Scientific
publications

BACKGROUND AND AIM: The association between colonic diverticulosis, diverticulitis, and mortality is controversial. This study evaluated the association between diverticular disease and mortality over a prolonged period in a GP cohort. METHODS: GP records were sourced from the United Kingdom medical database (THIN). Diverticulosis and diverticulitis were identified via Read codes. The overall patient cohort (n = 1 274 260) included patients with colonic diverticula (n = 39 521 [3.1%], mean age 54) and no diverticula (control group) (n = 1 234 739 [96.9%], mean age 38). Poisson regression estimated relative rates, and durational time at risk and survival probability were calculated. RESULTS: Colonic diverticula are associated with an increased mortality risk when compared with nondiverticular patients (OR = 1.89, 95% CI 1.84-1.94; p < 0.001). However, controlling for age, sex, and potential confounding variables yielded a decreased mortality risk overall for colonic diverticula patients (HR = 0.66, 95% 0.64-0.6; p < 0.001). When the diverticulitis cohort is separated into uncomplicated and complicated, increased mortality is observed in both uncomplicated diverticulitis (HR = 0.64, 95% CI 0.61-0.66; p < 0.001) and complicated diverticulitis (HR = 1.14, 95% CI 1.02-1.28; p = 0.024), but on controlling for confounding, there is a decreased risk of mortality for uncomplicated diverticulitis (HR = 0.65, 95% CI 0.63-0.66; p < 0.001) but almost two times increased mortality risk for complicated diverticulitis (HR = 1.18, 95% CI 1.05-1.32; p = 0.006). CONCLUSIONS: In this large UK GP sample, controlling for age, sex, and comorbidities, patients with uncomplicated diverticula are associated with a lower mortality risk. However, complicated diverticulitis still carries two times the risk of mortality than those with no or uncomplicated colonic diverticula.

BACKGROUND: Many studies have investigated early predictors for Alzheimer's disease (AD), Parkinson's disease (PD), and amyotrophic lateral sclerosis (ALS). However, evidence is sparse regarding specific and common predictors for these diseases. We aimed to identify medication use, health conditions, and blood biomarkers that might be associated with the risk of AD, PD, and ALS ten years later. METHODS: We conducted population-based nested case-control studies of AD, PD, and ALS using electronic medical records in Europe (France, the UK, and Sweden) and Australia. We retrieved data on medication use, diagnosed health conditions, and measured blood biomarkers from electronic medical records or biomedical cohorts. Conditional logistic regression models and meta-analysis were applied to assess the associations between these factors and the risk of receiving a diagnosis of AD, PD, or ALS. FINDINGS: We included a total of 149,642 AD cases (mean age: 79.1-81.2 years), 252,696 PD cases (73.2-75.9 years), and 27,533 ALS cases (64.4-69.6 years). The prescription of psychoanaleptics and nasal preparations was consistently associated with an increased risk of AD, PD, and ALS 5-10 years later. Constipation and use of related medications were associated with an increased risk of AD and PD, while diabetes and use of antidiabetics were associated with a reduced risk of ALS. A higher level of triglycerides was associated with a lower risk of AD, whereas a higher level of Apolipoprotein B was associated with a lower risk of PD, 5-10 years later. INTERPRETATION: Psychoanaleptics and nasal preparations may serve as common predictors for diagnosis of AD, PD, and ALS 5-10 years later. Conversely, the increased prevalence of constipation is specific to AD and PD, while the decreased prevalence of diabetes and use of antidiabetics is specific to ALS. FUNDING: EU Joint Programme-Neurodegenerative Disease Research.

