Vertebral Fracture Risk Is Reduced in Women Who Lose Femoral Neck BMD With Teriparatide Treatment


  • Drs. Watts, Miller, Kohlmeier, and Sebba are consultants and speakers for Eli Lilly and Company. Drs. Chen, Wong, and Krohn are employees of Eli Lilly and Company.

  • Published online on January 29, 2008


Response to osteoporosis therapy is often assessed by serial BMD testing. Patients who lose BMD without secondary causes of bone loss may be considered to be “nonresponders” to treatment. We examined vertebral fracture (VF) risk, change in lumbar spine (LS) BMD, and change in amino-terminal extension peptide of procollagen type I (PINP) in postmenopausal women whose femoral neck (FN) BMD decreased, increased, or was unchanged after receiving teriparatide (TPTD) or placebo (PL) in the Fracture Prevention Trial. FN and LS BMD were measured at baseline and 12 mo. VFs were assessed by lateral spine radiographs at baseline and study endpoint. A BMD change from baseline of >4% was considered to be clinically significant. Decreases of >4% FN BMD were less common in women receiving TPTD (10%) versus PL (16%, p < 0.05), yet women on TPTD who lost FN BMD still had significant reductions in VF risk compared with PL (RR = 0.11; 95% CI = 0.03–0.45). VF risk reduction with TPTD compared with PL was similar across categories of FN BMD change from baseline at 12 mo (loss >4%, loss 0–4%, gain 0–4%, or gain >4%; interaction p = 0.40). Irrespective of FN BMD loss or gain, TPTD-treated women had statistically significant increases in LS BMD and PINP compared with PL. In both groups, losses or gains in FN BMD at 12 mo corresponded to losses or gains in BMC rather than changes in bone area. In conclusion, loss of FN BMD at 12 mo in postmenopausal women with osteoporosis treated with TPTD is nevertheless consistent with a good treatment response in terms of VF risk reduction.


Several therapeutic options are currently available for the treatment of osteoporosis. In clinical trials, these drugs have been shown to reduce the risk of fracture and increase BMD, which is considered by some to be a surrogate marker for fracture risk reduction. In clinical practice, individual patient response to osteoporosis therapies is often monitored by serially measuring BMD.(1) Whereas many patients experience an increase in BMD with therapy, some do not show a significant change and some may actually have a decrease in BMD, particularly when BMD is measured early in the course of therapy. Decrease in BMD may be a cause for clinical concern. Several recent reviews have discussed the considerations that may contribute to a poor BMD response or “nonresponse” while on osteoporosis therapy and the potential relevance of this finding.(1–4) For a change in BMD from the baseline value to be considered clinically significant,(1,2,5) it must exceed the calculated least significant change (LSC), which reflects operator variability and precision of BMD measurements for that facility. In contrast to the established relationship between low BMD at baseline and increased fracture risk in untreated patients, the relationship between change in BMD and fracture risk in patients taking osteoporosis therapy is less well characterized.

Evidence on the fracture efficacy of osteoporosis therapies in patients with a significant loss of BMD (greater than the LSC) is sparse. As recently summarized,(2) several posthoc analyses showed that patients treated with alendronate or risedronate who gained BMD had significantly decreased vertebral fracture risks compared with patients on the same treatment who lost BMD. In a posthoc analysis of the Fracture Intervention Trial, Chapurlat et al.(6) showed that patients treated with alendronate who had a BMD decrease of 0–4% at the total hip still had significant reductions in vertebral fracture risk compared with placebo (PL) patients with a similar BMD loss. However, differences of <4% in serial BMD measurements may not be significant. In the analysis of Chapurlat et al.,(6) women who lost >4% femoral neck BMD had a numerically lower fracture incidence compared with PL, but the difference was not statistically significant, possibly because of low numbers of events. It is not known whether patients treated with other osteoporosis therapies who have a significant loss of BMD also have a decreased vertebral fracture risk compared with PL.

