Detection of medications associated with Alzheimer's disease using ensemble methods and cooperative game theory.

OBJECTIVE: To study the feasibility of evaluating feature importance with Shapley Values and ensemble methods in the context of pharmacoepidemiology and medication safety. METHODS: We detected medications associated with Alzheimer's disease (AD) by examining the additive feature attribution with combined approach of Gradient Boosting and Shapley Values in the Medication use and Alzheimer's disease (MEDALZ) study, a nested case-control study of 70,719 verified AD cases in Finland. Our methodological approach is to do binary classification using Gradient boosting (an ensemble of weak classifiers) in a supervised learning manner. Then we apply Shapley Values (from cooperative game theory) to analyze how feature combinations affect the classification result. Medication use with a five to one year time-window before AD diagnosis was ascertained from Prescription register. RESULTS: Antipsychotics with low or medium dose, antidepressants with medium to high dose, and cardiovascular medications with medium to high dose were identified as the contributing features for separating cases with AD from controls. Medium to high amount of irregularity in the purchase pattern were an indicating feature for separating AD cases from controls. The similarity of medication purchases between AD cases and controls made the feature evaluation challenging. CONCLUSIONS: The combined approach of Gradient Boosting and feature evaluation with Shapley Values identified features that were consistent with findings from previous hypothesis-driven studies. Additionally, the results from the additive feature attribution identified new candidates for future studies on AD risk factors. Our approach also shows promise for studies based on observational studies, where feature identification and interactions in populations are of interest; and the applicability of using Shapley Values for evaluating feature relevance in pattern recognition tasks.

Long-term thiazide use and risk of low-energy fractures among persons with Alzheimer's disease-nested case-control study.

We investigated the association between thiazide use and the risk of low-energy fractures among community dwellers with Alzheimer's disease. Longer use was associated with a decreased risk of low-energy fractures. This study extends the previous knowledge of reduced fracture risk of thiazides to persons with Alzheimer's disease. INTRODUCTION: To investigate the association between thiazide use and the risk of low-energy fractures (LEF), and hip fracture among community dwellers with Alzheimer's disease (AD). No prior study has evaluated the effect of thiazides on LEF risk of AD patients. METHODS: LEF cases were identified from the MEDALZ study, including all community-dwelling persons diagnosed with AD in Finland 2005-2011. During the follow-up from AD diagnoses until the end of 2015, cases with LEF (N = 10,416) and hip fracture (N = 5578) were identified. LEF cases were matched with up to three controls without LEF, according to time since AD diagnosis, age and gender. Thiazide use identified from the Prescription register data was modeled with PRE2DUP method. Current use was defined in 0-30 days' time window before the fracture/matching date, and duration of current use was assessed. The association between thiazide exposure and LEFs was assessed with conditional logistic regression. RESULTS: Current thiazide use was observed in 10.5% of LEF cases and 12.5% of controls. Current thiazide use was associated with a decreased risk of LEF (adjusted OR [aOR] 0.83, 95% CI 0.77-0.88). In terms of the duration of use, no association was observed with short-term use (< 1 year or 1-3 years), while longer use (> 3 years) was associated with a reduced risk of LEF (aOR 0.77, 95% CI 0.71-0.83) and hip fracture (aOR 0.68, 95% CI 0.60-0.78). CONCLUSIONS: Our study extends the previous knowledge of reduced fracture risk of thiazides to persons with AD, a population with significantly increased background risk of fractures.

The risk of Alzheimer's disease associated with benzodiazepines and related drugs: a nested case-control study.

