The Fracture Intervention Trial (FIT) and other clinical trials demonstrated that 3 to 4 years of alendronate (ALN) treatment prevents bone loss and reduces the risk of X-ray-defined and clinical vertebral fracture in women with low bone density (femoral neck T-scores of less than −1.6) or a vertebral fracture.1, 2 ALN also was shown to reduce the risk of nonvertebral fracture (NVF) in women with femoral neck bone mineral density (BMD) T-scores of −2.5 or less who do not have a prevalent vertebral fracture3 and appears to reduce the risk in those with prevalent vertebral fracture.1 However, whether patients should continue treatment beyond 4 years is uncertain. To address this issue, the FIT Long Term Extension (FLEX) Trial compared the effects of continuing and discontinuing ALN after 5 years.4 Ten years of ALN did not significantly reduce the risk of NVFs compared with 5 years of ALN followed by 5 years of placebo. Continuing ALN reduced the risk of clinical vertebral fractures but not the risk of vertebral fractures identified on the basis of radiographs alone. These results did not differ when participants were stratified by femoral neck T-score, vertebral fracture status, or osteoporosis status at FLEX baseline.4, 5 It is not known whether further stratification by both T-score and vertebral fracture status might identify subgroups of women that would benefit from continued ALN use. Additionally, it is possible that a woman's BMD after 5 years of therapy might influence the antifracture efficacy of continuing ALN during the next 5 years. For example, women who have already had substantial gains in BMD may not benefit as much from additional therapy. Clarification of these issues could help clinicians to determine which patients should continue bisphosphonate therapy after 5 years. Therefore, we undertook a post hoc analysis of FLEX results to determine if the antifracture efficacy of continued ALN differed in subgroups defined by femoral neck T-score and vertebral fracture status at FLEX baseline and in subgroups defined by BMD changes during ALN use prior to FLEX.
Subjects and Methods
The design as well as results of the FIT and the FLEX Trial have been reported previously.1, 3, 4, 6, 7 In the FIT, postmenopausal women aged 55 to 81 years with low femoral neck BMD (≤0.68 g/cm2, equivalent to a T-score of −1.6) were eligible to participate. Of 6549 enrolled participants, 2027 women with at least one prevalent vertebral deformity were enrolled in the Vertebral Fracture Arm (VFA), and 4432 women with no existing vertebral deformity were enrolled in the Clinical Fracture Arm (CFA). In each arm, women were randomized to ALN (5 mg/day for 2 years, 10 mg/day thereafter; n = 3236) or placebo (n = 3223). Average follow-up in the FIT VFA was 2.9 years and the in CFA 4.2 years. One year of ALN (10 mg/day) was offered at no cost to all participants at the end of the FIT.
Eligibility in the FLEX Trial was limited to women assigned to ALN during the FIT who completed a total of at least 3 years of treatment during the trial plus the subsequent open-label period. Women whose total-hip BMD at FLEX baseline was less than 0.515 g/cm2 (T-score of less than −3.5)8 or whose total-hip BMD was lower than at the FIT baseline were ineligible. Ten of the 11 original clinical centers in the FIT participated in the FLEX Trial. Of the 3236 women enrolled in the FIT and assigned to ALN treatment at these 10 centers, 2852 surviving women were contacted about the possibility of participating in the FLEX Trial. Of these, a total of 1099 women (39%) were enrolled in the FLEX Trial. All women provided written informed consent, and the protocol was approved by the appropriate institutional review boards.
At FLEX baseline, participants were randomly assigned to receive ALN 10 mg/day (n = 333, 30%), ALN 5 mg/day (n = 329, 30%), or placebo (n = 437, 40%) for 5 years. Each participant also was offered a daily supplement containing 500 mg of calcium and 250 IU of vitamin D3.
At FLEX baseline, BMD was measured at the anteroposterior (AP) lumbar spine, hip (femoral neck, trochanter, and total), and total body using dual-energy X-ray absorptiometry (DXA). Measurements were made with the same Hologic QDR 2000 densitometers used for the FIT (Hologic, Inc., Bedford, MA, USA). Hip BMD was measured annually; spine and total-body BMD were measured at the 36-, 48-, and 60-month visits.
