RESEARCH ARTICLE Estimating the Incidence and Key Risk Factors of Cardiovascular Disease in Patients at High Risk of Imminent Fracture Using Routinely Collected Real-World Data From the UK Marta Pineda-Moncusí,1† Leena El-Hussein,1† Antonella Delmestri,1 Cyrus Cooper,1,2 Alireza Moayyeri,3 Cesar Libanati,3 Emese Toth,3 Daniel Prieto-Alhambra,1,2,4,5 and Sara Khalid1 1 Centre for Statistics in Medicine, Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, Nuffield Orthopaedic Centre, University of Oxford, Oxford, UK 2 MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton General Hospital, Southampton, UK 3 UCB Biopharma Sprl, Brussels, Belgium 4 Fundacio Institut Universitari per a la recerca a l’Atenci o Primària de Salut Jordi Gol i Gorina (IDIAPJ Gol), CIBERFES, Barcelona, Spain 5 Universitat Autònoma de Barcelona, Bellaterra (Cerdanyola del Vallès), Cerdanyola del Vallès, Spain ABSTRACT The objective of this work was to estimate the incidence rate of cardiovascular disease (CVD) events (myocardial infarction, stroke, or CVD death) at 1 year among three cohorts of patients at high risk of fracture (osteoporosis, previous fracture, and anti-osteoporosis medication) and to identify the key risk factors of CVD events in these three cohorts. To do so, this prospective cohort study used data from the Clinical Practice Research Datalink, a primary care database from United Kingdom. Major adverse cardiovascular events (MACE, a composite outcome for the occurrence of either myocardial infarction [MI], stroke, or CVD death) were identified in patients aged 50 years or older at high or imminent fracture risk identified in three different cohorts (not mutually exclusive): recently diag- nosed with osteoporosis (OST, n = 65,295), incident fragility fracture (IFX, n = 67,065), and starting oral bisphosphonates (OBP, n = 145,959). About 1.90%, 4.39%, and 2.38% of the participants in OST, IFX, and OBP cohorts, respectively, experienced MACE events. IFX was the cohort with the higher risk: MACE incidence rates (cases/1000 person-years) were 19.63 (18.54–20.73) in OST, 52.64 (50.7– 54.5) in IFX, and 26.26 (25.41–27.12) in OBP cohorts. Risk of MACE events at 1 year was predicted in the three cohorts. Models using a set of general, CVD, and fracture candidates selected by lasso regression had a good discrimination (≥70%) and internal validity and generally outperformed the models using only the CVD risk factors of general population listed in QRISK tool. Main risk factors com- mon in all MACE models were sex, age, smoking, alcohol, atrial fibrillation, antihypertensive medication, prior MI/stroke, established CVD, glomerular filtration rate, systolic blood pressure, cholesterol levels, and number of concomitant medicines. Identified key risk factors highlight the differences of patients at high risk of fracture versus general population. Proposed models could improve pre- diction of CVD events in patients with osteoporosis in primary care settings. © 2022 The Authors. Journal of Bone and Mineral Research published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research (ASBMR). KEY WORDS: OSTEOPOROSIS; ORAL BISPHONATES; CARDIOVASCULAR RISK ASSESSMENT; INCIDENCE RATES; PROGNOSTIC MODEL Introduction have been developed for cardiovascular events (heart attacks or stroke), including the Framingham Heart Study,(5-7) CHADS2 tool,(8-10) and QRISK tool.(11,12) These tools have been developed C ardiovascular disease (CVD) and osteoporosis are both lead- ing causes of morbidity and mortality(1,2) worldwide and their prevalence increases with age.(3,4) Several prediction tools in general, usually younger, populations and have not been val- idated in patients with osteoporosis. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Received in original form January 31, 2022; revised form May 31, 2022; accepted July 7, 2022. Address correspondence to: Daniel Prieto-Alhambra, MD, PhD, Botnar Research Centre, Old Road, Headington, Oxford OX3 7LD, UK. E-mail: daniel.prietoalhambra@ndorms.ox.ac.uk Additional Supporting Information may be found in the online version of this article. † MP-M and LE-H contributed equally to this work. Journal of Bone and Mineral Research, Vol. 37, No. 10, October 2022, pp 1986–1996. DOI: 10.1002/jbmr.4648 © 2022 The Authors. Journal of Bone and Mineral Research published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research (ASBMR). n 1986 PINEDA-MONCUSÍ ET AL. Journal of Bone and Mineral Research Identification of CVD risk factors is particularly challenging for 1. Patients with an incident diagnosis of osteoporosis (read or patients with osteoporosis because the association between ICD-10 codes). We refer to this group as the osteoporosis fracture risk or anti-osteoporosis treatment and cardiovascular cohort (OST). events remains unclear.(13) Some established risk factors for oste- 2. Patients with a first incident fracture at an osteoporotic site oporosis and fractures,(14) such as female sex and low weight, (all except face, skull, and digits), diagnosed either through have been found to be protective against CVD. Conversely, some read codes or ICD-10 codes. This cohort is referred to as the other risk factors, including age, low bone mineral density,(15) imminent fracture risk cohort (IFX). prior fracture, obesity,(16-18) or type 2 diabetes,(19) are associated 3. Incident users of oral bisphosphonates (BP) without BP use in with an increased risk of CVD. Patients with a history of CVD have the prior year, referred to as the oral BP treatment been shown to be at increased risk of osteoporotic fractures,(20) cohort (OBP). whereas higher risk of stroke and coronary artery disease is observed among patients with osteoporotic fracture or low bone Index date was defined as time of recorded incident diagno- mineral density.