Development and external validation of a 1- and 5-year fracture prediction tool based on electronic medical records data: The EPIC risk algorithm

OBJECTIVES

We aimed to develop and validate a fracture risk algorithm for the automatic identification of subjects at high risk of imminent and long-term fracture risk.

RESEARCH, DESIGN, AND METHODS

A cohort of subjects aged 50-85, between 2007 and 2017, was extracted from the Catalan information system for the development of research in primary care database (SIDIAP). Participants were followed until the earliest of death, transfer out, fracture, or 12/31/2017. Potential risk factors were obtained based on the existing literature. Cox regression was used to model 1 and 5-year risk of hip and major fracture. The original cohort was randomly split in 80:20 for development and internal validation purposes respectively. External validation was explored in a cohort extracted from the Spanish database for pharmaco-epidemiological research in primary care.

RESULTS

A total of 1.76 million people were included from SIDIAP (50.7 % women with mean age of 65.4 years). Hip and major fracture incidence rates were 3.57 [95%CI 3.53 to 3.60] and 11.61 [95%CI 11.54 to 11.68] per 1000 person-years, respectively. The derived model included 19 risk factors. Internal validity showed good results on calibration and discrimination. The 1-year C-statistic for hip and major fracture were 0.851 (95%CI 0.853 to 0.864), and 0.717 (95%CI 0.742 to 0.749) respectively. The 5-year C-statistic for hip and major fracture were 0.849 (95%CI 0.847 to 0.852) and 0.724 (95%CI 0.721 to 0.727) respectively. External validation showed good performance for hip and major fracture risk prediction.

CONCLUSIONS

We have developed and validated a clinical prediction tool for 1- and 5-year hip and major osteoporotic fracture risks using electronic primary care data. The proposed algorithm can be automatically estimated at the population level using the available primary care records. Future work is needed on the cost-effectiveness of its use for population-based screening and targeted prevention of osteoporotic fractures.

Assessment of thyroiditis risk associated with HPV vaccination among girls aged 9-18 years: A time-varying cohort study

BACKGROUND

Previous studies have suggested a relationship between human papillomavirus vaccine and autoimmune diseases, including thyroiditis. Thus, we aimed to evaluate the risk of thyroiditis associated with HPV vaccination among girls using the Primary Care Database For Pharmacoepidemiological Research (BIFAP) in Spain.

METHODS

In this retrospective cohort study, girls in BIFAP aged 9-18 years from 2007 to 2016, free of past thyroiditis and HPV vaccination, were included. Hazard Ratios (HRs; 95% CI) of thyroiditis were calculated within exposed periods (up to 2 years of vaccination) and post-exposed periods (from 2 years after vaccination onwards) compared with non-exposed periods, overall, by dose and by type of vaccine, adjusted for potential confounders collected at different times. In a post-hoc analysis, we moved back the thyroiditis date (30 days) as a theoretical delay in diagnosis.

RESULTS

Out of the 388,411 girls included in the cohort, 153,924 were vaccinated against HPV and 480 thyroiditis (253 autoimmune) cases were identified (334 non-exposed; 103 exposed; 43 post-exposed). Adjusted HR was 1.18 [95% CI: 0.79-1.76] for exposed (1.25 [0.77-2.04] for bi- and 1.15 [0.76-1.76] for quadri-valent vaccines) and 1.26 [0.74-2.14] for post-exposed periods. HR was 1.50 [0.87-2.59] for the 1^st dose, 1.13 [0.66-1.91] for the 2^nd and 1.11 [0.71-1.72] for the 3^rd one. When the diagnosis date was moved back, the risk was 1.14 [0.76-1.70] for exposed period, being 1.80 [0.86-3.76] and 1.40 [0.74-2.66] after 1^st dose of bi- and quadri-valent, respectively.

CONCLUSIONS

We did not observe an increased risk of thyroiditis following HPV vaccination (whether bi- or quadri-valent). Even though the point estimate was higher after 1^st HPV vaccination dose than after subsequent doses, a dose-effect was not confirmed. Results remained similar after applying a lag time.