INTRODUCTION: Systemic inflammation is recognised as a critical driver of atherosclerotic cardiovascular disease (ASCVD), especially in patients with comorbid chronic kidney disease (CKD). This study aims to assess the prevalence of systemic inflammation in the ASCVD population in Spain. METHODS: Outpatient electronic medical records from The Health Improvement Network (THIN®) database were used to identify patients with ASCVD and a C-reactive protein (CRP) measurement ≥1 between January 2014 and July 2023 in Spain. The proportion of patients with systemic inflammation (defined as CRP ≥ 2 mg/L) was estimated at the first CRP measurement (index date) and at the end of the study. The patients' characteristics, comorbidities, and drug dispensation in the prior 12 months were reported by systemic inflammation status at the index date. RESULTS: Overall, 15,798 patients with ASCVD were included in the study (mean age: 71.1 years; 57% men), of whom 34% had CKD. The proportion of patients with systemic inflammation at the index date was 58% (65% among CKD patients) and 56% (62% among CKD patients) at the end of the study. Patients with systemic inflammation were more frequently smokers, obese, with comorbidities, and had higher low-density lipoprotein cholesterol and triglycerides levels than patients without systemic inflammation. Overall, patients with ASCVD and systemic inflammation used statins and aspirin less frequently compared to patients without systemic inflammation, while they used antibiotics, anticoagulants, and antihypertensives more frequently. CONCLUSION: Systemic inflammation prevalence is high among patients with ASCVD in Spain, especially among patients with comorbid CKD. Therapeutic strategies focused on targeting systemic inflammation may have beneficial effects in reducing the burden of ASCVD.

Non-adherence and non-persistence of direct oral anticoagulants (DOAC) in patients with non-valvular atrial fibrillation remain major concerns. Most studies on non-persistence and non-adherence during the implementation phase (i.e., from the first to the last dose of DOAC) focus only on new DOAC users. We conducted a retrospective cohort study in France in 2018-2019 using the real-world THIN database to determine non-adherence during the implementation phase, non-persistence, and their associated factors among all DOAC users. Non-adherence was defined as the proportion of days covered < 80% and non-persistence as a supply gap > 30 days. Factors associated with non-adherence or non-persistence were identified using logistic regressions. We included a total of 5,059 DOAC users: 1,358 new users and 3,701 prevalent users. Non-adherence was 18.7% at 6 months and 24.2% at 1 year and non-persistence was 16.1% and 28.4%, respectively. Non-adherence was almost the same in new and prevalent DOAC users and non-persistence was slightly higher in new users. Factors associated with non-adherence and non-persistence were: age < 65 years and having < 5 comedications. Further studies will be needed to quantify the impact of non-adherence or non-persistence.

INTRODUCTION: Patients with inflammatory bowel disease (IBD) may experience delays in their diagnosis. This study aimed to develop and validate a risk prediction tool for IBD. METHODS: A retrospective cohort study was conducted using primary care data from 2010 to 2019, including symptomatic patients aged ≥18. UK-based primary care databases linked to hospital records were utilized for model development and validation. Cox proportional hazards models were used to derive risk equations for IBD, ulcerative colitis (UC), and Crohn's disease (CD) in men and women. Candidate predictors included demographics, comorbidities, symptoms, extraintestinal manifestations, and laboratory results. Model performance was evaluated using measures of fit, discrimination, and calibration at 1, 2, 3, and 5 years after symptom onset. RESULTS: In total, 2 054 530 patients were included in the derivation cohort and 673 320 in the validation cohort. In the derivation cohort, 0.7% were diagnosed with IBD (66.3% UC and 33.7% CD). Predictors in the final IBD model included age, smoking, body mass index, gastrointestinal symptoms, extraintestinal manifestations, comorbidities, family history of IBD, and laboratory investigations. The model demonstrated good discrimination and calibration; C-statistic 0.78 (95% confidence interval [CI], 0.77-0.79) in men and 0.78 (95% CI, 0.77-0.79) in women. In the validation cohort, the model tended to slightly overestimate IBD risk at higher risk thresholds. CONCLUSIONS: A risk model using patient demographics, symptoms, and laboratory results accurately predicted IBD, UC, and CD at 1, 2, 3, and 5 years after symptom onset, potentially aiding in prioritizing patients for a referral or fecal calprotectin testing in primary care.