(21) sis, first incident fracture, and incident use of BP, for the OST, The effects of anti-osteoporosis medications on CVD risk is IFX, and OBP cohorts, respectively. Participants were followed inconclusive: despite no evidence from clinical trials that oral from index date up to a maximum of 2 years. We censored par- bisphosphonates (BP) have an impact on cardiovascular ticipants at the earliest of first: study outcome, death, transfer risk,(22,23) some publications suggest a protective effect.(24) out of practice, or the end of follow-up period. Included patients Meanwhile, the European Medicines Agency (EMA) has advised were at least 50 years old and had at least 1 year of data available contraindications to patients with a history of prior myocardial before index date. Participants could potentially be present in infarction (MI) or stroke regarding romosozumab, the most more than one of the cohorts above, with different index dates. recent medication option for osteoporosis,(25) in addition to the For OBP cohort, subjects with use of any anti-osteoporotic drug preexisting restrictions to menopausal hormone therapy(26) (except calcium and vitamin D supplements) in the previous year and strontium ranelate.(27) This variable impact of anti- were excluded. osteoporotic medication on CVD risk highlights the clinical utility In the IFX cohort, high risk of imminent fracture was defined of identifying patients who are being considered for osteoporo- following Kanis and colleagues(31) designation: imminent risk sis treatment and might be at elevated risk of CVD. period is the following 2-year period after a fracture.(32) To address this issue, our overarching aim was to assess the absolute risk of CVD experienced by elderly patients at higher frac- Candidate risk factors ture risk in the UK, as well as to identify key CVD risk factors (both The overall set of variables considered for inclusion in the predic- generic and specific ones) for these patients. We estimate inci- tion model contained risk factors from the QRISK model,(12) as dence rates of major adverse cardiovascular events (MACE) well as additional risk factors identified in the literature as being among cohorts newly diagnosed with osteoporosis, those with potentially associated with CVD.(15-17,20,21,33-41) These included first recorded fracture, and oral BP therapy initiators obtained sociodemographic and lifestyle factors, laboratory measure- from the UK general population, and developed and internally val- ments, medications, and comorbidities (Table 1). idated models that predict 1-year MACE in these cohorts of high- risk patients. Additionally, a secondary analysis of 2-year MACE and a sensitivity analysis for MI/stroke prediction is reported. Outcomes The main outcome of the study was 1-year occurrence of MACE, Subjects and Methods a composite outcome of the first occurrence of either stroke, MI, or death due to CVD (recorded as the primary cause of death Data source in ONS). Additionally, secondary analysis of 2-year occurrence of MACE Data for this study were obtained from the Clinical Practice and sensitivity analysis excluding death (MI/stroke) at 1 and Research Datalink (CPRD) GOLD, which contains anonymized 2 years is reported in the Supplemental Data: MI/stroke, a com- electronic primary care records for the UK (www.cprd.com/ posite outcome of the first occurrence of either stroke or MI. primarycare). In addition to demographic information, the data included medication prescriptions, clinical events, tests, referrals, Statistical analysis and hospital admissions along with their major outcomes in a sample of >16 million patients (including deceased and trans- Baseline characteristics of all three cohorts were described. ferred out; 2.3 million are current patients, covering approxi- mately 3.6% of UK population).(28,29) For this study, an extract Estimation of incidence rates from January 1, 1995, to January 31, 2017, was used. We used Incidence rates (IR) and their 95% confidence interval of each CPRD GOLD data linked to the Hospital Episode Statistics (HES) outcome at 1 and 2 years after index date were calculated Admitted Patient Care, which contains clinical diagnoses during (cases/1000 person-years) through ERIC Notebook person-time hospital admissions in England, to the Office for National Statis- methodology.(42) tics (ONS) mortality records, and to the Index of Multiple Depriva- tion (IMD) data set. We reconciled CPRD GOLD and ONS mortality Construction of the prediction models dates of death following published guidelines.(30) Performance of QRISK variables to estimate the 1-year risk of MACE was assessed (QRISK variables are listed in Supple- Participants mental Table S1). Finally, all the available candidate risk The study population included patients at high or imminent frac- factors described in the above were combined into a pre- ture risk as identified from literature, divided into three cohorts: diction model (henceforth referred to as “ALL”). Lasso Journal of Bone and Mineral Research MACE IN IMMINENT-FRACTURE-RISK PATIENTS 1987 n Table 1. Baseline Characteristics of Patients in the Development Set and the Validation Set OST IFX OBP Variable n = 65,295 n = 67,065 n = 145,959 Sex, male (%) 8616 (13.2) 16,386 (24.4) 29,547 (20.2) Age >75 years 31,105 (47.6) 48,726 (72.7) 77,199 (52.9) Age group, years (%) 50–59 8315 (12.7) 4764 (7.1) 16,388 (11.2) 60–69 15,949 (24.4) 7646 (11.4) 31,665 (21.7) 70–79 21,165 (32.4) 15,573 (23.2) 45,679 (31.3) 80–89 16,586 (25.4) 27,546 (41.1) 42,053 (28.8) >89 3280 (5.0) 11,536 (17.2) 10,174 (7.0) SES (%) 1 15,953 (24.4) 14,980 (22.3) 35,837 (24.6) 2 15,643 (24.0) 15,990 (23.8) 35,738 (24.5) 3 13,794 (21.1) 14,322 (21.4) 30,877 (21.2) 4 11,921 (18.3) 12,892 (19.2) 26,566 (18.2) 5 7918 (12.1) 8820 (13.2) 16,828 (11.5) Smokinga (%) Ex 20,487 (31.4) 