SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Appendix 1

Although a growing body of evidence suggests a link between diabetes and cancer, it is not clear whether diabetes independently increases the risk of cancer. We conducted a comprehensive assessment of the association between pre-existing diabetes and total and site-specific cancer risk based on a pooled analysis of eight cohort studies in Japan (>330 000 subjects). We estimated a summary hazard ratio by pooling study-specific hazard ratios for total and site-specific cancer by using a random-effects model. A statistically increased risk was observed for cancers at specific sites, such as colon (hazard ratio; HR = 1.40), liver (HR = 1.97), pancreas (HR = 1.85) and bile duct (HR = 1.66; men only). Increased risk was also suggested for other sites, and diabetes mellitus was associated with an overall 20% increased risk in total cancer incidence in the Japanese population. The association between these two diseases has important implications for reiterating the importance of controlling lifestyle factors and may suggest a possible strategy for cancer screening among patients with diabetes. Studies continuously investigating the risk factors for diabetes are also important.

In Japan, as in other countries, the increasing prevalence of diabetes presents a serious public health problem. The estimated numbers of persons with diabetes in 1997, 2002 and 2007 were 6.9 million (prevalence 5.5%), 7.4 million (prevalence 5.4%) and 8.9 million (prevalence 7.1%), respectively.[1]

A growing body of evidence suggests a link between diabetes and cancer. A recent meta-analysis showed that people with diabetes are at elevated risk for cancers of the liver,[2] biliary tract,[3] pancreas,[4] stomach,[5] colorectum,[6] kidney,[7] bladder,[8] breast[9] and endometrium,[10] but at decreased risk for prostate cancer.[11] Research suggests that hyperinsulinemia acting through aberrations in the insulin-like growth factor pathways or steroid hormone metabolism is involved in mitogenic actions. However, whether diabetes independently increases the risk of these cancers or whether cancer and diabetes simply share common risk factors, such as obesity or physical inactivity, is not clear; many of the studies included in the meta-analysis did not necessarily control for other lifestyle factors. Moreover, evidence from other cancer sites and the impact of diabetes on total cancer have not been elucidated completely.[12-14]

In the present study, we conducted a comprehensive assessment of the association between pre-existing diabetes and total and site-specific cancer by means of a pooled analysis of eight cohort studies in Japan (>330 000 subjects).

Patients and Methods

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Appendix 1

Study population

In 2006, the Research Group for the Development and Evaluation of Cancer Prevention Strategies in Japan initiated a pooling project using original data from major cohort studies to evaluate the association between lifestyle and major forms of cancer and mortality in Japanese people. The following a priori inclusion criteria were set for the present purpose: the study had to be a population-based cohort study conducted in Japan starting in the mid-1980s to the mid-1990s, it had to include >30 000 participants, it had to have collected information on the history of diabetes in a questionnaire at baseline, and it had to have collected cancer incidence data during the follow-up period. Eight ongoing studies that met these criteria were identified: (i) the Japan Public Health Center-Based Prospective Study, Cohort I (JPHC-I);[15] (ii) the Japan Public Health Center-Based Prospective Study, Cohort II (JPHC-II);[15] (iii) the Japan Collaborative Cohort Study (JACC);[16] (iv) the Miyagi Cohort Study (MIYAGI);[17] (v) the Ohsaki National Health Insurance Cohort Study (OHSAKI);[18] (vi) the Three-Prefecture Cohort Study, Miyagi (3-pref MIYAGI);[19] the (vii) Three-Prefecture Cohort Study, Aichi (3-pref AICHI);[19] and (viii) the Takayama Study (TAKAYAMA).[20] When analyzing the individual results of each study, we excluded subjects who had a previous history of cancer and those for whom information on diabetes mellitus was missing. Table 1 profiles the studies included in the analysis. Each study was approved by the appropriate institutional review board.

Table 1. Characteristics of the eight cohort studies included in a pooled analysis of diabetes mellitus and the risk of cancer incidence
StudyPopulationAge (years) at baseline surveyYear(s) of baseline surveyPopulation sizeRate of response (%) to baseline questionnaireMethod of follow upFor the present pooled analysis
Age (years) RangeLast follow-up timeMean duration of follow up (years)Size of cohortNumber of cancer cases
MenWomenMenWomen
  1. JACC, Japan Collaborative Cohort Study; JPHC, Japan Public Health Center-based prospective Study; MIYAGI, Miyagi Cohort Study; Ohsaki, Ohsaki National Health Insurance Cohort Study; 3-pref MIYAGI, Three Prefecture Study – Miyagi portion; 3-pref AICHI, Three Prefecture Study – Aichi portion; TAKAYAMA, Takayama Cohort Study.

