This manuscript was presented in part as an oral presentation at 86th Annual Meeting of The Endocrine Society, New Orleans, LA, USA, June 16-19, 2004.
Drs Chen, Satterwhite, Sipos, Misurski, and Wagman are employees and own stock in Eli Lilly and Company. Dr Licata received funding and serves as a speaker for Eli Lilly and Company. Dr Lewiecki owns stock in General Electric, received funding and serves as a consultant for Aventis, Eli Lilly and Company, GE Lunar, Kyphon, Merck, Novartis, Procter & Gamble, Roche, and Wyeth. He also received funding from Amgen and Pfizer.
The relationship between early changes in biochemical markers of bone turnover and the subsequent BMD response to daily teriparatide therapy in women with postmenopausal osteoporosis was studied. Changes in five biochemical markers, obtained from a subset of women enrolled in the Fracture Prevention Trial, were examined. Early increases in the PICP and the PINP were the best predictors of BMD response to teriparatide in this analysis.
Introduction: Early reductions in biochemical markers of bone turnover with antiresorptive therapy negatively correlate with subsequent increases in BMD. We undertook this analysis to determine if early changes in biochemical markers with teriparatide therapy predict subsequent increases in BMD.
Materials and Methods: In the Fracture Prevention Trial, 1637 postmenopausal women with osteoporosis were randomized to receive daily, self-administered, subcutaneous injections of placebo, teriparatide 20 μg/day, or teriparatide 40 μg/day. Serum concentrations of two bone formation markers (bone-specific alkaline phosphatase [bone ALP] and the carboxy-terminal extension peptide of procollagen type 1 [PICP]) and urinary concentrations of two bone resorption markers (free deoxypyridinoline [DPD] and N-terminal telopeptide [NTX]) were assessed in a trial population subset (n = 520) at baseline and at 1, 3, 6, and 12 months. We also assessed serum concentrations of another bone formation marker, the amino-terminal extension peptide of procollagen type 1 (PINP), in a subset of 771 women at baseline and 3 months. Lumbar spine (LS) BMD was measured by DXA at baseline and 18 months. Femoral neck BMD was measured at baseline and 12 months.
Results and Conclusion: Baseline bone turnover status correlated positively and significantly with BMD response. The highest correlations occurred for the LS BMD response to teriparatide 20 μg/day. Among all studied biochemical markers, increases in PICP at 1 month and PINP at 3 months correlated best with increases in LS BMD at 18 months (0.65 and 0.61, respectively; p < 0.05). The relationships between these two biochemical markers and the LS BMD response were stronger than the corresponding relationships for the femoral neck BMD response. Using receiver operator curve analysis, we determined that the increases in PICP at 1 month and PINP at 3 months were the most sensitive and accurate predictors of the LS BMD response.
High bone turnover and low bone mass in women with postmenopausal osteoporosis reflect an increase in the number of active bone remodeling units, with elevated osteoclast activity within each unit.(1) As a result, products of osteoblast and osteoclast activity measured as biochemical markers in serum and urine may be elevated.(2–6) Therapy with antiresorptive agents reduces bone turnover, resulting in stabilization or an increase in BMD. Early antiresorptive-induced reductions in biochemical markers have been correlated with subsequent increases in BMD, thereby helping to predict BMD response to therapy. (1,7–12)
The recombinant human parathyroid hormone fragment, rhPTH(1-34) (teriparatide), has been available over 1 year for the treatment of men and postmenopausal women with osteoporosis at high risk for fracture. In contrast to antiresorptive agents, teriparatide increases bone remodeling, thereby promoting an increase in the production of bone formation and resorption markers.(13–16) No studies have examined the association between biochemical markers and the BMD response in postmenopausal women who have initiated teriparatide monotherapy for osteoporosis. We undertook this analysis to quantify the relationships between biochemical markers and the BMD responses to teriparatide at the lumbar spine (LS; 18 months) and femoral neck (FN; 12 months) in a subset of participants in the Fracture Prevention Trial.
