Osteoporos Int (2014) 25:1659–1666 DOI 10.1007/s00198-013-2433-3
REVIEW
How long should we treat? C. Roux & K. Briot
Received: 11 January 2013 / Accepted: 14 June 2013 / Published online: 26 March 2014 # International Osteoporosis Foundation and National Osteoporosis Foundation 2013
Abstract Osteoporosis is a chronic disease, for which effective drugs are available. These drugs have reduced the risk of osteoporosis-related fractures in robust trials of 3– 5 years duration. There is no evidence of anti-fracture efficacy for treatments of longer duration. The consequences of stopping treatments are very different for the different molecules. Bisphosphonates can be safely discontinued after 3– 5 years of treatment if there was optimal adherence and if patients are no longer osteoporotic. This discontinuation cannot be applied in patients with recent fractures or for other treatments. Safety of prolonged treatment is a huge concern which must be managed appropriately. The decision of a prolonged treatment is driven by the underlying risk of fracture. This risk must be assessed regularly in order to share with the patient the benefit-risk ratio of prolonged treatment. Keywords Atypical fractures . Benefit-risk ratio . Bisphosphonates . Fractures . Long-term treatments . Osteoporosis . Osteonecrosis of the jaw . Stopping treatment “La question ne se pose pas : il y a trop de vent (Boris VIAN)”
Introduction Osteoporosis is a medical condition characterized by an increased risk of fractures. Hip fractures are the most devastating fractures, but attention must be paid also to vertebral fractures [1, 2] and to some, nonvertebral nonhip fractures which carry an excess mortality in the years following the fracture [3–7]. C. Roux (*) : K. Briot Department of Rheumatology, Cochin Hospital, Paris-Descartes University, Paris, France e-mail: [email protected]
Moreover, these fractures are strong risk factors for other fractures [8, 9] and the 5-year mortality in patients with subsequent fractures is higher than in those with only one event [1]. Patients with a high risk of fracture can be identified by bone mineral density (BMD) measurements and fracture risk assessment tools, such as FRAX®, which integrates clinical risk factors, with or without BMD, to predict an individual 10year risk of sustaining a fracture. For patients who have a high risk of first or recurrent fracture, both nonpharmacological and pharmacological approaches must be implemented. Several pharmaceutical treatments exist for the management of osteoporosis that effectively reduce the risk of vertebral, nonvertebral, and hip fractures. Proof of their efficacy is robust, as demonstrated in postmenopausal women in clinical trials of optimal methodology, (prospective placebo-controlled design), assessing fractures as the principal outcome. However, duration of these studies is 3–5 years; a chronic condition such as osteoporosis requires long-term treatment. The proportion of frail elderly persons at high risk of falls and fractures is expected to increase dramatically in the next decades. Indeed, human senescence has been delayed: the age at which remaining life expectancy is 10 years has been postponed by 8 years in Sweden, 12 years in Japan; in some countries, the number of centenarians climbs vertically [10]. This has profound implications not only for societies but also individuals, who expect to reach old age in good health. Moreover, severe osteoporosis and fractures are not only a problem for elderly people but can occur in 60-year-old patients [11], and data suggest that the consequences of fractures can be more severe in the younger rather than in the older patients. In the Study of Osteoporotic Fractures, a total of 5,580 women were observed prospectively for two decades: increased short-term mortality after hip fracture was in fact driven by the excess mortality in those aged 65–69 years [odds ratio (OR)=5.0 (2.6–9.5)] and 70–79 years [OR=2.4 (1.8–3.3)] [12]. Thus, treatment can be implemented in 60-year-old patients. How long these patients should be treated is a key issue.
1660
In this review, we summarize available data, with a special focus on safety, and discuss indications for prolonged antiosteoporotic treatment. For hypothetical purposes, we will consider that patients have optimal adherence to the prescribed drug, to calcium and vitamin D supplements if they are necessary, and to nonpharmacological treatments. Fall prevention programs are beneficial in elderly people with high risk of falling, by reducing the number of falls and the number of fallers [13]. Assessment and correction of postural instability, excess of antihypertensive drugs or hypnotics, poor visual acuity, etc. [14] can help decrease the risk of falls and fractures. In a clinical setting, it is intuitive that this approach participates in the optimization of patients'care.
