RESUMEN
INTRODUCTION: German epidemiologic cancer registries may store only encrypted personal identifiers. Thus, record linkage with secondary databases needs to be performed via procedures that are based on encrypted identifiers. In this paper, we describe the linkage of patient data from a statutory health insurance company (AOK NordWest) and from the Disease Management Program for diabetes mellitus type 2 with the database of the cancer registry. We report the cancer incidence in patients with type 2 diabetes (T2D). METHODS: Personal identifying variables of the patient cohort were encrypted before being sent electronically to the cancer registry and submitted to a probabilistic record linkage with registry data. The study included T2D patients who were residents of the Münster, Detmold, or Arnsberg districts and who were aged 40-79 years. Only primary cancers occurring between the date of enrolment and the censoring date (31 December 2010) were included. The standardized incidence ratio (SIR) was calculated relative to the number of incident cases expected on the basis of the averaged incidence rates in the general population. RESULTS: The record linkage took about 3 weeks of processing time. A total of 67,447 T2D (49.2 % men) cases were included for analyses. Incident cancer was diagnosed in 2,086 men and 1,578 women. Cohort members showed an elevated risk for cancer of the liver (SIR =1.86; 95% CI =1.47-2.31), pancreas (SIR = 1.62; 95 % CI =1.36-1.91), lung (SIR = 1.21; 95% CI 1.11-1.32), and uterus (SIR = 1.34; 95 % CI 1.08-1.65), and they were less likely to be diagnosed with prostate cancer (SIR =0.72; 95% CI = 0.65-0.79). DISCUSSION: The findings of this study suggest that record linkage of secondary databases with cancer registry data for research purposes can be effectively carried out in compliance with strict data-protection regulations.
Asunto(s)
Seguridad Computacional/estadística & datos numéricos , Minería de Datos/estadística & datos numéricos , Diabetes Mellitus Tipo 2/epidemiología , Registro Médico Coordinado/métodos , Sistemas de Registros Médicos Computarizados/estadística & datos numéricos , Neoplasias/epidemiología , Sistema de Registros , Adulto , Anciano , Estudios de Cohortes , Comorbilidad , Femenino , Alemania/epidemiología , Humanos , Incidencia , Masculino , Persona de Mediana Edad , Factores de RiesgoRESUMEN
AIMS/HYPOTHESIS: Recent prospective studies found an elevated cancer risk shortly after diabetes diagnosis, and this was probably due to increased ascertainment. This study investigated whether site-specific cancer risks are also raised following enrolment in a disease management programme for type 2 diabetes mellitus (DMP-DM2). METHODS: We linked records from a DMP-DM2 to population cancer registry data. The study period was from June 2003 to December 2009. Standardised incidence ratios (SIRs) were calculated for time intervals following DMP enrolment using the cancer incidence rates of the general source population. Additionally, Poisson regression with natural splines was used to assess time-dependent cancer incidence by diabetes duration. RESULTS: There were 2,034 first invasive cancer cases identified over 163,738 person-years of follow-up. Pancreatic cancer risk was significantly increased mainly in the first year after enrolment (SIR 1.62); the increment was only seen for patients in whom diabetes had been diagnosed less than 1 year before DMP-DM2 enrolment. Risk of endometrial cancer was similarly raised in the first year after DMP-DM2 enrolment among individuals newly diagnosed with diabetes but decreased rapidly thereafter. There was no time dependence in the incidence of cancers of the liver, lung, colon, breast and prostate. CONCLUSIONS/INTERPRETATION: Enrolment in a DMP-DM2 did not appear to induce ascertainment bias for most cancers. Cancer risks were initially increased, especially for pancreatic cancer, potentially as a result of reverse causality. Ascertainment bias and time-dependent incidence of cancer appear to be less of a problem in settings using DMP-like structures for the study of the association between diabetes duration, glucose-lowering medication and cancer incidence.
Asunto(s)
Diabetes Mellitus Tipo 2 , Neoplasias/diagnóstico , Anciano , Manejo de la Enfermedad , Femenino , Humanos , Masculino , Persona de Mediana Edad , Estudios ProspectivosRESUMEN
AIMS: We evaluated the patterns and determinants that influence the selection, timing and duration of first-line antihyperglycaemic drug (AHD) treatment in patients with type 2 diabetes in Germany, focusing specifically on treatment-naive AHD initiators. METHODS: Pharmacy dispensing claims data were linked with a cohort of patients newly enrolled in a German Disease Management Program for type 2 diabetes (DMP-DM2) between 2003 and 2009. We examined uptake of first-line pharmacotherapy in previously unmedicated patients and identified predictors of receiving AHD therapy in general and metformin in particular using multivariable regression analyses. RESULTS: There were 27,138 unmedicated patients with type 2 diabetes and 47.0% of them were started on AHD treatment within 5 years after enrollment. Initial severity of diabetes was the major predictor of receiving first-line pharmacotherapy. Metformin accounted for 63% of newly prescribed AHD in 2003 and more than 80% in 2009 while sulfonylureas accounted for only 10%. Initiating metformin as first-line AHD was associated with younger age, higher BMI, lower HbA1c, and shorter diabetes duration (multivariate p<0.001 for all). Therapy switch or step-up was less frequent among metformin initiators than sulfonylurea initiators. CONCLUSIONS: The majority of patients were not started on AHD therapy within 5 years after enrollment. In line with recent therapy guidelines, current first-line antihyperglycaemic treatment was increasingly based on metformin. AHD initiators started on sulfonylurea were generally more advanced in their disease and were started later on primary pharmacotherapy.