JPHC-IJapanese residents of 5 public health center areas in Japan40–59199061 59582Cancer registries and death certificates40–592008/12/3116.420 28821 80629151949
JPHC-IIJapanese residents of 6 public health center areas in Japan40–691993–199478 82580Cancer registries and death certificates40–692008/12/3113.329 21732 48440032370
JACCResidents from 45 areas throughout Japan40–791988–1990110 79283Cancer registries (24 selected areas) and death certificates40–792009/12/3112.923 26133 26034322436
MIYAGIResidents of 14 municipalities in Miyagi Prefecture, Japan40–64199047 60592Cancer registries and death certificates40–642003/12/3112.522 39524 06423351531
OhsakiBeneficiaries of National Health Insurance among residents of 14 municipalities in Miyagi Prefecture, Japan40–79199454 99695Cancer registries and death certificates40–792005/12/318.923 00325 08032351786
3-pref MIYAGIResidents of 3 municipalities in Miyagi Prefecture, Japan40–98198431 34594Cancer registries and death certificates40–981992/12/317.513 73417 0701136786
3-pref AICHIResidents of 2 municipalities in Aichi Prefecture, Japan40–103198533 52990Cancer registries and death certificates40–1032000/12/3111.510 84612 2311048754
TAKAYAMAResidents of Takayama, Gifu, Japan35–199231 55285Cancer registries and death certificates35–1012008/3/3113.314 17316 54719741514
Total         156 917182 54220 07813 126

Follow up

Subjects were followed from the baseline questionnaire (JPHC-I, 1990; JPHC-II, 1993–1994; JACC, 1988–1990; MIYAGI, 1990; OHSAKI, 1994; 3-pref MIYAGI, 1984; 3-pref AICHI, 1985; TAKAYAMA, 1992) to the last date of follow up in each study (JPHC-I, 2008; JPHC-II, 2008; JACC, 2009; MIYAGI, 2003; OHSAKI, 2005; 3-pref MIYAGI, 1992; 3-pref AICHI, 2000; TAKAYAMA, 2008). Residence status, including survival, was confirmed through the residential registry. Migration from a study area was treated as censoring at the date of migration. Among the eight cohorts, the percentage ranged from 5.1% to 6.1% for five cohorts, from 14.0% to 19.0% for two cohorts, and was 28.4% for one cohort.

Assessment of exposure

All the studies included were population-based, and blood data were available for only a part of one study. Therefore, we used self-reported past history of diabetes. Information on a history of diabetes in the baseline questionnaire was obtained by using one of the following questions: “Has a doctor ever told you that you have any of the following diseases? – diabetes mellitus (JPHC-I, JPHC-II, 3-pref MIYAGI, TAKAYAMA: yes/no; 3-pref AICHI: current/past/never)” or “Have you ever suffered from any of the following diseases? – diabetes mellitus (JACC, MIYAGI and OHSAKI: no/yes-under medication/yes-cured/yes-not under medication).” Having diabetes currently or in the past, with or without medication, was defined as “diabetes” in the analysis.

Assessment of outcome

Study outcome was defined as the incidence of cancer (total and site-specific) during the follow-up period of each study. In all cohorts in the present study, cancer diagnoses were identified through population-based cancer registries and/or active patient notification from major local hospitals. Indices of data quality were available for seven cohorts. Although the quality and completeness of the case ascertainment varied by cohort in the range of 4.7–11.3% for Death Certificate Only (five cohorts), 23.0% for Death Certificate Notification (one cohort), and 41% for Mortality and Incidence ratio (one cohort), we believe that the overall quality of cancer ascertainment was high enough to conduct the present analysis.

Statistical analysis

Follow-up time was calculated as the duration from the date of the baseline questionnaire in each study until the date of cancer incidence or the end of follow up, whichever came first. In each individual study, we estimated sex-specific hazard ratios (HR) and their 95% confidence intervals (CI) for total and site-specific cancer incidence for diabetes using the Cox proportional hazards model. In each study, two types of adjustment were performed for estimation of HR: age and area (applicable for JPHC-I, JPHC-II and JACC only) (HR1). We conducted further multivariate adjustments by including in the model covariates that were either known or suspected confounding factors: history of cerebrovascular disease, coronary heart disease, cigarette smoking, alcohol consumption, body mass index, leisure-time sports or physical exercise, green leafy vegetable consumption and coffee intake (HR2). We conducted an analysis excluding early diagnosis (within 3 years) from both the numerator and the denominator (HR3). The cut-off points of each covariate are listed in the footnotes of Tables 2-4. An indicator term for missing data was created for each covariate. sas (version 9.1; SAS Institute, Cary, NC, USA) and stata (version 11; Stata Corporation, College Station, TX, USA) statistical software packages were used for the HR estimations.