MATERIALS AND METHODS
Results of the Fracture Prevention Trial have been previously published. (17) Briefly, 1637 ambulatory, postmenopausal women ranging in age from 42 to 86 years were randomized to receive daily, self-administered, subcutaneous injections of placebo (n = 544), teriparatide 20 μg/day (n = 541), or teriparatide 40 μg/day (n = 552) with daily calcium (1000 mg) and vitamin D (400-1200 IU) supplementation. Whereas the study was planned for 36 months, the finding of osteosarcoma in a concurrent rat carcinogenicity study prompted early trial discontinuation.(18) Subsequently, the median duration of exposure to teriparatide was 19 months.(17)
Biochemical markers of bone turnover
Serum concentrations of two bone formation markers (bone-specific alkaline phosphatase [bone ALP] and the carboxy-terminal extension peptide of procollagen type 1 [PICP]) and urinary concentrations of two bone resorption markers (free deoxypyridinoline [DPD] and N-terminal telopeptide [NTX]) were assessed in a trial subset (n = 520). Blood and urine specimens were collected in the morning at baseline and at 1, 3, 6, and 12 months after study initiation, as well as at study discontinuation or at the study close-out visit. In a separate, partially overlapping subset of 771 participants (all five bone markers were evaluated in 445 participants), serum collected at baseline and again at 3 months was analyzed retrospectively to assess the concentration of amino-terminal extension peptide of procollagen type 1 (PINP), a bone formation marker for which a commercial assay was newly available. Specimens were stored at −20°C at the study site for 2-4 weeks and sent to a central laboratory for processing (Covance Laboratories, Indianapolis, IN, USA). Batch testing was performed because specimens were received on a weekly basis. At −20°C, samples of all four biochemical markers were stable during the processing interval (bone ALP = 4 months, PICP = 6 months, NTX = 1 year, DPD = 2 years). For PINP, serum samples were stored at −20°C at Covance Central Laboratories and later shipped to SUPREME SA (Liege, Belgium) for testing. All PINP samples were processed as one batch. The manufacturer did not provide the stability for PINP.
Bone ALP was measured by a two-site immunoradiometric assay (interassay CV, 7.4-7.9%; Hybritech, Beckman-Coulter, Brea, CA, USA), PICP was measured by an equilibrium radioimmunoassay (interassay CV, 5.4-8.5%; DiaSorin, Stillwater, MN, USA), PINP was measured by radioimmunoassay (interassay CV, 3.1-8.2%; Orion Diagnostica, Espoo, Finland), NTX was measured by a competitive-inhibition ELISA (interassay CV, 6.7-14.8%; Ostex, Seattle, WA, USA), and DPD was measured by a competitive enzyme immunoassay (interassay CV, 8.6-10.1%; Pyrilinks-D; Metra Biosystems, Mountain View, CA, USA). NTX and DPD were normalized for creatinine excretion.
Assessment of BMD
BMD was assessed by DXA using Hologic (Bedford, MA, USA), Norland (Ft Atkinson, WI, USA), and Lunar (Madison, WI, USA) equipment. LS BMD was assessed at baseline, 18 months, and the last visit of the study. Femoral neck BMD was assessed at baseline and 12 months. To eliminate differences attributable to densitometer manufacturer, spine and FN BMD values were converted to standardized units (expressed as milligrams per square centimeter). (19) Percent change in BMD from baseline to endpoint(s) was calculated for each patient. Regions of severe scoliosis and vertebral fracture sites were excluded from LS BMD measurements. Serial measurements of a spine phantom at each center were used to adjust for minor changes in densitometer performance and to assess the consistency of measurements between centers.(20) In this analysis, the relationships between early changes in biochemical markers and FN BMD response at 12 months and LS BMD response at 18 months were evaluated. Twelve months was chosen to measure the relationship at the FN because this was the last time-point at which BMD was measured at this site.
Demographics and baseline characteristics were summarized across treatment groups using descriptive statistics. The distribution of the changes in biochemical markers was skewed; therefore, ranked ANOVA was used to compare the changes in biochemical markers from baseline between groups at each time-point. Ranked ANOVA included the terms of treatment and country. Medians are presented as the measure of central tendency. The relationship between baseline biochemical markers and the percent change in BMD were evaluated by Spearman rank correlation analysis. The relationship between changes in biochemical markers, measured as both absolute and percent values, and the percent change in BMD were also evaluated by Spearman rank correlation analysis. Because of the different pattern of change of the biochemical markers, we performed a confirmatory analysis that evaluated the correlation between the area under the curve of the markers at 1, 3, and 6 months and the percent change in BMD. The PINP subset included 445 women who were in the four biochemical marker subset and an additional 326 women who were not part of the four biochemical marker analysis. To assess potential differences between these subsets, we repeated the Spearman rank correlation analysis for women included only in the four biochemical marker subset.
The study closeout visit occurred 5-6 weeks after the last dose of injectable study drug was administered. Marker status at this time could have changed substantially, and the biochemical marker endpoint analysis had the potential to underestimate the effect of teriparatide treatment. Therefore, marker values from this time-point are not included in this analysis.
