SEARCH

SEARCH BY CITATION

Abstract

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

Diabetes mellitus (DM) has been reported to be associated with an increased risk of site-specific cancers; however, few studies have assessed associations of DM with both total and site-specific cancers in Japan. We examined the association of a history of DM with cancer incidence in a population-based prospective cohort study in Japan. A total of 14 173 men and 16 547 women over 35 years old, who completed a self-administered baseline questionnaire in 1992, were followed up for cancer incidence from September 1992 to March 2008. At baseline, 6.3% men and 2.9% women had a history of diabetes. A total of 1974 men and 1514 women were identified as newly diagnosed with cancer. Hazard ratios (HR) and 95% confidence intervals (CI) were determined using Cox proportional hazards models. After controlling for potential confounders, men with DM had a modest risk increase of total cancer occurrence compared with those without DM (HR, 1.09; 95% CI, 0.93–1.29). Increased risk of cancer of the liver (HR, 2.18; 95% CI, 1.27–3.74), bile duct (HR, 2.17; 95% CI, 1.01–4.66), and larynx (HR, 3.61; 95% CI, 1.16–11.2) in diabetic men were observed. In women, significant increased risk of total cancer (HR, 1.35; 95% CI, 1.06–1.73) and stomach cancer (HR, 2.15; 95% CI, 1.30–3.54) were observed among diabetic subjects. These data suggest that people with DM may be at increased risk of both total and some site-specific cancers.

Diabetes mellitus (DM) is well-known to be a major risk factor of cardiovascular disease.[1] The potential association of DM with cancer risk has been surmised for decades.[2, 3] Diabetic patients have several metabolic and immune abnormalities, such as persistent hyperinsulinemia, increase in insulin-like growth factor (IGF-1),[4] hyperglycemia,[5] and chronic inflammation.[6] These systemic characteristics are highly likely to promote carcinogenesis.

Previous meta-analyses of epidemiological studies have generally suggested that DM increases risks of site-specific cancer of the liver,[7, 8] pancreas,[9-11] and colorectum.[12-14] Diabetes mellitus was significantly positively associated with cancer of the biliary tract,[15, 16] breast,[17-19] endometrium,[20-22] and stomach[23, 24] and significantly inversely associated with cancer of prostate[25-27] in some but not all systematic meta-analyses. Most previous epidemiological studies were carried out in North America and Western Europe. Previous reviews showed that a small number of studies have assessed the association of DM with all cancer as well as site-specific cancers in Asian countries.[7-27] The associations of DM with total and site-specific cancers have not been fully investigated in this area, including Japan.

As seen in other nations, the number of people with DM is increasing and becoming a large public health problem in Japan; the number of diabetic persons was estimated at 6.9 million in 1997, 7.4 million in 2002, and 8.9 million in 2007 based on the National Health and Nutrition Survey Report.[28, 29] Elucidation of the association between this epidemic disease and cancer is essential to public health. Therefore, we carried out a prospective study to assess the associations between pre-existing DM and the development of both total and site-specific cancer in a cohort of men and women in Japan.

Materials and Methods

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

Study group

Participants in the present study were from the Takayama Study cohort. In September 1992, that population-based prospective study was initiated among 36 990 residents of Takayama, Gifu Prefecture, Japan, who were over 35 years of age. Details of the study design have been described elsewhere.[30] At the baseline, participants filled in a self-administered questionnaire on socioeconomic and various lifestyle factors, including demographic characteristics, smoking and drinking habits, habitual diet, physical activity, and reproductive and medical histories, including DM. A total of 14 427 men and 17 125 women, 85.3% of the target population, responded to the questionnaire. For the present analysis, we excluded the subjects who reported on the baseline questionnaire that they had any a previous history of cancer (254 men and 578 women). Thus, the final cohort for the current study consisted of 14 173 men and 16 547 women.