21,484 (32.0) 50,162 (34.4) No 34,049 (52.1) 35,272 (52.6) 75,838 (52.0) Yes 10,759 (16.5) 10,309 (15.4) 19,959 (13.7) Drinkinga (%) Ex 3564 (5.5) 3945 (5.9) 7226 (5.0) No 17,062 (26.1) 21,092 (31.5) 41,791 (28.6) Yes 44,669 (68.4) 42,028 (62.7) 96,942 (66.4) Diabetes type Ib (%) 152 (0.2) 164 (0.2) 297 (0.2) Diabetes type IIb (%) 3311 (5.1) 4171 (6.2) 8340 (5.7) Chronic obstructive pulmonary diseaseb (%) 4939 (7.6) 3745 (5.6) 11251 (7.7) Chronic kidney diseaseb (%) 5790 (8.9) 8117 (12.1) 13396 (9.2) Rheumatoid arthritisb (%) 5045 (7.7) 2891 (4.3) 19746 (13.5) Lupusb (%) 144 (0.2) 55 (0.1) 328 (0.2) Systemic heart diseasea (%) 2341 (3.6) 563 (0.8) 5222 (3.6) Anti-osteoporosis usea (%) 11752 (18.0) 5356 (8.0) Heparin usea (%) 372 (0.6) 282 (0.4) 1035 (0.7) Beta-blocker usea (%) 10564 (16.2) 10392 (15.5) 22753 (15.6) Hypertensiona (%) 5091 (7.8) 3334 (5.0) 9514 (6.5) Deep vein thrombosis or pulmonary embolisma (%) 522 (0.8) 439 (0.7) 1334 (0.9) Anticoagulant usea (%) 3285 (5.0) 3536 (5.3) 7546 (5.2) Antidepressants TCAa (%) 7039 (10.8) 5352 (8.0) 14,451 (9.9) Antidepressants SSRIa (%) 5990 (9.2) 7293 (10.9) 12651 (8.7) Hypercholesterolemiaa (%) 1561 (2.4) 667 (1.0) 2543 (1.7) Statin usea (%) 15571 (23.8) 14084 (21.0) 33,988 (23.3) Osteoporosis historyb NA 5521 (8.2) 46215 (31.7) Family history of cardiovascular disease (%) 6755 (10.3) 4036 (6.0) 13204 (9.0) Family history of cardiovascular disease before age 60 years (%) 99 (0.2) 47 (0.1) 170 (0.1) Heart failureb (%) 2323 (3.6) 3405 (5.1) 5040 (3.5) Migraineb (%) 10103 (15.5) 6294 (9.4) 19961 (13.7) Severe mental illnessb (%) 10203 (15.6) 8836 (13.2) 19256 (13.2) Vascular diseaseb (%) 839 (1.3) 1078 (1.6) 1765 (1.2) Atrial fibrillationb (%) 3664 (5.6) 4784 (7.1) 8022 (5.5) On antihypertensive drug (%) 37053 (56.7) 38036 (56.7) 79044 (54.2) Antipsychotic usea (%) 371 (0.6) 867 (1.3) 773 (0.5) Steroid usea (%) 9367 (14.3) 5181 (7.7) 37201 (25.5) Erectile dysfunctiona (%) 871 (1.3) 1065 (1.6) 2797 (1.9) Charlson score (%) 0 37,782 (57.9) 40,024 (59.7) 80,717 (55.3) 1 13,274 (20.3) 11,209 (16.7) 31,132 (21.3) 2 7960 (12.2) 7388 (11.0) 18,141 (12.4) ≥3 6279 (9.6) 8444 (12.6) 15,969 (10.9) (Continues) n 1988 PINEDA-MONCUSÍ ET AL. Journal of Bone and Mineral Research Table 1. Continued OST IFX OBP Variable n = 65,295 n = 67,065 n = 145,959 Cardiovascular disease (%) No 58,034 (88.9) 58,381 (87.1) 128,516 (88.0) Ever >1 year before index date 5616 (8.6) 6996 (10.4) 13,474 (9.2) 1 year before index 719 (1.1) 743 (1.1) 1551 (1.1) 6 months before index 723 (1.1) 767 (1.1) 1763 (1.2) 1 month before index 203 (0.3) 178 (0.3) 655 (0.4) MI or stroke (%) No 61,350 (94.0) 60,046 (89.5) 135,335 (92.7) Ever >1 year before index date 2850 (4.4) 5083 (7.6) 7481 (5.1) 1 year before index 1095 (1.7) 1936 (2.9) 3143 (2.2) Established CVDb = ever (%) 7428 (11.4) 11,506 (17.2) 18,774 (12.9) Any fracture history (%) No 49,542 (75.9) 67,065 (100) 114,386 (78.4) Ever >1 year before index date 5813 (8.9) 0 10,598 (7.3) 1 year before index 9940 (15.2) 0 20,975 (14.4) Hip fracture history (%) No 62,040 (95.0) 67,065 (100) 136,610 (93.6) Ever >1 year before index date 1118 (1.7) 0 1720 (1.2) 1 year before index 2137 (3.3) 0 7629 (5.2) Shoulder fracture history (%) No 64,836 (99.3) 67,065 (100) 145,199 (99.5) Ever >1 year before index date 232 (0.4) 0 333 (0.2) 1 year before index 227 (0.3) 0 427 (0.3) Spine fracture history (%) No 63,880 (97.8) 67,065 (100) 143,554 (98.4) Ever >1 year before index date 326 (0.5) 0 356 (0.2) 1 year before index 1089 (1.7) 0 2049 (1.4) Wrist fracture history (%) No 60,237 (92.3) 67,065 (100) 137,118 (93.9) Ever >1 year before index date 2363 (3.6) 0 4040 (2.8) 1 year before index 2695 (4.1) 0 4801 (3.3) BMIa (%) <18.5 5615 (8.6) 9265 (13.8) 10,877 (7.5) 18.6–24.9 32,564 (49.9) 33,554 (50.0) 66,542 (45.6) 25–29.9 18,343 (28.1) 16,720 (24.9) 44,109 (30.2) 30–39.9 8164 (12.5) 6983 (10.4) 22,452 (15.4) ≥40 609 (0.9) 543 (0.8) 1979 (1.4) No. of GP visitsa (%) 0 3219 (4.9) 15,017 (22.4) 15,177 (10.4) 1–5 16,176 (24.8) 15,528 (23.2) 36,213 (24.8) 6–10 17,452 (26.7) 13,677 (20.4) 32,344 (22.2) 11–15 11,938 (18.3) 9140 (13.6) 24,466 (16.8) ≥16 16,510 (25.3) 13,703 (20.4) 37,759 (25.9) No. of GP emergency visitsa (%) 0 53,095 (81.3) 53,221 (79.4) 119,698 (82.0) 1 6570 (10.1) 6392 (9.5) 14,083 (9.6) 2 2439 (3.7) 2844 (4.2) 5282 (3.6) 3–5 2282 (3.5) 3163 (4.7) 4993 (3.4) ≥6 909 (1.4) 1445 (2.2) 1903 (1.3) eGFRa (%) ≤29 708 (1.1) 1974 (2.9) 1777 (1.2) 30–44 3470 (5.3) 7099 (10.6) 9749 (6.7) 45–59 11,950 (18.3) 15,335 (22.9) 30,290 (20.8) 60–89 45,497 (69.7) 39,774 (59.3) 96,838 (66.3) ≥90 3670 (5.6) 2883 (4.3) 7305 (5.0) SBPa (%) <120 9076 (13.9) 9184 (13.7) 18,549 (12.7) (Continues) Journal of Bone and Mineral Research MACE IN IMMINENT-FRACTURE-RISK PATIENTS 1989 n Table 1. Continued OST IFX OBP Variable n = 65,295 n = 67,065 n = 145,959 120–139 26,605 (40.7) 25,872 (38.6) 57,685 (39.5) 140–159 22,588 (34.6) 23,213 (34.6) 51,589 (35.3) ≥160 7026 (10.8) 8796 (13.1) 18,136 (12.4) DBPa (%) <80 33,729 (51.7) 36,247 (54.0) 75,132 (51.5) 80–89 24,505 (37.5) 23,468 (35.0) 53,798 (36.9) 90–99 5764 (8.8) 5732 (8.5) 13,678 (9.4) ≥100 1297 (2.0) 1618 (2.4) 3351 (2.3) No. of concomitant medicinesa (%) 0 5171 (7.9) 16,072 (24.0) 20,662 (14.2) 1–3 12,858 (19.7) 10,222 (15.2) 26,686 (18.3) 4–6 14,446 (22.1) 12,572 (18.7) 29,125 (20.0) 7–9 12,375 (19.0) 10,930 (16.3) 26,049 (17.8) 10–12 8788 (13.5) 7675 (11.4) 18,868 (12.9) ≥13 11,657 (17.9) 9594 (14.3) 24,569 (16.8) Cholesterol measurementa (HDL/LDL) (%) ≤3.5 39,143 (59.9) 42,892 (64.0) 86,247 (59.1) 3.6–5 20,792 (31.8) 18,630 (27.8) 46,654 (32.0) >5 5360 (8.2) 5543 (8.3) 13,058 (8.9) No. of previous fracturesb (%) 0 46,551 (71.3) NA 112,120 (76.8) 1 9763 (15.0) NA 18,129 (12.4) ≥2 8981 (13.8) NA 15,710 (10.8) OST = patients with incident diagnosis of osteoporosis; IFX = patients with incident fragility fracture; OBP = incident users of oral bisphosphonates; SES = socioeconomic status; MI = myocardial infarction; BMI = body mass index; eGFR = estimated glomerular filtration rate; SBP = cholesterol, systolic blood pressure; DBP = diastolic blood pressure. a In the year before start. b Ever. regression selected the key risk factors, which were then Patient and public involvement entered into a final logistic regression equation. Model perfor- mance was evaluated for this final equation and model coeffi- Used data were previously collected and all participant records cients and intercept terms reported. Missing data were were linked-anonymized. Hence, no patients or members of handled using multiple imputation and combined using the public were directly implicated in the design or analysis of Rubin’s rules as required.