To determine which biochemical marker characteristics offered the greatest sensitivity and accuracy for predicting LS BMD at 18 months, we conducted a receiver operator characteristic (ROC) analysis. For this analysis, we defined women with a ⩾3% increase in LS BMD at 18 months as responders, women with a decrease in LS BMD >3% as nonresponders, and women with a response in the range of DXA precision error (i.e., greater than or equal to −3% but <3%) as indeterminate responders. ROC curves were constructed by varying the cut-off point (measured as the unit change in bone turnover marker concentration). Corresponding cut-offs for a given specificity were derived based on the logistic regression model.(1) The overall diagnostic accuracy for predicting LS BMD changes was assessed by the areas under the ROC curves (AUC). To test the robustness of this analysis, we applied the same model evaluating the change in LS BMD as a dichotomous variable. A change in BMD >3% was used to define responders, and a change in BMD of <3% was defined as nonresponders. A similar analysis was also performed using a change in LS BMD >0% to define responders, and a change in BMD of ⩽0% to define nonresponders, allowing for inclusion of all women in the trial subset.
To examine the relationship between changes in biochemical markers and fracture risk, we conducted a logistic regression analysis. This analysis included a treatment dummy variable term (pooled teriparatide groups and placebo), a change in biochemical marker continuous variable term, and a combined vertebral and nonvertebral fracture dependent variable (yes/no) term. The diagnosis of vertebral and nonvertebral fracture has been previously described.(17)
The age, years since menopause, height, weight, baseline BMD values at the LS and FN for the entire Fracture Prevention Trial population, the four biochemical marker subset, and the PINP subset are presented in Table 1. There were no material differences between the cohorts. In Table 2, age, years since menopause, height, weight, and baseline BMD values at the LS and FN are presented for the placebo and teriparatide groups of the four biochemical marker cohort. There were no significant differences between the teriparatide and placebo groups (p > 0.05). In the PINP cohort, the baseline PINP concentrations for the placebo, teriparatide 20 μg/day, and teriparatide 40 μg/day groups were 49.6 ± 20.3 (n = 260), 46.9 ± 18.3 (n = 257), and 47.7 ± 22.0 ng/ml (n = 254), respectively. There were no significant differences in baseline characteristics in the PINP cohort (p > 0.05).
Table Table 1.. Baseline Demographic and BMD for the Fracture Prevention Trial Cohort, the Four-Biochemical Marker Cohort, and the PINP Cohort (Mean ± SD)
Table Table 2.. Baseline Demographic BMD and Bone Turnover Marker Data for Each Study Group (Mean ± SD)
Bone turnover marker and BMD response to teriparatide
In response to teriparatide, concentrations of bone formation markers increased promptly (Fig. 1). PICP concentrations peaked 1 month after initiation of therapy. Teriparatide also induced increases in bone ALP concentrations after 1 month, which remained elevated at 12 months. The median percent increases in bone resorption markers were generally smaller than the elevations in bone formation markers at 1 month; however, from 3 months forward, all increases in bone resorption markers were significant compared with placebo (Fig. 1; p < 0.05).
For the subset of women in which four biochemical markers were evaluated, the mean percent increase in LS BMD at 18 months was 11.6 ± 7.4% (SD) in response to teriparatide 20 μg/day, 14.6 ± 8.6% in response to teriparatide 40 μg/day, and 0.63 ± 4.6% with placebo (p < 0.001 for both teriparatide 20 μg/day and teriparatide 40 μg/day versus placebo). The mean percent change in FN BMD at 12 months was +1.3 ± 4.4% in response to teriparatide 20 μg/day, +3.3 ± 5.0% in response to teriparatide 40 μg/day, and −0.3 ± 4.6% with placebo (p < 0.01 for both teriparatide 20 μg/day and teriparatide 40 μg/day versus placebo). Similar BMD response results were found for the PINP subset (data not shown). Of all participants in the four biochemical marker subset, 68% were identified as responders (i.e., LS BMD increase of ⩾3% at 18 months), and 8% were identified as nonresponders (i.e., BMD decrease >3% at 18 month). An indeterminate response (i.e., greater than or equal to −3% but <3%) occurred in 24% of the participants in this subset. Using these criteria, similar BMD response results were found for the PINP subset (data not shown). For reference, in the biochemical marker cohort of the Fracture Prevention Trial, 92% of women treated with teriparatide 20 μg/day and 94% of women treated with teriparatide 40 μg/day had a BMD response ⩾3% at 18 months. In the placebo group, 22.7% were categorized as responders according to these criteria.