Data collection

Information on a history of DM was obtained through the question in the baseline questionnaire, “Has your doctor ever told you that you have any of the following diseases?” Diabetes mellitus and other major diseases, including cancer and hypertension, were listed for possible selection in the questionnaire. When participants chose the DM option, pre-existing DM was defined as affirmative. The validity of self-reported diabetes was assessed in a subsample of participants who attended blood sampling (n = 214). The sensitivity and specificity of self-reported DM compared with the reference standard defined by hemoglobin A1c (≥6.1% based on the diagnostic criteria set by the Japan Diabetes Society before 2012) were 57.4% and 96.5%, respectively. Participants who had smoked a total of over 20 packs of cigarettes in their life were defined as smokers. Those who had never smoked were defined as never smokers; those who still smoked at baseline as current smokers; and those who had quit smoking as former smokers. Body mass index (BMI) was calculated as weight at baseline divided by height squared and presented in kg/m2. Physical activity was evaluated by asking subjects the average number of hours they had engaged in activities such as strenuous sports, vigorous work, and moderate exercise during the past year. Time per week spent at each activity was multiplied by the corresponding energy expenditure, which was defined as a metabolic equivalent (MET) and summed up to a physical activity score (MET-h/week). The details and validity of this score are described elsewhere.[31] Diet and alcohol were evaluated by a 169-item semiquantitative food frequency questionnaire. Subjects were asked to report average intake frequency and usual serving sizes for each listed food item during the last year. Each intake of food and nutrient was estimated from the Standard Tables of Food Composition in Japan.[32] Details of this questionnaire were validated in a previously published report.[33]

Outcome measurement

The primary study endpoint was defined as cancers that newly developed during the follow-up period from September 1, 1992 until March 31, 2008. Cancer occurrence was mostly ascertained by means of a linkage with the records of two local population-based cancer registries. Hospital-based cancer registries and medical records in two local base hospitals were also reviewed and used to collect information on cancer incidence in the study area. The mortality-to-incidence rate for all cancer was 0.51. The site of cancer was coded following the rules of the International Statistical Classification of Diseases and Related Health Problems, 10th Revision. The earliest date of diagnosis was adopted for subjects with multiple primary cancer cases at the different times for total cancer incidence. Residence and survival of the participants were confirmed using the residential registers of the municipalities or family registers. Registration of death is required by the Family Registration Law in Japan. During the follow-up period, 942 (6.5%) men and 974 (5.7%) women moved out of Takayama. In regard to 104 (0.7%) men and 147 (0.9%) women, the date of emigration of was not known. This study was approved by the local ethics board of the Gifu University Graduate School of Medicine (Gifu, Japan).

Data analysis

Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated by the Cox proportional hazards model to describe the relative risk of cancer development for those with DM at baseline compared with those without DM. The person-years of follow-up for each participant were accumulated from the date of baseline survey until the date of occurrence of any cancer, the date of death, the date of moving out of Takayama, or the end of the study period, March 31, 2008, whichever came first. Participants who moved away on a date unknown were determined to be censored at the latest point when they were confirmed to live in the city. Age at baseline was used in the model for adjustment because of a potent risk factor for cancer. Additional adjustment was carried out for potential confounding factors such as smoking status (never, former, or current); BMI in quintile; physical activity (MET-h/week); education level (≤11, 12–14, or ≥15 years); history of hypertension, ischemic heart disease, and stroke (yes); alcohol intake (g/day); total fat intake (g/day); total energy intake (kcal/day); and coffee intake (frequency per month). These variables are known or suspected through previous studies as risk factors for cancer. Because latent cancer might have altered subjects' lifestyle at baseline and this might have contributed to the improvement of DM, we further repeated the analysis after excluding the subjects who had cancer during the first 3 years of follow-up. The data were managed using the sas version 9.2 software (SAS Institute, Cary, NC, USA). The analyses were carried out separately for men and women. Significance was defined as two-sided < 0.05.

Results

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

During the follow-up of 407 310 person-years (average duration of follow-up, 13.2 years), a total of 3488 participants (1974 men and 1514 women) were identified as newly diagnosed with cancer. Baseline characteristics of the study population by the diabetes status in the Takayama Study are shown in Table 1. At baseline, 889 (6.3%) men and 487 (2.9%) women reported that they had DM. Men and women with a history of DM were more likely to be old, short, less educated, and physically inactive, and their BMI was higher than those without DM. Those with DM had a lower consumption of alcohol and total energy and a higher intake of total vegetables. Those with DM also had more frequently reported that they had hypertension, stroke, and ischemic heart diseases. In men, those with a history of DM were more likely to be past smokers but less likely to be current smokers. Women with DM were less likely to be currently married and premenopausal.

Table 1. Baseline characteristics of men and women by self-reported history of DM in the Takayama Study
 Men Women 
Without DMWith DM a Without DMWith DM a
Mean (SD)Mean (SD)
Age54.3 (12.3)59.3 (11.2)<0.000155.5 (13.2)64.0 (11.8)<0.0001
Body mass index (kg/m2)22.5 (2.8)23.0 (2.9)<0.000122.0 (2.9)22.5 (3.3)0.0004
Height (cm)164.6 (6.9)163.4 (6.8)<0.0001152.0 (6.4)149.9 (6.0)<0.0001
Physical activity (MET-h/week)26.9 (41.4)22.3 (37.2)0.001218.6 (29.5)13.5 (22.7)0.0002
Total energyl intake (kcal/day)2611.4 (873.8)2496.7 (870.7)0.00022133.2 (788.0)1905.1 (770.4)<0.0001
Total vegetable intake (g/day)370.4 (259.0)411.5 (288.3)<0.0001394.3 (266.3)423.6 (286.1)0.0172
Total alcohol intake (g/day)41.4 (41.0)38.1 (42.9)0.02007.8 (16.9)4.2 (11.5)<0.0001
 No. (%) a No. (%) a
  1. a

    Student's t-test for continuous variables and χ2-test for categorical variables.