(43) The data were split 50-50 into the reported data. training and validation datasets. All models were developed using training data and tested using the validation data. Sup- Results plemental Fig. S1 describes the steps used to build the final model. Further explanations of the prediction model develop- A total of 65,295, 67,065, and 145,959 participants were included ment are described in Supplemental Materials. in the OST, IFX, and OBP cohorts, respectively (Fig 1). Most of the The same steps were repeated for 2-year MACE (secondary cohorts were populated by women (OST: 86.80%; IFX: 76.7%; analysis), for 1- and 2-year MI/stroke outcomes (sensitivity analy- OBP: 79.8%). The IFX cohort had the older population (mean sis) and for sex-based models, included in Supplemental age, years [standard deviation]: 79.52 [11.01]), followed by the Materials. OBP (74.35 [10.88]) and OST (73.05 [10.67]). Baseline characteris- tics of each cohort are shown in Table 1. Baseline characteristics stratified by outcome for the development and validation data sets are provided in Supplemental Table S2A–C. Model performance At 1-year follow-up, MACE IR (95% confidence interval [CI], in We assessed the models internally using the validation data sets. units of cases/1000 person-years) was 19.6 (18.5–20.7) in OST, Discrimination was evaluated by calculating the area under the 26.3 (25.4–27.1) in OBP, and 52.6 (50.7–54.5) in the IFX cohort curve (AUC). The AUC was produced for all 20 validation data (Fig. 2). IFX cohort also had the highest incidences when strati- sets, then pooled using Rubin’s rules. Calibration was assessed fied by age groups (Supplemental Fig. S2). Those who experi- by producing calibration plots of observed versus predicted enced MACE were older in general (higher proportion aged probabilities, in tenths of predicted risk. Calibration plots were >75 years) with higher comorbidity (Charlson score) and higher also produced for 10-year age groups and sex. prevalence of drug use (eg, antihypertensives and beta- All statistical analyses took place in R version 3.6.0, including blockers). MICE, glmnet, rpart, gbm, caret, flextable, pROC, and officer Supplemental Fig. S3 display the IRs of 2-year MACE, 1- and packages. 2-year stroke/MI, and MACE and stroke/MI outcomes stratified n 1990 PINEDA-MONCUSÍ ET AL. Journal of Bone and Mineral Research Fig. 1. Study flow chart. OST = patients with incident diagnosis of osteoporosis; IFX = patients with incident fragility fracture; OBP = incident users of oral bisphosphonates; CPRD = clinical practice research datalink. Fig. 2. Incidence rates of MACE after 1 year of follow-up. Incidence rate is reported with 95% confidence intervals. OST = patients with incident diagnosis of osteoporosis; IFX = patients with incident fragility fracture; OBP = incident users of oral bisphosphonates; MACE = composite outcome for the occur- rence of either myocardial infarction, stroke, or cardiovascular disease death. by sex. Overall, despite the considerably lower proportion of Supplemental Figs. S6 and S7 report 1-year MACE calibration males in this study, they suffered higher incidence rates of both plots by age and by age and sex, respectively. Supplemental outcomes in the 1- and 2-year follow-up periods. Figs. S8–S10 present the calibration plots of 2-year MACE and MI/stroke models. Generally, models were well-calibrated, with Development and performance of the prediction models an overpredicting risk for the population <60 years old and underpredicting for those aged >80 years, probably caused by Predictive models using risk factors identified from lasso reach the lower proportion of participants belonging to either AUC values above 70% for the three models, and these equalled category. or outperformed the models using just QRISK factors (Supple- mental Fig. S4). MACE models for OST and OBP populations Selected risk factors for patients at high risk of fracture had higher discrimination values compared with the stroke/MI models (Supplemental Fig. S5), whereas AUC values in IFX model Table 2 displays the risk factors selected from lasso for the overall were better for the stroke/MI outcomes. Differences between 1- models of 1-year MACE along with their odd ratios (OR) and con- and 2-year prediction models were minimal with no apparent fidence intervals, and Supplemental Table S3 lists its beta coeffi- pattern; however, sex-based models (Supplemental Fig. S5B, C) cients. Sex, age, smoking, and drinking in the prior year, atrial show lower AUC values for men-only models. It could be related fibrillation diagnosis, use of antihypertensive medication, prior to the considerably lower sample size of those cohorts. MI or stroke, CVD history, number of concomitant medicines in Journal of Bone and Mineral Research MACE IN IMMINENT-FRACTURE-RISK PATIENTS 1991 n Table 2. Predictors of 1-Year MACE Overall Models (Risk Factors Selected by Lasso Regression) Predictor OST OR (95% CI) IFX OR (95% CI) OBP