The Effect on Bone Mineral Density and Bone Turnover Markers

Cessation of osteoporosis treatment has been shown to have a range of effects on BMD as well as BTM.

Bisphosphonates

Clinical trials demonstrate significant differences in the bisphosphonates regarding BMD and BTM following discontinuation of therapy. The effect of 1 year (1 dose) of zolendronate appears to persist for at least 3 years in women with osteopenia. In patients with osteoporosis, alendronate for at least 5 years results in a greater persistence of effect than 2 to 3 years of therapy. Discontinuation of risedronate following 3 years of treatment results in a more rapid loss of bone density and BTM than 2 years of alendronate or 3 years of zolendronate.

Alendronate

When individuals were followed off of drug therapy following treatment with alendronate for 2 years or less, there were fewer robust effects on suppression of bone turnover and maintenance of BMD when compared with those for whom treatment with alendronate had been continued for 3 years or longer before discontinuation.

Two cohorts of patients treated with alendronate for up to 10 years have been reported. Both include an arm that saw placebo for 5 years following as many as 5 years of alendronate. In the Alendronate Phase III Osteoporosis Treatment Study Group, postmenopausal women with osteoporosis were randomized to alendronate 20 mg/d for 2 years followed by 5 mg/d for 1 year, 5 mg daily for 3 years, 10 mg daily for 3 years, or placebo. Two 2-year extensions and then a 3-year extension were performed. The 5-mg and 10-mg daily arms continued on the same respective daily alendronate dose, whereas the placebo arm was discontinued after the first 2-year extension. The 20 mg/d × 2-years, 5 mg/d × 1-year original arm received 5 mg/d in the first 2-year extension (alendronate × 5 years) and then saw placebo for 5 years. Tonino and colleagues reported the results of the second 2-year extension at year 7. For the alendronate × 5-years placebo × 2-years arm, no significant decline in BMD at the lumbar spine or hip was seen, but significant declines in the forearm were noted. Urinary N-telopeptide increased during the first year off therapy but then remained stable well below baseline (year 5 level −73%, year 7 level −57.9%). The 10-year results of this study were reported by Bone and colleagues, in which this same group had now been followed off therapy for 5 years. BMD was maintained in the spine. Declines in BMD were noted in the total hip after year 7 (off therapy for 2 years) but at year 10 were comparable to those on alendronate 5 mg daily for the entire 10 years. Urine N-telopeptide remained suppressed at more than 50% of baseline levels.

The second alendronate cohort comes from FLEX, the Fracture Intervention Trial (FIT) Long Term Extension. In FIT, all patients received alendronate 5 mg/d for 2 years followed by 10 mg/d. Average follow-up was 2.9 years in the vertebral fracture arm and 4.2 years in the clinical fracture arm. Patients were offered up to 1 year of alendronate 10 mg/d on completion of FIT. FLEX enrolled 1099 patients who were then rerandomized to placebo (n = 437), alendronate 5 mg/d (n = 329), or alendronate 10 mg/d (n = 333), and followed for an 5 additional years. Patients in the placebo arm saw a 1.52% increase in lumbar spine BMD, with declines of 1.48% in the femoral neck, 3.38% in the total hip, and 3.21% in the forearm. The decline in BMD of the total hip exceeded the baseline value at the start of FLEX by 0.16%. BTM gradually increased over 5 years compared with those who remained on alendronate. When compared with pretreatment levels in FIT 10 years earlier, serum C-terminal telopeptide of type 1 collagen was −7% and N-propeptide of type 1 collagen was −24% in those who had been randomized to the placebo arm.

Risedronate

In the VERT-NA (Vertebral Efficacy with Risedronate Therapy—North America) trial, 599 of the 818 who completed the study were enrolled in a 1-year extension in which risedronate and placebo patients stopped therapy. Lumbar spine BMD decreased 0.83% and femoral neck BMD dropped 1.23% in those who had stopped risedronate, although these values were still greater than the control group who had been on placebo in the VERT-NA trial. Urine N-telopeptide increased significantly from a median of 30.3 nmol bone collagen equivalents (BCE)/mmol creatinine at end of treatment to 50.9 nmol BCE/mmol creatinine after 1 year off risedronate. This increase in the BTM represented a complete resolution of effect when compared with the control group.