Smoking status
Never smoker2155 (16.7)133 (15.5)<0.000111 854 (82.4)338 (81.1)0.7568
Current smoker6997 (54.2)407 (47.3) 1865 (13.0)59 (14.2) 
Former smoker3754 (29.1)320 (37.2) 663 (4.6)20 (4.8) 
Currently married12 022 (91.4)794 (90.0)0.150511 796 (74.7)277 (57.7)<0.0001
Education 12 years or more5534 (42.2)328 (37.2)0.00345259 (33.3)92 (19.6)<0.0001
History of hypertension2622 (19.7)295 (33.2)<0.00012965 (18.5)171 (35.1)<0.0001
History of stroke225 (1.7)26 (2.9)0.0071134 (0.8)20 (4.1)<0.0001
History of ischemic heart disease586 (4.4)80 (9.0)<0.0001682 (4.3)52 (10.7)<0.0001
Postmenopausal women   9358 (59.7)412 (86.7)<0.0001
Current hormone replacement therapy   1078 (7.2)31 (7.0)0.8991

In men, the HR of total cancer occurrence modestly increased for those with a history of DM compared with those without DM (HR, 1.09; 95% CI, 0.93–1.29) after controlling for covariates (Table 2), although this association was marginally significant. We observed a statistically significant increase in the risk of cancer in the liver (HR, 2.18; 95% CI, 1.27–3.74), bile duct (HR, 2.17; 95% CI, 1.01–4.66), and larynx (HR, 3.61; 95% CI, 1.16–11.2) in those with a history of DM in the multivariate model. A non-significantly increased HR was found for pancreatic cancer (HR, 1.63; 95% CI, 0.81–3.29), in contrast with a non-significantly decreased HR for prostate cancer (HR, 0.56; 95% CI, 0.28–1.15). In women (Table 3), a significant risk increase of total cancer was observed among those with a history of DM (HR, 1.35; 95% CI, 1.06–1.73). We observed that the risk of stomach cancer increased significantly in women with a history of DM in the multivariate model (HR, 2.15; 95% CI, 1.30–3.54). Tobacco smoking has been reported to be a major risk factor for laryngeal cancer.[34] Therefore, we repeated the analysis including pack-years (0, 1–15, >15), instead of smoking status as a covariate. The HR of laryngeal cancer in men remained significant (HR, 3.89; 95% CI, 1.24–12.2). After excluding the subjects who had been diagnosed with cancer in the first 3 years of follow-up, the increased risk of pancreatic cancer in men obtained marginal significance (pancreatic cancer in men: HR, 2.01; 95% CI, 1.00–4.09). After the exclusion, the HR of prostate cancer for those with a history of DM was 0.48 (95% CI, 0.21–1.10, = 0.08). For others, exclusion of the subjects with previous cancer incidence in the first 3 years from baseline did not alter the results.

Table 2. Hazard ratios (HRs) of cancer incidence in accordance with self-reported history of diabetes mellitus (DM) among men in the Takayama Study, September 1992–March 2008
 ICD10No. of CasesPerson- yearsAge-adjustedMultivariate
HR95%CIHRa95%CI
  1. a

    Adjusted for age at baseline, smoking status, body mass index, physical activity, length of education in years, history of hypertension, history of stroke, history of ischemic heart disease, total energy intake, and intake of fat, ethanol, and coffee. CI, confidence interval.