Table 2. Summary hazard ratios (HR) and 95% confidence intervals (CI) of history of diabetes for total and site-specific cancers: Men and women combined
Cancer siteNumber of studiesNo diabetesDiabetesHR1 (95% CI)HR2§ (95% CI)HR3 (95% CI)Between studies
Person-years of follow-upNumber of casesPerson-years of follow-upNumber of casesQPheteroI2 (%)
  1. †Results given in this column are those pertaining to model HR3. ‡Adjusted for age (years, continuous) and area (applicable for JPHC-I, JPHC-II and JACC only). §Further adjusted for history of cerebrovascular disease (no, yes), coronary heart disease (no, yes), cigarette smoking (pack-years, 0/1–19/20–29/30–39/40 or more), alcohol consumption (ethanol equivalent g/week, continuous), body mass index (continuous), leisure-time sports or physical exercise (JPHC-I and II: less than monthly/1–3 days per month/more than weekly; JACC, MIYAGI and OHSAKI: almost none/more than 1 h per week; TAKAYAMA: none/vigorous exercise or activity, or moderate exercise 1 or more hours per week; 3-pref MIYAGI and AICHI: no information), green leafy vegetables (TAKAYAMA: <4 days per week/4–6 days per week/almost daily; other cohorts: <3 days per week/3–4 days per week/almost daily) and coffee intake (JPHC-I and II: almost none/1–2 days per week/3–4 days per week/1–2 cups per day/3–4 cups per day/5 or more cups per day; JACC: <2 cups per month/1–2 cups per week/3–4 cups per week/almost daily 1–2 cups/almost daily 3–4 cups/almost daily 5 or more cups; MIYAGI, OHSAKI, 3-pref MIYAGI and 3-pref AICHI: none/occasionally/1–2 cups per day/3–4 cups per day/5 or more cups per day; TAKAYAMA: less than once per week/1 day per week/2–6 times per week/daily/2–3 times per day/more than 4 times per day). ¶Adjusted for same covariates as HR2 and excluding early diagnosis within 3 years from the baseline.

All sites83 963 57030 634192 691.623881.23 [1.18–1.28]1.20 [1.15–1.26]1.19 [1.12–1.25]7.720.369.38
All sites excluding the liver83 963 57028 891192 691.621201.18 [1.12–1.23]1.15 [1.09–1.21]1.14 [1.07–1.22]9.560.2226.76
All sites excluding the liver and pancreas83 963 57027 716192 691.619871.14 [1.09–1.20]1.12 [1.06–1.19]1.11 [1.04–1.18]8.780.2720.24
Esophagus73 572 768841176 490.6661.12 [0.87–1.44]1.04 [0.74–1.46]1.03 [0.75–1.43]5.230.510.00
Stomach73 572 7685939176 490.64381.09 [0.95–1.25]1.05 [0.87–1.28]1.06 [0.91–1.22]7.690.2621.95
Colon62 874 0632586144 532.62321.43 [1.25–1.64]1.34 [1.15–1.54]1.40 [1.19–1.64]1.980.850.00
Rectum62 874 0631395144 532.6951.13 [0.92–1.39]1.10 [0.86–1.41]1.14 [0.87–1.50]5.620.3411.03
Liver73 572 7681593176 490.62512.36 [1.89–2.95]2.15 [1.76–2.62]1.97 [1.65–2.36]2.850.830.00
Bile duct73 572 768832176 490.6721.32 [1.03–1.69]1.29 [0.97–1.70]1.35 [0.99–1.85]5.340.500.00
Pancreas73 572 7681052176 490.61201.80 [1.49–2.18]1.86 [1.50–2.30]1.85 [1.46–2.34]3.360.760.00
Lung73 572 7683359176 490.62331.00 [0.85–1.17]1.01 [0.85–1.20]1.00 [0.82–1.22]7.790.2523.02
Kidney62 874 063326144 532.6281.49 [0.78–2.85]1.45 [0.76–2.77]1.57 [0.62–3.94]10.970.0363.55
Bladder73 572 768701176 490.6561.16 [0.82–1.65]1.22 [0.85–1.76]1.28 [0.89–1.86]5.610.3510.92
Lymphoma72 655 459282132 392.9221.28 [0.83–1.98]1.33 [0.83–2.13]1.35 [0.82–2.22]0.151.000.00
Table 3. Summary hazard ratios (HR) and 95% confidence intervals (CI) of history of diabetes for total and site-specific cancers in men
Cancer siteNumber of studiesNo diabetesDiabetesHR1 (95% CI)HR2§ (95% CI)HR3 (95% CI)Between studies
Person-years of follow-upNumber of casesPerson-years of follow upNumber of casesQPheteroI2 (%)
  1. †Results given in this column are those pertaining to model HR3. ‡Adjusted for age (years, continuous) and area (applicable for JPHC-I, JPHC-II and JACC only). §Further adjusted for history of cerebrovascular disease (no, yes), coronary heart disease (no, yes), cigarette smoking (pack-years, 0/1–19/20–29/30–39/40 or more), alcohol consumption (ethanol equivalent g/week, continuous), body mass index (continuous), leisure-time sports or physical exercise (JPHC-I and II: less than monthly/1–3 days per month/more than weekly; JACC, MIYAGI and OHSAKI: almost none/more than 1 h per week; TAKAYAMA: none/vigorous exercise or activity, or moderate exercise 1 or more hours per week; 3-pref MIYAGI and AICHI: no information), green leafy vegetables (TAKAYAMA: <4 days per week/4–6 days per week/almost daily; other cohorts: <3 days per week/3–4 days per week/almost daily) and coffee intake (JPHC-I and II: almost none/1–2 days per week/3–4 days per week/1–2 cups per day/3–4 cups per day/5 or more cups per day; JACC: <2 cups per month/1–2 cups per week/3–4 cups per week/almost daily 1–2 cups/almost daily 3–4 cups/almost daily 5 or more cups; MIYAGI, OHSAKI, 3-pref MIYAGI and 3-pref AICHI: none/occasionally/1–2 cups per day/3–4 cups per day/5 or more cups per day; TAKAYAMA: less than once per week/1 day per week/2–6 times per week/daily/2–3 times per day/more than 4 times per day). ¶Adjusted for same covariates as HR2 and excluding early diagnosis within 3 years from baseline.

