Retrospective, longitudinal, population-based cohort study in Barcelona City Primary Care, managed by the Catalan Health Institute and made up of 51 primary care teams, staffed by 754 family and community physicians in the city of Barcelona. The target population is 1,257,695 inhabitants (74.6% of the population of the city of Barcelona), of which 51.5% (648,067 inhabitants) are women.16

The study included all women assigned to the 51 teams of Barcelona City Primary Care who had an active prescription for bisphosphonates in January 2014 and had received bisphosphonate treatment for at least five. The cohort was followed for five years, from January 2014 to January 2019. No new patients were included, and attrition was due to mortality or loss to follow-up when the women were transferred to other healthcare providers or territories. Patients lost to follow-up were not excluded but were censored in the survival analysis. Patients with less than two years of follow-up and those who began treatment with a drug other than bisphosphonate for osteoporosis during the follow-up period were excluded from the study.

Included women were classified according to the risk of developing new fractures. Women at high risk of fracture were defined as those who had a history of osteoporotic fracture17 (fractures that occurred in the previous five years) and/or who received treatment with an aromatase inhibitor during the follow-up period.18,19 We considered women who did not have a history of osteoporotic fracture17 nor were receiving treatment with an aromatase inhibitor to be at low risk of fracture.

The continuity or deprescription of the bisphosphonate was assessed according to clinical criteria during the follow-up period (Table 1). We analyzed the number of days that the woman was treated with bisphosphonate for each year of follow-up; if the woman had been prescribed the bisphosphonate for at least 150 consecutive days, we considered that she had been taking bisphosphonates for that year. Based on this definition, the women included in the study were classified into patients who were either continuing bisphosphonate treatment (those with an active prescription for ≥60% of the follow-up period) or who were deprescribed bisphosphonate treatment (active prescription for <60% of the follow-up period). Supplementary Figure 1 shows an explanatory scheme of the study design.

At the beginning of each year, family physicians were sent a list of their patients in treatment with bisphosphonates for at least five years, together with an information letter and evidence sheets justifying their withdrawal. The physicians themselves then made the final determination on the appropriateness of deprescribing the medication, based on the individual patient's clinical characteristics. Every three months the doctors received feedback on the drugs withdrawn.

All data were extracted anonymously from the electronic medical records for primary care of the Catalan Institute of Health (ECAP). The ECAP database contains longitudinal information on demographics, socioeconomic status, diagnoses, symptoms, and prescriptions, among other data.

The following variables were collected:

• •

  Demographic variables: age, sex and the MEDEA deprivation index.20 This index, used in the ECAP, categorizes census tracts into quintiles based on their socioeconomic situations. People residing in quintile 1 (Q1) have the lowest level of deprivation, while those living in Q5 areas have the highest level of deprivation.

• •

  Diagnostic variables: diseases were coded in the ECAP using the International Classification of Diseases, 10th revision (ICD-10). Diagnostic codes for osteoporotic fractures were identified (Supplementary Table 1). We analyzed vertebral, hip, or femur fractures, and total osteoporotic fractures. Due to the possible adverse effects on the bone produced by inhaled glucocorticoids in postmenopausal women, we also included diagnoses for chronic obstructive pulmonary disease (COPD, ICD-10 codes J41, J42, J43 and J44); and asthma (J45).

• •

  Prescription drug-related variables: the active prescription database in the primary care medical records system includes the drugs dispensed by community pharmacies, thus excluding drugs dispensed in hospitals or by hospital pharmacy services. Drugs were classified according to the Anatomical, Therapeutic, Chemical (ATC) Classification System. We analyzed bisphosphonates (ATC codes M05BA, except zolendronic acid, a drug for hospital use [ATC M05BA08] and M05BB as well as aromatase inhibitors, ATC code L02BG), linked to an increased risk of frailty fractures and associated with secondary osteoporosis.

Patients’ demographic and clinical profile was analyzed using descriptive statistics. The comparative analysis was made between the group of women who discontinued versus the group of women who continued bisphosphonate treatment. The cumulative incidence of fractures and incidence density were calculated, and logistic regression analysis was used to determine the association of each variable with the fracture event (vertebral, hip/femur, and total fractures). Cox models were applied to calculate the hazard ratio (HR) and its confidence interval (CI), which showed the strength of the association of each variable analyzed with the appearance of the event over time.

All analyses were performed with the SPSS statistical package for Windows, version 25 (SPSS Inc., New York, IL, USA). The level of significance was set at 0.05.

The variables collected from the electronic medical record were linked to a unique and anonymous personal identifier, thus guaranteeing the confidentiality of the patients in accordance with the provisions of Organic Law 3/2018, of December 5, on personal data protection and guarantee of digital rights and Regulation (EU) 2016/679 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data.

The study protocol was approved by the Clinical Research Ethics Committee of the Foundation University Institute for Primary Health Care Research Jordi Gol i Gurina (IDIAPJGol) (Protocol no: CEI: 21/248-P).

