Individuals with diabetes, specifically those with type 2 diabetes mellitus (T2DM) have a higher risk for many cancers such as the breast, colon and rectum, endometrium, liver, and pancreatic cancers as compared to those without diabetes [
In the case of prostate cancer, an inverse relationship between diabetes and cancer risk has been observed [
The main modality of treatment for diabetes is pharmacotherapy with antidiabetes drugs such as metformin, thiazolidinediones, sulfonylureas, and insulin. All classes of antidiabetes drugs may indirectly affect the risk of prostate cancer by controlling hyperglycemia. Of special interest is the use of metformin for diabetes management because of its unique actions on insulin resistance and hyperinsulinemia [
Therefore, the current study is conducted to investigate the association between metformin use and cancer stage at diagnosis among elderly fee-for-service Medicare beneficiaries with T2DM and incident prostate cancer.
A cohort study design was adopted with a baseline and an index date as depicted in Figure
Schematic presentation of study design to examine the relationship between metformin use and cancer stage at diagnosis.
Data were derived from the SEER-Medicare linked database. The SEER data comprised 18 population-based cancer registries having precise and accurate information on all newly diagnosed cancer cases since 1973. At present, the data consisted of a total of 1,847,363 cases of all cancers and 340,769 cases of prostate cancer among the elderly population whose age was at least 65 at the time of diagnosis of cancer presenting. With 98% ascertainment of cases with the medical records and the highest level of certification of data quality from the North American Association of Central Cancer Registries, the SEER data are considered to be the most comprehensive and high-quality population-based data on cancer incidence and their treatment and outcomes. The Patient Entitlement and Diagnosis Summary File (PEDSF) provides information on the cancer diagnosis up to ten cancers, types of cancers, cancer stage, individual’s demographic attributes, marital status, and tumor characteristics at the time of cancer diagnosis. The Medicare is the primary health insurer for 97% of the US population aged 65 years and older [
The study cohort comprised 45,618 men with incident prostate cancer diagnosed between 2008 and 2009. We excluded 42,966 cases for the following reasons: those diagnosed with prostate cancer during the autopsy; those who had multiple cancers; those who had carcinoma in situ; those aged 65 years and younger; those who died during the study period; those enrolled in Medicare Health Maintenance Organizations; those not continuously enrolled in Medicare Parts A, B, and D during the study period, and those missing cancer stage at diagnosis. The details on the study cohort selection process are provided in Figure
Study cohort development flow diagram for study population of elderly Medicare beneficiaries diagnosed with prostate cancer and diabetes.
After all the exclusions, the final study cohort consisted of 2,652 elderly men with type 2 diabetes mellitus (T2DM) and incident prostate cancer. To note, every year nearly half of enrolled Medicare beneficiaries in Parts A and B are also enrolled in Medicare Part D since 2006. Therefore, our study population with prostate cancer and T2DM reduced to 7,424 enrolled in Medicare Parts A and B to 2,652 with enrollment Part D. To reduce the selection bias, we compared the characteristics of elderly men with T2DM and incident prostate cancer enrolled in Medicare Parts A and B to those enrolled in Parts A, B, and D as shown in the Appendix. These two sets of population shared similar characteristics.
The American Joint Committee on Cancer (AJCC) Tumor-Node-Metastases (TNM) classification was used to identify the stage of prostate cancer from the PEDSF file. Based on the AJCC-TNM systems, men were classified as having localized cancer stage if they had T1 or T2 clinical stage with no regional lymph node (NX-N0) involvement and absence of any distant metastasis (M0) [
Metformin use was identified using the Medicare Part D files. Metformin prescriptions were identified using the national drug codes (NDCs). Men with at least one prescription for metformin during the baseline period were considered as metformin users and men without any prescriptions for metformin were considered as nonusers.
We utilized the Anderson Healthcare Behavior and Utilizations Model (ABM) model [
Significant group differences in the study population characteristics by metformin use were examined with chi-square tests. A binary logistic regression was used to determine the associations between predisposing, enabling, need, and external environment characteristics and metformin use. C-statistics and area under the curve were used to assess the model fit. The logistic regression was used to derive inverse probability treatment weights (IPTWs) and these standardized IPTWs were used to control for the observed selection bias in regressions on the cancer stage.