Is anti-fracture efficacy sustained with long-term treatment? Some data in the placebo-controlled trials suggest that the anti-fracture effect of some drugs is stronger in the first year of treatment: the relative risk of vertebral fracture is decreased by 45 % over 1 year, 39 % over 3 years and 24 % over 5 years with strontium ranelate [15]; this risk is decreased by 65 % over 1 year but 41 % over 3 years with risedronate [16]. This apparent reduction of effect can be explained by a higher proportion of placebo-treated subjects who experience incident fractures and are censored. Thus, the level of risk of fracture within the placebo group could decrease with time. Such a result has not been observed with two injectable drugs, zoledronic acid [17] and denosumab [18]; in these studies, the incidence of fractures in the placebo groups was lower than in trials cited above; moreover, completion may be higher in studies with parenteral drugs. To provide long-term data, extension studies have been conducted with most of the treatments. The reliability of these extension studies is low, because of the absence of placebo, (precluding any definitive conclusion about anti-fracture efficacy), and selection of patients, i.e., those with optimal adherence, good tolerance, and low risk of fracture. However, having these limitations in mind, these studies give information for the management of patients who require long-term treatment. The Multiple Outcomes of Raloxifen (MORE) trial was conducted over 3 years, and had a 12-month blinded extension [19], confirming the anti-vertebral fracture effect. Continuing Outcomes Relevant to Evista study was a randomized placebo-controlled extension of MORE, providing 8year data for raloxifene. From the 7,705 patients included in MORE at baseline, 4,011 chose to enroll, and the median time between the two studies was 10.6 months. Raloxifene had no effect on nonvertebral fracture risk after 8 years. However, the main objective of the trial was to assess the incidence of breast cancer, and no spine radiographs were performed [20].
Osteoporos Int (2014) 25:1659–1666
The anti-fracture effect of strontium ranelate has been demonstrated over 5 years against placebo; the long-term effect of strontium ranelate has been explored up to 10 years in an open-label extension study [21]. From the 3,082 patients who received strontium ranelate in the pivotal studies, 893 were included in the first extension study after 5 years of placebo-controlled period, and 237 patients, who received strontium ranelate for 10 years, are the basis of this observation. These patients had baseline characteristics similar to the whole population. The cumulative incidence of new fractures was similar during the 5-year extension period and during the first 5-year placebo-controlled period, 20 % on average. An effort was made to approach the anti-fracture efficacy of strontium ranelate in the absence of a placebo group by comparing the results to the ones predicted by using the FRAX value (without BMD; mean, 26 %) in patients at the time of their inclusion in the extension study. Fracture Intervention Trial Long-Term Extension (FLEX) is an extension of the pivotal Fracture Intervention Trial (FIT) study of alendronate with 10-year follow-up. From the initial 3,236 patients assigned to receive alendronate in the FIT trial, 1,099 were randomized in FLEX: 437 in the placebo group and 329 and 333 in the alendronate group, 5 and 10 mg per day, respectively [22]. Because women could receive a placebo after randomization, those with a low hip BMD (T−2.0. No significant difference between placebo and alendronate (two doses pooled) was observed for all clinical fractures or nonvertebral fractures. There was a statistically significant lower risk of clinical vertebral fractures, without difference in the risk of morphometric vertebral fractures [22]. The longest follow-up of patients in risedronate trials is 7 years, in a group of 68 patients [23]. It was observed a similar annualized incidence of new vertebral fractures during years 6–7 and years 0–3 (which was the initial placebocontrolled study). In a cohort from a large USA medical claims database, 287,099 patients with oral bisphosphonate prescription were identified: with the methodological limits of such studies, data suggest persistence of overall hip fracture protection with long-term, up to 8 years, use of bisphosphonates [24]. From the initial 2,629 patients who received three infusions in the Reduced Incidence with Zoledronic Acid (HORIZON) pivotal fracture trial, 1,233 were randomized to a 3-year extension study: 470 completed follow-up in the placebo group and 451 in the yearly zoledronic acid group (407 received three doses). At baseline of the extension period, the mean age of participants was 75.5 years, 53–57 % of them had a femoral neck T-score≤−2.5, and 30–36 % had at least two prevalent vertebral fractures. No significant differences
