Scientific
publications

BACKGROUND: Obesity substantially increases the risk of the development of chronic kidney disease. Adipose tissue expresses all of the components of the renin-angiotensin system (RAS), contributing to the high prevalence of hypertension in obese patients and driving renal hyperfiltration and subsequent glomerular injury. METHODS: We performed a retrospective cohort study using a United Kingdom primary care database, evaluating the effect of time-updated exposure to RAS blockade versus all other antihypertensive medications in obese, hypertensive, non-diabetic patients. We used Cox proportional hazards modeling with and without marginal structural modeling to assess the hazards of developing a primary outcome of 50% reduction in estimated glomerular filtration rate (eGFR) (across 2 consecutive values), end stage renal disease or death. RESULTS: A total of 219,701 patients met inclusion criteria, with a median 7.2 years of follow-up. Median baseline eGFR was 72.6 ml/min/1.73 m2. Compared to other antihypertensive medications, patients treated with RAS blockade had a modestly elevated hazard of adverse renal outcomes using traditional Cox regression (hazard ratio (HR) 1.04, 95% CI 1.01-1.07) and no significantly increased hazard by marginal structural modeling (HR 1.02, 95% CI 0.97-1.08). Patients treated with RAS blockade had a significantly reduced hazard of incident diabetes, but no significant difference in mortality. CONCLUSION: This study, conducted in a large real-world cohort, provides evidence that RAS blockade may not provide benefit with regard to longitudinal renal outcomes in obese, hypertensive patients. Further research is needed to elucidate the hemodynamic and renoprotective role of antihypertensive medications in obese patients.

BACKGROUND: UK Dementia Strategies prioritise fair access to mental and physical healthcare. We investigated whether there are inequalities by deprivation or gender in healthcare received by people with dementia, and compared healthcare received by people with and without dementia. METHODS: we investigated primary care records of 68,061 community dwelling dementia patients and 259,337 people without dementia (2002-13). We tested hypotheses that people with dementia from more deprived areas, and who are women receive more psychotropic medication, fewer surgery consultations, are less likely to receive annual blood pressure, weight monitoring and an annual review, compared with those from less deprived areas and men. RESULTS: only half of people with dementia received a documented annual review. Deprivation was not associated with healthcare received. Compared to men with dementia, women with dementia had lower rates of surgery consultations (adjusted incidence rate ratio (IRR) 0.90, 95% CI 0.90-0.91), of annual blood pressure monitoring (adjusted IRR 0.96, 95% CI 0.95-0.97) and of annual weight monitoring (adjusted IRR 0.91, 95% CI 0.90-0.93). Men with dementia were less likely to be taking psychotropic medication than women with dementia. People with dementia had fewer surgery consultations and were less likely to have their weight and blood pressure monitored at least annually, compared to the non-dementia group. CONCLUSIONS: people with dementia, in particular women, appear to receive less primary healthcare, but take more psychotropic medication that may negatively impact their physical health. Reducing these inequalities and improving access of people with dementia to preventative healthcare could improve the health of people with dementia.

BACKGROUND: Various definitions of hyperkalaemia have been used in clinical research, and data from routine clinical practice on its incidence are sparse. We aimed to establish the incidence of hyperkalaemia in patients with newly diagnosed heart failure in the UK general population using different definitions for the condition. METHODS: We conducted a large retrospective cohort study using data from The Health Improvement Network primary care database. Patients with newly diagnosed heart failure (N = 19,194) were identified and followed until the first occurrence of hyperkalaemia. Different serum potassium (K(+)) thresholds were evaluated as possible definitions for hyperkalaemia, and incidence rates (IRs) calculated using a final operational definition both overall and among patient sub-groups. RESULTS: IRs of hyperkalaemia ranged from 0.92-7.93 per 100 person-years according to the definition. Based on considerable differences in the serum K(+) normal range used between practices, 2176 (11.3 %) individuals were identified with a record of hyperkalaemia using our operational definition of a proportional increase of ≥10 % above the upper bound of the normal range: IR 2.90 per 100 person-years (95 % CI 2.78-3.02) over a mean follow-up of 3.91 years. Incidence rates were higher in older patients, and in those with diabetes or renal impairment. CONCLUSIONS: Hyperkalaemia is a common finding in heart failure patients in primary care, but its incidence can vary nearly ten-fold depending on its definition. Since assessment of hyperkalaemia risk is essential for therapeutic decision making in heart failure patients, this finding warrants consideration in future epidemiological studies.