All sites81 763 65318 227115 620.517481.24 [1.18–1.30]1.21 [1.15–1.28]1.19 [1.12–1.27]5.200.640.00
All sites excluding the liver81 763 65317 062115 620.515321.16 [1.10–1.22]1.15 [1.08–1.22]1.13 [1.06–1.21]5.090.650.00
All sites excluding the liver and pancreas81 763 65316 466115 620.514531.14 [1.07–1.22]1.13 [1.06–1.20]1.12 [1.05–1.19]5.450.610.00
Esophagus71 592 174740105 219.5631.14 [0.87–1.47]1.07 [0.79–1.44]1.02 [0.73–1.42]4.510.610.00
Stomach71 592 1744089105 219.53401.05 [0.89–1.24]1.03 [0.84–1.25]1.02 [0.88–1.17]4.540.600.00
Colon61 309 293149488 263.51751.55 [1.33–1.82]1.58 [1.32–1.89]1.58 [1.31–1.90]3.340.650.00
Rectum61 309 29388588 263.5701.11 [0.86–1.43]1.05 [0.80–1.36]1.12 [0.84–1.50]3.830.570.00
Liver71 592 1741078105 219.52012.40 [1.94–2.97]2.25 [1.83–2.76]2.07 [1.70–2.53]2.790.830.00
Bile duct61 498 79436298 182.7441.52 [1.08–2.16]1.52 [1.07–2.15]1.66 [1.14–2.41]1.830.870.00
Pancreas71 592 174533105 219.5701.75 [1.36–2.25]1.72 [1.30–2.28]1.58 [1.15–2.17]3.010.810.00
Lung71 592 1742429105 219.51890.97 [0.83–1.14]1.01 [0.83–1.22]1.01 [0.82–1.25]7.630.2721.39
Prostate61 309 293127388 263.5981.02 [0.79–1.31]0.98 [0.70–1.36]0.96 [0.64–1.43]11.330.0555.88
Kidney41 027 10815467 821.9201.55 [0.82–2.94]1.48 [0.67–3.29]2.25 [0.82–6.14]8.020.0562.58
Bladder61 483 12048593 216.8461.22 [0.83–1.79]1.30 [0.89–1.91]1.32 [0.90–1.96]5.400.377.39
Lymphoma3764 79711854 004.7131.33 [0.76–2.32]1.73 [0.94–3.18]1.60 [0.82–3.10]0.410.810.00
Table 4. Summary hazard ratios (HR) and 95% confidence intervals (CI) of history of diabetes for total and site-specific cancers in women
Cancer siteNumber of studiesNo diabetesDiabetesHR1 (95% CI)HR2§ (95% CI)HR3 (95% CI)Between studies
Person-years of follow upNumber of casesPerson-years of follow upNumber of casesQPheteroI2 (%)
  1. †Results given in this column are those pertaining to model HR3. ‡Adjusted for age (years, continuous) and area (applicable for JPHC-I, JPHC-II and JACC only). §Further adjusted for history of cerebrovascular disease (no, yes), coronary heart disease (no, yes), cigarette smoking (pack-years, 0/1–19/20–29/30–39/40 or more), alcohol consumption (ethanol equivalent g/week, continuous), body mass index (continuous), leisure-time sports or physical exercise (JPHC-I and II: less than monthly/1–3 days per month/more than weekly; JACC, MIYAGI and OHSAKI: almost none/more than 1 h per week; TAKAYAMA: none/vigorous exercise or activity, or moderate exercise 1 or more hours per week; 3-pref MIYAGI and AICHI: no information), green leafy vegetables (TAKAYAMA: <4 days per week/4–6 days per week/almost daily; other cohorts: <3 days per week/3–4 days per week/almost daily) and coffee intake (JPHC-I and II: almost none/1–2 days per week/3–4 days per week/1–2 cups per day/3–4 cups per day/5 or more cups per day; JACC: <2 cups per month/1–2 cups per week/3–4 cups per week/almost daily 1–2 cups/almost daily 3–4 cups/almost daily 5 or more cups; MIYAGI, OHSAKI, 3-pref MIYAGI and 3-pref AICHI: none/occasionally/1–2 cups per day/3–4 cups per day/5 or more cups per day; TAKAYAMA: less than once per week/1 day per week/2–6 times per week/daily/2–3 times per day/more than 4 times per day). ¶Adjusted for same covariates as HR2 and excluding early diagnosis within 3 years from baseline.