OBJECTIVE: To assess the association between benzodiazepine and related drug (BZDR) use and risk of Alzheimer's disease (AD) with cumulative consumption and duration of use based models. METHOD: A nationwide nested case-control study of all Finnish community-dwelling persons who received clinically verified AD diagnosis in 2005-2011 (N = 70 719) and their matched controls (N = 282 862). AD diagnosis was based on DSM-IV and NINCDS-ADRDA criteria. BZDR purchases were extracted from the Prescription Register since 1995. The association between BZDR use and AD was assessed using conditional logistic regression with 5-year lag time between exposure and outcome. RESULTS: Benzodiazepine and related drug use was associated with modestly increased risk of AD (adjusted OR 1.06, 95% CI 1.04-1.08). A dose-response relationship was observed with both cumulative consumption and duration. Adjustment for other psychotropics removed the cumulative dose-response relationship by attenuating the ORs in the highest dose category. CONCLUSION: Benzodiazepine and related drug use in general was associated with modestly increased risk of AD. No major differences were observed between different subcategories of BZDRs (i.e. benzodiazepines, Z drugs, short-/medium-acting or long-acting BZDRs). As dose-response relationship abolished after adjustment for other psychotropics, it is possible that the association may partially be due to antidepressants and/or antipsychotics, or concomitant use of these medications.

Proton pump inhibitor use and risk of hip fractures among community-dwelling persons with Alzheimer's disease-a nested case-control study.

BACKGROUND: Hip fractures are a major health concern among older persons with Alzheimer's disease, who usually use many concomitant drugs for several diseases. Evidence of the association between proton pump inhibitor use and risk of hip fracture is contradictory. AIM: To investigate whether the long-term use of proton pump inhibitor is associated with risk of hip fractures among community-dwelling persons with Alzheimer's disease. METHODS: In this nested case-control study, the nationwide MEDALZ data were utilised. Community-dwelling persons with Alzheimer's disease who encountered incident hip fracture (N = 4818; mean age 84.1) were included as cases. Four controls were matched for each case at the date of hip fracture (N = 19 235; mean age 84.0). The association between hip fracture and duration of current PPI use (ongoing use during 0-30 days before the index date), and cumulative duration of use during 10 years before was investigated with conditional logistic regression. RESULTS: Long-term or cumulative proton pump inhibitor use was not associated with an increased risk of hip fracture. Current proton pump inhibitor use was associated with an increased risk of hip fracture (adjusted OR 1.12, 95% CI 1.03-1.22). The risk was increased in short-term current use (<1 year) (adjusted OR 1.23, 95% CI 1.10-1.37). CONCLUSIONS: The increased risk of hip fracture was evident only in short-term proton pump inhibitor use, but no association was found for long-term or cumulative use. Thus, our findings do not support previous assumptions that long-term proton pump inhibitor use would be associated with an increased risk of hip fractures.

Hospital-treated mental and behavioral disorders and risk of Alzheimer's disease: A nationwide nested case-control study.

BACKGROUND: Studies investigating psychiatric disorders as Alzheimer's disease (AD) risk factors have yielded heterogeneous findings. Differences in time windows between the exposure and outcome could be one explanation. We examined whether (1) mental and behavioral disorders in general or (2) specific mental and behavioral disorder categories increase the risk of AD and (3) how the width of the time window between the exposure and outcome affects the results. METHODS: A nationwide nested case-control study of all Finnish clinically verified AD cases, alive in 2005 and their age, sex and region of residence matched controls (n of case-control pairs 27,948). History of hospital-treated mental and behavioral disorders was available since 1972. RESULTS: Altogether 6.9% (n=1932) of the AD cases and 6.4% (n=1784) of controls had a history of any mental and behavioral disorder. Having any mental and behavioral disorder (adjusted OR=1.07, 95% CI=1.00-1.16) or depression/other mood disorder (adjusted OR=1.17, 95% CI=1.05-1.30) were associated with higher risk of AD with 5-year time window but not with 10-year time window (adjusted OR, 95% CI 0.99, 0.91-1.08 for any disorder and 1.08, 0.96-1.23 for depression). CONCLUSIONS: The associations between mental and behavioral disorders and AD were modest and dependent on the time window. Therefore, some of the disorders may represent misdiagnosed prodromal symptoms of AD, which underlines the importance of proper differential diagnostics among older persons. These findings also highlight the importance of appropriate time window in psychiatric and neuroepidemiology research.

Differences in analgesic use in community-dwelling persons with and without Alzheimer's disease.