Correlations between biochemical markers of bone turnover and BMD response
Bone turnover status at baseline correlated significantly with subsequent BMD responses. The highest correlation coefficient values were found with the LS BMD response in the group that received teriparatide 20 μg/day (Table 3). Specifically, from largest to smallest, the correlation coefficients were 0.41 for PINP, 0.40 for NTX, 0.36 for PICP, 0.28 for bone ALP and 0.23 for DPD. All of these correlation coefficient values were significant (p < 0.05) and positive.
Table Table 3.. Spearman Correlation Coefficients for Baseline Bone Turnover Markers and BMD Response at the LS and the FN at 18 and 12 Months, Respectively
The correlation coefficients between absolute changes in biochemical markers and BMD responses at the FN and LS at 12 and 18 months, respectively, are listed in Table 4. The highest correlation coefficient value between a bone turnover marker and BMD response was observed between the change in PICP at 1 month and the 18-month LS BMD response (0.65; p < 0.05). For teriparatide 20 μg/day, the correlation coefficient value for the LS BMD response and PICP at 1 month was higher than for the change in any other marker at any other time-point (0.39; p < 0.05). Similarly, for teriparatide 40 μg/day, the correlation coefficient value for the LS BMD response and the PICP at 1 month was higher than for the change in any other marker at any other time-point (0.53; p < 0.05). The second highest correlation coefficient value was observed between the change in PINP at 3 months and the LS BMD response (0.62; p < 0.05). The relationships between these two biochemical marker changes and the LS BMD response at 18 months are shown in Fig. 2. The results were similar whether percent change (PICP at 1 month = 0.63, p < 0.0001; PINP at 3 months = 0.58, p < 0.0001) or AUC (PICP at 1 month = 0.65, p < 0.0001; PINP at 3 months = 0.62, p < 0.0001) of the biochemical markers of bone turnover was used in the correlation analysis. The results for the cohort that included only women who had all five markers measured were similar (PICP at 1 month = 0.67, p < 0.0001; PINP at 3 months = 0.65, p < 0.0001).
Table Table 4.. Spearman Correlation Coefficients Between Changes in Bone Turnover Markers and BMD Response at the FN and LS at 12 and 18 Months, Respectively
Correlations between biochemical markers of bone turnover and fracture risk
In the four biochemical marker subset, 26 patients experienced a vertebral fracture, 17 experienced a nonvertebral fracture, and 2 experienced both. In the PINP subset, which was redundant with the four biochemical marker subset, 49 patients experienced a vertebral fracture, 22 experienced a nonvertebral fracture, and 3 experienced both. No significant relationships between the changes in biochemical markers at any time-point and the overall fracture risk were found (p > 0.05).
Prediction of LS BMD response using biochemical markers of bone turnover
Results of the ROC analyses, which defined responders as those with a >3% increase in LS BMD, are shown in Table 5. Given 90% specificity, PICP at 1 month and PINP at 3 months had the highest sensitivities (59% and 69%, respectively) for predicting the 18-month LS BMD response. AUC values, representing diagnostic accuracy, were highest for PICP at 1 month (0.83) and PINP at 3 months (0.81). For PICP at 1 month, the probability of a positive LS BMD response was 0.93, provided that the individual had a change in serum concentration of at least 46.0 ng/ml. For PINP at 3 months, the probability of a positive LS BMD response was 0.88, provided that the individual had a change in serum concentration of at least 17.2 ng/ml. Similar results were obtained if a response was defined as a change in LS BMD >3%, and a nonresponse was defined as a change in BMD of <3%. If a response was defined as any increase in BMD (i.e., BMD increase >0%), the results were also similar (data not shown).
Table Table 5.. Results of Bone Marker Changes in Predicting LS BMD at 18 Months in Women Who had a LS BMD Increase >3% and the Sensitivity, Probability, and Corresponding Cut-off Values of Bone Turnover Markers for 90% Specificity
Biochemical markers can rapidly provide information on the early response to osteoporosis therapy. The increases in biochemical markers seen with teriparatide show an early and profound effect on bone turnover. PICP increased promptly, peaking 1 month after initiation of teriparatide. Concentrations of bone ALP were significantly elevated as early as 1 month and remained elevated across the duration of the treatment period. The increases in bone formation marker concentrations were followed by increases in bone resorption markers, signifying overall activation of the remodeling process. Because teriparatide also induces early and sustained increases in LS BMD in this population,(21) a continuously positive coupling balance consistent with net bone formation predominates.