OR (95% CI) Sex, male (%) 1.61 (1.3, 2) 1.29 (1.12, 1.49) 1.42 (1.25, 1.61) Age group, years (%) 50–59 Ref Ref Ref 60–69 1.36 (0.79, 2.35) 7.66 (3.05, 19.23) 1.58 (1.14, 2.19) 70–79 3.07 (1.85, 5.11) 12.56 (5.01, 31.5) 2.46 (1.79, 3.36) 80–89 4.9 (2.9, 8.28) 20.39 (7.84, 53) 4.26 (3.09, 5.87) >89 7.88 (4.44, 13.99) 27.17 (9.92, 74.38) 5.79 (4.06, 8.24) SES (%) x x 1 x x Ref 2 x x 1.17 (1.02, 1.36) 3 x x 1.28 (1.11, 1.48) 4 x x 1.34 (1.15, 1.56) 5 x x 1.29 (1.08, 1.54) Smokinga x Ex Ref Ref Ref No 0.93 (0.73, 1.18) 1 (0.85, 1.18) 0.98 (0.79, 1.23) Yes 1.42 (1.02, 1.98) 1.16 (0.89, 1.5) 1.14 (0.89, 1.47) Drinkinga Ex Ref Ref Ref No 1.19 (0.7, 2) 1.14 (0.85, 1.54) 1.03 (0.71, 1.49) Yes 0.84 (0.51, 1.39) 1.11 (0.79, 1.57) 0.92 (0.65, 1.31) Diabetes type Ib x x x Diabetes type IIb x x 1.11 (0.9, 1.36) Chronic obstructive pulmonary diseaseb x x 1.05 (0.87, 1.27) Chronic kidney diseaseb x x 0.81 (0.61, 1.06) Rheumatoid arthritisb x x 0.88 (0.74, 1.04) Lupusb x x x Systemic heart diseasea x x 0.62 (0.37, 1.03) Anti-osteoporosis usea x x x Heparin usea x x x Beta-blocker usea x 1.11 (0.95, 1.29) 1.16 (1.02, 1.33) Hypertensiona x x x Deep vein thrombosis or pulmonary embolisma x x x Anticoagulant usea x x 0.93 (0.75, 1.14) Antidepressants TCAa x x x Antidepressants SSRIa x x 1.28 (1.09, 1.49) Hypercholesterolemiaa x x x Statin usea x x x Osteoporosis historyb x x 0.74 (0.66, 0.83) Family history of cardiovascular disease x x 0.94 (0.78, 1.14) Family history of cardiovascular disease before age 60 years x x x Heart failureb x x 1.03 (0.84, 1.27) Migraineb x x x Severe mental illnessb x x x Vascular diseaseb x x x Atrial fibrillationb 1.61 (1.27, 2.05) 1.2 (1.01, 1.43) 1.29 (1.08, 1.55) On antihypertensive drug 1.23 (0.97, 1.55) 1.08 (0.87, 1.33) 1.05 (0.9, 1.22) Antipsychotic usea X X x Steroid usea X x Erectile dysfunctiona X X x Charlson score x 0 x x Ref 1 x x 1.01 (0.87, 1.17) 2 x x 0.94 (0.79, 1.13) ≥3 x x 1.06 (0.85, 1.31) Cardiovascular disease x x No x x Ref Ever >1 year before index date x x 1.16 (1, 1.36) (Continues) n 1992 PINEDA-MONCUSÍ ET AL. Journal of Bone and Mineral Research Table 2. Continued Predictor OST OR (95% CI) IFX OR (95% CI) OBP OR (95% CI) 1 year before index x x 1.18 (0.84, 1.66) 6 months before index x x 1.52 (1.14, 2.03) 1 month before index x x 2.18 (1.49, 3.21) MI or stroke No Ref Ref Ref Ever >1 year before index date 0.99 (0.7, 1.4) 1.23 (0.98, 1.55) 1.34 (1.09, 1.64) 1 year before index 2.03 (1.38, 2.99) 1.56 (1.19, 2.03) 2.52 (2, 3.18) Established CVDb 1.9 (1.46, 2.48) 1.7 (1.42, 2.04) 1.49 (1.25, 1.79) Any fracture history x x x No x x x Ever >1 year before index date x x x 1 year before index x x x Hip fracture history x x x No x x x Ever >1 year before index date x x x 1 year before index x x x Shoulder fracture history x x x No x x x Ever >1 year before index date x x x 1 year before index x x x Spine fracture history x x x No x x x Ever >1 year before index date x x x 1 year before index x x x Wrist fracture history x x x No x x x Ever >1 year before index date x x x 1 year before index x x x BMIa x <18.5 Ref x Ref 18.6–24.9 0.72 (0.49, 1.05) x 0.74 (0.57, 0.96) 25–29.9 0.56 (0.35, 0.9) x 0.58 (0.41, 0.81) 30–39.9 0.42 (0.2, 0.85) x 0.55 (0.36, 0.84) ≥40 0.66 (0.22, 1.93) x 0.39 (0.17, 0.91) No. of GP visitsa 0 Ref x Ref 1–5 1.01 (0.61, 1.66) x 0.9 (0.73, 1.1) 6–10 0.94 (0.56, 1.59) x 0.78 (0.61, 1.01) 11–15 0.95 (0.55, 1.62) x 0.78 (0.6, 1.01) ≥16 1.07 (0.63, 1.81) x 0.86 (0.67, 1.12) No. of GP emergency visitsa x 0 x Ref Ref 1 x 1.25 (1.05, 1.47) 1.15 (0.99, 1.35) 2 x 1.43 (1.15, 1.78) 1.4 (1.13, 1.74) 3–5 x 1.21 (0.97, 1.5) 1.47 (1.2, 1.8) ≥6 x 1.13 (0.82, 1.56) 2.53 (1.96, 3.26) eGFRa ≤29 Ref Ref Ref 30–44 1.11 (0.44, 2.81) 0.87 (0.61, 1.24) 0.91 (0.56, 1.5) 45–59 0.89 (0.37, 2.15) 0.7 (0.38, 1.27) 0.83 (0.49, 1.4) 60–89 0.76 (0.31, 1.86) 0.59 (0.17, 2.01) 0.58 (0.27, 1.27) ≥90 0.79 (0.3, 2.07) 0.52 (0.13, 2.13) 0.5 (0.21, 1.22) SBPa <120 Ref Ref Ref 120–139 1.19 (0.88, 1.61) 1.04 (0.85, 1.28) 1.09 (0.92, 1.31) 140–159 1.32 (0.98, 1.77) 1.2 (0.95, 1.51) 1.24 (1.02, 1.52) ≥160 1.32 (0.91, 1.89) 1.54 (1.11, 2.14) 1.39 (1.09, 1.77) (Continues) Journal of Bone and Mineral Research MACE IN IMMINENT-FRACTURE-RISK PATIENTS 1993 n Table 2. Continued Predictor OST OR (95% CI) IFX OR (95% CI) OBP OR (95% CI) a DBP x <80 x Ref Ref 80–89 x 1.04 (0.89, 1.21) 0.94 (0.82, 1.08) 90–99 x 1.18 (0.91, 1.54) 1.05 (0.84, 1.31) ≥100 x 1.13 (0.71, 1.79) 1.31 (0.89, 1.93) No. of concomitant medicinesa 0 Ref Ref Ref 1–3 0.69 (0.42, 1.12) 1.14 (0.9, 1.44) 0.87 (0.7, 1.07) 4–6 0.83 (0.51, 1.36) 1.29 (1.03, 1.63) 0.75 (0.59, 0.96) 7–9 1.01 (0.61, 1.66) 1.12 (0.87, 1.43) 0.88 (0.68, 1.14) 10–12 1.12 (0.67, 1.87) 1.26 (0.96, 1.65) 1 (0.76, 1.31) ≥13 1.11 (0.66, 1.88) 1.25 (0.95, 1.66) 0.98 (0.74, 1.31) Cholesterol measurementa (HDL/LDL) x ≤3.5 Ref Ref Ref 3.6–5 1.34 (0.92, 1.95) 1.07 (0.56, 2.04) 1.2 (0.98, 1.46) >5 1.74 (0.94, 3.2) 1.54 (0.38, 6.31) 1.35 (0.89, 2.03) No. of previous fracturesb x 0 Ref x x 1 1.24 (1.01, 1.53) x x ≥2 0.99 (0.79, 1.25) x x OST = patients with incident diagnosis of osteoporosis; IFX = patients with incident fragility fracture; OBP = incident users of oral bisphosphonates; OR = odds ratio; CI = confidence intervals; MACE = composite outcome for the occurrence of either myocardial infarction, stroke, or cardiovascular dis- ease death; SES = socioeconomic status; MI = myocardial infarction; BMI = body mass index; eGFR = estimated glomerular filtration rate; SBP = cholesterol, systolic blood pressure; DBP = diastolic blood pressure. a In the year before start. b Ever. the prior year, and estimated glomerular filtration rate (eGFR), Prior studies using CPRD show that the general population aged systolic blood pressure (SBP), and cholesterol measurements in 70+ years had an IR of MACE of 15.1 (per 1000 person-years),(32) the