BACKGROUND: There are conflicting findings about analgesic use among persons with cognitive impairment compared to cognitively intact older persons. The objective of our study was to investigate the prevalence of analgesic use in community-dwelling persons with and without Alzheimer's disease (AD), within six months after AD diagnosis and to find out factors associated with the use of analgesics and specific analgesic groups. METHOD: We utilized data from register based MEDALZ (Medication use and Alzheimer's disease) cohort consisting of all community-dwelling persons diagnosed with AD during 2005-2011 in Finland and their matched comparison persons without AD. Altogether, 67,215 persons with AD and one comparison person for each case were included. Drug use data were collected from the Prescription Register and comorbidities from Special Reimbursement and Hospital Discharge Registers. RESULTS: Statistically significant (p < 0.001) yet mostly small differences were found for analgesics use: analgesics were used by 34.9% and 33.5% of persons with and without AD, respectively. Paracetamol was the most frequently used analgesic both among persons with (25.0%) and without AD (19.1%). Persons with AD used less frequently NSAIDs (Nonsteroidal Anti-inflammatory Drugs) (13.2% vs. 17.3%) and mild opioids (5.0% vs. 7.1%), while the use of strong opioids was more common in comparison to persons without AD (1.3% vs. 1.1%, respectively). Analgesic users were more likely women, aged ≥80 years, had asthma/COPD, cardiovascular disease, diabetes, cancer, hip fracture, osteoporosis, rheumatoid arthritis, and lower socioeconomic position. CONCLUSION: Further studies are needed to evaluate the adequateness of pain relief in older persons with and without AD. SIGNIFICANCE: Persons with Alzheimer's disease (AD) used more frequently paracetamol and less frequently NSAIDs and mild opioids. A decreasing trend of NSAID use was observed among persons with AD during the study period.

Serum 25-hydroxy-vitamin D and ionised calcium in relation to lung function and allergen skin tests.

BACKGROUND: Evidence suggests that higher levels of vitamin D and calcium are associated with greater lung function and that vitamin D is inversely associated with atopic sensitisation. It is unknown whether the associations of vitamin D and calcium with lung function are independent of each other or mediated by atopic sensitisation. OBJECTIVE: To study the associations of 25-hydroxyvitamin D (25(OH)D) and ionised calcium levels with lung function and specific allergen sensitisation in adolescents (12-19 years) and adults (20-59 years) and to assess whether the associations with lung function are due to altered atopic sensitisation. METHODS: This is a cross-sectional analysis of the data from the third National Health and Nutrition Examination Survey. RESULTS: The 25(OH)D levels were positively associated with forced vital capacity in adolescents (0.035 (95% confidence interval (CI): 0.007-0.064) s.d.; s.d. in model adjusted for multiple confounders). This association and the previously reported association between higher serum levels of 25(OH)D and better lung function in adults were independent of serum calcium levels, which were not associated with lung function. In adults, calcium was associated with sensitisation to grass allergens (odds ratio per s.d., 1.17 (95% CI: 1.03-1.32), 1.15 (95% CI: 1.01-1.31) and 1.18 (95% CI: 1.06-1.32) for white oak, Bermuda grass and short ragweed, respectively) and peanut odds ratio 1.21 (95% CI: 1.02-1.43) after adjusting for age, gender and race/ethnicity, but these associations attenuated towards the null after adjusting for additional confounders. The associations were independent of 25(OH)D levels, which were not associated with allergen sensitisation. CONCLUSIONS: Circulating levels of 25(OH)D are positively associated with lung function and this does not seem to be driven by allergen sensitisation or influenced by calcium levels.

Association of the FTO gene variant (rs9939609) with cardiovascular disease in men with abnormal glucose metabolism--the Finnish Diabetes Prevention Study.