Osteoporos Int (2014) 25:1659–1666
between treatment groups were observed for all fractures, except morphometric vertebral fractures; there was no difference for clinical vertebral fractures [25]. Of the 7,808 participants to the pivotal 3-year study of denosumab (FREEDOM trial), 4,550 were enrolled in an extension study with an open-label design for an additional 7 years. Data have been reported for the two first years of extension, making available fracture data in patients exposed to 5 years of denosumab (2,343 patients); in the extension, fracture incidence rates remained low, similar to the yearly incidence observed during FREEDOM. The expected fracture rates in a hypothetical cohort of long-term placebo controls was estimated by the method of “virtual twins”, and these calculated incidences were higher than the observed incidences in the treated patients [26]. From a methodological point of view, there is no evidence that anti-fracture efficacy exists after 3–5 years. From a practical point of view, there is no evidence that bisphosphonates stop working beyond 3–5 years of treatment, for a period of time. Provided data are reassuring, although there is a bias related to selection of patients. Whether or not a switch to another treatment could help for anti-fracture efficacy is unknown.
Is there a risk related to stopping treatment? Cessation of some treatments results in decrease in BMD. This is well known for hormone replacement therapy (HRT). After 2 years of treatment with conjugated estrogen 0.625 mg/day, 81 women were switched to placebo, and during the next year they experienced a 4.5 and 2.4 % decrease at the spine and trochanter, respectively [27]. More importantly, the post intervention follow-up of the Women’s Health Initiative trial including 15,730 women, which demonstrated that HRT reduces hip and clinical vertebral fractures, showed that the risk of fractures was comparable among women in the discontinued HRT group and placebo group; this suggests a greater increase in the annualized risk of fractures in women after HRT therapy [28]. One year of discontinuation of raloxifene (after 5 years of administration) results in 2.4 % decrease in lumbar spine BMD [29]. Such accelerated bone loss has also been observed after withdrawal of denosumab. In an off-treatment extension of a randomized, double-blind study, 128 patients were followed for 2 years after a 2-year treatment period. BMD decreased at all sites; most of the decrease occurred during the first year after cessation of treatment. After 2 years off treatment, BMD was not different to pretreatment value at the spine and hip but lower at the radius [30]. After denosumab discontinuation, the bone resorption marker increases immediately, above the baseline values, peaks 6 months after discontinuation, and returns to pretreatment values. There is a
1661
statistically significant correlation between resorption marker value before treatment and after 2 years of treatment cessation. The key issue of potential effect of these variations on fracture risk has been assessed in 797 subjects in FREEDOM who discontinued treatment after receiving 2 to 5 doses of either denosumab or placebo. They were followed for a median of 0.8 years per subject; 42 and 28 % of the previous placebo and denosumab groups, respectively, received an anti-osteoporotic treatment during the follow-up. There was no difference in fracture occurrence pattern between the groups during the off-treatment period. These data have been obtained in a population with a mean baseline Tscore of −2.8 and −2.1 at the lumbar spine and hip, respectively, and only 26 % of patients had prevalent vertebral fractures [31]. Thus, these reassuring data do not apply to a higher-risk population. Long-term bone retention is a well-known pharmacological property of bisphosphonates, although not similar for all the molecules. Patients are still exposed to these drugs after they stop taking them. The fracture risk after discontinuation of risedronate has been assessed after 3 years of treatment. Initially, 1,628 patients were randomized to receive either placebo or risedronate 5 mg and 759 entered a 1-year follow-up study [32]. The relative risk of vertebral fractures was reduced by 46 % in the former risedronate group compared to the former placebo