This analysis confirms a significant and positive relationship between baseline bone turnover marker status and BMD response to therapy. However, the highest correlations were observed between the change in PICP at 1 month and PINP at 3 months and the 18-month LS BMD response. Secondary analyses where changes in biochemical markers were represented as both percentages and AUCs confirmed the results of the primary analysis. These results agree with those of previous analyses using pharmacodynamic modeling, where PICP and PINP were shown to be strong predictors of the absolute increase in LS BMD in women as well as in men. (22,23) PICP at 1 month and PINP at 3 months also emerged as the most sensitive and accurate predictors of the LS BMD response. This was true regardless of how a LS BMD response was defined (i.e., ⩾3% or >0%). The relationships between these two biochemical markers and the LS BMD response to therapy were stronger than the corresponding relationships for the FN BMD response. There is biologic rationale for this finding. The FN is composed of cortical and trabecular bone and has a slower rate of remodeling than the LS, which is relatively rich in trabecular bone. Teriparatide-mediated early increases in biochemical markers may not be associated with as rapid or as large a BMD response in the FN compared with the LS. The modest BMD gains at the FN may also be attributable to a shorter duration of exposure—12 months—compared with the LS, and therefore do not represent the effects of a full course of teriparatide therapy.
To establish the true relationship between biochemical markers and the BMD response, subjects from the placebo and treated groups were analyzed as a single group. Combining these groups reflects the poor adherence to therapy that can be encountered in clinical practice,(1,10,14) yielding practical information for the clinician. Combining these groups, however, affects the correlation coefficients for baseline biochemical markers and early marker changes differently. For the relationships between baseline biochemical markers and subsequent BMD response, values of biochemical markers are similar across all groups, but substantial increases in BMD will only occur in response to therapy. Therefore, combining the placebo and treatment groups results in increased variance and reduced correlation coefficient values. However, when evaluating the relationship between changes in biochemical markers of bone turnover and subsequent BMD response, combining placebo and treated groups expands the number of data points and the data range.(14) Both of these factors favor higher correlation coefficient values.
A few large studies have evaluated correlations between markers and fracture risk. Reductions in serum osteocalcin, bone ALP, and PINP have been shown to be predictive of decreased vertebral fracture risk in postmenopausal women with osteoporosis on raloxifene.(24,25) Moreover, the decrease in vertebral fracture risk at 3 years was better predicted by the 1-year percent decrease in serum osteocalcin than the 1-year percent increase in FN BMD.(26) Alendronate-mediated reductions in C-terminal telopeptide, bone ALP, and PINP have also been associated with a reduction in fracture risk,(27) and a recent study reported that risedronate-mediated decreases in concentrations of urinary C-telopeptide of type I collagen and NTX were significantly associated with a reduction in vertebral fracture risk.(28)
In contrast to antiresorptive agents, teriparatide increases bone remodeling—both bone formation and bone resorption—resulting in a net gain of new bone. Whereas the ability of teriparatide to reduce fracture risk was clearly shown in the Fracture Prevention Trial,(17) no significant relationships between changes in biochemical markers and fracture risk were found in this analysis. The small number of fractures in the biochemical marker cohorts limited the interpretation of this relationship. Therefore, the clinical need to define future fracture risk based on biochemical marker responses to teriparatide persists.
Because bone strength likely reflects an improvement in a number of parameters, including bone microstructure, composition, and BMD, it would seem prudent to consider biochemical markers as one tool in the evaluation of a patient on teriparatide therapy. Results from the sensitivity analysis indicate that a portion of individuals who did not have increases in biochemical markers concentrations that exceeded the cut-off point still had increases in BMD. This points to the limitations of relying exclusively on biochemical markers to identify responders and nonresponders to therapy. Another limitation of this analysis was that serum PINP was assayed at only one time-point (3 months). In addition, comparisons between the PINP correlation coefficients and those of the other markers must be made cautiously, because the subset population was larger for PINP than the subset population for the four other biochemical markers. This could enhance correlation coefficients quantifying the relationship between changes in biochemical markers and the BMD response. In contrast, the increased number of data points may have decreased the correlation coefficient quantifying the relationship between baseline biochemical markers and the BMD response. However, a secondary correlation analysis in women who had all five biochemical markers collected yielded results similar to those of the larger subset.
The authors thank the investigators of the Fracture Prevention Trial and Nadine Baker for technical support. This study was supported by Eli Lilly and Company. Data were analyzed at Lilly Research Laboratories, Eli Lilly and Company.