Effect of regulatory interventions on agomelatine use in Spain: A multiple intervention time-series analysis in a nationwide electronic healthcare record database

BACKGROUND

Liver injury is an important identified risk for agomelatine and several measures were put in place to prevent and minimize such risk. The study aims to assess the impact of four interventions on the incidence of agomelatine use, particularly among patients aged ≥75 in Spain between 2011 and 2018.

METHODS

Quasi-experimental interrupted time-series analysis to examine data from a nationwide electronic healthcare record database (BIFAP). Quarterly cumulative incidence of agomelatine use per 100 000 patients was calculated and the impact of four regulatory interventions was quantified.

RESULTS

The incidence of agomelatine use decreased by 85% and 87% from first quarter 2011 to last quarter 2018 in patients below and above 75 years old, respectively. Regulatory actions taken were not associated with an immediate and significant falling level of use or slope. The incidence was less than expected 6 months after the first and third intervention for patients below and above 75 years old, and more than expected after the second and fourth intervention for both populations, though these analyses were underpowered to observe significant results. The downward trend became less pronounced, reaching a residual level of use, which remained stable in the last segment of the study period.

CONCLUSION

New users of agomelatine decreased throughout the study period, starting before interventions took place. The effect of specific interventions might be masked by the progressive decrease tendency, constant over the study period. The effects of external factors that might overlap, unintended consequences, and issues concerning statistical modeling in situations where rates are already falling, should be considered when interpreting the results.

Papillomavirus vaccination and Guillain-Barre Syndrome among girls: A cohort study in Spain

BACKGROUND

Studies of the association of Guillain-Barré Syndrome (GBS) with papillomavirus vaccination (HPVv; scheduled from 2007) have provided contradicting results, probably due to the low frequency of this disease. We aimed at estimating that risk relative to non-vaccination among girls, by using the Spanish Primary Care Database for Pharmacoepidemiological Research (BIFAP).

METHODS

A cohort study of girls aged 9-18 years during 2007-2016 free of GBS or HPVv was selected and followed up to GBS diagnosis. Follow-up time was divided by time-varying HPVv exposure and confounders. Crude Incidence rates (IR per 1,000,000 person-years (py)) and adjusted Hazard Ratios (HR) of GBS were estimated anytime after vaccination compared to non-exposed periods. HRs were also estimated for the first 90 days after HPVv (risk-window) and thereafter.

RESULTS

Out of 388,849 girls, of which 154,255 were vaccinated, 6 'confirmed' GBS cases occurred during non-exposure periods (IR of 5.83 per million person-years; 95% CI: 2.62-12.97) and 3 'confirmed' cases anytime after vaccination (IR of 7.87; 95% CI: 2.54-24.39). The resulting adjusted HR anytime after vaccination was 1.24 (95% CI: 0.19-8.00). All three cases occurred after the risk window of 90 days with an HR of 1.77 (95% CI: 0.25-12.54) for post-exposure periods as compared with non-exposure. Since zero cases occurred during the risk window, no HR could be estimated for exposed periods.

CONCLUSIONS

Incidences of GBS were in line with the range previously reported for young people, supporting the potential of BIFAP for performing studies on GBS. However, a lack of power may be present for quantifying the relative risk of such a rare disease after the vaccination among the study cohort, where we can only exclude an increased risk of 8-times relative to no vaccination.

Use of hydrochlorothiazide and risk of skin cancer in a large nested case-control study in Spain

PURPOSE

Hydrochlorothiazide (HCTZ) use has been linked to skin cancer in northern European countries. We assessed the association between HCTZ exposure and risk of malignant melanoma (MM) and keratinocyte carcinoma (KC) in a European Mediterranean population.

METHODS

Two parallel nested case-control studies were conducted in Spain using two electronic primary healthcare databases, each one providing data on both exposure and outcomes: SIDIAP and BIFAP. Cancer cases were matched to 10 controls by age and gender through risk-set sampling. The ORs and 95% CI for MM and KC associated with previous HCTZ use were estimated using conditional logistic regression. In BIFAP, KC cases were further identified as basal cell carcinoma (BCC) or squamous cell carcinoma (SCC).