BACKGROUND: frailty is an especially problematic expression of population ageing. International guidelines recommend routine identification of frailty to provide evidence-based treatment, but currently available tools require additional resource. OBJECTIVES: to develop and validate an electronic frailty index (eFI) using routinely available primary care electronic health record data. STUDY DESIGN AND SETTING: retrospective cohort study. Development and internal validation cohorts were established using a randomly split sample of the ResearchOne primary care database. External validation cohort established using THIN database. PARTICIPANTS: patients aged 65-95, registered with a ResearchOne or THIN practice on 14 October 2008. PREDICTORS: we constructed the eFI using the cumulative deficit frailty model as our theoretical framework. The eFI score is calculated by the presence or absence of individual deficits as a proportion of the total possible. Categories of fit, mild, moderate and severe frailty were defined using population quartiles. OUTCOMES: outcomes were 1-, 3- and 5-year mortality, hospitalisation and nursing home admission. STATISTICAL ANALYSIS: hazard ratios (HRs) were estimated using bivariate and multivariate Cox regression analyses. Discrimination was assessed using receiver operating characteristic (ROC) curves. Calibration was assessed using pseudo-R(2) estimates. RESULTS: we include data from a total of 931,541 patients. The eFI incorporates 36 deficits constructed using 2,171 CTV3 codes. One-year adjusted HR for mortality was 1.92 (95% CI 1.81-2.04) for mild frailty, 3.10 (95% CI 2.91-3.31) for moderate frailty and 4.52 (95% CI 4.16-4.91) for severe frailty. Corresponding estimates for hospitalisation were 1.93 (95% CI 1.86-2.01), 3.04 (95% CI 2.90-3.19) and 4.73 (95% CI 4.43-5.06) and for nursing home admission were 1.89 (95% CI 1.63-2.15), 3.19 (95% CI 2.73-3.73) and 4.76 (95% CI 3.92-5.77), with good to moderate discrimination but low calibration estimates. CONCLUSIONS: the eFI uses routine data to identify older people with mild, moderate and severe frailty, with robust predictive validity for outcomes of mortality, hospitalisation and nursing home admission. Routine implementation of the eFI could enable delivery of evidence-based interventions to improve outcomes for this vulnerable group.

BACKGROUND: Little is known about the relative risk of common bacterial, viral, fungal, and parasitic infections in the general population of individuals exposed to systemic glucocorticoids, or about the impact of glucocorticoid exposure duration and predisposing factors on this risk. METHODS AND FINDINGS: The hazard ratios of various common infections were assessed in 275,072 adults prescribed glucocorticoids orally for ≥15 d (women: 57.8%, median age: 63 [interquartile range 48-73] y) in comparison to those not prescribed glucocorticoids. For each infection, incidence rate ratios were calculated for five durations of exposure (ranging from 15-30 d to >12 mo), and risk factors were assessed. Data were extracted from The Health Improvement Network (THIN) primary care database. When compared to those with the same underlying disease but not exposed to glucocorticoids, the adjusted hazard ratios for infections with significantly higher risk in the glucocorticoid-exposed population ranged from 2.01 (95% CI 1.83-2.19; p < 0.001) for cutaneous cellulitis to 5.84 (95% CI 5.61-6.08; p < 0.001) for lower respiratory tract infection (LRTI). There was no difference in the risk of scabies, dermatophytosis and varicella. The relative increase in risk was stable over the durations of exposure, except for LRTI and local candidiasis, for which it was much higher during the first weeks of exposure. The risks of infection increased with age and were higher in those with diabetes, in those prescribed higher glucocorticoid doses, and in those with lower plasma albumin level. Most associations were also dependent on the underlying disease. A sensitivity analysis conducted on all individuals except those with asthma or chronic obstructive pulmonary disease produced similar results. Another sensitivity analysis assessing the impact of potential unmeasured confounders such as disease severity or concomitant prescription of chemotherapy suggested that it was unlikely that adjusting for these potential confounders would have radically changed the findings. Limitations of our study include the use of electronic medical records, which could have resulted in some degree of misclassification of the infectious outcomes; a possible reporting bias, as general practitioners could be more prone to record an infection in those exposed to glucocorticoids; and a low number of events for some outcomes such as scabies or varicella, which may have led to limited statistical power. CONCLUSIONS: The relative risk of LRTI and local candidiasis is very high during the first weeks of glucocorticoid exposure. Further studies are needed to assess whether low albumin level is a risk factor for infection by itself (e.g., by being associated with a higher free glucocorticoid fraction) or whether it reflects other underlying causes of general debilitation.