All-sites Without DM1824171 6381.00 1.00 
With DM15010 4161.090.92, 1.281.090.93, 1.29
EsophagusC15Without DM65177 4311.00 1.00 
With DM210 8510.430.10, 1.740.410.10, 1.68
StomachC16Without DM402175 9341.00 1.00 
With DM3610 7541.170.83, 1.641.200.85, 1.69
ColonC18Without DM238176 4911.00 1.00 
With DM1510 7800.880.52, 1.480.820.48, 1.39
RectumC20Without DM133177 1051.00 1.00 
With DM1210 8211.280.71, 2.311.270.70, 2.30
LiverC22Without DM90177 3681.00 1.00 
With DM1610 8392.351.38, 4.012.181.27, 3.74
Bile ductC21, 23Without DM45177 5061.00 1.00 
With DM810 8512.171.02, 4.612.171.01, 4.66
PancreasC25Without DM69177 5281.00 1.00 
With DM910 8481.670.83, 3.341.630.81, 3.29
LarynxC32Without DM15177 4931.00 1.00 
With DM410 8363.911.29, 11.93.611.16, 11.2
LungC34Without DM263177 1481.00 1.00 
With DM1910 8310.910.57, 1.441.060.66, 1.70
ProstateC61Without DM171177 0111.00 1.00 
With DM810 8270.610.30, 1.230.560.28, 1.15
KidneyC64Without DM38177 4221.00 1.00 
With DM210 8340.680.16, 2.830.680.16, 2.84
BladderC67Without DM111177 0741.00 1.00 
With DM810 8100.920.45, 1.880.920.45, 1.90
Table 3. Hazard ratios (HRs) of cancer incidence in accordance with self-reported history of diabetes mellitus (DM) among women in the Takayama Study, September 1992–March 2008
 ICD-10No. of CasesPerson-yearsAge-adjustedMultivariate
HR95%CIHRa95%CI
  1. a

    Adjusted for age at baseline, smoking status, body mass index, physical activity, length of education in years, history of hypertension, history of stroke, history of ischemic heart disease, total energy intake, and intake of fat, ethanol, and coffee. CI, confidence interval.

All-sites Without DM1446219 4601.00 1.00 
With DM6857971.371.07, 1.741.351.06, 1.73
StomachC16Without DM217224 3121.00 1.00 
With DM1759242.091.27, 3.432.151.30, 3.54
ColonC18Without DM224224 2691.00 1.00 
With DM859640.970.48, 1.970.950.47, 1.93
RectumC20Without DM93224 7641.00 1.00 
With DM559751.690.68, 4.171.810.73, 4.49
LiverC22Without DM68225 1141.00 1.00 
With DM259740.750.18, 3.050.650.16, 2.69
Bile ductC23, 24Without DM67225 1691.00 1.00 
With DM459761.420.52, 3.911.320.48, 3.65
PancreasC25Without DM65225 1991.00 1.00 
With DM459741.720.63, 4.751.530.55, 4.26
LungC34Without DM111225 0121.00 1.00 
With DM159780.230.03, 1.680.220.03, 1.56
BreastC50Without DM177224 2511.00 1.00 
With DM559591.220.50, 2.971.150.47, 2.83
Uterus cervixC53Without DM44224 9851.001 1.00 
With DM259671.950.47, 8.132.130.51, 8.95
Uterus corpusC54Without DM38225 0161.00 1.00 
With DM259602.650.63, 11.22.490.59, 10.6
OvaryC56Without DM50225 0621.00 1.00 
With DM159670.730.10, 5.320.700.10, 5.11
KidneyC64Without DM18225 1901.00 1.00 
With DM159721.440.19, 10.81.340.18, 10.3
BladderC67Without DM38225 0601.00 1.00 
With DM159780.660.09, 4.850.690.09, 5.10

Discussion

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

In this prospective cohort study, we found that a self-reported history of DM was significantly associated with a 35% increase in the risk of total cancer in women. A modest risk increase of total cancer for those with pre-existing DM was observed in men, although this association was not significant. The relationship between a history of DM and total cancer incidence is expressed as the sum of risks of individual site-specific cancer for a history of DM. To date, four prospective studies[35-38] and one case–control study[39] have been carried out in Japan, three of which adopted cancer incidence[35, 36, 39] and the others adopted cancer mortality.[37, 38] Except for one study,[36] these studies reported significant increase in the risk of total cancer among Japanese men and women; the relative risks of total cancer for those with DM were 1.27–1.44 among men and 1.21–1.88 among women.[35, 37-39] Previous studies carried out in North America and Western Europe have shown similar magnitudes of associations; the RRs ranged from 1.00 to 1.20 in men and from 1.10 to 1.51 in women.[40] Our findings in the present study did not contradict the results of previous studies and could contribute to the accumulation of evidence regarding the association between a history of DM and cancer in Japan.

Previous meta-analyses of prospective cohort studies on DM and liver cancer revealed a 1.87- to 2.23-fold significant increase in the risk of liver cancer associated DM.[7, 8] Two meta-analyses reported that DM is associated with increased risk of biliary tract cancer, which resulted in summary relative risks of 1.43 and 1.60.[15, 16] The present study confirmed a significantly increased risk of liver and biliary tract cancer incidence by DM in men. A few reports of meta-analyses assessing DM and pancreatic cancer found that individuals with a history of DM have a significant risk increase; the summary relative risks were 1.73–2.60.[9-11] In the present study, the positive association between DM and pancreatic cancer incidence in men reached statistical significance after excluding subjects who had developed pancreatic cancer in the first 3 years of the follow-up. This result supports the possibility that a history of DM is causally related to pancreatic cancer.