prior year, appeared in all models. Predictors common in whereas our study populations including younger individuals (ie, two of the three cohort models were body mass index (BMI), age 50+ patients) have higher IR. Specifically the IFX cohort had beta-blocker use, number of general practitioner (GP) visits, the highest incidence for MACE (51.1/1000 person-years), which number of GP emergency visits, and diastolic blood pressure could be explained by this cohort having an older age (71% were (DBP) measurements in the year before start. older than 75 years) and the largest proportion of men (23.3%) Supplemental Table S4A, B summarizes risk factors selected among the three cohorts, followed by OBP (26.3–20.2%) and from lasso and ORs for 2-year MACE and for MI/stroke models, OST (19.6–13.2%) cohorts. The observation of higher incidences respectively, and Supplemental Table S5A, B lists its beta coeffi- in IFX cohort was consistent when IR of each study cohort was cients. Supplemental Table S6A–C reports all sex-based models stratified by age groups. and Supplemental Table S7A–C lists its beta coefficients. Fitting the list of risk factors from QRISK into a prediction Detailed explanation and an example of how to obtain an esti- model for 1-year MACE events, we obtained AUC values of mate for an individual is reported in Supplemental Materials. 0.73, 0.67, and 0.71 in OST, IFX, and OBP cohorts, respectively. However, starting from the list of “ALL” risk factors for CVD avail- able in CPRD and selecting the most important through lasso Discussion regression, we obtained model equations that exceed QRISK (AUC in selected risk factors from ALL set: 0.75 in OST, 0.70 in In this study, we evaluated the incidence of major adverse car- IFX, and 0.75 in OBP). This list included generic features as well diovascular events though a composite outcome, MACE,(44) and as those specific to the study population, and all of them can assessed risk factors of CVD to predict this outcome at 1 year in be found readily in primary care data. Among them, age had three different cohorts. The IFX cohort can be used for a sec- the largest statistically significant effect size. ondary fracture prevention program, the OBP cohort has the Comparing our models to existing cardiovascular prediction potential to be used in primary prevention because it approxi- tools, we found that the performances of Framingham and mates patients newly diagnosed and treated for osteoporosis, QRISK studies were higher in the general population: AUC and the OST cohort can be used as a general screening in pri- >0.76 and >0.86, respectively.(5,11) Framingham equations have mary care. been validated and recalibrated multiple times using different We observed that incidence of MACE was slightly higher at populations,(46) whereas QRISK has a higher accuracy for UK pop- 1 year than at 2 years, especially for IFX cohort. When stratifying ulation than the Framingham tool.(47) However, neither tool was by sex, men had higher incidence rates than women, which developed for the osteoporotic/fracture risk population, and nei- agrees with results published by the British Health Foundation, ther includes specific risk factors (eg prior fractures and alcohol where male incidences at UK in 2017 were higher than female.(45) consumption) for these particular patients, in whom short-term n 1994 PINEDA-MONCUSÍ ET AL. Journal of Bone and Mineral Research cardiovascular risk might be over- or underestimated. In fact, Fra- information reported in this article was presented as a poster mingham only permits risk calculation over long periods, and abstract (ID: P1129) by LE-H, SK, MA, ET, CL and, DP-A at the there are no studies extrapolating to shorter risk intervals. WCO-IOF-ESCEO 2020 congress (available at https://doi.org/10. The need of specific CVD tools for populations at higher frac- 1007/s00198-020-05696-3). ture risk and at short-term can be rationalised by the lower per- The study was sponsored by Amgen and UCB Pharma. The formance of the models using the QRISK list (ie, using the study funders play a role in the conceptualization, study design, predictors selected for general population). This may be particu- and preparation of the manuscript. larly interesting to note in our sensitivity analysis, which uses an The study was approved by the Independent Scientific Advi- outcome closer to the QRISK tool: AUC values of MI/Stroke sory Committee with protocol number 18_116R. models decrease to a range of 0.62 to 0.70 when applying the Authors’ roles: Study design: DP-A, SK, MA, CL, and ET. CPRD QRISK factors to osteoporotic/fracture risk population. data management and advanced curation: AD. Data analysis: The proposed predictive models have good predictive power LE-H and SK. Data interpretation: MP-M, LE-H, DP-A, and and internal validity (discrimination and calibration) in OBP and SK. Drafting manuscript: MP-M, LE-H, CC, MA, CL, and OST cohorts for 1-year MACE events (the obtained equations SK. Revising manuscript content: all authors. Approving final ver- are included in this article), and the IFX models reach the 70% sion of manuscript: all authors. SK takes responsibility for the AUC threshold, considered as the minimum acceptable integrity of the data analysis. discrimination.