group. In contrast, there was a −0.8 and −1.23 % decrease in lumbar spine and femoral neck BMD, in the previous risedronate treated patients. Urinary NTX as a marker of bone resorption was available at the end of the follow-up in 89 patients of the former risedronate group: it increased from 30 to 51 nmol BCE/nmol creatinine. These data suggest a discrepancy between the short-term changes of surrogate markers and vertebral fracture risk. This limits their use in clinical practice for predicting vertebral fracture risk in the year after cessation of risedronate treatment. Whether or not bone turnover marker changes can help to decide when to resume treatment is unknown. In FLEX [22], patients switched to placebo had a 1.5 % increase in lumbar spine BMD and a 3.38 % decrease in hip BMD over 5 years. Biochemical markers were assessed in a subgroup of 87 patients, and the first assessment was performed 3 years after cessation of alendronate. A gradual rise in markers was measured over 5 years, with a value 50 to 60 % higher than in patients who continued alendronate. This study does not give information on changes in markers in the first years after cessation of alendronate [22]. In the extension study of HORIZON [25], patients switching to placebo after 3 years of zoledronic acid treatment had after 3 years of follow-up, a slightly lower BMD than patients who continued treatment (with differences of 1.36 and 2.06 % at the femoral neck and lumbar spine, respectively). There was no difference in markers of bone
1662
turnover at year 6 between patients continuously treated over 6 years or switched to placebo after 3 years of treatment. Post hoc analyses have been conducted in order to assess the determinants of long-term anti-fracture effects. Data are limited by methodological issues precluding any definitive conclusion on fracture effect. These data suggest that BMD measurements of the femoral neck could be used to select patients for prolonged therapy. In FLEX, among women without vertebral fracture at baseline, continuation of alendronate reduced nonvertebral fractures in those having a femoral neck T-score of −2.5 or less. Such a benefit was not observed in those having a higher T-score. These results cannot be applied to patients with vertebral fractures at baseline, though it is the most likely population to treat [33]. The predictive value of a T-score
REVIEW
How long should we treat? C. Roux & K. Briot
Received: 11 January 2013 / Accepted: 14 June 2013 / Published online: 26 March 2014 # International Osteoporosis Foundation and National Osteoporosis Foundation 2013
Abstract Osteoporosis is a chronic disease, for which effective drugs are available. These drugs have reduced the risk of osteoporosis-related fractures in robust trials of 3– 5 years duration. There is no evidence of anti-fracture efficacy for treatments of longer duration. The consequences of stopping treatments are very different for the different molecules. Bisphosphonates can be safely discontinued after 3– 5 years of treatment if there was optimal adherence and if patients are no longer osteoporotic. This discontinuation cannot be applied in patients with recent fractures or for other treatments. Safety of prolonged treatment is a huge concern which must be managed appropriately. The decision of a prolonged treatment is driven by the underlying risk of fracture. This risk must be assessed regularly in order to share with the patient the benefit-risk ratio of prolonged treatment. Keywords Atypical fractures . Benefit-risk ratio . Bisphosphonates . Fractures . Long-term treatments . Osteoporosis . Osteonecrosis of the jaw . Stopping treatment “La question ne se pose pas : il y a trop de vent (Boris VIAN)”
Introduction Osteoporosis is a medical condition characterized by an increased risk of fractures. Hip fractures are the most devastating fractures, but attention must be paid also to vertebral fractures [1, 2] and to some, nonvertebral nonhip fractures which carry an excess mortality in the years following the fracture [3–7]. C. Roux (*) : K. Briot Department of Rheumatology, Cochin Hospital, Paris-Descartes University, Paris, France e-mail: [email protected]