Women were stratified into those who lost and those who gained BMD at the femoral neck during the period from FIT baseline to FLEX baseline. Femoral neck (FN) BMD was measured at FIT baseline using the same procedures as those used during the FLEX Trial.6 Participants with an FN BMD at FLEX baseline that was greater than their FN BMD at FIT baseline were categorized as gaining BMD; others were categorized as losing BMD.
Fracture incidence in ALN and placebo groups was an exploratory objective in the FLEX Trial. Potential clinical fractures were identified by self-report and confirmed by radiology or surgical reports. Fractures of the skull and those owing to cancer or excessive trauma were excluded. If a vertebral fracture was diagnosed by a participant's physician, usually following reported back pain, the spine radiograph used to diagnose the fracture was requested and compared with the radiograph obtained at FLEX baseline. Clinical vertebral fractures were confirmed if the semiquantitative grading of vertebral fractures9 had increased by at least one category from baseline.
Lateral spine radiographs were obtained at FLEX baseline and 36 and 60 months for morphometric vertebral fracture ascertainment. Six points were placed on each vertebra to define anterior, posterior, and middle heights (Synarc, Inc., San Francisco, CA, USA), and radiographs with suspected incident morphometric fractures also were given semiquantitative grades.9, 10 A new fracture was defined as a decrease of 20% or more and at least 4 mm in any vertebral height with a semiquantitative confirmation. Prevalent vertebral fracture at baseline was defined as either randomization to the FIT VFA or presence of deformity on baseline FLEX radiograph as assessed by quantitative morphometry.11, 12
Clinical fractures were analyzed using Cox proportional hazard models to compare the two treatment groups within strata defined by FLEX baseline FN T-scores and vertebral fracture status. Comparison of the proportion of participants with morphometric vertebral fractures was made using the odds ratio from logistic regression. As specified in the protocol, the primary analysis of fractures for the main FLEX Trial results used pooled results for the two ALN groups, and we used the same approach for these subgroup analyses. These analyses were performed using SAS Version 9.1 (SAS Institute Inc., Cary, NC, USA).
For absolute risk reduction, we estimated the between-group risk difference by the difference in fitted risks for the groups under the logistic model. The confidence interval of the risk difference was calculated from the robust covariance matrix provided from the nonlinear combinations of logistic regression coefficients using the delta method, or first-order Taylor series expansion, using Stata Version 9 (Stata Corp., College Station, TX, USA).
To test for multiplicative interaction between ALN continuation and FN T-score, we included an interaction term between treatment assignment and FN T-score as a continuous variable in models with fracture outcomes. To test for additive interaction, the difference in risk differences from nonlinear combinations of logistic regression coefficients, we used Wald-type tests, employing the delta method to calculate p values.
The average age of women enrolled in the FLEX Trial was 73 years, and 61% reported very good or excellent health (Table 1). The average duration of prior ALN use was 5 years. At FLEX baseline, 376 (34%) women had a prevalent vertebral fracture, whereas 723 (66%) did not. Among the women without a prevalent vertebral fracture, 25% had an FN T-score of −2.5 or less and 20% had a lumbar spine (LS) T-score of −2.5 or less. Among women with a prevalent vertebral fracture (n = 376), 37% had an FN T-score of −2.5 or less and 35% had an LS T-score of −2.5 or less. Most women gained BMD at the FN during the FIT, but about 20% experienced bone loss. The treatment groups within the vertebral fracture categories were similar with respect to mean age, body mass index (BMI), race, self-reported health, current smoking, and fracture history.
Table 1. Characteristics of Participants With and Without a Vertebral Fracture at FLEX Baseline
Women assigned to placebo in FLEX Trial after ∼5 years of ALN use in the FIT.
Women assigned to an additional 5 years of ALN in the FLEX Trial after ∼5 years of ALN use in the FIT. Results for the 5 and 10 mg/day groups are combined.