Ibandronate

There are no published data on the cessation of ibandronate and subsequent effects on BMD and BTM.

Zolendronate

Zolendronate, 5 mg is approved for every 24-month dosing in patients with osteopenia. In a study of 50 postmenopausal women with osteopenia, the effect of a single intravenous dose of 5 mg zolendronate persisted for the duration of a 3-year trial when compared with placebo control. BMD was higher in the zolendronate group by 6.8% at the lumbar spine and 4.0% at the total hip. Mean levels of serum C-telopeptide were 44% lower than the placebo group.

The HORIZON PFT (Health Outcomes and Reduced Incidence with Zolendronic Acid Once Yearly Pivotal Fracture Trial) was a 3-year double-blind, placebo-controlled trial that enrolled 7736 postmenopausal women. A 3-year extension of HORIZON PFT randomized 1233 women who had received intravenous zolendronate 5 mg yearly in the core trial to receive either 3 additional years of zolendronate (Z6) or 3 years of placebo (Z3P3). The Z3P3 group saw declines of 2.03% in the lumbar spine, 1.2% in the total hip, and 1.04% in the femoral neck, values that were significant when compared with the Z6 group but still well above pretreatment levels. Markers of bone turnover “rose slightly” in the Z3P3 group but procollagen type 1 N-terminal propeptide remained “about 47% below” pretreatment values.

Hormones

Estrogens have a waning effect on bone in postmenopausal women whether after short-term or long-term use. Short-term hormone therapy has been reported not to protect against bone loss after 1 or 2 years of discontinuation. Long-term hormone therapy also failed to protect against bone loss. Evidence from the Framingham Study indicated that even after 7 years of estrogen therapy, there was little residual effect on bone density 10 to 20 years after estrogen withdrawal.

The rate of bone loss from studies after estrogen withdrawal is still controversial. Some studies have reported identical bone loss rates compared with placebo, whereas others have reported accelerated bone loss versus placebo. Some of the controversy is due to differences in study population (healthy or osteoporotic), duration of treatment and discontinuation periods, method of measuring bone mass changes, different treatment regimens, age of study population, and menopausal status.

Raloxifene

The benefit of raloxifene for bone appears to be short lived after cessation of treatment. Naylor and colleagues showed that the bone turnover response was lost 6 months after treatment cessation despite nearly 2 years of therapy. In addition, bone loss that was greater than in the control group, suggesting accelerated bone loss. Hip BMD was 2% less than baseline 192 weeks after treatment. Neele and colleagues showed that 5 years of treatment with raloxifene did not protect against bone loss 1 year after withdrawal of therapy, and that the rate of bone loss was not significantly different from that of placebo-treated women.

Denosumab

Denosumab treatment has reversible characteristics that have been demonstrated in several clinical trials. In a study of osteopenic postmenopausal women, after treatment with denosumab (60 mg every 6 months × 4), cessation of therapy resulted in a return to baseline in serum C-telopeptide within 9 months followed by a compensatory overshoot at 12 months and return to baseline by 30 months after stopping denosumab. BMD returned to baseline within 18 months of stopping the drug but remained above the BMD of the placebo treatment cohort.

Teriparatide

BMD data were collected over an 18-month period after the conclusion of the Teriparatide Fracture Prevention trial. The BMD at the total hip and lumbar spine was maintained during this time, although it was somewhat confounded by the use of some osteoporosis medications such as bisphosphonates in 47% of those followed. The use of bisphosphonates for 12 months or more was associated with a greater maintenance of BMD gains than for those who did not use osteoporosis drugs. Similar results were seen in the lumbar spine BMD of men 30 months after stopping teriparatide.

Bisphosphonates

Studies suggest that a minimum of 2 to 3 years of bisphosphonate therapy with good compliance is needed to demonstrate fracture benefit. Furthermore, treatment with bisphosphonates for 3 to 5 years and subsequent discontinuation for 3 to 5 years is not associated with an increased risk of hip fracture (alendronate; zolendronate), but when compared to those who continued on bisphosphonate, is associated with increased risk of vertebral fractures (alendronate; zolendronate), and an increased risk of nonvertebral fractures in women without prevalent vertebral fractures and BMD T-scores of −2.5 or less (alendronate).