(32) Secondary and sensitivity analysis show no differences using Author Contributions 2-year models and better performance of MACE than MI/Stroke models. Marta Pineda-Moncusí: Writing – original draft; writing – review The proposed study is observational in nature and hence can- and editing. Leena El-Hussein: Formal analysis; writing – original not address causality but rather describe associations. There is draft; writing – review and editing. Antonella Delmestri: Data no guarantee that all possible risk factors are included, but for curation; writing – review and editing. Cyrus Cooper: Writing – all those factors that are, multivariable regression ensures that original draft; writing – review and editing. Alireza Moayyeri: they are adjusted for (and hence reducing the risk of confound- Conceptualization; writing – original draft; writing – review and ing). The three presented cohorts are not mutually exclusive but editing. Cesar Libanati: Conceptualization; writing – original encompass the diversity of the population at high risk of fracture, draft; writing – review and editing. Emese Toth: Conceptualiza- and the different criteria used to evaluate them. Another limita- tion; writing – review and editing. Sara Khalid: Conceptualiza- tion is the lack of external validity, which can be assessed in tion; formal analysis; writing – original draft; writing – review future studies to ensure the validity of the models across differ- and editing. Daniel Prieto-Alhambra: Conceptualization; writ- ent populations. The enhanced performance observed in female ing – review and editing. population was expected due to the higher representation of females in our cohorts. The main strengths of this study are the Peer Review large sample size and the wide selection of routinely collected potential risk factors included.(48) The peer review history for this article is available at https:// To summarize, incidence rate of MACE events in the studied publons.com/publon/10.1002/jbmr.4648. populations ranged from 19.6 to 52.6, with IFX as the cohort with the higher risk. Efforts in predicting the study events outline the differences between general and the osteoporotic/fracture risk Data Availability Statement population. The resulting algorithms include risk factors specific to the study population as well as more generic features that can Data that supports the findings of this study was provided by UK be found easily in primary care data. Further work will focus on CPRD database. Availability of data is subject to protocol validating these models in external cohorts. approval by CPRD’s Research Data Governance Process. Disclosures References 1. Benjamin EJ, Blaha MJ, Chiuve SE, et al. Heart disease and stroke All authors have completed the ICMJE disclosure form at http:// Statistics-2017 update: a report from the American Heart Association. www.icmje.org/disclosure-of-interest/ and declare the following Circulation. 2017;135(10):e146-e603. interests: MP-M, LE-H, AD, CC, and SK have no conflicts to declare. 2. NCD Countdown 2030 Collaborators. NCD countdown 2030: world- DP-A reports an institutional grant from NIHR, grants from Chesi- wide trends in non-communicable disease mortality and progress Taylor and Novartis, grants and other support from Amgen and towards sustainable development goal target 3.4. Lancet. 2018; 392(10152):1072-1088. UCB Biopharma, and other support from Astellas, AstraZeneca, Johnson and Johnson, Janssen—on behalf of IMI-funded EHDEN 3. Bhatnagar P, Wickramasinghe K, Wilkins E, et al. Trends in the epide- miology of cardiovascular disease in the UK. Heart. 2016;102(24): and EMIF consortiums—and Synapse Management Partners. 1945-1952. MA, CL, and ET are current employees of UCB Biopharma and 4. Svedbom A, Hernlund E, Ivergard M, et al. Osteoporosis in the hold stock shares of the company. European Union: a compendium of country-specific reports. Arch Osteoporos. 2013;8:137. Acknowledgments 5. Wilson PW, D’Agostino RB, Levy D, et al. Prediction of coronary heart disease using risk factor categories. Circulation. 1998;97(18):1837- 1847. The authors acknowledge Mahkameh Mafi for project manage- 6. Wilson PWF. Framingham risk score for hard coronary heart disease ment; CPRD for their support; and Amgen and UCB Pharma for [Internet]. MDCalc. Available at: https://www.mdcalc.com/framingham- funding this study. A subsection of an earlier version of risk-score-hard-coronary-heart-disease2021 Journal of Bone and Mineral Research MACE IN IMMINENT-FRACTURE-RISK PATIENTS 1995 n 7. Andersson C, Johnson AD, Benjamin EJ, et al. 70-year legacy of the 29. CPRD. Data highlights [updated page last reviewed March 29, 2021]. Framingham Heart Study. Nat Rev Cardiol. 2019;16(11):687-698. Available at: https://cprd.com/data-highlights. Accessed April 14, 2021. 8. Gage BF, Waterman AD, Shannon W, et al. Validation of clinical clas- sification schemes for predicting stroke: results from the National 30. Delmestri A, Prieto-Alhambra D. CPRD GOLD and linked ONS mortal- Registry of Atrial Fibrillation. JAMA. 2001;285(22):2864-2870. ity records: reconciling guidelines. Int J Med Inform. 2020;136: 104038. 9. Gage BF, van Walraven C, Pearce L, et al. Selecting patients with atrial fibrillation for anticoagulation: stroke risk stratification in patients 31. Kanis, JA, Johansson H, Odén A, et al. “Characteristics of recurrent taking aspirin. Circulation. 2004;110(16):2287-2292. fractures.” Osteoporosis International 2018;29(8):1747-1757. 10. Gage B. CHADS2 score for atrial fibrillation stroke risk [Internet]. 32. Debray TP, Damen JA, Snell KI, et al. A guide to systematic review and MDCalc. Available at: https://www.mdcalc.com/chads2-score-atrial- meta-analysis of prediction model performance. BMJ. 2017;356: fibrillation-stroke-risk2021 i6460. 11. Hippisley-Cox J, Coupland C, Brindle P. Development and validation 33. Lewington S, Clarke R, Qizilbash N, et al. Age-specific relevance of of QRISK3 risk prediction algorithms to estimate future risk of cardio- usual blood pressure to vascular mortality: a meta-analysis of individ- vascular disease: prospective cohort study. BMJ. 2017;357:j2099. ual data for one million adults in 61 prospective studies. Lancet. 2002; 360(9349):1903-1913. 12. QRISK.org. QRISK®3–2018 risk calculator [Internet]. 