Health care professionals in the field of postmenopausal osteoporosis were described by the article of Bone et al.(1) Since then, this publication has been spread all over the world and used as a promotion material to show the long-term effects of alendronate.
The primary mission in osteoporosis treatment is to reduce the fracture risk, irrespective of BMD changes (while admitting that BMD is rather strongly associated with bone strength or fracture risk at the population level). Nevertheless, it is well established that BMD does not equate bone fragility of an individual patient, and moreover, BMD itself is subject to substantial uncertainty.(2, 3) Because the quality rather than the quantity of bone has gained great attention, our colleagues are more interested in the antifracture data of alendronate rather than its favorable effects on BMD.
Bone et al.(1) accepted fractures and stature as the safety data, and they suggested that prolonged treatment did not result in any loss of benefit. In the extension periods of the original trial, none of the data can be accepted as any primary endpoint. Because this trial was not designed for analytic purposes and the drop-out rate was extremely high (>50% according to the authors), all of the results at the tenth year should be accepted as secondary endpoints. Therefore, the fracture rates are as important as BMD changes and have to be discussed in detail.
The authors emphasized that BMD increased substantially between 6 and 10 years of the extension period at the lumbar vertebrae. Although this increase in both alendronate groups reached statistical significance in comparison with initial values, and the discontinuation group gained very little amount of bone in this period (statistically not significant), the fracture rates were similar in the three groups as pointed out by the authors. These data again indicate that BMD is not a sole predictor of fracture rate.
Furthermore, the estimated vertebral fracture counts from the given total number (n = 228) and fracture rates were 10% and 13.9%, 5% and 6.6%, and 4% and 5%, in the 5 mg, discontinuation, and 10 mg alendronate groups, respectively, all of which were morphometric fractures detected between 6 and 10 years of the extension period of the trial (Table 1). From this point, one can calculate the fracture statistics among the groups. Even proportionally quite different, the fracture rates among the groups were computed by Pearson's χ2 test, and there was no statistically significant difference (p = 0.112), which may be subject to a type 2 error (fail to reject null hypothesis) because of the small sample size. Even if these analyses were supposed to be correct, there would be no need to continue alendronate beyond 5 years, because the antifracture efficacies were the same up to 10 years whether alendronate was used or not. This statement was validated in the same study because the bone turnover marker, urinary type I collagen N-telopeptide, values did not return to pretreatment levels from 5 years after discontinuation. Decrease in height findings were also parallel to this confirmation.
Even not reaching statistical significance, the 5 mg group had surprisingly higher morphometric vertebral fractures than the discontinuation group (2.11 times higher) and the 10 mg group (2.78 times higher) at the 6- to 10-year interval. It is difficult to explain this disproportional result on a biologic basis. Because the Food and Drug Administration approved 5 mg alendronate as the prophylactic dose for osteoporosis, these results need further explanation. If the cumulative dose of drug is taken into account, in the discontinuation group, almost the same amount of alendronate has been accumulated in 2 (20 mg/day) and the next 3 (5 mg/day) years in comparison with the 5 mg/day group.
Detecting differences in the occurrence of infrequent events is problematic in small sample sizes such as those present in this trial. From a fracture point of view, to reach 80% power at the α = 0.05 level and to show at least a 50% fracture difference between groups, the study needs 1551 subjects in each arm for a minimum fracture rate of 5%. Because the number of subjects in each group is smaller than the above-mentioned figure, it can be concluded that this trial was quite underpowered in terms of antifracture efficacy of alendronate at 10 years. On the other hand, although the fracture rates in each group seem to be quite different from each other, the statistical tests were unable to show any difference that is subject to a type 2 error (fail to reject the null hypothesis). It can be theoretically calculated that if 308 patients could have been reached (instead of 228) at this third extension period (theoretically the same fracture rates supposed to stay the same), the statistical significance level (0.05) would have been reached to reject the null hypothesis. There is no need to say that 308 patients is pretty close to 228.
In the ten years of data, there was not a placebo group; therefore, a comparison of these fracture incidences with an age-matched osteoporotic group seems impossible. However, one can estimate the possible fracture rate from the recently existed data revealed from different placebo groups of trials. For instance, the recently published data of Silverman et al.(4) compared the event rates for osteoporotic fractures in postmenopausal women (n = 2565; mean age = 67 years) with osteoporosis. The data document that, in women without prevalent vertebral fractures (n = 1627), the event rates per 1000 patient-years were 45.4 for any fracture, 15.2 for vertebral fracture, 4.7 for clinical vertebral fracture, and 0.9 for hip fracture. In women with prevalent vertebral fractures (n = 938), the event rates per 1000 patient-years were 117.4 for any new fracture, 77.1 for new vertebral fracture, 25.7 for clinical vertebral fracture, and 5.8 for hip fracture. Figure 1 shows the different fracture rates of vertebrae in the first 3 years of the alendronate trial, the 6- to 10-year extension period of the same trial, and the placebo trial of Silverman et al.
Although Bone et al.(1) stated that prevalent fractures in the 5 mg group (30.8%), the 10 mg group (17.5%), and the discontinuation group (27.2%) were not different from each other, the difference was estimated, and a statistically significant difference was found between the 5 mg group (30.8%) and the 10 mg group (17.5%; Pearson χ2, p = 0.045). This may explain the occurrence of fewer fractures in the 10 mg group, because of less prevalent vertebral fractures and not because of alendronate.
In conclusion, complexity of the results in this trial has caused difficulty in its interpretation. Long-term effects of alendronate resulted in inconclusive evidence because of the small sample size, high drop-out rate, heterogeneous distribution of subjects in each arm, and biologically unexplainable results of fracture rates. Nevertheless and in line with my conclusion, it is certain that the study of Bone et al.(1) lacked adequate statistical power to reveal differences in fracture rate between different groups. This important goal was not, and could not, be on their agenda, unfortunately. The original study design and sample size principally dictate what can be done later with the data.
The need for further well-designed trials on long-term antifracture efficacy of anti-osteoporotic drugs is still remaining.