BACKGROUND AND AIM: The common single nucleotide polymorphism (SNP) in the FTO (fat mass and obesity associated) gene has been consistently associated with an increased risk of obesity. We investigated whether the SNP rs9939609 (T/A) of the FTO is associated with risk factors of cardiovascular diseases (CVD), including serum levels of C - reactive protein (CRP), the chemokine RANTES (Regulated on Activation, Normal T Cell Expressed and Secreted; CCL5), and serum and lipoprotein lipids in the Finnish Diabetes Prevention Study (DPS). Furthermore, we examined whether the rs9939609 increased the CVD risk in the DPS and if these results could be replicated in a larger cross-sectional population-based random sample of Finnish men (the METSIM). METHODS AND RESULTS: In the DPS, altogether 490 (BMI≥25kg/m(2)) subjects with impaired glucose tolerance were genotyped for rs9939609. Cardiovascular morbidity and mortality data were collected during the median follow-up of 10.2 years. The replication study was a population-based cross-sectional study of 6214 men. In the DPS, the AA genotype of rs9939609 was associated, independently of BMI, with increased RANTES (p=0.002) and decreased HDL cholesterol concentrations (p=0.007) in men. During the follow-up, the AA genotype was associated with an adjusted 2.09-fold risk (95% CI 1.17-3.73, p=0.013) of CVD in men. In the METSIM Study, the association with a history of myocardial infarction was replicated in the subgroup of men with type 2 diabetes. CONCLUSION: We suggest that the variation in the FTO gene may contribute to the development of CVD in men with an abnormal glucose metabolism.

Association of sequence variations in the gene encoding insulin-like growth factor binding protein 5 with adiponectin.

BACKGROUND: Insulin-like growth factor binding protein 5 (IGFBP5) binds to IGF and thus modulates IGF signaling pathway. We have shown earlier that the IGFBP5 gene was downregulated in the adipose tissue after 12-week carbohydrate diet with low insulinemic response. OBJECTIVE: The aim was to examine the putative contribution of genetic variation of the IGFBP5 gene to the characteristics of metabolic syndrome and incidence of type 2 diabetes (T2DM) in the Finnish Diabetes Prevention Study (DPS). METHODS: DPS is a longitudinal study where 522 subjects with impaired glucose tolerance were randomized to either lifestyle intervention group or control group. DNA was available from 507 subjects (mean body mass index (BMI) 31.2+/-4.5 kg/m(2), age 55+/-7 years). The eight single-nucleotide polymorphisms (SNPs) were selected from HapMap database and genotyped by Taqman allelic discrimination protocol. The main results were confirmed in a larger cross-sectional study population (METSIM). In addition, the gene expression of IGFBP5 was studied in two previously published study populations (FUNGENUT and GENOBIN) of 124 subjects with insulin resistance (BMI 32.2+/-3.5 kg/m(2), age 57.7+/-7.4 years). RESULTS: Three out of eight IGFBP5 markers (rs9341234, rs3276 and rs11575134) were significantly associated with circulating adiponectin concentrations in men. Furthermore, mRNA expression studies of subcutaneous adipose tissue showed that mRNA concentrations of IGFBP5 correlated with adiponectin concentrations in all subjects and in women. None of the IGFBP5 SNPs were associated with T2DM. CONCLUSIONS: Our findings show that IGFBP5 has a gender-specific association with adiponectin, which may modulate the development of metabolic syndrome.

The genetic variation in the tenomodulin gene is associated with serum total and LDL cholesterol in a body size-dependent manner.

We have reported that the sequence variation in the tenomodulin (TNMD) gene is associated with the risk of type 2 diabetes (T2DM), central obesity and serum levels of systemic immune mediators in the Finnish Diabetes Prevention Study (DPS), which is a longitudinal lifestyle intervention study on 522 middle-aged persons with impaired glucose tolerance (IGT). The aim of this study was to investigate whether the association with T2DM, observed in the DPS could be replicated in a larger, cross-sectional population-based random sample of 5298 men (3020 with normoglycaemia, 984 with impaired fasting glucose, 436 with IGT and 811 with T2DM) from the region of Kuopio, eastern Finland. To further explore the putative mechanisms linking TNMD to T2DM and metabolic syndrome, we studied the associations of TNMD sequence variation with lipid abnormalities characteristic to metabolic syndrome. The association with T2DM risk was not replicated, but significant associations were found with serum low-density lipoprotein and total cholesterol in a body mass index-dependent manner. These associations were also observed in the men of DPS, whereas in women these associations were not significant. These results from two independent study populations suggest that the genetic variation in TNMD could modulate cholesterol metabolism in obese men.