data on repeated measurements from 133,820 participants with a total of 406,364 observations. These data were used to estimate regression dilution ratios (RDR) or regression calibration (RC) based on linear mixed effect models [23]–[25]. RRs derived from quintile and standardised z-score analyses were then corrected by dividing the regression coefficient in the Cox model by the estimated regression dilution ratio (RDR) of exposure. RRs from the z-scores analyses which adjusted for all individual metabolic factors in one model were corrected by regression calibration.

Further Analytic Considerations

Since reproductive factors are important risk factor for gallbladder diseases in women [3], [4], we did additional risk estimation separately for women <50 years of age (n = 214,572) and ≥50 years of age (n = 72,748) using this age-cut-off as a proxy for pre- and postmenopausal status, respectively.

Our main analyses excluded the first year after baseline measurements in order to account for possible reverse causality between exposures and event. We consolidated the issue by performing further lag-time analyses that excluded the first 3 years of follow-up.

Statistical analyses were performed in Stata (version 10.0, StataCorp LP, College Station, Texas) and R (version 2.7.2, used for random error correction).

Ethics

The study was approved by The Research Review Board of Umeå, Sweden, the Regional Committee for Medical and Health Research Ethics, Southeast Norway and the Ethikkommission of the Land Vorarlberg, Austria. Participants from Sweden and Austria provided written informed consent to participate in this study. In Norway, the participants were invited to come to the health survey and a questionnaire was sent together with the invitation. An attendance to the health examination where the participants delivered their filled in questionnaire, has been accepted by the Data Inspectorate as an informed consent, but not a written consent. Written consent was obtained from 1994 onwards.

Strengths and Limitations

The main strengths of our study are the large number of participants and its prospective design. We used data from population-based surveys in three countries, with almost complete coverage of data for measured exposure factors. The large number of repeated measurements within the study population allowed us to adjust for random error in the individual MetS factors. We also used high quality national registers in Austria, Norway, and Sweden for the follow-up regarding cancer diagnoses [26]–[28].

An important limitation of the study is the lack of data on gallstone status - a well established risk factor for GBC [3], [4], [29]. Gallstone status is also highly linked to the presence of metabolic risk factors like obesity, diabetes and dyslipidemia [17], [30]. With this constellation we cannot exclude the possible mechanistic role of gallstones in the association between metabolic risk factors and GBC as has been elegantly presented in previous prospective studies on obesity and diabetes [31], [32] as well as case-control studies [15]–[16]. However, it is also evident that a considerable proportion of individuals with GBC show no sign of cholelithiasis [3], [4], signifying the presence of other factors that may play important role gallbladder carcinogenesis.

Another limitation of the study is the noticeably small number of events, despite the large number participants, which might have contributed to the large confidence intervals seen especially in the quintile analyses. Lack of information on the reproductive history [3] among women may also be a limitation. However, we did a further risk estimation for women <50 and ≥50 years of age as a proxy for pre- and postmenopausal age groups, respectively. Our study is also limited by lack of data on socioeconomic status as well as some other behavioural aspects like alcohol consumption and physical activity. Moreover we lack data on treatment history of the abnormal metabolic factors like hypertension and dyslipidemia which might, to some extent, have confounded our risk estimate. For the most recent definition of the MetS [5] we lack data on specific factors like waist circumference and high density lipoprotein cholesterol which we had to replace with BMI and total cholesterol respectively. Consequently we have presented our results based on a metabolic syndrome score which we used as a proxy for the syndrome [33].

Comparisons with the Literature

The observed significant association between metabolic syndrome and GBC in this study with large number of participants and adequately long follow-up period strengthens reports by a previous case-control study based on over 600 biliary tract cancers which also found significant role of Mets on GBC [17]. However, such studies may be limited by the timing of exposure and outcome not being able to exclude reverse causality. This may probably be the case with the lack of association between waist circumference and risk of GBC. Such deficiencies are better dealt with prospective studies of long follow up period. Although several prospective studies exist on individual metabolic factors [31], [32], literatures on the association between MetS as an entity and GBC are scarce [13]. In a previous prospective study [13] MetS was defined as simultaneous exposure to antihypertensive, hypoglycaemic and hypolypemic treatments which is a rather rough approximation of MetS. Compared to ours this previous study did not find statistically significant association between MetS and GBC which is a rather questionable finding in the face of evident significant association with most of the individual components.

Although several independent mechanisms are depicted to underpin the association between obesity and cancer, the mechanisms that link obesity with gallbladder cancer risk are unclear [34]–[36]. Many studies have identified obesity as a risk to be more pronounced in women, and suggested a possible role of sex hormones (mainly oestrogen) in the pathogenesis of GBC [8], [13], [29], [36]–[41]. There are even indications that sex hormone receptors exist on the tumour tissue [42]. Our results, that increasing BMI poses greater risk mainly in younger women of premenopausal age group, might be supportive of this mechanism. This finding is supported by another cohort study in Norway [13].

In tumours that depend on oestrogen for their growth, like breast and endometrial cancers, obesity is shown to be of greater risk in postmenopausal women [43]–[46]. In our study, however, the risk for GBC was higher for younger women below the age of 50. This observation might be due to chance, as we did not find a significant interaction between age and BMI. However, if confirmed in further studies, the clinicopathological mechanisms may be entirely different for GBC.

Blood glucose levels were shown to be associated with incidence of cancer overall and in several specific sites like the colon, pancreas, liver, and endometrium in previous studies [11], [28], [47]–[48]. Studies reporting specifically a link between GBC and blood glucose levels are almost inexistent [28]. The association between glucose and cancer risk in our study remained after adjustment for major putative confounders like BMI, smoking and age, indicating a possible causal link. However, biological mechanisms in the association between blood glucose and cancer are poorly understood. A large case-control study by associated with high blood pressure, none of these studies identified blood pressure as risk for GBC.

A case-control study on serum lipids and biliary tract cancers including gallbladder cancer showed that compared to controls, cases had significantly higher mean levels of serum triglyceride (STG) [15]. However, our study, based on prospective data analyses as well as other similar cohort studies did not confirm this finding [48], [52]. In the study by Andreotti et al serum measurement took place shortly after cancer diagnosis. In this constellation one cannot rule out a possible reverse causation due to disease effect [53].

In conclusion, our study showed that increasing BMI and blood glucose levels are possible risk factors for GBC. Obesity was seen to pose a greater risk among women in the premenopausal age. Beyond the individual factors, the results of our study show that the metabolic syndrome as an entity presents a risk constellation for the occurrence of gallbladder cancer. Considering the rise in temporal trend of BMI and blood glucose levels [48], [54], we would anticipate that the incidence ofShebl et al indicated that although diabetes could be a risk factor for gallstone formation, the association between diabetes and GBC can be explained only partly by the positive association between diabetes and gallstones [14].

The inverse association we observed between total cholesterol and GBC in women may be largely due to preclinical effects of the cancer on total serum cholesterol [49]. A lag-time sub-analysis excluding 3 years of follow up after baseline measurement, rendered the association non-significant although the direction of association persisted. This was also shown in another recently published Me-Can study on total serum cholesterol and cancer [50].

Studies on the association between blood pressure and GBC incidence are scarce [48], [51]. Although it was shown that several cancer sites might be significantly GBC might also increase.