BACKGROUND/PURPOSE: Because ankylosing spondylitis (AS) is uncommon, large medical record databases offer important opportunities for pharmacoepidemiologic research. However, the validity of AS diagnoses recorded by a general practitioner (GP) is unknown. We assessed the validity of algorithms for identifying AS in The Health Improvement Network (THIN). METHODS: THIN is a database of GP records for over 10 million persons in the UK. In 2014, we administered a questionnaire to GPs of 100 adults for whom an AS diagnosis had been recorded. As high positive predictive value (PPV) is critically important in AS research, we sought to determine the PPV of an AS diagnostic code relative to the GP's clinical impression as the gold standard. Other AS algorithms included: more than one AS diagnostic code, prescription of a nonsteroidal anti-inflammatory drug (NSAID), disease modifying anti-rheumatic drug (DMARD) or biologic. RESULTS: In 61 of 85 returned questionnaires, the GP's clinical impression confirmed AS yielding an overall PPV of 72%. PPV was 89% for two AS codes >7 days apart, and was 86% for an AS code plus a DMARD/biologic. Sensitivity was reduced with algorithms requiring two AS codes (64%) and a DMARD/biologic prescription (30%). Algorithms requiring prescription of an NSAID, or the absence of OA or RA had lower PPV (71-75%) and higher sensitivity (95-98%). CONCLUSION: An AS identification algorithm of two AS diagnoses separated by >7 days provided the highest PPV. This algorithm should be used for pharmacoepidemiologic studies in THIN.

PURPOSE: To validate the recorded diagnoses of colorectal cancer (CRC) and identify false negatives in The Health Improvement Network (THIN) primary care database. METHODS: We conducted a validation study of incident CRC cases in THIN among patients aged 40-89 years from 2000-2011. CRC Read code entries (N = 3805) were verified by manual review of patients' electronic medical records (EMRs) including free-text comments. Incident CRC cases in THIN ascertained following manual review were validated against two data sources deemed gold standards: (i) questionnaires sent to primary care practitioners (PCPs; for a random sample of 100 potential CRC cases), and (ii) Hospital Episode Statistics (HES) among linked practices. False negatives in THIN were identified by searching for International Classification of Diseases-10 codes related to CRC in HES. RESULTS: Of 3805 CRC cases identified in THIN via Read codes, 3033 patients (80.0%) were considered definite cases after manual review of EMRs. The positive predictive value (PPV) of CRC Read codes was 86.0% after removing patients identified from THIN via a Read code for 'fast track referral for suspected CRC'. The response rate from PCPs was 87.0% (n = 87), and the PPV of CRC in THIN was 100% based on PCP questionnaires. Using HES, the PPV for CRC in THIN was 97.9% (556/568), and false negative rate was 6.1% (36/592). CONCLUSIONS: CRC diagnostic Read codes in THIN have a high PPV, which is increased further following manual review of free-text comments. The false negative rate of CRC diagnoses in THIN is low.