Moreover, these fractures are strong risk factors for other fractures [8, 9] and the 5-year mortality in patients with subsequent fractures is higher than in those with only one event [1]. Patients with a high risk of fracture can be identified by bone mineral density (BMD) measurements and fracture risk assessment tools, such as FRAX®, which integrates clinical risk factors, with or without BMD, to predict an individual 10year risk of sustaining a fracture. For patients who have a high risk of first or recurrent fracture, both nonpharmacological and pharmacological approaches must be implemented. Several pharmaceutical treatments exist for the management of osteoporosis that effectively reduce the risk of vertebral, nonvertebral, and hip fractures. Proof of their efficacy is robust, as demonstrated in postmenopausal women in clinical trials of optimal methodology, (prospective placebo-controlled design), assessing fractures as the principal outcome. However, duration of these studies is 3–5 years; a chronic condition such as osteoporosis requires long-term treatment. The proportion of frail elderly persons at high risk of falls and fractures is expected to increase dramatically in the next decades. Indeed, human senescence has been delayed: the age at which remaining life expectancy is 10 years has been postponed by 8 years in Sweden, 12 years in Japan; in some countries, the number of centenarians climbs vertically [10]. This has profound implications not only for societies but also individuals, who expect to reach old age in good health. Moreover, severe osteoporosis and fractures are not only a problem for elderly people but can occur in 60-year-old patients [11], and data suggest that the consequences of fractures can be more severe in the younger rather than in the older patients. In the Study of Osteoporotic Fractures, a total of 5,580 women were observed prospectively for two decades: increased short-term mortality after hip fracture was in fact driven by the excess mortality in those aged 65–69 years [odds ratio (OR)=5.0 (2.6–9.5)] and 70–79 years [OR=2.4 (1.8–3.3)] [12]. Thus, treatment can be implemented in 60-year-old patients. How long these patients should be treated is a key issue.
1660
In this review, we summarize available data, with a special focus on safety, and discuss indications for prolonged antiosteoporotic treatment. For hypothetical purposes, we will consider that patients have optimal adherence to the prescribed drug, to calcium and vitamin D supplements if they are necessary, and to nonpharmacological treatments. Fall prevention programs are beneficial in elderly people with high risk of falling, by reducing the number of falls and the number of fallers [13]. Assessment and correction of postural instability, excess of antihypertensive drugs or hypnotics, poor visual acuity, etc. [14] can help decrease the risk of falls and fractures. In a clinical setting, it is intuitive that this approach participates in the optimization of patients'care.
Is anti-fracture efficacy sustained with long-term treatment? Some data in the placebo-controlled trials suggest that the anti-fracture effect of some drugs is stronger in the first year of treatment: the relative risk of vertebral fracture is decreased by 45 % over 1 year, 39 % over 3 years and 24 % over 5 years with strontium ranelate [15]; this risk is decreased by 65 % over 1 year but 41 % over 3 years with risedronate [16]. This apparent reduction of effect can be explained by a higher proportion of placebo-treated subjects who experience incident fractures and are censored. Thus, the level of risk of fracture within the placebo group could decrease with time. Such a result has not been observed with two injectable drugs, zoledronic acid [17] and denosumab [18]; in these studies, the incidence of fractures in the placebo groups was lower than in trials cited above; moreover, completion may be higher in studies with parenteral drugs. To provide long-term data, extension studies have been conducted with most of the treatments. The reliability of these extension studies is low, because of the absence of placebo, (precluding any definitive conclusion about anti-fracture efficacy), and selection of patients, i.e., those with optimal adherence, good tolerance, and low risk of fracture. However, having these limitations in mind, these studies give information for the management of patients who require long-term treatment. The Multiple Outcomes of Raloxifen (MORE) trial was conducted over 3 years, and had a 12-month blinded extension [19], confirming the anti-vertebral fracture effect. Continuing Outcomes Relevant to Evista study was a randomized placebo-controlled extension of MORE, providing 8year data for raloxifene. From the 7,705 patients included in MORE at baseline, 4,011 chose to enroll, and the median time between the two studies was 10.6 months. Raloxifene had no effect on nonvertebral fracture risk after 8 years. However, the main objective of the trial was to assess the incidence of breast cancer, and no spine radiographs were performed [20].