RESULTS

In adjusted analyses, both ever and cumulative high (≥50,000 mg) use of HCTZ were associated with an increased risk of KC. The risk estimates for high use were 1.30 (1.26-1.34) in SIDIAP and 1.20 (1.12-1.30) in BIFAP, with a lower risk for BCC (1.11 [1.02-1.21]) than for SCC (1.71 [1.45-2.02]). A dose-response relationship was observed between cumulative doses of HCTZ and KC risk. Inconsistent results were found for high use of HCTZ and risk of MM: 1.25 (1.09-1.43) in SIDIAP and 0.85 (0.64-1.13) in BIFAP.

CONCLUSIONS

In this European Mediterranean population, a high cumulative use of HCTZ was related to an increased risk of KC with a clear dose-response pattern.

Drug utilization study on immediate release Fentanyl in Spain. Prevalence, incidence, and indication

PURPOSE

We aimed to characterize the trends of immediate release fentanyl (IRF) use in Spain between 2012 and 2017 and indication for its use. IRF drugs are rapid-acting opioids approved to treat breakthrough cancer pain (BTCP) in patients already receiving maintenance opioid therapy for chronic cancer pain. A substantial increase in consumption of IRF has been observed with emerging cases of abuse and dependence, most of them in noncancer patients.

METHODS

An ecological descriptive consumption study with aggregated data from drug dispensed by community pharmacies and reimbursed by the National Health System in which Defined Daily Doses per 10 000 inhabitants (DID) were calculated and a retrospective cohort study using data from the Spanish Database for Pharmacoepidemiological Research in Primary Care in which participants entered the cohort study after 1 year with the Primary Care Practitioners were performed. Annual prevalence and incidence rate of IRF use were estimated by sex and calendar year. Potential indication was also assessed.

RESULTS

IRF use in Spain increased from 2.1 DID in 2012 to 3.8 DID in 2017. The incidence rate and prevalence increased in 53% and 74%, respectively. Patients without previous cancer or BCTP diagnosis represented 27% of incident users, predominantly women. Half of patients with noncancer-related diagnosis had a musculoskeletal disorder linked to the first IRF prescription.

CONCLUSIONS

National consumption and new IRF users in Spain increased over the study period and one quarter of patients did not have a BTCP or cancer diagnosis registered in their clinical record.

The recording and prevalence of Inflammatory bowel disease in girls' primary care medical Spanish records

PURPOSE

Inflammatory bowel disease (IBD) recording validation among girls in the Spanish Primary Care Database For Pharmacoepidemiological Research (BIFAP).

METHODS

In this observational study, girls aged 9 to 18 years registered in BIFAP between 2002 and 2016, were followed up until there was a recorded IBD diagnosis or a referral to specialist indicating IBD. Anonymized profiles were reviewed to retrieve diagnosis confirmation (a positive colonoscopy or biopsy, specialist, or physician's comments mentioning the IBD diagnosis) or discarding (negative procedure results, alternative diagnosis, or family history). "possible" IBD were profiles missing that evidence, or had suspected IBD. The prescriptions of intestinal anti-inflammatory agents, azatioprine, and mercaptopurine were collected. The prevalence of IBD was estimated after review.

RESULTS

Out of 480 634 girls, 323 had a first ever recorded IBD, of which, 37.8% (N = 122) were "confirmed" incident IBD diagnosis, 19.8% (N = 64) discarded and 38.7% (N = 125) "possible" IBD. Additionally, 12 IBD records (3.7%) referred to prevalent IBD. Prescriptions were recorded in 94.3% (confirmed), 63.2% (possible), 83.3% (prevalent), and 3.1% (discarded) IBD cases. Prevalence was 52.83 "confirmed" or 93.58/10⁵ girls when "possible" IBD were added.