Significant unadjusted associations between metformin use and cancer stage at diagnosis were examined with chi-square tests. The IPTW-adjusted multivariable logistic regressions were used to analyze the relationship between metformin use and cancer stage at diagnosis. As the odds ratios and relative risk are approximately similar for the events with low prevalence, such as advanced prostate cancer (≤10%) [
The study cohort consisted of 2,652 elderly men with preexisting T2DM and incident prostate cancer between 2008 and 2009. Table
Characteristics of the study cohort elderly Medicare beneficiaries with diabetes and incident prostate cancer by metformin use SEER-Medicare linked database, 2007–2010.
All | Overall | Metformin users | Non-metformin-user | Sig. | |||
---|---|---|---|---|---|---|---|
|
% |
|
% |
|
% | ||
2,652 | 100 | 948 | 35.7 | 1,704 | 64.3 | ||
|
|||||||
|
|||||||
|
|
||||||
66–74 | 1,579 | 59.5 | 609 | 64.2 | 970 | 56.9 | |
75+ | 1,073 | 40.5 | 339 | 35.8 | 734 | 43.1 | |
|
| ||||||
Whites | 1,914 | 72.2 | 680 | 71.7 | 1,234 | 72.4 | |
African American | 323 | 12.2 | 108 | 11.4 | 215 | 12.6 | |
Hispanic/Latino | 150 | 5.7 | 70 | 7.4 | 80 | 4.7 | |
Others | 265 | 10.0 | 90 | 9.5 | 175 | 10.3 | |
|
|||||||
Unmarried | 222 | 8.4 | 73 | 7.7 | 149 | 8.7 | |
Married | 1,550 | 58.4 | 581 | 61.3 | 969 | 56.9 | |
Divorced/separated | 393 | 14.8 | 136 | 14.3 | 257 | 15.1 | |
Others | 487 | 18.4 | 158 | 16.7 | 329 | 19.3 | |
|
|||||||
|
|||||||
|
|||||||
|
| ||||||
|
664 | 25.0 | 265 | 28.0 | 399 | 23.4 | |
|
664 | 25.0 | 243 | 25.6 | 421 | 24.7 | |
|
664 | 25.0 | 229 | 24.2 | 435 | 25.5 | |
|
660 | 24.9 | 211 | 22.3 | 449 | 26.3 | |
|
|||||||
0–8.52 | 666 | 25.1 | 222 | 23.4 | 444 | 26.1 | |
8.53–15.16 | 655 | 24.7 | 229 | 24.2 | 426 | 25.0 | |
15.17–26.09 | 664 | 25.0 | 248 | 26.2 | 416 | 24.4 | |
26.1–100 | 667 | 25.2 | 249 | 26.3 | 418 | 24.5 | |
|
|||||||
Yes | 2,415 | 91.1 | 872 | 92.0 | 1,543 | 90.6 | |
No | 237 | 8.9 | 76 | 8.0 | 161 | 9.4 | |
|
|||||||
Yes | 1,858 | 70.1 | 672 | 70.9 | 1,186 | 69.6 | |
No | 794 | 29.9 | 276 | 29.1 | 518 | 30.4 | |
|
|
||||||
Yes | 1,576 | 59.4 | 656 | 69.2 | 920 | 54.0 | |
No | 1,076 | 40.6 | 292 | 30.8 | 784 | 46.0 | |
|
|
||||||
Yes | 102 | 3.8 | 62 | 6.5 | 40 | 2.3 | |
No | 2,550 | 96.2 | 886 | 93.5 | 1,664 | 97.7 | |
|
|||||||
|
|||||||
|
|||||||
|
|
||||||
Northeast | 487 | 18.4 | 146 | 15.4 | 341 | 20.0 | |
South | 570 | 21.5 | 223 | 23.5 | 347 | 20.4 | |
North-central | 351 | 13.2 | 112 | 11.8 | 239 | 14.0 | |
West | 1,244 | 46.9 | 467 | 49.3 | 777 | 45.6 | |
|
|||||||
|
|||||||
|
|||||||
|
|
||||||
0 to 0 | 846 | 31.9 | 305 | 32.2 | 541 | 31.7 | |
1 to 1 | 476 | 17.9 | 195 | 20.6 | 281 | 16.5 | |
2 to 3 | 811 | 30.6 | 297 | 31.3 | 514 | 30.2 | |
4 to 13 | 519 | 19.6 | 151 | 15.9 | 368 | 21.6 |
Notes: based on the data of 2,652 elderly men aged 66 years and older diagnosed with prostate cancer between 2008 and 2009 using a surveillance, epidemiology, and end-results- (SEER-) Medicare linked Part D data. Significant group differences by metformin use are based on chi-square tests. % represented in the column is row percentages.