Information regarding the effects of discontinuation of bisphosphonates on fracture risk is limited, and drawing conclusions based on this is problematic. There are two sources of information that can be accessed: administrative databases and clinical trial extensions.

Gallagher and colleagues studied 36,164 women in the General Practice Research Database (GPRD) in the United Kingdom who received a new prescription for alendronate (74.1%) or risedronate (25.9%). There was no evidence of residual effect on fracture risk after stopping bisphosphonate when the mean follow-up was 2.39 years, and few used bisphosphonates for 5 years of longer. Patients who recently stopped bisphosphonates had a similar fracture risk to those who had discontinued further in the past and to patients who had just started therapy.

New users of alendronate (77%) and risedronate (23%) were also the subject of a study by Curtis and colleagues. In addition, they evaluated for the effect of compliance as measured by the medication possession ratio (MPR). A cohort of 9063 women who were members of “a large US healthcare organization” and compliant with drug therapy for at least 2 years were examined. Women who discontinued drug therapy had lower hip fracture rates than nonadherent women, if they had previously taken bisphosphonates for 2 or more years and had been compliant with drug therapy (MPR of 66%–100%). However, when time since discontinuation was examined, there was a twofold to threefold increased hazard ratio for hip fracture in women who had stopped more than 9 months ago. Women who discontinued drug therapy and who had higher degrees of compliance (MPR 88%–100%) and those previously having received bisphosphonates for 3 or more years had numerically lower rates of hip fracture than those who were less compliant or on shorter duration of drug therapy. Nevertheless, the fracture rates were still higher than those who remained on bisphosphonates.

A population-based, nested case-control study of women 68 years or older from Ontario, Canada recently looked at the incidence of subtrochanteric fracture in those on an oral bisphosphonate between 2002 and 2008. During this 6-year interval, short-term use of bisphosphonate (100 days to 3 years) did not reduce fracture risk, but use for 3 or more years did reduce fracture risk.

Alendronate Clinical Extension Studies

The Alendronate Phase III Osteoporosis Treatment Study Group obtained lateral radiographs of the spine at the end of each extension. Morphometric vertebral fractures and clinical fractures were reported as safety end points. At 10 years, there was no evidence of an increased rate of morphometric vertebral fractures in the ALN5P5 discontinuation group (6.6%) as compared with the ALN10 (10 mg/d × 10 years) group (5.0%). During years 8 to 10, 12.0% of the ALN5P5 group sustained a first nonvertebral fracture compared with 8.1% in the ALN10 (10 mg) group and 11.5% in the ALN10 (5 mg) group.

In the FLEX cohort, 437 patients who took alendronate 5 mg/d for 2 years then 10 mg/d for up to 3 years were randomized to placebo for 5 years. When compared with patients on 10 years of alendronate, clinical vertebral fractures were more common in the placebo group despite a slight increase in bone density in the spine over 5 years: placebo, 23 of 437 (5.3%) versus alendronate, 16 of 662 (2.4%) (95% confidence interval [CI] 0.24–0.85). A nonsignificant increase in morphometric fractures was also noted, but there was no difference in hip or nonvertebral fractures.

A post hoc analysis of FLEX reported by Schwartz and colleagues looked at those women without prevalent vertebral fractures at FLEX baseline, and found a greater risk of nonvertebral fractures in those who stopped alendronate after 5 years if their femoral neck T-score was −2.5 or less at FLEX baseline.

Risedronate Clinical Extension Studies

In the extension of VERT-NA, despite seeing significant declines in BMD and increases in BTM in the 1 year off risedronate, the incidence of new morphometric vertebral fractures was decreased by 46% compared with the placebo group (relative risk 0.54, 95% CI 0.34–0.86, P = .009). New vertebral fractures occurred in 26 of 398 (6.5%) patients who had previously taken risedronate and in 42 of 361 (11.6%) patients on placebo. By contrast, there was no significant difference in nonvertebral fractures, with 19 of 398 (4.8%) in the risedronate group and 18 of 361 (5.0%) in the placebo group (although these were collected as adverse events and radiologic confirmation was not required). This result suggests a residual protective effect of risedronate on morphometric fractures but, because there was not a comparator group that continued on risedronate, questions as to diminished fracture benefit associated with drug holiday cannot be addressed.