CONCLUSIONS

For a third of the girls, the first recorded IBD included evidence confirming the diagnosis while most of those with missing evidence had treatment indicated for IBD. For research focused in sensitivity, an algorithm including "possible" plus "confirmed" episodes is recommended, whereas only "confirmed" to guarantee higher predictive value. Prevalence suggests that IBD is not a rare disease among girls.

Cessation rate of anti-osteoporosis treatments and risk factors in Spanish primary care settings: a population-based cohort analysis

Among 95,057 patients ≥50 years with new anti-osteoporosis medications (AOM) (2001-2013) in primary care, 1-year cessation was 51% (28%-68%), higher in men, smokers, patients with missing lifestyle data, and out normal BMI, and lower in those aged 60-79, with recent fractures or other anti-osteoporotics, suggesting non-severe osteoporosis and less risk awareness.

PURPOSE

Low compliance to anti-osteoporosis medications (AOM) has been previously reported. We aimed to estimate 1-year cessation rates of different AOMs as used in Spanish healthcare settings, and to identify associated risk factors.

METHODS

A cohort study was performed using primary care records data (BIFAP). Patients entered the cohort when aged 50 years in 2001-2013, with ≥1 year of data available, and identified as incident users of AOM (1-year washout). Participants were divided into six cohorts: alendronate, other oral bisphosphonates, selective oestrogen receptor modulators, strontium ranelate, teriparatide, and denosumab. Patients were followed from therapy initiation to the earliest of cessation (90-day refill gap), switching (to alternative AOM), loss to follow-up, death, or end of 2013. One-year therapy cessation was estimated using life tables. Hazard ratios (of cessation) according to age, sex, lifestyle factors, morbidity, and co-medication were estimated after stepwise backwards selection.

RESULTS

A total of 95,057 AOM users were identified (91% women; mean age 68). One-year cessation was 51% overall, highest for strontium ranelate (68%), and lowest for denosumab (28%). Cessation probability was higher in men (14% to 2.1-fold), smokers (>6%), and patients with missing BMI (19-28%) or smoking (6-20%) data, and overweight/obese/underweight (7% to 2.6-fold increase compared to normal weight). Patients aged 60-79 years, with a recent fracture or other drugs used for osteoporosis, had better persistence.

CONCLUSIONS

Over half of the patients initiating AOM stopped therapy within the first year after initiation. The described risk factors for cessation could be proxies for non-severe osteoporosis, and/or disease/risk awareness, which could inform the targeting of high-risk patients for monitoring and/or interventions aimed at improving persistence.

Secular trends of use of anti-osteoporotic treatments in Spain: A population-based cohort study including over 1.5million people and more than 12years of follow-up

OBJECTIVE

Different regulatory actions for anti-osteoporotic medication (AOM) were taken during the last years, including marketing of new drugs, safety warnings, or restrictions on the indications. We aimed to characterise the secular trends of AOM use in Spain from 2001 to 2013.

METHODS

A cohort study using the Spanish Database for Pharmacoepidemiological Research in Primary Care (BIFAP), was performed. BIFAP includes anonym records for 4million patients. Participants entered the study when aged ≥50years in 2001-2013 and after 1year of data available, and were followed to an AOM prescription (including alendronate, other bisphosphonates, SERM (selective estrogen receptor modulators), strontium ranelate, teriparatide or denosumab), death, lost or the end of December 2013. Prevalence (%) and incidence rate (IR/1000person-years (py)) of AOM users were computed by years and sex.

RESULTS

Out of 1.5million participants, 135,410 received AOM treatment during 2001-2013. Prevalence was 6.1% (women) and 1.1% (men), that increased from 2001 (2.0%) to 2009 (7.6%) to decrease thereafter. Out of them, 95,057 were incident. The IR was 24.90 (women) and 2.77 (men), that increased from 2001 (21.25 and 1.96) to 2007 (35.84 and 3.64), and decreased to 12.48 and 1.81 (2013). IRs were highest for bisphosphonates along the years (ranging 3.70-14.73 and 0.57-1.75 in women and men respectively), followed by SERM up to 2005 (6.51-9.02 and 0.06-0.07), and strontium ranelate from 2006 (4.66 and 0.45) to 2012 (2.05 and 0.26). IR for teriparatide increased from marketing in 2004 (0.10-1.01 and 0.02-0.29), as was denosumab from marketing in 2011 (0.03-2.64 and 0.09-0.15).