The anabolic osteoporosis therapy teriparatide (TPTD) [human recombinant PTH(1-34)] 20 μg/d was shown to decrease the risk of new vertebral fractures by 65% and to increase lumbar spine (LS) BMD by 8.6% compared with PL after a median of 19 mo in the Fracture Prevention Trial.(7) At 1 yr of treatment with TPTD 20 μg/d in this trial, 87% and 38% of women were considered to be “BMD responders” at the LS and total hip, respectively.(8) Similar results were seen in women treated with the TPTD 40-μg/d dose.

The objective of these analyses was to determine the vertebral fracture risks in women with either no significant gain or even a loss of BMD at the femoral neck (FN) at 1 yr when treated with TPTD or PL in the Fracture Prevention Trial. Additional analyses were done to examine the corresponding changes in LS BMD and the bone formation marker amino-terminal extension peptide of procollagen type I (PINP) in groups with different degrees of change in FN BMD at 1 yr.


Study population

The randomized, double-blind, placebo-controlled Fracture Prevention Trial evaluated the efficacy and safety of TPTD in 1637 postmenopausal women with prior vertebral fractures.(7) The inclusion and exclusion criteria have been previously described in detail.(7) Women were randomly assigned to receive PL (N = 544), TPTD 20 μg/d (N = 541), or TPTD 40 μg/d (N = 552) and self-administered study drug subcutaneously. All enrolled women received supplements of 1000 mg/d calcium and 400–1200 IU/d vitamin D. The median duration of observation was 21 mo, with a median duration of exposure to TPTD of 19 mo. The trial was conducted in accordance with good clinical practices and the Declaration of Helsinki. The institutional review board at each site approved the protocol before study initiation. All enrolled patients provided written informed consent.

Assessment of fractures, BMD, and bone turnover markers

Lateral thoracic and LS radiographs were taken at baseline and at study endpoint (median, 19 mo).(7) Radiographs were assessed at a central location (Osteoporosis and Arthritis Research Group, San Francisco, CA, USA) by expert radiologists blinded to treatment assignment but not to the temporal sequence of the radiographs. Radiographic assessment was performed using the semiquantitative method of Genant et al.(9) A new morphometric vertebral fracture was reported if height loss of greater than ∼20% was seen in a vertebra that was not fractured at baseline. Clinical vertebral fractures were not assessed in the Fracture Prevention Trial.(7)

BMD at the LS and FN was measured at baseline and 12 mo by DXA using Hologic (Bedford, MA, USA), Norland (Fort Atkinson, WI, USA), or Lunar (Madison, WI, USA) equipment and analyzed at a central center.(7) All vertebral deformities identified by DXA were removed from the LS BMD calculation.

Serum concentrations of PINP were analyzed retrospectively from samples obtained from a subset of 589 women at baseline and 3 mo.(10) PINP was chosen because it gave the highest signal-to-noise ratio of all bone turnover markers assessed in the Fracture Prevention Trial.(11) PINP was measured by radioimmunoassay (interassay CV, 3.1–8.2%; Orion Diagnostica, Espoo, Finland).

Statistical analyses

This posthoc analysis was performed on a subset of 1216 women who had femoral neck BMD measurements at baseline and 12 mo, as well as paired baseline and postbaseline spine radiographs. FN BMD was used for categorization to have adequate statistical power to determine the primary outcome of vertebral fracture risk reduction. Because only a subset of women at some study sites had a total hip BMD measurement, the total hip site could not be used in this analysis.

Because vertebral fracture risk reduction was similar between the two doses of TPTD in the overall trial population,(7) these groups were pooled in these analyses. The distribution of change in FN BMD in the PL and TPTD groups was analyzed first. Each group (PL and TPTD) was divided into four categories based on change in BMD at the FN, using categories as defined by Chapurlat et al.(6): BMD loss >4%, BMD loss between 0% and 4%, BMD gain between 0% and 4%, and BMD gain >4%. A BMD loss or gain of >4% corresponds to a change that is beyond a clinically meaningful LSC for DXA measurements at most testing centers. In each of the four categories of percent change in FN BMD, the risk of new vertebral fractures was compared between the PL and TPTD treatment groups using the χ2 test. Because of the limited number of nonvertebral fractures within each category of BMD change, the relative risk of nonvertebral fractures was not determined.