Age, years (mean ± SD)
72.6 ± 6.1
71.9 ± 5.6
75.7 ± 5.1
74.4 ± 5.3
Body mass index, kg/m2 (mean ± SD)
25.5 ± 4.3
25.5 ± 4.0
26.3 ± 4.2
26.4 ± 4.9
General health (self-reported)
Very good or excellent
Fair or poor
Walk for exercise (%)
Fall in last 12 months (%)
History of clinical fracture (>45 years)
Duration alendronate use, years (mean ± SD)
5.1 ± 0.7
5.1 ± 0.7
4.6 ± 0.7
4.7 ± 0.7
Using alendronate at baseline (%)
Using HRT or raloxifene at baseline (%)
Time since start of FIT, years (mean ± SD)
5.8 ± 0.2
5.8 ± 0.2
5.5 ± 0.3
5.5 ± 0.4
Time from end of FIT, years (mean ± SD)
1.5 ± 0.2
1.5 ± 0.2
2.5 ± 0.4
2.4 ± 0.4
FLEX baseline BMD, g/cm2 (mean ± SD)
0.73 ± 0.09
0.74 ± 0.09
0.70 ± 0.09
0.70 ± 0.10
0.62 ± 0.07
0.62 ± 0.06
0.60 ± 0.07
0.60 ± 0.08
0.92 ± 0.14
0.91 ± 0.14
0.87 ± 0.15
0.87 ± 0.15
0.45 ± 0.06
0.46 ± 0.06
0.43 ± 0.05
0.43 ± 0.06
0.98 ± 0.09
0.99 ± 0.09
0.95 ± 0.09
0.95 ± 0.10
FLEX baseline BMD T-score at femoral neck (mean ± SD)
−2.1 ± 0.7
−2.1 ± 0.6
−2.3 ± 0.7
−2.3 ± 0.7
−2.0 < FN T-score
−2.5 < FN T-score ≤ −2.0
FN T-score ≤ −2.5
FLEX baseline BMD T-score at lumbar spine (mean ± SD)
−1.4 ± 1.3
−1.5 ± 1.2
−1.9 ± 1.3
−1.9 ± 1.3
−2.0 < LS T-score
−2.5 < LS T-score ≤ −2.0
LS T-score ≤ −2.5
Change in femoral neck BMD from FIT baseline to FLEX baseline
Women without a vertebral fracture at FLEX baseline
Nonvertebral Fracture Risk
As reported previously, continued ALN treatment did not reduce the risk of nonvertebral fracture (NVF) among all women without a baseline vertebral fracture [relative risk (RR) = 0.86, 95% confidence interval (CI) 0.59–1.3].4 In this analysis, we found a significant interaction between FN T-score as a continuous variable and ALN efficacy (p = .019), which implies that the reduction in NVFs becomes significant at lower BMD levels. When we stratified women based on standard cutpoints for FN T-score, those at the lowest levels of FN BMD (T-scores of −2.5 or less) had a significant reduction in NVFs (RR = 0.50, 95% CI 0.26–0.96) (Table 2). In those with an FN T-score of between −2 and −2.5 (RR = 0.79, 95% CI 0.37–1.66) and in women with an FN T-score of greater than −2 (RR = 1.41, 95% CI 0.75–2.66), NVF risk was not significantly reduced with ALN continuation (Fig. 1).
Table 2. Continuing or Discontinuing ALN Treatment and Risk of Fracture Stratified by Baseline Presence of Vertebral Fracture and Femoral Neck T-Score
Women with FN T-scores of less than −2.5 had a higher rate of NVF than women with higher BMDs. Combined with the greater effect of ALN, women with T-scores of −2.5 or less had a substantially greater reduction in absolute risk of NVF (risk difference −13.32%, 95% CI −25.46 to −1.18) than did women with higher BMDs. In those with an FN T-score of greater than 2.5 and −2 or less and in women with an FN T-score of greater than −2, absolute risk of NVFs was not significantly reduced. The p value for the interaction in absolute risk, comparing the three categories of FN T-score, was .049.
Clinical Vertebral Fracture
As reported previously, continuing ALN treatment reduced the risk of clinical vertebral fracture among all participants (RR = 0.45, 95% CI 0.24–0.85). Efficacy was similar among those without a baseline vertebral fracture (RR = 0.42, 95% CI 0.16–1.1) with no evidence of interaction by baseline vertebral fracture status (p = .86 for interaction).4 In these analyses, there was no evidence of differences across levels of FN T-scores for the effect of ALN on the risk of clinical vertebral fractures, considering relative risk reduction (test for multiplicative interaction, p = .47) or absolute risk reduction (test for additive interaction, p = .47). However, only 18 women without a baseline vertebral fracture experienced a clinical vertebral fracture in the FLEX Trial, and the number in each of the three T-score subgroups was small.