Our results suggest that deprescribing bisphosphonates in women who have already received five years of treatment does not increase the risk of developing new fractures. Indeed, in low-risk women, continuing this treatment might could even favor the appearance of new osteoporotic and vertebral fractures. Other factors associated with an increased risk of fracture include advanced age and polypharmacy.

These results are in line with the recommendations of large scientific societies to periodically reassess the need for continuing bisphosphonate treatment after five years.3,14,21 However, there is still little evidence to elucidate the optimal duration of treatment and the length of its suspension.

Studies of bisphosphonate have not demonstrated long-term efficacy for preventing hip fractures or non-vertebral fractures, and results have been contradictory for clinical vertebral fractures.10,11,22 Moreover, there are few observational and real-world data studies assessing the impact of bisphosphonate deprescription in terms of fracture, and their results are inconsistent.23–25

Our results are in line with those from other recent cohort studies, which have also failed to find an increased risk of fracture (vertebral fracture, hip fracture or any osteoporotic fracture) in women who take therapeutic holidays after two to three years of treatment with bisphosphonates.23,24 Adams et al. assessed the risk of fracture at one year from the therapeutic vacation; the authors did not observe any benefit associated with continuing bisphosphonates, whether in women with or without a history of fracture. Similarly to our findings, they also saw a reduction in the risk of vertebral fracture associated with bisphosphonate discontinuation among women with no history of fracture.23 In another cohort study, this time in women with a history of fracture who had received at least two years of bisphosphonates, the authors likewise reported no differences in the risk of vertebral or hip fractures in patients who discontinued bisphosphonate treatment after 2 years compared with those who continued it for 10–13 years.24

Our results are also consistent with those from different systematic reviews, which question the real efficacy of bisphosphonates in women at low risk of fracture.4–6 However, they differ from those reported by other authors, such as Curtis et al., who observed an increased risk of fracture associated with discontinuing alendronate for more than two years, although without differentiating between women at high versus low risk.25

This study draws data from a large database of real-world medical records pertaining to women previously exposed to long-term bisphosphonate treatment and attended in primary care. The large number of women analyzed makes it a representative cohort of women taking bisphosphonates in Catalonia. Likewise, the five-year follow-up period makes it possible to assess, with a certain degree of certainty, the impact of the intervention on the appearance of new fractures.

Retrospective cohort studies using population-based databases allow the study of a large number of patients and facilitate the timely attainment of results. On the other hand, underreporting by health professionals may lead to inaccuracies in the medical record. In addition, the variables related to the treatments are extracted from the electronic prescription database, precluding the assessment of therapeutic adherence. Furthermore, our database does not include zolendronic acid, commonly used in Spanish hospitals, so there is a risk of misclassification bias in the women included in the deprescription group. There are also different tools for predicting the risk of fracture according to other associated variables (e.g. bone mineral density, tobacco, alcohol), which were not contemplated in our study; however, recent reviews have concluded that the presence of the criteria used (advanced age, female sex, and history of fracture) are independently associated with an increased risk of fracture, regardless of other diagnostic criteria.26 In our study, treatment with calcium and/or vitamin D supplements has not been considered as a variable, due to the difficulty that exists in the studies based on real-world data to know the calcium and/or vitamin D supplementation done through diet and/or supplementation. On the other hand, in prospective, randomized, placebo-controlled trials of supplementation have not shown a benefit on fracture rates. Furthermore, we measured the exposure to glucocorticoids indirectly, according to the diagnosis of the most prevalent pathologies in primary care that are characterized by frequent chronic use of glucocorticoids. This decision was due to the complexity involved in quantifying the exposure time to glucocorticoids that is associated with an increased risk of fracture in terms of dose and duration of treatment.

Finally, in our study, treatment continuity was defined based on the fulfillment of two criteria: annual exposure, by which women had to be treated with bisphosphonate for at least five consecutive months, and exposure during follow-up, by which the women had to have been prescribed bisphosphonate for at least 60% of the time. Regarding the annual exposure criterion, despite having established a minimum cutoff of five months, most of the women in the continuation group were prescribed bisphosphonate for more than six months of the year. These data are supported by different studies23,27–29 assessing the risk of fracture in relation to the duration of exposure to bisphosphonate, which is considered effective if taken 50–80% of the time. However, there is no consensus on the methodology to assess the exposure time or the cutoff values to be used, and the literature uses heterogeneous criteria.23,27–30 Furthermore, in the specific case of bisphosphonates, their effect lasts for years after they are withdrawn,2 so the patient continues to perceive a benefit even after they stop taking them.

Our results suggest that deprescribing bisphosphonates in women who have already received five years of treatment does not increase fracture risk, calling into question whether the high number of women at low risk of fracture in our area benefit from prolonged bisphosphonate treatment. It would be interesting to further understand the advantages of extending treatment based on fracture risk stratification. In addition, this study manifests the role of the real world data in the risk benefit assessment in primary care patients.