DCSI: diabetes complication severity index; PCP: primary care physician; PSA: prostate specific antigen; Sig.: level of significance.
Table
Table
Adjusted odds ratios and 95% confidence intervals from logistic regressions on metformin use among elderly Medicare beneficiaries with diabetes incident prostate cancer surveillance, epidemiology, and end-results- (SEER-) Medicare linked data, 2007–2010.
AOR | 95% CI | Sig. | |
---|---|---|---|
|
|||
|
|||
|
|||
66–74 years | 1.31 |
|
|
75 years or above | Ref. | ||
|
|||
Whites | Ref. | ||
African American | 0.91 |
|
|
Latino | 1.62 |
|
|
Others | 0.80 |
|
|
|
|||
Married | Ref. | ||
Unmarried | 0.81 |
|
|
Divorced/separated | 0.94 |
|
|
Others | 0.83 |
|
|
|
|||
|
|||
|
|||
|
|||
|
1.72 |
|
|
|
1.34 |
|
|
|
1.18 |
|
|
|
Ref. | ||
|
|||
0–8.52 | 1.24 |
|
|
8.53–15.16 | 1.21 |
|
|
15.17–26.09 | 1.18 |
|
|
26.1–100 | Ref | ||
|
|||
Yes | 1.12 |
|
|
No | Ref. | ||
|
|||
Yes | 1.94 |
|
|
No | Ref | ||
|
|||
Yes | 2.95 |
|
|
No | Ref. | ||
|
|||
|
|||
|
|||
|
|||
0 to 1 | Ref. | ||
2 to 2 | 1.13 |
|
|
3 to 3 | 0.98 |
|
|
4 to 13 | 0.65 |
|
|
|
|||
|
|||
|
|||
|
|||
Northeast | Ref. | ||
South | 1.40 |
|
|
North-central | 1.07 |
|
|
West | 1.37 |
|
|
Notes: based on the data of 2,652 elderly men aged 66 years and older diagnosed with prostate cancer between 2008 and 2009 using a surveillance, epidemiology, and end-results- (SEER-) Medicare linked Part D data. Significant group differences are based on log-likelihood test for metformin use.
DCSI: diabetes complication severity index; PCP: primary care physician; Ref.: reference group; Sig.: level of significance;
Table
(a) Number and IPTW-adjusted percentage of men with localized versus advanced stage of prostate cancer by metformin use among elderly Medicare beneficiaries with diabetes and incident prostate cancer surveillance, epidemiology, and end-results- (SEER-) Medicare linked database, 2007–2010. (b) Unadjusted and adjusted odds ratios and 95% confidence intervals from (IPTW) logistic regressions for advanced stage at diagnosis of cancer.
Localized | Advance | Sig. | |||
---|---|---|---|---|---|
Overall |
|
Weighted% |
|
Weighted% | |
2,493 | 93.7 | 159 | 6.3 | ||
|
| ||||
Yes | 902 | 95.3 | 46 | 4.7 | |
No | 1,591 | 93.3 | 113 | 6.7 |
Advanced stage at diagnosis IPTW | |||||
|
|||||
|
|||||
OR | 95% CI | Sig. | |||
|
|||||
|
|||||
Yes | 0.69 | [0.49, 0.95] |
|
||
No | Ref. | ||||
|
|||||
|
|||||
AOR | 95% CI | Sig. | |||
|
|||||
|
|||||
Yes | 0.68 | [0.48, 0.97] |
|
||
No | Ref. |
Notes: based on the data of 2,652 elderly men aged 66 years and older diagnosed with prostate cancer between 2008 and 2009 using a surveillance, epidemiology, and end-results- (SEER-) Medicare linked data. % is weighted percentage for IPTW. Significant differences are based on the log-likelihood test using a logistic regression with IPTW weights. Adjusted model controlled for predisposing, enabling, need, and external environment related factors.