PURPOSE: We aimed to quantify the extent of over-the-counter (OTC) low-dose aspirin use among patients in The Health Improvement Network (THIN) in the UK. METHODS: In September 2013, a random sample of low-dose aspirin users (75 past users and 75 never users) was identified based on prescriptions recorded in THIN. Primary care practitioners (PCPs) were sent questionnaires to provide information on patients' use of OTC low-dose aspirin. RESULTS: One hundred and forty valid questionnaires were received (93.30% [95%CI: 88.16-96.34] response rate). Current use of low-dose aspirin was reported by PCPs in 4.23% (95%CI: 1.45-11.70) (n = 3) of past users (OTC use in one patient) and in 2.9% (95%CI: 0.78-9.70) (n = 2) of never users (OTC use in one patient). In addition, PCPs reported past use of low-dose aspirin in 88.70% (95%CI: 79.31-94.18) (n = 63) of past users (all prescribed; none as OTC) and in 2.82% (95%CI: 0.78-9.70) (n = 2) of never users (as OTC). Among past users, PCPs reported the indication for low-dose aspirin as primary cardiovascular disease (CVD) prevention in 63.16% (95%CI: 50.18-74.48) of patients and secondary CVD prevention in 31.58% (95%CI: 21.00-44.48) of patients. Corresponding percentages based on THIN were 78.95% (95%CI: 66.71-87.53) and 21.1% (95%CI: 12.47-33.29), respectively. CONCLUSION: Our findings show the small impact of potential misclassification of low-dose aspirin use in THIN due to unrecorded OTC use. The small proportion of false negatives confirms the utility of THIN for utilization and outcome studies of low-dose aspirin.

PURPOSE: Epidemiological data on adhesion-related complications following intra-abdominal surgery are limited. We tested the accuracy of recording of these surgeries and complications within The Health Improvement Network (THIN), a primary care database within the UK. METHODS: Individuals within THIN from 1995 to 2011 with an incident intra-abdominal surgery and subsequent bowel obstruction (SBO) or adhesiolysis were identified using diagnostic codes. To compute positive predictive values (PPVs), requests were sent to treating physicians of patients with these diagnostic codes to confirm the surgery, SBO, or adhesiolysis code. Completeness of recording was estimated by comparing observed surgical rates within THIN to expected rates derived from the Hospital Episode Statistics dataset within England. Cumulative incidence rates of adhesion-related complications at 5 years were compared with a previously published cohort within Scotland. RESULTS: Two hundred seventeen of 245 (89%) questionnaires were returned (180 SBO and 37 adhesiolysis). The PPV of codes for surgery was 94.5% (95%CI: 91-97%). The 88.8% of procedure types were correctly coded. The PPV for SBO and adhesiolysis was 86.1% (95%CI: 80-91%) and 89.2% (95%CI: 75-97%), respectively. Colectomy, appendectomy, and cholecystectomy rates within THIN were 99%, 95%, and 84% of rates observed in national Hospital Episode Statistics data, respectively. Cumulative incidence rates of adhesion related complications following colectomy, appendectomy, and small bowel surgery were similar to those published previously. CONCLUSIONS: Surgical procedures, SBO, and adhesiolysis can be accurately identified within THIN using diagnostic codes. THIN represents a new tool for assessing patient-specific risk factors for adhesion-related complications and long-term outcomes.