2018. Available at: https://qrisk.org/three/ 34. Lewington S, Whitlock G, Clarke R, et al. Blood cholesterol and vascu- lar mortality by age, sex, and blood pressure: a meta-analysis of indi- 13. Garcia-Gomez MC, Vilahur G. Osteoporosis and vascular calcification: vidual data from 61 prospective studies with 55,000 vascular deaths. a shared scenario. Clin Investig Arterioscler. 2020;32(1):33-42. Lancet. 2007;370(9602):1829-1839. 14. Yedavally-Yellayi S, Ho AM, Patalinghug EM. Update on osteoporosis. 35. Mach F, Baigent C, Catapano AL, et al. 2019 ESC/EAS guidelines for Prim Care. 2019;46(1):175-190. the management of dyslipidaemias: lipid modification to reduce car- 15. Tanko LB, Christiansen C, Cox DA, et al. Relationship between osteo- diovascular risk. Eur Heart J. 2020;41(1):111-188. porosis and cardiovascular disease in postmenopausal women. 36. Jha P, Ramasundarahettige C, Landsman V, et al. 21st-century haz- J Bone Miner Res. 2005;20(11):1912-1920. ards of smoking and benefits of cessation in the United States. N Engl 16. Prieto-Alhambra D, Premaor MO, Fina Aviles F, et al. The association J Med. 2013;368(4):341-350. between fracture and obesity is site-dependent: a population-based 37. Rosengren A, Smyth A, Rangarajan S, et al. Socioeconomic status and study in postmenopausal women. J Bone Miner Res. 2012;27(2): risk of cardiovascular disease in 20 low-income, middle-income, and 294-300. high-income countries: the Prospective Urban Rural Epidemiologic 17. Premaor MO, Compston JE, Fina Aviles F, et al. The association (PURE) study. Lancet Glob Health. 2019;7(6):e748-e760. between fracture site and obesity in men: a population-based cohort 38. Cosman F, de Beur SJ, LeBoff MS, et al. Clinician’s guide to prevention study. J Bone Miner Res. 2013;28(8):1771-1777. and treatment of osteoporosis. Osteoporos Int. 2014;25(10):2359- 18. Prieto-Alhambra D, Premaor MO, Aviles FF, et al. Relationship 2381. between mortality and BMI after fracture: a population-based study 39. Pisani P, Renna MD, Conversano F, et al. Major osteoporotic fragility of men and women aged >/=40 years. J Bone Miner Res. 2014; fractures: risk factor updates and societal impact. World J Orthop. 29(8):1737-1744. 2016;7(3):171-181. 19. Martinez-Laguna D, Tebe C, Javaid MK, et al. Incident type 2 diabetes 40. Barzilay JI, Davis BR, Pressel SL, et al. The impact of antihypertensive and hip fracture risk: a population-based matched cohort study. medications on bone mineral density and fracture risk. Curr Cardiol Osteoporos Int. 2015;26(2):827-833. Rep. 2017;19(9):76. 20. Collins TC, Ewing SK, Diem SJ, et al. Peripheral arterial disease is asso- 41. Hofbauer LC, Henneicke H. β-Blockers and bone health. J Clin Invest. ciated with higher rates of hip bone loss and increased fracture risk in 2018;128(11):4745-4747. older men. Circulation. 2009;119(17):2305-2312. 42. Alexander LK LB, Ricchetti-Masterson K, Yeatts KB. Calculating 21. Laroche M, Pecourneau V, Blain H, et al. Osteoporosis and ischemic person-time. ERIC Notebook. 2nd ed. Chapel Hill, NC: cardiovascular disease. Joint Bone Spine. 2017;84(4):427-432. NCIPH Training Website, University of North Carolina; 2015. Available at: https://sph.unc.edu/files/2015/07/nciph_ERIC4.pdf. Accessed 22. Sharma A, Einstein AJ, Vallakati A, et al. Risk of atrial fibrillation with September 28, 2021. use of oral and intravenous bisphosphonates. Am J Cardiol. 2014; 113(11):1815-1821. 43. Rubin DB. Underlying Bayesian theory. Multiple imputation for non- response in surveys. New York: John Wiley & Sons; 1987 pp 75-112. 23. Huang BT, Huang FY, Gui YY, et al. Association between bisphospho- nates therapy and incident myocardial infarction: meta-analysis and 44. Wilcox R, Kupfer S, Erdmann E. Effects of pioglitazone on major trial sequential analysis. J Cardiovasc Pharmacol. 2015;66(5):468-477. adverse cardiovascular events in high-risk patients with type 2 diabe- tes: results from PROspective pioglitAzone clinical trial in macro vas- 24. Kranenburg G, Bartstra JW, Weijmans M, et al. Bisphosphonates for cular events (PROactive 10). Am Heart J. 2008;155(4):712-717. cardiovascular risk reduction: a systematic review and meta-analysis. Atherosclerosis. 2016;252:106-115. 45. British Heart Foundation in collaboration with the Institute of Applied Health Research at the University of Birmingham. Heart 25. EMA. Evenity (romosozumab). 2020. Available at: https://www.ema. and circulatory disease statistics 2020. Statistics Compendium europa.eu/en/medicines/human/EPAR/evenity (Tables). Available from: https://www.bhf.org.uk/what-we-do/our- 26. EMA. Clinical investigation of medicinal products for hormone research/heart-statistics/heart-statistics-publications/cardiovascular- replacement therapy of oestrogen deficiency symptoms in postmen- disease-statistics-2020 opausal women. 2005. Available at: https://www.ema.europa.eu/en/ 46. Bitton A, Gaziano TA. The Framingham Heart Study’s impact on clinical-investigation-medicinal-products-hormone-replacement- global risk assessment. Prog Cardiovasc Dis. 2010;53(1):68-78. therapy-oestrogen-deficiency-symptoms 47. Collins GS, Altman DG. An independent and external validation of 27. EMA. Recommendation to restrict the use of Protelos / Osseor (stron- QRISK2 cardiovascular disease risk score: a prospective open cohort tium ranelate). 2013. Available at: https://www.ema.europa.eu/ study. BMJ. 2010;340:c2442. en/news/recommendation-restrict-use-protelos-osseor- 48. Li Y, Sperrin M, Ashcroft DM, et al. Consistency of variety of machine strontium-ranelate learning and statistical models in predicting clinical risks of individual 28. CPRD. Release notes: CPRD GOLD. 2018. Available at: https://cprdcw. patients: longitudinal cohort study using cardiovascular disease as cprd.com/_docs/Release_Notes_October2018.pdf exemplar. BMJ. 2020;371:m3919. n 1996 PINEDA-MONCUSÍ ET AL. Journal of Bone and Mineral Research