IPTW: inverse probabilities treatment weights; PSA: prostate specific antigen level; PCP: primary care physician visit; Sig.: level of significance;
Table
The current study is the first largest population-based study to examine whether the risk of advanced prostate cancer diagnosis is reduced with metformin use among elderly men with preexisting T2DM and incident prostate cancer in the US. Characteristics of metformin users were consistent with literature. We observed that a higher percentage of men living in the Western region of the united states were prescribed metformin as compared to those living in other regions. Census-track level income was also associated with metformin use. These study findings are consistent with one published study on geographical disparities in antidiabetes medications. Regional disparities in metformin use can reflect practice patterns in regions [
We found that the risk for advanced stage cancer diagnosis is reduced with metformin use among elderly men with T2DM and incident prostate cancer after controlling for the observed section bias between metformin users and nonusers and other independent variables. The current study addressed the limitations of the single population-based examination of metformin use and the cancer stage by incorporating a validated diabetes severity complications index and controlling for the observed selection bias. The current study findings are consistent with the preclinical evidence on the role of metformin in prevention of advanced prostate cancer [
It should be noted that the current study controlled for the observed selection bias because metformin users and nonusers were significantly different with respect to their predisposing, enabling, need, and external environment characteristics. Without adjustments for the observed selection bias, there was not a statistically significant difference in the cancer stage between metformin users and nonusers. Therefore, accounting for the observed selection bias is important in establishing an association between metformin use and the reduction in the risk of advanced prostate cancer diagnosis.
The current study has a number of strengths. The large cohort size and high-quality data on the clinical and pathological features of cancer at the time of diagnosis enabled us to examine not only the association between metformin use and the cancer stage diagnosis but also the initial choice of cancer treatment [
The current study has some limitations as well. The prescription claims for metformin and other drugs were used. Filling the prescriptions cannot be equated to the actual use of these drugs. We made an attempt to overcome this issue to some extent via measuring the one-year adherence to medication, and a lower proportion of those with adherence to metformin and nonadherent metformin had advanced prostate cancer as compared to nonusers. Due to lack of sufficient sample size, we did not present the results in tabular format. However, the direction of our findings is consistent with the previous studies suggesting no difference in the ever metformin users and adherent metformin users on grade of prostate cancer [
Metformin use was associated with a statistically significant reduction in the risk advanced prostate cancer diagnosis among elderly men with T2DM and incident prostate cancer. The current study findings highlight the need for additional studies in this area. Other population-based studies need to be conducted to confirm the study findings. If confirmed, randomized controlled trials can be carried out to examine the causal link between metformin use and the risk of advanced prostate cancer diagnosis.
See Tables
Codes and algorithms to identify diabetes complication severity index (DCSI) developed by Young et al. and modified by Chang et al.
Complications | ICD-9-CM code | DCSI score |
---|---|---|
|
||
Diabetic ophthalmologic disease | 250.5x | 1 |
Background retinopathy | 362.01 | 1 |
Other retinopathies | 362.1 | 1 |