Morphometric Vertebral Fracture
Morphometric vertebral fracture risk did not differ between those who continued and discontinued ALN (RR = 0.86, 95% CI 0.60–1.22).4 Relative risk reduction did not differ significantly across levels of FN T-scores (test for multiplicative interaction, p = .92) (Table 2). Similarly, there were no statistically significant differences in absolute risk reduction (test for additive interaction, p = .89).
Lumbar Spine T-Score
With T-scores based on LS BMD, there were no statistically significant interactions with baseline T-score for NVFs, clinical vertebral fractures, or morphometric vertebral fractures. For NVFs, the results appeared consistent with findings for FN T-scores, although the test for interaction was not statistically significant (p = .227). The relative risk of NVF appeared to be reduced with continued ALN treatment in women with LS T-scores of −2.5 or less (RR = 0.64, 95% CI 0.28–1.49), whereas there was no effect of continuing ALN on NVFs in those with LS T-scores of greater than −2.0 (RR = 1.00, 95% CI 0.62–1.62). Those with LS T-scores between −2.0 and −2.5 also appeared to have a reduced relative risk of NVF (RR = 0.65, 95% CI 0.27–1.61).
Women with a vertebral fracture at FLEX baseline
As reported previously, continued ALN treatment did not reduce the risk of NVF or morphometric vertebral fracture in all FLEX participants but did reduce the risk of clinical vertebral fracture. Among women with a vertebral fracture at baseline, there were no significant interactions between baseline FN BMD and the effect of continued ALN treatment on any of these fracture outcomes. Baseline FN T-score did not identify a subgroup of women with prevalent vertebral fracture with a different risk of NVF or morphometric vertebral fracture with continued ALN use (Table 2). Reduced risk of clinical vertebral fracture with continued ALN treatment did not appear to differ across baseline FN T-scores (data not shown). However, only 21 women with a prevalent vertebral fracture experienced a clinical vertebral fracture during the FLEX Trial. The results were similar for baseline T-scores at the lumbar spine (data not shown).
Change in BMD from FIT baseline to FLEX baseline
When FLEX participants were stratified into those who did and did not lose bone at the FN from FIT baseline to FLEX baseline, the relative risk reduction was similar in both groups for all the fracture outcomes. In models considering change in FN BMD as a continuous variable, we found no evidence that the effect of continued ALN treatment on fracture risk was modified by the extent of previous bone loss while using ALN for any of the fracture outcomes considered (data not shown).
In the original FLEX analysis, we found a significant reduction with continued ALN treatment only in clinical vertebral fractures but not in NVFs or in morphometric vertebral fractures. However, in this analysis, in women without prevalent vertebral fracture, we found that the effect of continuing ALN treatment on NVFs depended on baseline BMD. Risk was reduced among those with lower levels of FN BMD. These results are consistent with findings among women in the FIT without a baseline vertebral fracture. In the FIT, use of ALN reduced clinical fracture risk in those with baseline T-scores of −2.5 or less but not in those with T-scores greater than −2.3 The FLEX results indicate that ALN efficacy for prevention of NVF continues to depend on baseline BMD after 5 years of use. Women who have a T-score below −2.5 and do not have a vertebral fracture after 5 years of ALN treatment may benefit, in terms of NVF risk, from continued ALN use. On the other hand, above a T-score of −2 there was no indication of reduction in risk, suggesting that these patients may discontinue after 5 years without increasing their risk of NVF. Women with an FN T-score between −2 and −2.5, taken as a group, did not have a statistically significant reduction in NVF risk. We did not attempt to identify a T-score threshold for benefit more precisely in this study given the small number of fractures. Our analyses only provide evidence of NVF reduction with continued ALN use for women without a prevalent vertebral fracture who have an FN T-score below −2.5.