Logistic regression models were used to compare the relative effect of TPTD on the risk of new vertebral fractures across the categories of change in FN BMD. Effects in the logistic regression model included treatment, subgroup defined by FN BMD change, and treatment-by-subgroup interaction. A significant treatment-by-subgroup interaction (p < 0.10) would indicate that the relative treatment effects varied across the subgroups of FN BMD change. A nonsignificant treatment-by-subgroup interaction would indicate that no difference was detected in the relative treatment effects across the categories of FN BMD change. For continuous variables including change in LS BMD and PINP from baseline, subgroup analyses of FN BMD change categories were performed using ANOVA. Effects in the model included treatment, subgroup, and treatment-by-subgroup interaction. Within each subgroup, treatment difference was tested using ANOVA with treatment in the model. Ranked ANOVA was used to compare the actual change in PINP from baseline to 3 mo between groups (because the distribution of PINP changes was skewed). Pearson correlation was performed to assess the relationships between FN BMD change, BMC, and bone area to determine the relative contributions of BMC and bone area to FN BMD change.


There were no statistically significant differences in any baseline characteristic between the categories of change in FN BMD within the PL or TPTD treatment groups (Table 1). A greater proportion of women in the TPTD group (35%) compared with the PL group (17%) experienced a gain of >4% in FN BMD (p < 0.05), whereas 10% of women in the TPTD group had a >4% loss in FN BMD compared with 16% in the PL group (p < 0.05). The distribution curves (data not shown) for percent change in FN BMD were different, with mean changes of 0.00 ± 4.52 and 2.29 ± 5.25 (SD) for the PL and TPTD groups, respectively.

Table Table 1.. Baseline Characteristics of Women Treated With Placebo or Teriparatide by Category of Percent Change in FN BMD at 1 yr
original image

Vertebral fracture results by category of change in FN BMD are shown in Fig. 1. Women treated with TPTD had a significantly reduced risk of new vertebral fractures compared with the PL group irrespective of the change in FN BMD at 1 yr (treatment-by-subgroup interaction, p = 0.40). Consistent with published literature, PL-treated women who lost >4% in FN BMD had a significantly greater incidence of new vertebral fractures (p < 0.05) compared with PL-treated women with either no significant change or a gain in FN BMD. In TPTD-treated women with a similar significant FN BMD loss of >4%, the relative risk reduction for new vertebral fractures was 89% compared with PL, with an absolute risk reduction (ARR) of 20.5%.

Figure Figure 1.

The incidence of new vertebral fractures in women treated with PL or TPTD by category of percent change in FN BMD at 1 yr. Women with paired baseline and postbaseline spinal radiographs who had FN BMD measurements at baseline and 1 yr are included in this analysis. The risk of new vertebral fractures was significantly reduced with TPTD compared with PL, irrespective of the percent change in femoral neck BMD.

The TPTD treatment group had significantly greater increases in LS BMD compared with the PL group across all categories of change in FN BMD (Fig. 2). In women with a >4% loss in FN BMD, LS BMD increased by 7.5% from baseline to 1 yr in the TPTD group, significantly greater than the 0.14% increase in the PL group. Change in FN BMD with TPTD therapy was significantly but weakly correlated with change in LS BMD (r = 0.28, p < 0.01), whereas no correlation was seen in the PL group. The proportions of women with an increase in BMD above the least significant change of ≥3% at the LS were 67%, 74%, 84%, and 87% in those with a loss in FN BMD >4%, a loss between 0% and 4%, a gain between 0% and 4%, and a gain >4%, respectively.

Figure Figure 2.

Changes in LS BMD in women treated with PL or TPTD by category of percent change in FN BMD at 1 yr. All TPTD treatment groups had significant increases in LS BMD compared with PL and from baseline, irrespective of the percent change in FN BMD.