Zolendronate Clinical Extension Studies

In the 3-year extension of the HORIZON PFT trial, with more modest declines in BMD and slight increases in BTM, new morphometric vertebral fractures occurred more frequently in the Z3P3 group compared with those who continued on zolendronate (Z6). There were 30 fractures in 486 patients (6.2%) in the Z3P3 group and 14 fractures in 469 patients (3.0%) in the Z6 group (hazard ratio 2.07, P = .04). However, there were no reported differences in the rate of clinical vertebral, hip, and nonvertebral fractures between the Z3P3 and Z6 groups.

Hormone Therapy

Case-control studies have suggested a minimum of 5 years of estrogen treatment to reduce fracture risk. The Women’s Health Initiative (WHI) noted a 33% reduction in hip fractures in healthy postmenopausal women (not just women with osteoporosis) who had been treated with estrogen therapy for an average of 5.6 years. However, estrogens have been relegated to short-term use at the lowest possible dose to alleviate menopause symptoms, not for the prevention of chronic disease. Studies of fracture risk after cessation of estrogen have noted a loss in fracture risk protection. The FDA no longer recommends estrogen therapy for the treatment of osteoporosis.

Raloxifene

Raloxifene has been shown to be effective in preventing postmenopausal bone loss and vertebral fractures over a 3-year period in the randomized double-blind MORE trial (Multiple Outcomes of Raloxifene Evaluation) and the 3-year continuation with the CORE trial (Continuing Outcomes Relevant to Evista). This trial ultimately produced 8 years of safety data with raloxifene use and no significant change in fracture data, including no significant reduction in hip fracture risk compared with placebo. There is no limit on duration of treatment with raloxifene. Significant declines in bone density occur shortly after cessation of raloxifene therapy. However, there are no data on the change in fracture risk after discontinuing raloxifene.

Denosumab

Denosumab, given subcutaneously twice yearly for 3 years, was associated with a reduction in the risk of vertebral, nonvertebral, and hip fractures in women with osteoporosis. Six years of continuous treatment was associated with continued BMD accrual. The original pivotal fracture trial will be expanded through a 7-year extension, which will provide 10 years of data regarding long-term efficacy and safety. There is currently no recommended limit on duration of denosumab therapy.

There have not been sufficiently large studies to clarify fracture rates after stopping denosumab treatment in comparison with placebo. However, in the BTM and BMD study previously noted, there were no fracture rate differences in the 2-year period after 2 years of treatment compared with placebo: 3% nonvertebral fractures in each group and no clinical vertebral fractures were reported.

Teriparatide

In the Fracture Prevention Trial (FPT), treatment with once-daily teriparatide significantly reduced the risk of vertebral and nonvertebral fractures over a median duration of observation of 21 months. Lindsay and colleagues concluded that increased duration of teriparatide therapy resulted in a progressive decrease in the rate of nonvertebral fragility fractures and back pain, and adverse events occurred early in the treatment and not later, thus suggesting that patients may achieve improved outcomes by persisting on teriparatide therapy.

Expert review of the finding of osteosarcoma in teriparatide-treated Fischer 344 rats indicated that this finding was unlikely to predict an increased risk of osteosarcoma in human patients receiving teriparatide treatment for up to 2 years. Therefore in the United States, approval by the FDA was for 2 years of treatment. Further efficacy and safety was found in glucocorticoid-treated subjects considered at high risk of fracture when treated with 3 years of teriparatide therapy; nevertheless, the FDA has not altered the recommended treatment interval of 2 years.

There were 18 months of follow-up after the cessation of teriparatide in the FPT. Although there was confounding with bisphosphonate use, there was a continued significant fracture reduction effect that continued during this period, independent of bisphosphonate use, according to a logistic regression analysis. Again this was similar to the continued vertebral fracture efficacy in males up to 30 months after cessation of teriparatide treatment.