Retinal edema | 362.83 | 1 |
CSME | 362.53 | 1 |
Other retinal disorders | 362.81, 362.82 | 1 |
Proliferative retinopathy | 362.02 | 2 |
Retinal detachment | 361.xx | 2 |
Blindness | 369.xx.00–0.99 | 2 |
Vitreous hemorrhage | 379.23 | 2 |
|
||
|
||
Diabetic nephropathy | 250.4 | 1 |
Acute glomerulonephritis | 580 | 1 |
Nephrotic syndrome | 581 | 1 |
Hypertension, nephrosis | 581.81 | 1 |
Chronic glomerulonephritis | 582 | 1 |
Nephritis/nephropathy | 583 | 1 |
Chronic renal failure | 585 | |
Renal failure NOS | 586 | |
Renal insufficiency | 593.9 | |
|
||
|
||
Diabetic nephropathy | 356.9, 250.6 | 1 |
Amyotrophy | 358.1 | 1 |
Cranial nerve palsy | 951.0, 951.1, 951.3 | 1 |
Mononeuropathy | 354.0–355.9 | 1 |
Charcot’s arthropathy | 713.5 | 1 |
Polyneuropathy | 357.2 | 1 |
|
||
|
||
TIA | 435 | 1 |
Stroke | 431, 433, 434, 436 | 2 |
|
||
|
||
Atherosclerosis | 440.xx | 1 |
Other IHD | 411 | 1 |
Angina pectoris | 413 | 1 |
Other chronic IHD | 414 | 1 |
Myocardial infarction | 410 | 2 |
Ventricular fibrillation, arrest | 427.1, 427.3 | 2 |
Atrial fibrillation, arrest | 427.4, 427.5 | 2 |
Other ASCVD | 429.2 | 1 |
Old myocardial infarction | 412 | 2 |
Heart failure | 428 | 2 |
Atherosclerosis, severe | 440.23, 440.24 | 2 |
Aortic aneurysm/dissection | 441 | 2 |
|
||
|
||
Diabetic PVD | 250.7 | 1 |
Other aneurysms, LE | 442.3 | 1 |
PVD | 443.81, 443.9 | 1 |
Foot wound + complication | 892.1 | 1 |
Claudication, intermittent | 443.9 | 1 |
Embolism/thrombosis (LE) | 444.22 | 2 |
Gangrene | 785.4 | 2 |
Gas gangrene | 0.4 | 2 |
Ulcer of lower limbs | 707.1 | 2 |
|
||
|
||
Ketoacidosis | 250.1 | 2 |
Hyperosmolar | 250.2 | 2 |
Other comas | 250.3 | 1 |
Note: the table is adapted from the previous algorithm defined by Young et al. and modified by Change et al. to identify the severity of diabetes using claims database. Severity index was based on a scale ranging from 0 to 2 for each complication as follows: 0 = no abnormality, 1 = some abnormality, and 2 = severe abnormality.
ASCVD, atherosclerotic cardiovascular disease; CSME, cystoid macular edema/degeneration; DCSI, diabetes complications severity index; IHD, ischemic heart disease; ICD-9-CM, international classification of diseases, ninth revision, clinical modification; LE, lower extremity; NOS, not otherwise specified; PVD, peripheral vascular disease; TIA, transient ischemic attack.
Baseline characteristics of elderly men with prostate cancer with T2DM overall and enrolled in Part D program SEER-Medicare linked database, 2008-2009.
All | DM Part D ( |
DM overall ( |
---|---|---|
% | % | |
|
||
66–74 | 59.5 | 59.4 |
75+ | 40.5 | 40.6 |
|
||
|
||
Whites | 72.2 | 75.3 |
African American | 12.2 | 14.5 |
Hispanic/Latino | 5.7 | 3.0 |
Others | 10.0 | 7.1 |
|
||
|
||
Unmarried | 8.4 | 6.6 |
Married | 58.4 | 63.2 |
Divorced/separated | 14.8 | 14.0 |
Others | 18.4 | 16.2 |
|
||
|
||
|
25.0 | 24.9 |
|
25.0 | 25.1 |
|
25.0 | 25.0 |
|
24.9 | 25.0 |
|
||
|
||
0–8.52 | 25.1 | 25.0 |
8.53–15.16 | 24.7 | 25.0 |
15.17–26.09 | 25.0 | 25.1 |
26.1–100 | 25.2 | 25.0 |
|
||
|
||
Yes | 70.1 | 65.0 |
No | 29.9 | 35.0 |
|
||
|
||
Yes | 91.1 | 88.4 |
No | 8.9 | 11.6 |
|
||
|
||
Northeast | 18.4 | 19.3 |
South | 21.5 | 24.9 |
North-central | 13.2 | 11.8 |
West | 46.9 | 44.0 |
|
||
|
||
0 to 1 | 26.1 | 26.1 |
2 to 6 | 24.7 | 24.7 |
7 to 22 | 25.3 | 25.3 |
23 to 147 | 23.9 | 23.9 |
|
||
|
||
0 to 3 | 24.5 | 24.5 |
4 to 16 | 24.7 | 24.7 |
17 to 44 | 26.2 | 26.2 |
45 to 343 | 24.5 | 24.5 |
|
||
|
||
2008 | 50.9 | 51.6 |
2009 | 49.1 | 48.4 |
The authors have no competing interests to declare.
Amit D. Raval and Usha Sambamoorthi wrote the paper and researched data. Suresh Madhavan, Malcolm D. Mattes, Wenhui Wei, and Xiaoyun Pan reviewed/edited the paper.
This project was supported by Agency for Health Research and Quality (AHRQ) Grant no. R24HS018622-03 and National Institute of General Medicine Sciences (Grant U54GM104942). The content is solely the responsibility of the authors and does not necessarily represent the official views of AHRQ and NIH.