OBJECTIVES: We assessed the association of smoking at diagnosis of inflammatory bowel disease (IBD) on the need for an intestinal resection. METHODS: The Health Improvement Network was used to identify an inception cohort of Crohn's disease (n=1519) and ulcerative colitis (n=3600) patients from 1999-2009. Poisson regression explored temporal trends for the proportion of newly diagnosed IBD patients who never smoked before their diagnosis and the risk of surgery within 3 years of diagnosis. Cox proportional hazard models assessed the association between smoking and surgery, and effect modification was explored for age at diagnosis. RESULTS: The rate of never smokers increased by 3% per year for newly diagnosed Crohn's disease patients (incidence rate ratio (IRR) 1.03; 95% confidence interval (CI): 1.02-1.05), but not for ulcerative colitis. The rate of surgery decreased among Crohn's disease patients aged 17-40 years (IRR 0.96; 95% CI: 0.93-0.98), but not for ulcerative colitis. Smoking at diagnosis increased the risk of surgery for Crohn's disease patients diagnosed after the age of 40 (hazard ratio (HR) 2.99; 95% CI: 1.52-5.92), but not for those diagnosed before age 40. Ulcerative colitis patients diagnosed between the ages of 17 and 40 years and who quit smoking before their diagnosis were more likely to undergo a colectomy (ex-smoker vs. never smoker: HR 1.66; 95% CI: 1.04-2.66). The age-specific findings were consistent across sensitivity analyses for Crohn's disease, but not ulcerative colitis. CONCLUSIONS: In this study, the association of smoking and surgical resection was dependent on the age at diagnosis of IBD.

PURPOSE OF THE RESEARCH: Spasticity is one of the well-recognized complications of stroke which may give rise to pain and limit patients' ability to perform daily activities. The predisposing factors and direct effects of post-stroke spasticity also involve high management costs in terms of healthcare resources, and case-control designs are required for establishing such differences. Using 'The Health Improvement Network' (THIN) database, such a study would not provide reliable estimates since the prevalence of post-stroke spasticity was found to be 2%, substantially below the most conservative previously reported estimates. The objective of this study was to use predictive analysis techniques to determine if there are a substantial number of potentially under-recorded patients with post-stroke spasticity. METHODS: This study used retrospective data from adult patients with a diagnostic code for stroke between 2007 and 2011 registered in THIN. Two algorithm approaches were developed and compared, a statistically validated data-trained algorithm and a clinician-trained algorithm. RESULTS: A data-trained algorithm using Random Forest showed better prediction performance than clinician-trained algorithm, with higher sensitivity and only marginally lower specificity. Overall accuracy was 75% and 72%, respectively. The data-trained algorithm predicted an additional 3912 records consistent with patients developing spasticity in the 12months following a stroke. CONCLUSIONS: Using machine learning techniques, additional unrecorded post-stroke spasticity patients were identified, increasing the condition's prevalence in THIN from 2% to 13%. This work shows the potential for under-reporting of PSS in primary care data, and provides a method for improved identification of cases and control records for future studies.

BACKGROUND: Kawasaki disease is reported to be increasing in incidence and is the commonest childhood cause of acquired heart disease in the Western world. AIM: To determine the current UK incidence of Kawasaki disease across childhood and adolescence; and investigate trends over time and season. DESIGN AND SETTING: An observational, descriptive study in the UK. METHOD: The Health Improvement Network (THIN) database of primary healthcare records was searched for codes or text indicating Kawasaki disease. Identified records were compared with a study case definition and a date of onset was assigned to cases. The incidence, age/sex distribution, and trend in seasonal and temporal distribution were estimated (2008-2012). RESULTS: A total of 110 episodes of Kawasaki disease in 109 children were identified from 3.9 million person-years observation. The incidence of Kawasaki disease was 2.8 per 100 000 person-years (95% confidence interval [CI] = 2.3 to 3.4) when aged <20 years; 9.1 (95% CI = 7.3 to 11.2) aged <5 years, and 3.0 per 100 000 (95% CI = 2.0 to 4.3) across the age groups when possible cases were included. More cases were identified in males (55%) with one-fifth of cases diagnosed after 5 years of age. There was no statistically significant trend in incidence over the study years (P = 0.10 adjusted for sex and month), or between seasons (P = 0.65 adjusted for year and sex). CONCLUSION: Although the incidence of Kawasaki disease remains low and has stabilised in the UK, GPs should recognise that the condition occurs throughout childhood and across the seasons.

