A second opportunity for assessing predictive elements of treatment failure was offered by the GLOW stud y [64]. This is a prospective cohort study of more than 60,000 women from ten countries recruited from 723 primary care practices in 17 sites. These centers were located in six European countries, Canada, the USA, and Australia. Women received a self-administered posted questionnaire including personal characteristics, diagnostics, and treatments received, quality of life, and fracture history. The results of this analysis were those from baseline to the end of the third year of observation. According to the definition of the IOF working group [62], treatment failure was defined as the occurrence of two incident fractures after being on therapy for at least 1 year. During the 3-year study period, 6.5 % of women on continuous therapy with the same drug sustained one fracture, while 1.3 % suffered two or more. The variables associated with treatment failure (≥2 incident fractures) in univariate analyses were a lower score in the SF-36 scale (physical function and vitality) at baseline, higher FRAX© score, falls in the past 12 months, having more comorbid conditions, prior fracture, current use of glucocorticoids, need of arms to assist when standing, and unexplained weight loss. After multivariate analysis, a worse SF-36 vitality score, prior fracture, and two or more falls in the previous year remained as independent predictors. These results are summarized in Table 5.1 and Fig. 5.1.

A third recent investigation has been carried out using the SIDIAP database in Catalonia, Spain [65]. SIDIAP gathers clinical information from electronic records of primary care health providers, prescription and pharmacy invoices, laboratory tests, and hospital admissions. The ICD-10 coding system is used, and the database includes data of more than five million people (over 80 % of the population of Catalonia). The cases analyzed were those patients suffering incident fractures in spite of being persistent for at least 6 months and with a medication possession ratio of 80 % or above. Only 35 % of patients among the 21,385 got a prescription completed at least 6 months of treatment. In this subgroup of persistent individuals, the incidence of fractures while on treatment was of 3.4 per 100 person-years. The predictors of these fractures were older age, previous fracture, underweight, inflammatory arthritis, use of proton pump inhibitors, and vitamin D deficiency (Table 5.1). To note that when starting bisphosphonates, the time elapsed between a prior fracture and the date of treatment initiation was relevant in these data. When the fracture was recent (less than 6 months before starting therapy), the risk was higher than for “older” fractures (those that had occurred more than 6 months before starting the drugs) (Fig. 5.2).