Clinical vertebral fractures were reduced among those who continued ALN treatment in the FLEX cohort as a whole (RR = 0.45, 95% CI 0.24–0.85), and there was no evidence of interaction with vertebral fracture prevalence.4 In this analysis, we did not find evidence that this protective effect varied significantly across levels of FN T-scores in women without a vertebral fracture at FLEX baseline. However, the ability to identify differences was limited by the small number of these fractures in the FLEX Trial. Morphometric vertebral fracture was not reduced with continued ALN use in the FLEX cohort.4 and we did not find evidence of any differences across levels of FN T-scores in women without a prevalent vertebral fracture.
Among women with a prevalent vertebral fracture, we did not identify any subgroup with a reduction in NVFs with continued ALN use. Those with FN T-scores of −2.5 or less had an RR for NVF of 1.11 (95% CI 0.61–2.02) for continued ALN use. As reported previously, this group as a whole did not have a reduction in NVF risk during the FLEX Trial (RR = 1.20, 95% CI 0.80–1.8).4 In contrast, after 3 years of ALN use in the FIT, NFV risk appeared to be reduced among women with prevalent vertebral fracture, although the result did not quite achieve statistical significance (RR = 0.80, 95% CI 0.63–1.01).1 A subgroup analysis of women with prevalent vertebral fracture in the FIT found no interaction for the outcome of all clinical fractures for those with baseline FN BMDs above and below the median (0.59 g/cm2).13
For women with a prevalent vertebral fracture, the protective effect of continued ALN treatment for clinical vertebral fractures, previously identified in the FLEX cohort as a whole,4 did not differ significantly across levels of FN T-scores, although the numbers were small. We have reported previously that morphometric vertebral fracture was not reduced with continued ALN use among women with prevalent vertebral fracture,4 and we did not find evidence of differences across levels of FN T-scores. As we noted in our report on the main FLEX Trial results, women with prevalent vertebral fracture have a higher absolute risk of clinical vertebral fracture and therefore may benefit from continued ALN use.4
In models comparing those who did or did not lose bone during an average of 5 years of ALN use (from FIT to FLEX baseline), we found no differences in the risk of fracture for those continuing or discontinuing ALN. This suggests that BMD changes during ALN use, and therefore monitoring of BMD, are not useful in identifying women most likely to benefit from continued ALN use.
Women with low BMDs but without a prevalent vertebral fracture have a higher risk of NVFs than of vertebral fractures. In an analysis of this subset of women in the placebo arm of the FIT, NVFs accounted for over 90% of the total days of limited activity owing to fractures.14 Thus the effect of treatment on the risk of NVFs should be a central consideration in deciding whether to continue treatment in these patients. Our results suggest that continuing treatment for at least 5 more years will benefit those who have not yet achieved a femoral neck BMD T-score of greater than −2.5, whereas those with an FN T-score of greater than −2 do not appear to benefit.
The interaction between baseline BMD and efficacy for prevention of NVFs in women without a prevalent vertebral fracture has been observed in other trials of bisphosphonates,3, 15 but the underlying mechanism is not understood. It is possible that inhibition of resorption is most effective in women with the lowest BMDs because the strength of thinner trabecular struts and cortical bone is more strongly influenced by resorption pits.16 Biomechanical studies are needed to clarify the underlying mechanism of action.
These analyses are limited by the small numbers of fractures in the FLEX Trial, so our ability to detect differences in subgroups is limited. Conclusions also must be tempered by the post hoc nature of these analyses. However, the FLEX Trial provides a unique set of data on the effects of continuing ALN use, an issue that is important to patients and clinicians who often face decisions on long-term treatment.
Among women who have used ALN for 5 years, those who do not have a vertebral fracture but have low BMD may benefit in terms of prevention of NVFs from continuing ALN use for an additional 5 years. However, continuing ALN in women with T-score of greater than −2 does not appear to provide benefit in the prevention of NVFs. This is a new factor for clinicians and patients to consider in making treatment decisions, in addition to the benefit from prevention of clinical vertebral fracture in all participants that was reported from the main FLEX Trial.