All sites82 199 91712 40777 070.46401.19 [1.10–1.29]1.18 [1.08–1.30]1.19 [1.07–1.31]5.520.600.00
All sites excluding the liver82 199 91711 82977 070.45881.16 [1.07–1.26]1.16 [1.05–1.28]1.16 [1.03–1.31]8.410.3016.78
All sites excluding the liver and pancreas82 199 91711 25077 070.45341.12 [1.02–1.22]1.10 [0.99–1.22]1.08 [0.97–1.22]6.320.500.00
Esophagus2790 7552423 229.234.28 [0.80–22.90]4.70 [1.12–19.71]5.28 [1.48–18.86]0.920.340.00
Stomach71 980 594185071 270.4981.14 [0.90–1.44]1.22 [0.95–1.57]1.29 [0.97–1.72]6.830.3412.21
Colon61 564 770109256 268.4571.13 [0.87–1.48]0.92 [0.66–1.29]0.99 [0.69–1.42]1.900.860.00
Rectum61 564 77051056 268.4251.35 [0.81–2.25]1.48 [0.76–2.89]1.44 [0.66–3.14]9.700.0848.45
Liver71 980 59451571 270.4501.99 [1.41–2.81]1.84 [1.30–2.60]1.71 [1.14–2.57]4.530.610.00
Bile duct71 980 59443971 270.4261.28 [0.85–1.91]1.38 [0.85–2.24]1.44 [0.77–2.70]1.860.870.00
Pancreas71 980 59451971 270.4501.98 [1.33–2.94]2.27 [1.33–3.85]2.48 [1.48–4.16]10.870.0944.81
Lung71 980 59493071 270.4441.09 [0.80–1.47]1.08 [0.76–1.54]1.02 [0.68–1.51]3.260.780.00
Breast61 564 770138056 268.4430.95 [0.70–1.29]0.98 [0.69–1.38]1.03 [0.69–1.56]5.550.359.87
Cervix51 131 52920641 589.4111.59 [0.90–2.80]2.08 [1.02–4.27]2.63 [1.20–5.80]2.420.660.00
Uterine corpus51 439 54722451 165.5121.81 [1.01–3.27]1.69 [0.87–3.31]1.84 [0.90–3.76]1.700.790.00
Ovary3773 230.712732 026.771.32 [0.41–4.22]1.68 [0.69–4.07]1.22 [0.44–3.37]0.320.850.00
Kidney3785 630.45633 27541.52 [0.60–3.86]1.28 [0.46–3.55]1.26 [0.30–5.28]0.140.930.00
Bladder3944 409.69436 264.271.14 [0.58–2.24]1.45 [0.65–3.22]1.63 [0.69–3.87]0.010.990.00
Lymphoma41 301 92410844 814.482.00 [0.91–4.38]2.16 [0.88–5.32]2.43 [0.93–6.37]4.180.2428.16

For each of the three developed models (HR1, HR2 and HR3), summary HR estimates were obtained using the DerSimonian and Laird random-effects model. Briefly, the summary estimates were calculated as a weighted sum of the study-specific HR, the weights being taken to be the sum of the study-specific HR variance estimates and an estimated between-study variance component. Statistical heterogeneity among studies was assessed by means of the I2 statistics, which corresponds to the proportion of total variation in study estimates accounted for by between-study variation. The sas and R software packages were used for meta-analysis.