Osteoporos Int (2014) 25:1659–1666
The anti-fracture effect of strontium ranelate has been demonstrated over 5 years against placebo; the long-term effect of strontium ranelate has been explored up to 10 years in an open-label extension study [21]. From the 3,082 patients who received strontium ranelate in the pivotal studies, 893 were included in the first extension study after 5 years of placebo-controlled period, and 237 patients, who received strontium ranelate for 10 years, are the basis of this observation. These patients had baseline characteristics similar to the whole population. The cumulative incidence of new fractures was similar during the 5-year extension period and during the first 5-year placebo-controlled period, 20 % on average. An effort was made to approach the anti-fracture efficacy of strontium ranelate in the absence of a placebo group by comparing the results to the ones predicted by using the FRAX value (without BMD; mean, 26 %) in patients at the time of their inclusion in the extension study. Fracture Intervention Trial Long-Term Extension (FLEX) is an extension of the pivotal Fracture Intervention Trial (FIT) study of alendronate with 10-year follow-up. From the initial 3,236 patients assigned to receive alendronate in the FIT trial, 1,099 were randomized in FLEX: 437 in the placebo group and 329 and 333 in the alendronate group, 5 and 10 mg per day, respectively [22]. Because women could receive a placebo after randomization, those with a low hip BMD (T−2.0. No significant difference between placebo and alendronate (two doses pooled) was observed for all clinical fractures or nonvertebral fractures. There was a statistically significant lower risk of clinical vertebral fractures, without difference in the risk of morphometric vertebral fractures [22]. The longest follow-up of patients in risedronate trials is 7 years, in a group of 68 patients [23]. It was observed a similar annualized incidence of new vertebral fractures during years 6–7 and years 0–3 (which was the initial placebocontrolled study). In a cohort from a large USA medical claims database, 287,099 patients with oral bisphosphonate prescription were identified: with the methodological limits of such studies, data suggest persistence of overall hip fracture protection with long-term, up to 8 years, use of bisphosphonates [24]. From the initial 2,629 patients who received three infusions in the Reduced Incidence with Zoledronic Acid (HORIZON) pivotal fracture trial, 1,233 were randomized to a 3-year extension study: 470 completed follow-up in the placebo group and 451 in the yearly zoledronic acid group (407 received three doses). At baseline of the extension period, the mean age of participants was 75.5 years, 53–57 % of them had a femoral neck T-score≤−2.5, and 30–36 % had at least two prevalent vertebral fractures. No significant differences
Osteoporos Int (2014) 25:1659–1666
between treatment groups were observed for all fractures, except morphometric vertebral fractures; there was no difference for clinical vertebral fractures [25]. Of the 7,808 participants to the pivotal 3-year study of denosumab (FREEDOM trial), 4,550 were enrolled in an extension study with an open-label design for an additional 7 years. Data have been reported for the two first years of extension, making available fracture data in patients exposed to 5 years of denosumab (2,343 patients); in the extension, fracture incidence rates remained low, similar to the yearly incidence observed during FREEDOM. The expected fracture rates in a hypothetical cohort of long-term placebo controls was estimated by the method of “virtual twins”, and these calculated incidences were higher than the observed incidences in the treated patients [26]. From a methodological point of view, there is no evidence that anti-fracture efficacy exists after 3–5 years. From a practical point of view, there is no evidence that bisphosphonates stop working beyond 3–5 years of treatment, for a period of time. Provided data are reassuring, although there is a bias related to selection of patients. Whether or not a switch to another treatment could help for anti-fracture efficacy is unknown.