Bisphosphonates

Studies suggest that a minimum of 2 to 3 years of bisphosphonate therapy with good compliance is needed to demonstrate fracture benefit. Furthermore, treatment with bisphosphonates for 3 to 5 years and subsequent discontinuation for 3 to 5 years is not associated with an increased risk of hip fracture (alendronate; zolendronate), but when compared to those who continued on bisphosphonate, is associated with increased risk of vertebral fractures (alendronate; zolendronate), and an increased risk of nonvertebral fractures in women without prevalent vertebral fractures and BMD T-scores of −2.5 or less (alendronate).

Information regarding the effects of discontinuation of bisphosphonates on fracture risk is limited, and drawing conclusions based on this is problematic. There are two sources of information that can be accessed: administrative databases and clinical trial extensions.

Gallagher and colleagues studied 36,164 women in the General Practice Research Database (GPRD) in the United Kingdom who received a new prescription for alendronate (74.1%) or risedronate (25.9%). There was no evidence of residual effect on fracture risk after stopping bisphosphonate when the mean follow-up was 2.39 years, and few used bisphosphonates for 5 years of longer. Patients who recently stopped bisphosphonates had a similar fracture risk to those who had discontinued further in the past and to patients who had just started therapy.

New users of alendronate (77%) and risedronate (23%) were also the subject of a study by Curtis and colleagues. In addition, they evaluated for the effect of compliance as measured by the medication possession ratio (MPR). A cohort of 9063 women who were members of “a large US healthcare organization” and compliant with drug therapy for at least 2 years were examined. Women who discontinued drug therapy had lower hip fracture rates than nonadherent women, if they had previously taken bisphosphonates for 2 or more years and had been compliant with drug therapy (MPR of 66%–100%). However, when time since discontinuation was examined, there was a twofold to threefold increased hazard ratio for hip fracture in women who had stopped more than 9 months ago. Women who discontinued drug therapy and who had higher degrees of compliance (MPR 88%–100%) and those previously having received bisphosphonates for 3 or more years had numerically lower rates of hip fracture than those who were less compliant or on shorter duration of drug therapy. Nevertheless, the fracture rates were still higher than those who remained on bisphosphonates.

A population-based, nested case-control study of women 68 years or older from Ontario, Canada recently looked at the incidence of subtrochanteric fracture in those on an oral bisphosphonate between 2002 and 2008. During this 6-year interval, short-term use of bisphosphonate (100 days to 3 years) did not reduce fracture risk, but use for 3 or more years did reduce fracture risk.

Alendronate Clinical Extension Studies

The Alendronate Phase III Osteoporosis Treatment Study Group obtained lateral radiographs of the spine at the end of each extension. Morphometric vertebral fractures and clinical fractures were reported as safety end points. At 10 years, there was no evidence of an increased rate of morphometric vertebral fractures in the ALN5P5 discontinuation group (6.6%) as compared with the ALN10 (10 mg/d × 10 years) group (5.0%). During years 8 to 10, 12.0% of the ALN5P5 group sustained a first nonvertebral fracture compared with 8.1% in the ALN10 (10 mg) group and 11.5% in the ALN10 (5 mg) group.

In the FLEX cohort, 437 patients who took alendronate 5 mg/d for 2 years then 10 mg/d for up to 3 years were randomized to placebo for 5 years. When compared with patients on 10 years of alendronate, clinical vertebral fractures were more common in the placebo group despite a slight increase in bone density in the spine over 5 years: placebo, 23 of 437 (5.3%) versus alendronate, 16 of 662 (2.4%) (95% confidence interval [CI] 0.24–0.85). A nonsignificant increase in morphometric fractures was also noted, but there was no difference in hip or nonvertebral fractures.

A post hoc analysis of FLEX reported by Schwartz and colleagues looked at those women without prevalent vertebral fractures at FLEX baseline, and found a greater risk of nonvertebral fractures in those who stopped alendronate after 5 years if their femoral neck T-score was −2.5 or less at FLEX baseline.

Risedronate Clinical Extension Studies

In the extension of VERT-NA, despite seeing significant declines in BMD and increases in BTM in the 1 year off risedronate, the incidence of new morphometric vertebral fractures was decreased by 46% compared with the placebo group (relative risk 0.54, 95% CI 0.34–0.86, P = .009). New vertebral fractures occurred in 26 of 398 (6.5%) patients who had previously taken risedronate and in 42 of 361 (11.6%) patients on placebo. By contrast, there was no significant difference in nonvertebral fractures, with 19 of 398 (4.8%) in the risedronate group and 18 of 361 (5.0%) in the placebo group (although these were collected as adverse events and radiologic confirmation was not required). This result suggests a residual protective effect of risedronate on morphometric fractures but, because there was not a comparator group that continued on risedronate, questions as to diminished fracture benefit associated with drug holiday cannot be addressed.