Results

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Appendix 1

The present study included 339 459 subjects (156 917 men and 182 542 women) from eight ongoing large-scale population-based prospective studies in Japan (Table 1). During 4 156 262 person-years of follow up (mean 9.9 years/person), 33 204 incidences of cancer were identified (20 078 men and 13 126 women).

Results for men and women combined are shown in Table 2. Individuals who had a history of diabetes had a statistically significant increased risk of total cancer; when cases diagnosed within 3 years of the baseline were excluded, the hazard ratio was estimated as 1.19 (95% CI: 1.12–1.25). The association remained statistically significant even after exclusion of liver and pancreatic cancers, for which the risk was increased in individuals with diabetes (HR3 = 1.97 [1.65–2.36] and 1.85 [1.46–2.34], respectively).

The impact of diabetes on total cancer risk was similar in separate analyses of men and women. In men (Table 3), diabetes was associated with an increased risk of total cancer (HR3 = 1.19 [1.12–1.27]). Diabetes was also associated with a statistically significant increased risk of liver, pancreatic, colon and bile duct cancers; the HR were estimated as 2.07, 1.58, 1.58 and 1.66, respectively. An increased risk was also suggested for rectal, kidney and bladder cancers as well as for lymphoma, although these associations were not statistically significant. Diabetes was not associated with esophageal, stomach, lung or prostate cancer.

In women (Table 4), as in men, diabetes was associated with an increased risk of total cancer (HR3 = 1.19 [1.07–1.31]). This association became borderline significant when both liver and pancreatic cancers were excluded. HR3 values for liver and pancreatic cancers were 1.71 and 2.48, respectively. For the uterine corpus, a statistically significant increased risk of cancer was observed (HR1 = 1.81); however, the association did not remain statistically significant when further variables were adjusted for (HR2 = 1.69 [0.87–3.31]) and when early cases were omitted (HR3 = 1.84 [0.90–3.76]). Alternatively, for cancers of the cervix and esophagus, HR were statistically significant only in HR2 and HR3. An increased risk was also suggested for stomach, rectal, bile duct, ovarian, kidney and bladder cancers, as well as lymphoma, although without statistical significance. No association was seen for colon, lung or breast cancer.

Discussion

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Appendix 1

To the best of our knowledge, this is the first examination of the association between diabetes and cancer incidence by means of pooled analysis, which allows for stable summary quantitative estimates. When we pooled eight ongoing prospective cohort studies (which included >330 000 subjects), we found that diabetes was moderately associated with an increase in total cancer risk. Studies conducted to date have tended to investigate the relationship between diabetes and cancer site specifically, and only a few studies have focused on total cancer.[12, 21-24] Our results are in line with a recent systematic review and meta-analysis that revealed that diabetes was associated with a moderately increased risk of cancer incidence.[13, 14]

The increased risks of liver cancer and pancreatic cancer seen in the present study are consistent with the increased risk observed for both sexes in previous studies:[2, 4] on the basis of a meta-analysis of cohort studies, the summary estimates for heptatocellular carcinoma and pancreatic cancer were 2.01[2] and 1.94,[4] respectively, which are similar to our quantitative estimate. From a combination of 30 cohort studies, diabetes was shown to be associated with a 27% increase in the risk of colorectal cancer incidence with evident heterogeneity among studies.[6] Our results for colon cancer in men support this finding, although data for women and for rectal cancer did not show a clear association.

The association of diabetes with cancer at other sites of the gastrointestinal tract was unclear. Studies on the association of diabetes with biliary tract cancer have shown mixed results. The first systematic review, which was published recently, showed that diabetic individuals may have an approximately 50% increased risk of bile tract cancer.[3] Our results for men support this finding, but the association did not reach statistical significance for women. An analysis of 21 studies found that nondiabetic and diabetic individuals have similar risks of gastric cancer; however, a subgroup analysis found that diabetic women have an 18% increased risk of gastric cancer.[5] In line with this finding, we observed increased risk only among women, although the association did not reach the level of statistical significance (HR3 = 1.29 [0.97–1.72]). Some authors have suggested that the progression from diabetes to cancer may have different etiologies in men and women, perhaps owing to hormonal differences.[25] Helicobacter pylori may also play a key role in the association: in a 9-year cohort study of 2466 Japanese, fasting plasma glucose level was positively associated with the risk of developing gastric cancer;[26] however, the excess risk was observed only among H. pylori-positive subjects, which suggests that hyperglycemia may be a cofactor for both diabetes and gastric cancer. Evidence regarding the association between diabetes and the risk of esophageal cancer is contradictory.[27] Histologically, the major type of esophageal cancer in Japan and Taiwan is squamous cell carcinoma, but a recent case–control study conducted in Taiwan did not show any significant association between diabetes and esophageal cancer.[28] We observed no association among men, whereas a statistically significant excess risk was observed among women (HR2, HR3). However the 95% CI was wide, suggesting that the excess risk may have been a chance finding due to the small number of cases in women.