TPTD-treated women had statistically significantly higher PINP concentrations (Fig. 3), consistent with bone formation, compared with PL-treated women across all categories of change in FN BMD (treatment-by-subgroup interaction, p > 0.10). PINP concentrations at 3 mo were significantly higher by 36.7 ng/ml in women treated with TPTD who lost >4% BMD at the FN. Therefore, with TPTD treatment, even women who experienced a decrease in FN BMD had significant increases in LS BMD and higher PINP compared with baseline and with PL.

Figure Figure 3.

Median change in the bone formation marker PINP at 3 mo in women treated with PL or TPTD by category of percent change in FN BMD at 1 yr. Compared with PL and baseline, all groups treated with TPTD had significant increases in PINP, irrespective of loss or gain in FN BMD. *p < 0.05 between groups.

Within each category of loss or gain in FN BMD, the corresponding change in BMC and the calculated area of the region of interest as estimated by DXA were determined separately (Fig. 4A). Irrespective of PL or TPTD treatment, change in FN BMD was significantly correlated with change in FN BMC, which was ∼10-fold greater than the change in area. The minimal change in area did not significantly contribute to the change in FN BMD in both treatment groups (Fig. 4B).

Figure Figure 4.

Changes in BMC and area at the FN in women treated with PL or TPTD by category of percent change in FN BMC at 1 yr. (A) Change in FN BMD was apparently caused by change in BMC, because the area was unchanged. (B) In both the PL and TPTD groups, individual FN BMD values were significantly correlated with BMC but not with area.

The relationships between change in trochanter BMD and vertebral fracture risk, spine BMD, and PINP were similar to that described for FN BMD (data not shown). When a loss or gain of 3% in FN BMD was used for categorization, the results for vertebral fracture risk and changes in spine BMD and PINP were similar to those reported here using a 4% loss or gain in FN BMD (data not shown).


BMD is sometimes measured at 1 yr in clinical practice to try to assess the efficacy of TPTD therapy, which has a treatment duration of up to 2 yr. These analyses examined the association between FN BMD changes at 1 yr and prediction of new vertebral fracture risk at 19 mo. Irrespective of the observed loss or gain in FN BMD, postmenopausal women with osteoporosis treated with TPTD in the Fracture Prevention Trial nonetheless experienced similar reductions in vertebral fracture risk and consistent increases in LS BMD and PINP compared with PL. Clinicians will often initiate a workup for secondary causes of bone loss in TPTD-treated patients who lose BMD and are often labeled as “nonresponders.” With TPTD treatment, the incidence of new vertebral fractures was found to be similar in women who lost >4%, gained >4%, or had no significant change in BMD at the FN (Fig. 1). Despite the reassuring vertebral fracture risk reduction shown here, it is still important to study potential secondary causes of bone loss in TPTD-treated patients who have a significant decline in BMD.

Because the proximal femur consists of both cortical and trabecular bone, several possible mechanisms could explain the FN BMD decrease seen in some women with TPTD treatment. In ovariectomized monkeys treated with TPTD, increased cortical porosity, width, and area was observed at the FN along with improved biomechanical strength of the proximal femur.(12) Other preclinical studies showed that TPTD treatment increased cortical thickness and area with no change in bone strength.(13,14) This potential for increased cortical porosity may partly explain the decreases in FN BMD observed in some women treated with TPTD. Given the limitations of DXA in detecting small changes in the FN area, the observed nonsignificant change does not exclude the possibility of some change in FN geometry with TPTD therapy.

Results in the cortical compartment of TPTD-treated patients varied between clinical trials, depending on the skeletal site and bone assessment technique used.(15–17) TPTD-treated women in the Fracture Prevention Trial had an increased cortical thickness and no change in cortical porosity in endpoint biopsies compared with baseline.(15) Other clinical studies showed increased cortical bone area and periosteal circumference but no difference in cortical thickness in women treated with TPTD compared with PL.(16) FN volumetric BMD was increased in the trabecular compartment yet slightly decreased in the cortical compartment.(17) Increased diameter of a tubular structure such as the FN may theoretically give rise to an apparent decrease in BMD when analyzed using DXA. However, the current DXA analysis found no change in bone area in the FN with TPTD treatment, although this may partly be caused by limitations of DXA technology. It is possible that increased bone formation seen as increased PINP produces newly formed, undermineralized bone(12,15) that may result in an apparent decrease in BMD.(18) Taken together, the observed decreases in FN BMD with TPTD therapy in some women may be explained by a combination of changes in cortical structure or porosity or by hypomineralization. However, TPTD treatment primarily increases the amount of trabecular bone, which accounts for the increased LS BMD and results in decreased fracture risk in the Fracture Prevention Trial,(8) irrespective of the apparent gains or losses in FN BMD as measured using DXA.