Is there a risk related to stopping treatment? Cessation of some treatments results in decrease in BMD. This is well known for hormone replacement therapy (HRT). After 2 years of treatment with conjugated estrogen 0.625 mg/day, 81 women were switched to placebo, and during the next year they experienced a 4.5 and 2.4 % decrease at the spine and trochanter, respectively [27]. More importantly, the post intervention follow-up of the Women’s Health Initiative trial including 15,730 women, which demonstrated that HRT reduces hip and clinical vertebral fractures, showed that the risk of fractures was comparable among women in the discontinued HRT group and placebo group; this suggests a greater increase in the annualized risk of fractures in women after HRT therapy [28]. One year of discontinuation of raloxifene (after 5 years of administration) results in 2.4 % decrease in lumbar spine BMD [29]. Such accelerated bone loss has also been observed after withdrawal of denosumab. In an off-treatment extension of a randomized, double-blind study, 128 patients were followed for 2 years after a 2-year treatment period. BMD decreased at all sites; most of the decrease occurred during the first year after cessation of treatment. After 2 years off treatment, BMD was not different to pretreatment value at the spine and hip but lower at the radius [30]. After denosumab discontinuation, the bone resorption marker increases immediately, above the baseline values, peaks 6 months after discontinuation, and returns to pretreatment values. There is a
1661
statistically significant correlation between resorption marker value before treatment and after 2 years of treatment cessation. The key issue of potential effect of these variations on fracture risk has been assessed in 797 subjects in FREEDOM who discontinued treatment after receiving 2 to 5 doses of either denosumab or placebo. They were followed for a median of 0.8 years per subject; 42 and 28 % of the previous placebo and denosumab groups, respectively, received an anti-osteoporotic treatment during the follow-up. There was no difference in fracture occurrence pattern between the groups during the off-treatment period. These data have been obtained in a population with a mean baseline Tscore of −2.8 and −2.1 at the lumbar spine and hip, respectively, and only 26 % of patients had prevalent vertebral fractures [31]. Thus, these reassuring data do not apply to a higher-risk population. Long-term bone retention is a well-known pharmacological property of bisphosphonates, although not similar for all the molecules. Patients are still exposed to these drugs after they stop taking them. The fracture risk after discontinuation of risedronate has been assessed after 3 years of treatment. Initially, 1,628 patients were randomized to receive either placebo or risedronate 5 mg and 759 entered a 1-year follow-up study [32]. The relative risk of vertebral fractures was reduced by 46 % in the former risedronate group compared to the former placebo group. In contrast, there was a −0.8 and −1.23 % decrease in lumbar spine and femoral neck BMD, in the previous risedronate treated patients. Urinary NTX as a marker of bone resorption was available at the end of the follow-up in 89 patients of the former risedronate group: it increased from 30 to 51 nmol BCE/nmol creatinine. These data suggest a discrepancy between the short-term changes of surrogate markers and vertebral fracture risk. This limits their use in clinical practice for predicting vertebral fracture risk in the year after cessation of risedronate treatment. Whether or not bone turnover marker changes can help to decide when to resume treatment is unknown. In FLEX [22], patients switched to placebo had a 1.5 % increase in lumbar spine BMD and a 3.38 % decrease in hip BMD over 5 years. Biochemical markers were assessed in a subgroup of 87 patients, and the first assessment was performed 3 years after cessation of alendronate. A gradual rise in markers was measured over 5 years, with a value 50 to 60 % higher than in patients who continued alendronate. This study does not give information on changes in markers in the first years after cessation of alendronate [22]. In the extension study of HORIZON [25], patients switching to placebo after 3 years of zoledronic acid treatment had after 3 years of follow-up, a slightly lower BMD than patients who continued treatment (with differences of 1.36 and 2.06 % at the femoral neck and lumbar spine, respectively). There was no difference in markers of bone
1662
turnover at year 6 between patients continuously treated over 6 years or switched to placebo after 3 years of treatment. Post hoc analyses have been conducted in order to assess the determinants of long-term anti-fracture effects. Data are limited by methodological issues precluding any definitive conclusion on fracture effect. These data suggest that BMD measurements of the femoral neck could be used to select patients for prolonged therapy. In FLEX, among women without vertebral fracture at baseline, continuation of alendronate reduced nonvertebral fractures in those having a femoral neck T-score of −2.5 or less. Such a benefit was not observed in those having a higher T-score. These results cannot be applied to patients with vertebral fractures at baseline, though it is the most likely population to treat [33]. The predictive value of a T-score