To test whether an association exists between radon gas concentration in the home and increased multiple sclerosis (MS) incidence, a retrospective study was undertaken of MS incidence in known areas of raised domestic radon concentration in England and Wales, using The Health Improvement Network (THIN) clinical research database. The study population comprised 20,140,498 person-years of clinical monitoring (males: 10,056,628: 49.93%; females: 10,083,870: 50.07%), representing a mean annual population of 2.5 million individuals. To allow for the possible latency of MS initiation following exposure, data extraction was limited to patients with at least five years registration history with the same GP practice before first diagnosis. Patient records were allocated to one of nine radon concentration bands depending on the average radon level in their postcode sector. MS incidence was analysed by searching for patients with first MS diagnosis over the eight calendar years 2005-2012 inclusive. 1512 new MS cases were diagnosed, 1070 females, 442 males, equivalent to raw incidence rates of 7.51, 10.61 and 4.40 per 10(5) person-years respectively, comparable to previously reported results. Of these new cases, 115 could be allocated to one of the radon bands representing high radon areas. Standardising to the UK 2010 population, excess relative risk (ERR) figures for MS were calculated for each radon band. Linear regression of ERR against mean band radon concentration shows a positive gradient of 0.22 per 100 Bq·m(-3) (R(2) = 0.25, p = 0.0961) when forced through the origin to represent a linear-no-threshold response. The null hypothesis falls inside the 95% confidence interval for the linear fit and therefore this fit is not statistically significant. We conclude that, despite THIN sampling around 5% of the population, insufficient data was available to confirm or refute the hypothesised association between MS incidence and radon concentration.

BACKGROUND: Several prior studies have found an association between statin use and reduced risk of colorectal cancer. We hypothesized that these findings may be due to systematic bias and examined the independent association of colorectal cancer risk with statin use, serum cholesterol, and change in cholesterol concentration. METHODS AND FINDINGS: 22,163 colorectal cancer cases and 86,538 matched controls between 1995 and 2013 were identified within The Health Improvement Network (THIN) a population-representative database. Conditional logistic regression models estimated colorectal cancer risk with statin use, serum total cholesterol (mmol/L), and change in total cholesterol level. We confirmed a decreased risk of colorectal cancer with statin use (long-term: odds ratio [OR], 0.95; 95% confidence interval [CI], 0.91-0.99; short-term: OR, 0.92; 95% CI, 0.85-0.99). However, to assess whether the observed association may result from indication bias, a subgroup analysis was conducted among patients prescribed a statin. In this subgroup (n = 5,102 cases, n = 19,032 controls), 3.1% of case subjects and 3.1% of controls discontinued therapy. The risk of colorectal cancer was not significantly different among those who continued statin therapy and those who discontinued (OR, 0.98; 95% CI, 0.79-1.22). Increased serum cholesterol was independently associated with decreased risk of colorectal cancer (OR, 0.89 per mmol/L increase; 95% CI, 0.87-0.91); the association was only present if serum cholesterol was measured near the cancer diagnosis (<6 mo: OR, 0.76; 95% CI, 0.47-0.61; >24 mo: OR, 0.98; 95% CI, 0.93-1.03). Decreases in serum total cholesterol >1 mmol/L ≥1 year prior to cancer diagnosis were associated with subsequent colorectal cancer (statin users: OR, 1.25; 95 CI%, 1.03-1.53; nonusers: OR, 2.36; 95 CI%, 1.78-3.12). As an observational study, limitations included incomplete data and residual confounding. CONCLUSIONS: Although the risk of colorectal cancer was lower in statin users versus nonusers, no difference was observed among those who continued versus discontinued statin therapy, suggesting the potential for indication bias. The association between decreased serum cholesterol and colorectal cancer risk suggests a cholesterol-lowering effect of undiagnosed malignancy. Clinical judgment should be used when considering causes of cholesterol reduction in patients, including those on statin therapy.