For other site-specific cancers, we observed a positive association of DM with laryngeal cancer in men and stomach cancer in women in the present analyses. Significant increase of laryngeal cancer risk in men did not change after controlling for pack-years of smoking instead of smoking status. However, we have to keep in mind that it is difficult to control residual confounding effects. To date, a few case–control studies addressing the association between DM and laryngeal cancer have provided inconsistent results.[39, 41, 42] Our finding of gender difference in the association between DM and the risk of stomach cancer was consistent with two previous studies in Japan.[35, 39] Women have been reported to have a susceptibility to stomach cancer by Helicobacter pylori,[43] which may explain our observations. Among the other studies carried out in Asia, however, four prospective studies revealed a significant positive association in both men and women[44-47] and one reported no association.[36] None of the studies carried out in North America and Western Europe observed a significant risk increase of stomach cancer in relation to DM.[23, 24] Different etiology according to sex and geographical area may exist in the progression of stomach cancer with DM.

Numerous reports suggest that the underlying pathological status of DM is an insulin resistance, that is, it induces a compensatory hyperinsulinemia, abnormal carbohydrate, and lipid metabolism and an increase in the serum level of IGF-1.[4-6] The effects of IGF-1 and insulin on tumor progression in in vitro studies have been observed in the tissues of the colon, breast, lung, bladder, and prostate.[48] However, relationships between a history of DM and lung and bladder cancers were inconsistent in previous epidemiological studies.[49-56] Meta-analyses of epidemiological studies on the association between DM and cancer of the colon, rectum, and breast concluded that a history of DM is associated with increased risk of these cancers; the summary HR ranged from 1.20 to 1.38 for cancer of the colon and rectum and from 1.20 to 1.27 for breast cancer.[12-14, 17-19] In the present study, however, we failed to find significant associations of a history of DM with occurrence of these cancers. We observed a non-significant decreased risk of prostate cancer among men with DM after excluding subjects diagnosed within 3 years from the beginning of the study. Our results were similar to those in other studies.[25-27]

The strength of the present study is its prospective design, which can prevent recall bias. Other advantages are that this study was carried out in a community-based cohort from the general Japanese population, the participation rate was relatively high, and long follow-up was achieved. Several potential confounders of the association between DM and cancer incidence were accounted for in the analysis, although the possibility of residual confounding factors cannot be denied. Several limitations of the current study should be considered. Statistical power was limited to estimate incident risks of the small number of cancers. Given the size of our population, the statistical powers to detect an HR of 1.09 among men and 1.35 among women for total cancer, at the two-sided 5% significance, were 73% and 90%, respectively. We did not have sufficient power for analyses of site-specific cancers except for liver in men (88%) and stomach in women (90%). For example, if the HR was over 1.42 for breast cancer, the statistical power could have reached over 80%. Definition of a history of DM was based on self-reports without information about the use of medications for diabetes and the validity and reliability of the self-reports were not confirmed. However, a previous study in Japan[57] confirmed that the self-reported history of DM from 96% of subjects in the questionnaire was consistent with the diagnosis in their medical records, indicating a substantial agreement with high positive predictive value. Nonetheless, the sensitivity of a self-reported history of DM based on HbA1c were low in our subsample. Similar sensitivity (46%) and specificity (98%) were reported from the study using the definition by hyperglycemia.[58] It is likely that a certain portion of subjects who had DM were classified as non-cases in the present study. This misclassification may have attenuated the association. We should consider the possibility that subjects who had DM and were classified as DM cases may tend to possess more risk factors of cancer than those who had DM but were classified as non-cases. However, numerous potential confounders were controlled in the present study. We did not obtain information about the use of some antidiabetic drugs that could improve insulin resistance and reduce the risk of cancer.[59] Some diabetic subjects may have started therapy after the initiation of the present study, which may have attenuated the real associations.

In summary, the current study suggests that a pre-existing DM is positively associated with the incidence of total cancer and certain types of cancers for men and women in Japan. Given the increasing epidemic of DM in recent years, the promotion of primary and secondary prevention programs aimed at those with DM will be important for cancer prevention.

Acknowledgments

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

This study was supported by grants from the Ministry of Education, Culture, Sports, Science, and Technology of Japan.

References

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