OBJECTIVE: To investigate the prescribing trends of indapamide, a thiazide-like diuretic, and the long term comparative effectiveness of modified release versus immediate release indapamide. DESIGN: Cohort study. SETTING: IQVIA Medical Research Data UK database, incorporating data from The Health Improvement Network database, 1 January 2005 to 31 December 2020. PARTICIPANTS: Of 1 904 289 patients with hypertension, 86 388 started indapamide treatment during the study period. 30 021 patients received modified release and 56 367 immediate release indapamide. MAIN OUTCOME MEASURES: Monthly prescribing trends of modified release and immediate release indapamide are described. A pragmatic trial was emulated to compare the five year risks of composite cardiovascular events (myocardial infarction and stroke) and death between modified release and immediate release indapamide. Intention-to-treat and per protocol effects of treatment were estimated with pooled logistic regression models. Confounding and selection bias were accounted for by multivariable adjustments and inverse probability weights. RESULTS: 1 38 414 patients who used indapamide were identified among 1 904 289 patients with hypertension. A greater increase was seen in the proportion of users of immediate release indapamide (from 0.43% in 2005 to 2.31% in 2020) than in users of modified release indapamide (from 0.71% to 0.79%). 86 388 patients (30 021 and 56 367 who started modified release and immediate release indapamide, respectively) were eligible for the trial emulation. In the intention-to-treat analysis, no difference was found in the risk of cardiovascular events (hazard ratio 0.99, 95% confidence interval (CI) 0.90 to 1.08) or death (hazard ratio 0.97, 0.92 to 1.02) between modified release and immediate release indapamide. In the per protocol analysis, a lower risk of cardiovascular events was found with modified release indapamide than with immediate release indapamide (risk difference -0.39%, 95% CI -0.71% to -0.06%; hazard ratio 0.81, 95% CI 0.68 to 0.98), which was mainly driven by myocardial infarction (risk difference -0.36%, 95% CI -0.64% to -0.08%; hazard ratio 0.80, 95% CI 0.64 to 1.01). Similar risks of death (hazard ratio 1.03, 95% CI 0.90 to 1.17) were found for the two formulations. CONCLUSIONS: In patients treated with indapamide for hypertension, starting treatment with modified release or immediate release indapamide had similar risks for cardiovascular events or all cause mortality. In an exploratory secondary analysis, sustained treatment with modified release preparations was associated with a lower the risk of cardiovascular events but not all cause mortality compared with immediate release preparations. These findings need to be confirmed in prospective studies.

IMPORTANCE: The 2023 American College of Rheumatology and American Association of Hip and Knee Surgeons Clinical Practice Guideline concluded that obesity alone should not delay joint replacement. Therefore, a substantially increased utilization of joint replacement among patients with obesity could be expected. However, patients with obesity are at increased risk of revision, posing unique challenges as the surgery is complex and costly, and it remains unknown whether postoperative weight loss could decrease the risk of revision. OBJECTIVE: To examine the association of the proportion of postoperative weight loss following antiobesity medication use with the risk of revision among patients with obesity undergoing hip or knee replacement. DESIGN, SETTING, AND PARTICIPANTS: Using a target trial emulation, a causal inference framework, this retrospective cohort study investigated patients with obesity who underwent hip or knee replacement. Data were from the IQVIA Medical Research Database (2000-2023). Statistical analysis was performed from October 2023 to June 2024. MAIN OUTCOMES AND MEASURES: Emulated analyses of a hypothetical target trial were assessed for the association of small-to-moderate (2%-10%) or large (≥10%) weight loss after initiating antiobesity medications (orlistat, sibutramine, glucagon-like peptide-1 receptor agonists, and rimonabant) within 1 year with the risk of 5-year and 10-year revision after initiation of antiobesity medications. RESULTS: Among 3691 qualified participants (mean [SD] age, 64.7 [9.3] years; 2322 [62.9%] women), the 5-year risks of revision were 5.6%, 4.4%, and 3.7% for weight gain or stable, small-to-moderate weight loss, and large weight loss groups, respectively. Compared with the weight gain or stable group, the hazard ratios (HRs) were 0.75 (95% CI, 0.55-1.04) for the small-to-moderate weight loss group and 0.57 (95% CI, 0.36-0.91) for the large weight loss group. Similar results were observed when the analyses were performed separately for hip or knee replacement. The HRs for revision were 0.55 (95% CI, 0.32-0.93) for small-to-moderate weight loss and 0.49 (95% CI, 0.25-0.97) for large weight loss groups compared with the weight gain or stable group in patients undergoing knee replacement; the corresponding HRs for revision were 0.82 (95% CI, 0.54-1.25) and 0.53 (95% CI, 0.30-0.93) in patients undergoing hip replacement. Consistent findings were obtained regarding the association of weight loss with the 10-year risks after initiating antiobesity medications. CONCLUSIONS AND RELEVANCE: In this cohort study using a target trial emulation, a higher proportion of weight loss after initiating antiobesity medications within 1 year was associated with a lower risk of 5-year and 10-year revision among patients with obesity undergoing joint replacement. These results suggest that antiobesity medication use, with relatively safe and sustainable weight loss, may be an effective strategy for improving implant survivorship of hip and knee replacements in the obese population.