Dr. Schwartz receives research support from Merck & Co. and Amgen, and serves as a consultant for Merck & Co. Dr. Bauer receives research support from Merck & Co. and serves as a consultant for Novartis and Amgen. Dr. Cummings receives research support from Eli Lilly & Co. and Amgen and serves as a consultant for Amgen, Eli Lilly & Co. and Novartis. Dr. Cauley receives research support from Novartis, Merck and Co., and Pfizer and serves as a consultant for Novartis. Mrs. Palermo serves as a consultant for Nycomed. Dr. Hochberg serves as a consultant for Amgen, Merck and Co., Novartis, Roche, and Wyeth. Dr. Feldstein receives research support from Merck and Co., Amgen, and Sanofi-Aventis. Dr. Lombardi is an employee of Merck and Co. Dr. Black receives research support from Novartis, Merck and Co. and Roche, and serves as a consultant for Zosano Pharma. All other authors have no conflicts of interest.
Author contributions: AVS had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Members of the Fracture Intervention Trial Long-Term Extension (FLEX) research groups: Clinical centers: Wake Forest University, Winston-Salem/Greensboro, SC: SA Quandt, PhD (coprincipal investigator), CD Furberg, MD, PhD (principal investigator), J Spangler, MD (medical director), S Marion, RN (project director/clinic coordinator), J Stough (clinic coordinator); Kaiser Permanente Center for Health Research, Portland, OR: A Feldstein, MD (principal investigator), E Harris, PhD (coinvestigator), M Rix, RN (clinic coordinator); Stanford Medical Center, Palo Alto, CA: WL Haskell, PhD (coprincipal investigator), A Laws, MD (principal investigator), J Fair, NP, MS, PhD (project director/clinic coordinator); University of California, San Diego, CA: E Barrett-Connor, MD (principal investigator), D Schneider, MD (coinvestigator), ML Carrion-Petersen, RN (project director), N Kamantique, RN (clinic coordinator), K Kadlec, RN (clinic coordinator); University of Iowa, Iowa City/Davenport, IA: J Tomer, PhD (coprincipal investigator), RB Wallace, MD (principal investigator), D Staub, MD (subinvestigator), D Meyerholtz (project director/clinic coordinator), K Canady (project director/clinic coordinator); University of Maryland, Baltimore, MD: MC Hochberg, MD (principal investigator), R Flores, MD (coinvestigator), K Roney (project director); University of Miami, Miami, FL: S Levis, MD (principal investigator), A Herrin (clinic coordinator); University of Minnesota, Minneapolis, MN: K Ensrud, MD (principal investigator), S Diem, MD (coinvestigator), C Quinton (clinic coordinator); University of Pittsburgh, Pittsburgh, PA: J Cauley, DrPH (principal investigator), R McDonald, MD (principal investigator), L Harper (project director), M Nasim (clinic coordinator), L Prebehalla (clinic coordinator); University of Tennessee, Memphis, TN: S Satterfield, MD, DrPH (principal investigator), K Johnson, MD, MPH (coinvestigator), LD Burch (project director). Bone biopsy: Creighton University Osteoporosis Research Center: R Recker, MD. Coordinating center and radiology group: University of California, San Francisco, CA: D Black, PhD (principal investigator), SR Cummings, MD (coprincipal investigator, SC chair), MC Nevitt, PhD (coinvestigator, director morphometry and BMD QC), D Bauer, MD (coinvestigator, endpoint coordinator), A Schwartz, PhD (coinvestigator, project director), L Palermo (senior statistical programmer), C Fox (senior programmer), R Scott, (radiology group), C Yeung (radiology group/X-rays), L Nusgarten (fracture coordinator), L Denton (project assistant). Sponsor: Merck Research Laboratories: A Lombardi, MD (medical monitor), A Rybak-Feiglin (medical program coordinator), D Cohn (associate medical program coordinator), S Holk (associate medical program coordinator), C Dave (assistant medical program coordinator), J DiBona (assistant medical program coordinator), S Suryawanshi, PhD (statistician). Data and safety monitoring board: C Rosen, MD (chair), D DeMets, PhD, A Santora, MD, S Suryawanshi, PhD
The study was supported by contracts with Merck & Co. and was designed jointly by the non-Merck investigators and Merck employees. The study drug is manufactured and packaged by Merck.