As recently reviewed,(18) there is considerable controversy regarding the relationships between change in BMD and fracture risk reduction, which may differ between antiresorptive and anabolic osteoporosis therapies. Some analyses suggested that larger BMD increases with antiresorptive osteoporosis therapies may result in greater fracture risk reduction,(19–21) whereas other analyses found that only 4–18% of the fracture risk reduction could be explained by changes in BMD.(21–23) In contrast, increases in LS BMD with TPTD therapy account for 30–41% of the vertebral fracture risk reduction after accounting for baseline BMD values.(24)

The strengths of this analysis include a randomized, double-blind clinical trial design, with large numbers of patients enrolled with specific inclusion/exclusion criteria, who were highly compliant with TPTD therapy, with assessment of BMD and vertebral fractures at specified time points according to stringent criteria.(7) However, the clinical trial setting also leads to a major limitation of this posthoc analysis, because patients in clinical practice may have other medical conditions, use other concomitant medications, and have worse compliance with calcium, vitamin D, and osteoporosis therapy compared with clinical trial subjects. The consistency of DXA measurements in this trial was assessed using a spine phantom among the clinical study centers, many of which specialize in osteoporosis. BMD changes in subsequent time points of this trial may be underestimated because these analyses examined BMD changes at 1 yr. This analysis only provides evidence on vertebral fracture risk for changes in FN BMD, because total hip BMD data were not available for the entire study cohort.

In clinical practice, BMD measurements are usually taken about every 1–2 yr to assess the “response” to a specific osteoporosis treatment.(25) An apparent BMD “loss” within the least significant change, which is dependent on the precision error, is not considered to be clinically significant.(1,3,26) Clinicians may be concerned when a TPTD-treated patient has no change or a decrease in FN BMD on follow-up; however, such patients may have increased LS BMD and should not be considered as “nonresponders.” Before making any treatment decisions on such patients, suggested approaches include evaluating measurement precision error for each DXA facility(27) and ensuring that the patient is compliant with therapy, has adequate calcium and vitamin D intake, and does not have any secondary causes of osteoporosis.(1,3,26)

Based on this analysis, for a TPTD-treated patient with a significant decline in FN BMD, one would expect to see increases in LS BMD (if their spine measurement is valid) and/or increases in PINP, which would provide reassuring evidence to confirm that the patient is likely responding to TPTD. Previous analyses showed that changes in PINP at 3 mo were significantly correlated with changes in FN BMD at 12 mo and LS BMD at 18 mo.(10) Furthermore, between 77% and 97% of TPTD-treated patients had an increase in PINP of >10 μg/ml after 3 mo of therapy.(11) Whereas osteoporosis clinical trials often use bone turnover markers to monitor subjects, challenges such as availability and reimbursement limit the use of markers to monitor patients in clinical practice.

This analysis suggests that DXA technology has limitations for follow-up monitoring of the FN response in patients using TPTD. Until better tools such hip structural assessment(28) or QCT finite element analysis(29) to estimate bone strength become clinically available, BMD measurement with DXA remains the current “gold standard” to assess the beneficial changes with TPTD therapy.

In conclusion, this analysis showed that postmenopausal women with osteoporosis will nonetheless experience a significantly decreased risk of new vertebral fractures, a gain in LS BMD, and a significant increase in PINP, compared with PL patients, regardless of the gain or loss in FL BMD. Based on these results, clinical decisions for TPTD-treated patients should not be based on 1-yr changes in FN BMD alone, and other endpoints could be measured, such as LS BMD by DXA or markers of osteoblast activity such as PINP.


Eli Lilly and Company sponsored the Fracture Prevention Trial.