Association of prodromal Parkinson's disease (PD) with risk of PD and risk of mortality in individuals with PD warrant investigation through large-scale prospective study. We included 501,475 participants without PD at baseline. Eight prodromal features were measured. Incident PD cases were identified via hospital admission, death register, and self-report. Cox regression models were used to compute hazard ratios (HRs) and 95% confidence intervals (CIs). Multivariable-adjusted HRs and 95%CIs were 3.12 (2.58-3.78) for men and 2.71 (2.11-3.47) for women. Prodromal PD predicted only PD onset occurred during the first 6 years of follow-up (HR  = 10.5; 95% CI: 8.60-12.9), but not after 6 years (HR = 1.00; 95%CI: 0.76-1.32). The presence of prodromal PD conferred a higher risk of mortality among participants with PD. Having prodromal PD were associated with higher probability of developing PD in short-term and higher risk of mortality among individuals with PD.

BACKGROUND: Leveraging machine learning on electronic health records offers a promising method for early identification of individuals at risk for dementia and neurodegenerative diseases. Current risk algorithms heavily rely on age, highlighting the need for alternative models with strong predictive power, especially at age 65, a crucial time for early screening and prevention. METHODS: This prospective study analyzed electronic health records (EHR) from 76,427 adults (age 65, 52.1% women) using the THIN database. A general risk algorithm for Alzheimer's disease, Parkinson's disease, and dementia was developed using machine learning to select predictors from diagnoses, and medications. RESULTS: Medications (e.g., laxatives, urological drugs, antidepressants), along with sex, BMI, and comorbidities, were key predictors. The algorithm achieved a 38.4% detection rate at a 5% false-positive rate for 2-year dementia prediction. CONCLUSION: The validated prediction algorithms, easy to implement in primary care, identify high-risk 65-year-olds using medication records. Further refinement and broader validation are needed.

BACKGROUND: Nearly 3% of adults have attention-deficit and hyperactivity disorder (ADHD), although in the UK, most are undiagnosed. Adults with ADHD on average experience poorer educational and employment outcomes, worse physical and mental health and are more likely to die prematurely. No studies have yet used mortality data to examine the life expectancy deficit experienced by adults with diagnosed ADHD in the UK or worldwide. AIMS: This study used the life-table method to calculate the life-expectancy deficit for people with diagnosed ADHD using data from UK primary care. METHOD: A matched cohort study using prospectively collected primary care data (792 general practices, 9 561 450 people contributing eligible person-time from 2000-2019). We identified 30 039 people aged 18+ with diagnosed ADHD, plus a comparison group of 300 390 participants matched (1:10) by age, sex and primary care practice. We used Poisson regression to estimate age-specific mortality rates, and life tables to estimate life expectancy for people aged 18+ with diagnosed ADHD. RESULTS: Around 0.32% of adults in the cohort had an ADHD diagnosis, ~1 in 9 of all adults with ADHD. Diagnoses of common physical and mental health conditions were more common in adults with diagnosed ADHD than the comparison group. The apparent reduction in life expectancy for adults with diagnosed ADHD relative to the general population was 6.78 years (95% CI: 4.50 to 9.11) for males, and 8.64 years (95% CI: 6.55 to 10.91) for females. CONCLUSIONS: Adults with diagnosed ADHD are living shorter lives than they should. We believe that this is likely caused by modifiable risk factors and unmet support and treatment needs in terms of both ADHD and co-occurring mental and physical health conditions. This study included data from adults with diagnosed ADHD; the results may not generalise to the entire population of adults with ADHD, the vast majority of whom are undiagnosed.