Zolendronate Clinical Extension Studies

In the 3-year extension of the HORIZON PFT trial, with more modest declines in BMD and slight increases in BTM, new morphometric vertebral fractures occurred more frequently in the Z3P3 group compared with those who continued on zolendronate (Z6). There were 30 fractures in 486 patients (6.2%) in the Z3P3 group and 14 fractures in 469 patients (3.0%) in the Z6 group (hazard ratio 2.07, P = .04). However, there were no reported differences in the rate of clinical vertebral, hip, and nonvertebral fractures between the Z3P3 and Z6 groups.

Hormone Therapy

Case-control studies have suggested a minimum of 5 years of estrogen treatment to reduce fracture risk. The Women’s Health Initiative (WHI) noted a 33% reduction in hip fractures in healthy postmenopausal women (not just women with osteoporosis) who had been treated with estrogen therapy for an average of 5.6 years. However, estrogens have been relegated to short-term use at the lowest possible dose to alleviate menopause symptoms, not for the prevention of chronic disease. Studies of fracture risk after cessation of estrogen have noted a loss in fracture risk protection. The FDA no longer recommends estrogen therapy for the treatment of osteoporosis.

Raloxifene

Raloxifene has been shown to be effective in preventing postmenopausal bone loss and vertebral fractures over a 3-year period in the randomized double-blind MORE trial (Multiple Outcomes of Raloxifene Evaluation) and the 3-year continuation with the CORE trial (Continuing Outcomes Relevant to Evista). This trial ultimately produced 8 years of safety data with raloxifene use and no significant change in fracture data, including no significant reduction in hip fracture risk compared with placebo. There is no limit on duration of treatment with raloxifene. Significant declines in bone density occur shortly after cessation of raloxifene therapy. However, there are no data on the change in fracture risk after discontinuing raloxifene.

Denosumab

Denosumab, given subcutaneously twice yearly for 3 years, was associated with a reduction in the risk of vertebral, nonvertebral, and hip fractures in women with osteoporosis. Six years of continuous treatment was associated with continued BMD accrual. The original pivotal fracture trial will be expanded through a 7-year extension, which will provide 10 years of data regarding long-term efficacy and safety. There is currently no recommended limit on duration of denosumab therapy.

There have not been sufficiently large studies to clarify fracture rates after stopping denosumab treatment in comparison with placebo. However, in the BTM and BMD study previously noted, there were no fracture rate differences in the 2-year period after 2 years of treatment compared with placebo: 3% nonvertebral fractures in each group and no clinical vertebral fractures were reported.

Teriparatide

In the Fracture Prevention Trial (FPT), treatment with once-daily teriparatide significantly reduced the risk of vertebral and nonvertebral fractures over a median duration of observation of 21 months. Lindsay and colleagues concluded that increased duration of teriparatide therapy resulted in a progressive decrease in the rate of nonvertebral fragility fractures and back pain, and adverse events occurred early in the treatment and not later, thus suggesting that patients may achieve improved outcomes by persisting on teriparatide therapy.

Expert review of the finding of osteosarcoma in teriparatide-treated Fischer 344 rats indicated that this finding was unlikely to predict an increased risk of osteosarcoma in human patients receiving teriparatide treatment for up to 2 years. Therefore in the United States, approval by the FDA was for 2 years of treatment. Further efficacy and safety was found in glucocorticoid-treated subjects considered at high risk of fracture when treated with 3 years of teriparatide therapy; nevertheless, the FDA has not altered the recommended treatment interval of 2 years.

There were 18 months of follow-up after the cessation of teriparatide in the FPT. Although there was confounding with bisphosphonate use, there was a continued significant fracture reduction effect that continued during this period, independent of bisphosphonate use, according to a logistic regression analysis. Again this was similar to the continued vertebral fracture efficacy in males up to 30 months after cessation of teriparatide treatment.