CONCLUSIONS

Population-based estimates of AOM use in Spain peaked in 2007-2009 and decreased thereafter, irrespective of age and sex. New treatments were ten times higher in women than men. Bisphosphonates were the most frequently prescribed class, followed by SERM in women before 2006, strontium otherwise till 2012, and denosumab in women or teriparatide in men in 2013. Changes in the osteoporosis criteria, fracture risk assessment strategies, and regulatory actions for AOM around the time, may explain that trend.

Cardiovascular and Renal Outcomes of Renin-Angiotensin System Blockade in Adult Patients with Diabetes Mellitus: A Systematic Review with Network Meta-Analyses

BACKGROUND

Medications aimed at inhibiting the renin-angiotensin system (RAS) have been used extensively for preventing cardiovascular and renal complications in patients with diabetes, but data that compare their clinical effectiveness are limited. We aimed to compare the effects of classes of RAS blockers on cardiovascular and renal outcomes in adults with diabetes.

METHODS AND FINDINGS

Eligible trials were identified by electronic searches in PubMed/MEDLINE and the Cochrane Database of Systematic Reviews (1 January 2004 to 17 July 2014). Interventions of interest were angiotensin-converting enzyme (ACE) inhibitors, angiotensin receptor blockers (ARBs), and direct renin (DR) inhibitors. The primary endpoints were cardiovascular mortality, myocardial infarction, and stroke-singly and as a composite endpoint, major cardiovascular outcome-and end-stage renal disease [ESRD], doubling of serum creatinine, and all-cause mortality-singly and as a composite endpoint, progression of renal disease. Secondary endpoints were angina pectoris and hospitalization for heart failure. In all, 71 trials (103,120 participants), with a total of 14 different regimens, were pooled using network meta-analyses. When compared with ACE inhibitor, no other RAS blocker used in monotherapy and/or combination was associated with a significant reduction in major cardiovascular outcomes: ARB (odds ratio [OR] 1.02; 95% credible interval [CrI] 0.90-1.18), ACE inhibitor plus ARB (0.97; 95% CrI 0.79-1.19), DR inhibitor plus ACE inhibitor (1.32; 95% CrI 0.96-1.81), and DR inhibitor plus ARB (1.00; 95% CrI 0.73-1.38). For the risk of progression of renal disease, no significant differences were detected between ACE inhibitor and each of the remaining therapies: ARB (OR 1.10; 95% CrI 0.90-1.40), ACE inhibitor plus ARB (0.97; 95% CrI 0.72-1.29), DR inhibitor plus ACE inhibitor (0.99; 95% CrI 0.65-1.57), and DR inhibitor plus ARB (1.18; 95% CrI 0.78-1.84). No significant differences were showed between ACE inhibitors and ARBs with respect to all-cause mortality, cardiovascular mortality, myocardial infarction, stroke, angina pectoris, hospitalization for heart failure, ESRD, or doubling serum creatinine. Findings were limited by the clinical and methodological heterogeneity of the included studies. Potential inconsistency was identified in network meta-analyses of stroke and angina pectoris, limiting the conclusiveness of findings for these single endpoints.

CONCLUSIONS

In adults with diabetes, comparisons of different RAS blockers showed similar effects of ACE inhibitors and ARBs on major cardiovascular and renal outcomes. Compared with monotherapies, the combination of an ACE inhibitor and an ARB failed to provide significant benefits on major outcomes. Clinicians should discuss the balance between benefits, costs, and potential harms with individual diabetes patients before starting treatment.

REVIEW REGISTRATION

PROSPERO CRD42014014404.