AIMS/HYPOTHESIS: UK standard care for type 2 diabetes is structured diabetes education, with no effects on HbA , small, short-term effects on weight and low uptake. We evaluated whether remotely delivered tailored diabetes education combined with commercial behavioural weight management is cost-effective compared with current standard care in helping people with type 2 diabetes to lower their blood glucose, lose weight, achieve remission and improve cardiovascular risk factors. METHODS: We conducted a pragmatic, randomised, parallel two-group trial. Participants were adults (≥18 years) with overweight or obesity (BMI≥25 kg/m ) and recently diagnosed with type 2 diabetes (≤3 years), recruited from 159 primary care practices in England. We randomised participants to a tailored diabetes education and behavioural weight management programme (DEW; delivered by Weight Watchers) or to current standard care diabetes education (DE; Diabetes Education and Self Management for Ongoing and Newly Diagnosed [DESMOND] programme), using a computer-generated randomisation sequence in a 1:1 allocation stratified by gender and diabetes duration, unknown to those collecting and analysing the data. Participants could not be blinded due to the nature of the interventions. Participants completed assessments at 0, 6 and 12 months. The primary outcome was 12 month change from baseline in HbA . We also assessed bodyweight, blood pressure, cholesterol (total, HDL, LDL), glucose-lowering medication, behavioural measures (physical activity, food intake), psychosocial measures (eating behaviour, diabetes-related quality of life, wellbeing) and within-trial and modelled lifetime cost effectiveness. RESULTS: We randomised 577 participants (DEW: 289, DE: 288); 398 (69%) completed 12 month follow-up. We found no evidence for an intervention effect on change in HbA from baseline to 12 months (difference: -0.84 [95% CI -2.99, 1.31] mmol/mol, p=0.44) or 6 months (-1.83 [-4.05, 0.40] mmol/mol). We found an intervention effect on weight at 6 (-1.77 [-2.86, -0.67] kg) and 12 months (-1.38 [-2.56, -0.19] kg). Participants in DEW had a higher likelihood of achieving diabetes remission than participants in DE (6 months: RR 2.10 [95% CI 1.03, 4.47]; 12 months: RR 2.53 [1.30, 5.16]). DEW was cost-effective compared with DE in within-trial and lifetime analyses, in the latter generating an incremental cost effectiveness ratio of £2290 per quality-adjusted life year gained. CONCLUSIONS/INTERPRETATION: A commercial behavioural weight management programme combined with remote dietary counselling after diagnosis of type 2 diabetes did not improve HbA up to 12 months post intervention in this trial. The intervention could help people achieve weight loss and be cost-effective compared with current standard National Health Service care. TRIAL REGISTRATION: ISRCTN 18399564 FUNDING: National Institute for Health and Care Research (NIHR; RP-PG-0216-20010), Medical Research Council (MC_UU_00006/6), NIHR Cambridge Biomedical Research Centre (NIHR203312).

OBJECTIVE: To assess the characteristics, risk factors and clinical impact of penicillin and other antibiotic allergy labels in general practice in the UK. DESIGN: Population-based cohort study. SETTING: Primary care in the UK, 2000-2018. PARTICIPANTS: Adults aged 18-100 years who were registered with their general practice for at least 12 months between 01-Jan-2000 and 31-Dec-2018 and followed until 25-Sep-2019. MAIN OUTCOME MEASURES: The main outcomes include the annual prevalence and incidence of penicillin and other antibiotic allergy labels. Multinominal logistic regression was used to examine the characteristics associated with receiving an allergy label to different antibiotics. Cox regression modelling was used to compare the risk of resistant infections (methicillin-resistant Staphylococcus aureus [MRSA] and vancomycin-resistant enterococci) as well as Clostridioides difficile (C.difficile) infection between patients with and without allergy labels. The monthly proportion of patients who had a penicillin allergy test, either before their allergy label was recorded or within one year, was calculated to assess any impact of NICE penicillin allergy assessment recommendations (Clinical guideline [CG183]) in September 2014. RESULTS: Both the prevalence and incidence of penicillin allergy label showed a pattern of initial growth followed by a decline. The prevalence reached a maximum of 8.25% in 2011, and the incidence peaked at 0.46% in 2004. Older age, being female, living in less deprived areas, belonging to a larger general practice, and having co-morbidities were associated with a higher chance of receiving a penicillin or other antibiotic allergy label. Patients with antibiotic allergy labels were more likely to receive alternative broad-spectrum antibiotics and had a higher risk of MRSA and C.difficile infections. The introduction of NICE drug allergy guideline did not alter the proportion of patients undergoing penicillin allergy assessment. CONCLUSION: Penicillin and other antibiotic allergy labels are common and lead to radical change in the antibiotic prescribing practices and are associated with resistant and healthcare associated infections.