PURPOSE: Studies using primary care databases often censor follow-up at the date data are last collected from clinical computer systems (last collection date (LCD)). We explored whether this results in the selective exclusion of events entered in the electronic health records after their date of occurrence, that is, backdated events. METHODS: We used data from The Health Improvement Network (THIN). Using two versions of the database, we identified events that were entered into a later (THIN14) but not an earlier version of the database (THIN13) and investigated how the number of entries changed as a function of time since LCD. Times between events and the dates they were recorded were plotted as a function of time since the LCD in an effort to determine appropriate points at which to censor follow-up. RESULTS: There were 356 million eligible events in THIN14 and 355 million eligible events in THIN13. When comparing the two data sets, the proportion of missing events in THIN13 was highest in the month prior to the LCD (9.6%), decreasing to 5.2% at 6 months and 3.4% at 12 months. The proportion of missing events was largest for events typically diagnosed in secondary care such as neoplasms (28% in the month prior to LCD) and negligible for events typically diagnosed in primary care such as respiratory events (2% in the month prior to LCD). CONCLUSIONS: Studies using primary care databases, particularly those investigating events typically diagnosed outside primary care, should censor follow-up prior to the LCD to avoid underestimation of event rates.

PURPOSE: Results from observational studies on the same exposure-outcome association may be inconsistent because of variations in methodological factors, clinical factors or health care systems. We evaluated the consistency of results assessing the association between antidepressant use and the risk of hip/femur fractures in three European primary care databases using two different study designs. METHODS: Cohort and nested case control studies were conducted in three European primary care databases (Spanish BIFAP, Dutch Mondriaan and UK THIN) to assess the association between use of antidepressants and hip/femur fracture. A common protocol and statistical analysis plan was applied to harmonize study design and conduct between data sources. RESULTS: Current use of antidepressants was consistently associated with a 1.5 to 2.5-fold increased risk of hip/femur fractures in all data sources with both designs, with estimates for SSRIs generally higher than those for TCAs. In general, risk estimates in Mondriaan, the smallest data source, were higher compared to the other data sources. This difference may be partially explained by an interaction between SSRI and age in Mondriaan. Adjustment for GP-recorded lifestyle factors and matching on general practice had negligible impact on adjusted relative risk estimates. CONCLUSION: We found a consistent increased risk of hip/femur fracture with current use of antidepressants across different databases and different designs. Applying similar pharmacoepidemiological study methods resulted in similar risks for TCA use and some variation for SSRI use. Some of these differences may express real (or natural) variance in the exposure-outcome co-occurrences.

Following the FDA black box warning in 2004, substantial reductions in antidepressant (ATD) use were observed within 2 years in children and adolescents in several countries. However, whether these reductions were sustained is not known. The objective of this study was to assess more recent trends in ATD use in youth (0-19 years) for the calendar years 2005/6-2012 using data extracted from regional or national databases of Denmark, Germany, the Netherlands, the United Kingdom (UK), and the United States (US). In a repeated cross-sectional design, the annual prevalence of ATD use was calculated and stratified by age, sex, and according to subclass and specific drug. Across the years, the prevalence of ATD use increased from 1.3% to 1.6% in the US data (+26.1%); 0.7% to 1.1% in the UK data (+54.4%); 0.6% to 1.0% in Denmark data (+60.5%); 0.5% to 0.6% in the Netherlands data (+17.6%); and 0.3% to 0.5% in Germany data (+49.2%). The relative growth was greatest for 15-19 year olds in Denmark, Germany and UK cohorts, and for 10-14 year olds in Netherlands and US cohorts. While SSRIs were the most commonly used ATDs, particularly in Denmark (81.8% of all ATDs), Germany and the UK still displayed notable proportions of tricyclic antidepressant use (23.0% and 19.5%, respectively). Despite the sudden decline in ATD use in the wake of government warnings, this trend did not persist, and by contrast, in recent years, ATD use in children and adolescents has increased substantially in youth cohorts from five Western countries.