Balancing science and informed choice in decisions about vertebroplasty.

Worldwide there are approximately 1.4 million per­ sons with vertebral compression fractures.1 In the United States, there are approximately 750,000; only one third receive treatment.2 Prevalence es­ timates are imprecise because of heterogeneity in how vertebral fracture is defined.3 Annualized di­ rect care expenditures for osteoporotic fractures in the United States were estimated to range from $12 billion to $18 billion in 2002.2 Several treatments (e.g., bed rest, bracing, and pain medications) have been designed to foster pain relief and allow for increased activities. Per­ cutaneous vertebroplasty — the injection of poly­ methylmethacrylate directly into the vertebral com pression fracture — has emerged as a treat­ ment option.4 It is proposed that cement injected directly into the fracture provides immediate sta­ bility and relief of pain. Despite several studies suggesting a positive treatment effect for verte­ broplasty, as compared with control treatments, there have been no blinded or placebo­controlled, randomized trials. During the past 6 years, the number of verte­ broplasty procedures performed in the United States has doubled, from 0.43 to 0.89 per 1000 persons, with variation across geographic areas by nearly a factor of 30 (Fig. 1, and the figure in the Supplementary Appendix, available with the full text of this article at NEJM.org). The first two blinded, randomized, placebo­controlled tri­ als of vertebroplasty are reported in this issue of the Journal,5,6 and the results may change verte­ broplasty from a procedure that is virtually al­ ways considered to be successful to one that is considered no better than placebo. In investigating the treatment of subjective re­ ports of pain and decreased function, the only way to distinguish the effect of expectations by the patients from the intervention itself is to conceal the treatment from the patients. Two unblinded trials have compared vertebroplasty with medical management: one showed better pain relief with vertebroplasty at 1 day and 2 weeks,7 and the other showed no significant difference at 3 months.8 In the Investigational Vertebroplasty Safety and Efficacy Trial (INVEST) (ClinicalTrials.gov num­ ber, NCT00068822), Kallmes et al.5 report that pain and disability outcomes at 1 month in a group of patients who underwent vertebroplasty were similar to those in a control group that un­ derwent a sham procedure, with a trend toward a higher rate of clinically meaningful improve­ ment in the vertebroplasty group. Both study groups showed improvement within 3 days after the procedure. The cutoff for the primary out­ come was 1 month, but data were collected for 3 months. Because of crossover to the other group, the intention­to­treat analysis at 3 months prob­ ably underestimated the true treatment effect, whereas an as­treated analysis probably would have overestimated the effect. Interestingly, patients in the two groups (who were still unaware of which procedure they had received) crossed over to the other group after 1 month. The patients who crossed over had worse outcomes than those who did not request the other intervention. Since more patients than pre­ dicted were able to guess which treatment they had received, there might have been a benefit in understanding the treatment effect in those who guessed their treatment accurately. Did those who received vertebroplasty and guessed correctly do better or worse than those who did not guess correctly?