BACKGROUND: There is some evidence that perinatal anxiety (PNA) is associated with lower rates of infant vaccinations and decreased access to preventative infant healthcare, but results across studies have not been conclusive. OBJECTIVE: To investigate the relationship between maternal PNA and infant primary care use. METHODS: Cohort study of mother-infant pairs identified between 1998 and 2016 using IQVIA Medical Research Database (IMRD). PNA was identified through prescription, diagnosis and symptom records from start of pregnancy up to 1 year after birth. Outcomes include primary care consultation rate, attendance at the 6-8 week infant check and uptake of the infant 5-in-1 vaccination, comparing unadjusted rates of consultations and using logistic regression to analyse other outcomes. FINDINGS: Of the 248 618 women, 11 558 (4.7%) had a record of PNA. Infants of mothers with PNA had, on average, one more primary care consultation/person-year compared with those without (9.7 vs 8.7) in the year after birth. Mothers with PNA were more likely to have an infant who was vaccinated (adjusted OR (aOR) 1.33, 95% CI 1.20 to 1.48) but were less likely to have a record of attendance at the 6-8 week infant check (aOR 0.88, 95% CI 0.81 to 0.95). CONCLUSIONS: Infants of mothers with PNA had, on average, a slightly higher rate of primary care consultations and were more likely to be vaccinated but less likely to have a record of an infant check. CLINICAL IMPLICATIONS: Midwives and General Practitioners (GPs) providing care should consider how PNA may impact on infant health and how infant health may impact on maternal anxiety.

BACKGROUND AND AIMS: Observational healthcare data are an important tool for delineating patients' inflammatory bowel disease (IBD) journey in real-world settings. However, studies that characterize IBD cohorts typically rely on a single resource, apply diverse eligibility criteria, and extract variable sets of attributes, making comparison between cohorts challenging. We aim to longitudinally describe and compare IBD patient cohorts across multiple geographic regions, employing unified data and analysis framework. METHODS: We conducted a descriptive cohort study, using routinely collected healthcare data, from a federated network of data partners in sixteen databases from seven countries (USA, UK, France, Germany, Japan, Korea, and Australia); and computed the prevalence of thousands of attributes, across multiple baseline and follow-up time windows, for full disease cohorts and various strata. RESULTS: Characterizing the disease trajectory of 462,502 Crohn's disease (CD) and 589,118 ulcerative colitis (UC) subjects, we observed a decline over time in the average age at CD diagnosis in Europe and North America but less pronounced shifts in Japan and Korea; an uptick in the proportion of patients with anxiety diagnosis prior to CD diagnosis in European and US datasets; and stable rates of segmental colonic and small bowel resections within one and three years following UC and CD diagnosis, respectively, in most US databases. CONCLUSIONS: The study provides a comprehensive characterization of IBD patient cohorts from various countries including insights into disease trends, demographics, and pre-diagnosis symptoms. All characteristics and outcomes are publicly available, providing an unprecedented, comprehensive open resource for clinicians and researchers.

INTRODUCTION: Iron overloading disorders are associated with decreased bone mineral density. However, evidence on fracture risk is scarce. Therefore, we evaluated the risk of fracture associated with iron overload disorders compared to matched controls. METHODS: Using The Healthcare Improvement Network, a Cegedim database of UK general practice data, we identified patients >18 years with elevated iron (ferritin value >1000 µg/L) or an eligible diagnosis code for iron overloading disorders between 2010 and 2022. The first date of elevated iron or a diagnosis code defined the index date for iron overload patients, who were matched with up to 10 controls. Time-varying confounder-adjusted Cox proportional hazard models estimated the hazard ratios (HRs) and 95% confidence intervals. Analyses were stratified by osteoporotic fracture site (hip, vertebral, humerus, forearm) and evidence of elevated serum ferritin at baseline (ferritin >1000 µg/L), and sex. RESULTS: We identified 20 264 eligible patients and 192 956 controls. Overall, there was a 55% increased risk of any fracture among iron overload patients (HR 1.55 [1.42-1.68]). Fracture risk was increased at all sites, with the highest risk observed for vertebral fractures (HR 1.97 [1.63-2.10]). Patients with ferritin >1000 µg/L had a 91% increased risk of any fracture (HR 1.91 [1.73-2.16]) and a 2.5-fold increased risk of vertebral fractures (HR 2.51 [2.01-3.12]). There was no increased risk among patients without elevated serum ferritin at any site. Fracture risk was similar between sexes. DISCUSSION: This large population-based cohort study found a 55% increased risk of fracture associated with iron overload. The risk was highest among patients with laboratory-confirmed iron overload, highlighting the importance for clinicians to consider initiating osteoporosis therapy in patients with serum ferritin >1000 µg/L to minimize fracture risk.