A recently published meta-analysis showed that diabetes is associated with increased risk of kidney and bladder cancers.[7, 8] However, when studies were restricted to those with adjustments for body mass index or obesity, the association failed to reach the level of statistical significance for kidney cancer.[7] The evidence for a relationship between obesity and bladder cancer risk is limited and inconsistent.[8] Our pooled analysis of eight studies of populations of Japanese people, who are relatively lean compared with US and European populations, suggested a statistically insignificant elevated risk for kidney and bladder cancers. This result suggests that there may be some underlying mechanism common to both diseases that cannot totally be explained by obesity. In the first systematic review to evaluate the relationship between type 2 diabetes and non-Hodgkin lymphoma, Chao and Page[29] showed that the two diseases were positively associated, on the basis of 13 studies, including three prospective studies. However, the authors conclude that the evidence is inconclusive, owing to the methodological limitations of the included case–control studies, and note the need for more prospective studies with improved control of confounding. The elevated risk was more evident in prospective cohort studies, among women, in East Asian populations and in studies with adjustment for body mass index. This situation all meets to our present analysis with adjustment for body mass index and HR larger among women, although without statistical significance.

Increased exposure to estrogen as a result of diabetes is considered to be another factor affecting the relationship between site-specific cancers and diabetes. Our findings with regard to cancer of the cervix and uterine corpus contradict the findings reported for previous studies.[10] This difference may be due to the small sample size in our study, especially for HR2 and HR3. Our results for prostate cancer are in line with the results of previous studies showing a negative or null association.[11] Previous studies have been conducted mainly among white men. Race is reported to be one of the strongest risk factors for prostate cancer, and our pooled analysis adds important evidence from an Asian population. The suggested mechanism for the inverse association between diabetes and the risk of prostate cancer is the reduced level of testosterone, which is commonly seen in diabetic men or with obesity secondary to low levels of sex hormone-binding globulin. We found no association between breast cancer and diabetes, whereas previous studies have shown that diabetes is associated with an increased risk of breast cancer.[9] Epidemiologic studies have generally indicated a positive association between estrogen level and breast cancer risk in postmenopausal women. In this study, when women were stratified by menopausal status, similar results were observed; HR3 = 1.39 (0.57–3.40) and HR3 = 1.01 (0.63–1.60) for premenopausal and postmenopausal women, respectively. In a previous meta-analysis,[9] the relation between diabetes and breast cancer appeared to be confined to postmenopausal women, but the number of studies of premenopausal breast cancer was limited, and a test for difference in association by menopausal status was not statistically significant. To clarify whether the association varies by menopausal status, further investigations are warranted.

The most supported of the mechanisms suggested for the association between diabetes and cancer is insulin resistance with hyperinsulinemia, which may have a mitogenic effect by activating insulin-like growth factor.[30-32] Hyperinsulinemia and hyperglycemia have also been reported to promote tumor cell proliferation and metastasis in type 2 diabetes.[33, 34] These mechanisms are supported by the fact that treatment with metformin, an insulin sensitizer, is associated with a lower risk of cancer among diabetic patients, compared to patients treated with insulin or sulfonylurea.[35, 36] Furthermore, inflammatory cytokines produced by adipose tissues, such as interleukin-6, monocyte chemoattractant protein, and plasminogen activator inhibitor-1, may play important roles in carcinogenesis, cancer progression and poor prognosis.

It should be noted, however, that the relationship between diabetes and cancer may not be causal. First, confounding factors may obscure the relationship between these diseases. Although in the present analysis potential confounding factors were adequately adjusted across the study, it is possible that the effect of unadjusted (unmeasured, unknown) common factors cannot be totally excluded. Second, it is possible that cancer and diabetes simply share common risk factors, such as obesity or physical inactivity. As presented in Tables 2-4, for those HR showing significant results in model 1 (HR1), further adjustment for covariates including body mass index (relative marker of obesity) and physical activity, which are known risk factors for DM, slightly attenuated the results but remained statistically significant (HR2). This means that increased risk of cancer among diabetes is partially, but not fully, explained by these shared risk factors. Although several mechanism have been suggested for the association between diabetes and cancer, further studies using blood glucose or insulin level are needed to clarify the etiology. Third, detection bias may arise because diabetic subjects may receive medical care more frequently than nondiabetic subjects, leading to more frequent detection of cancer among diabetic subjects. Fourth, reverse causality may also exist. Cancer generally causes insulin resistance, and the resulting hyperglycemia may produce cytokines, such as tumor necrosis factor α.[37, 38] In the present analysis, HR3 was calculated by removing early diagnosis within 3 years. Removing early diagnosis within 5 years from the analysis also did not alter the findings essentially, and the possibility of reverse causality might be minimized.

The present study has several limitations. First of all, we cannot exclude the possibility that there may be some chance findings, so caution is needed in interpreting our results. Indeed, the probability of mistakenly concluding that a particular association is different from nil increases with the number of hypotheses tested. A correction for multiple testing, such as the Bonferroni procedure, is often used to control the overall probability of incorrectly rejecting at least one null hypothesis under the assumption that all null hypotheses (e.g. absence of effect) are simultaneously true.[39, 40] This procedure was not conducted because this would not answer our research question; that is, the assessment of the separate relationships between diabetes and each cancer site. The diagnosis of diabetes was based on self-report in all studies. According to a validation study conducted as part of one of the studies, self-reported diabetes exhibits fairly good agreement with diabetes documented in medical records (94%).[41] Case ascertainment based on self-reporting might result in either overreporting or underreporting, and these misclassifications would bias the association toward the null. In addition, type 1 and type 2 diabetes were not distinguished. However, because type 1 diabetes is less frequent than type 2 diabetes, especially in adult populations, it would be reasonable to suppose that most of the subjects had type 2 diabetes. Despite the large number of person-years, the number of cases of some site-specific cancers was only moderate. Finally, there may be some concerns about observational studies not providing an appropriate time reference for estimating the time-at-risk of participants. Korn et al.[42] recommend the use of age as a more natural and appropriate time scale that allows one to take into account the left truncated nature of the data. However, Pencina et al.[43] show that provided age is properly taken into account in the Cox model, the choice of one or the other time scale has no meaningful impact on the parameter estimates. In addition, Chalise et al.[44] conducted a simulation study which showed that Cox models using time-on-study as the time scale were robust to misspecification of the true underlying time scale. An analysis conducted on two of the studies included in our pooled analysis using age as the time scale showed that results in terms of relative risk of the association between diabetes and each of the considered cancer sites were very similar to those obtained using time-on-study. Therefore, we decided to present results based on time-on-study as time-scale Cox models because this allowed us to preserve coherence with already published results.[12, 22, 23]

The strength of the present study is that it includes most of the ongoing prospective studies in Japan, with overlapping birth generations and a similar survey period. Therefore, pooling of these studies allows for a stable quantitative estimate of the impact of diabetes among Japanese people. In the included studies, diabetes was measured before cancer diagnosis, which precludes the possibility of selection and recall bias. In addition, as mentioned above, the covariates were adequately controlled across studies, which removes a potential source of heterogeneity that can occur in a meta-analysis of published literature. The use of incidence rather than mortality as an end point is advantageous because it enables us to directly determine the contribution of diabetes to cancer risk.

In summary, by pooling data from eight cohort studies with a considerable number of subjects, statistically increased risk was observed for cancers at specific sites, such as colon (HR = 1.40), liver (HR = 1.97), pancreas (HR = 1.85) and bile duct (HR = 1.66; men only). Increased risk was also suggested for other sites and, as a whole, diabetes mellitus was associated with a 20% increase in the risk of total cancer incidence in the Japanese population.

The association between these two diseases has important implications for reiterating the importance of controlling lifestyle factors, and it may suggest a possible strategy for cancer screening among patients with diabetes. Furthermore, considering the increasing prevalence of diabetes worldwide and its association with cancer, studies continuously investigating the risk factors for diabetes are important.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Appendix 1

This study was supported by the National Cancer Center Research and Development Fund and Grant-in-Aid for the Third-Term Comprehensive Ten-Year Strategy for Cancer Control from the Ministry of Health, Labor and Welfare of Japan.

References

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Appendix 1

Appendix 1

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Appendix 1

Research group members: Shizuka Sasazuki (principal investigator), Shoichiro Tsugane, Manami Inoue, Motoki Iwasaki, Tetsuya Otani (until 2006), Norie Sawada (since 2007), Taichi Shimazu (since 2007), Taiki Yamaji (since 2007, National Cancer Center, Tokyo), Ichiro Tsuji (since 2004), Yoshitaka Tsubono (2003, Tohoku University, Sendai), Yoshikazu Nishino (until 2006, Miyagi Cancer Research Institute, Natori, Miyagi), Akiko Tamakoshi (since 2010, Hokkaido University, Sapporo), Keitaro Matsuo (–2010, 2012–), Hidemi Ito (2010–2011, Aichi Cancer Center, Nagoya), Kenji Wakai (Nagoya University, Nagoya), Chisato Nagata (Gifu University, Gifu), Tetsuya Mizoue (National Center for Global Health and Medicine, Tokyo) and Keitaro Tanaka (Saga University, Saga).