Reduction of New Heterotopic Ossification (HO) in the Open‐Label, Phase 3 MOVE Trial of Palovarotene for Fibrodysplasia Ossificans Progressiva (FOP)

Fibrodysplasia ossificans progressiva (FOP) is an ultra‐rare, severely disabling genetic disorder of progressive heterotopic ossification (HO). The single‐arm, open‐label, phase 3 MOVE trial (NCT03312634) assessed efficacy and safety of palovarotene, a selective retinoic acid receptor gamma agonist, in patients with FOP. Findings were compared with FOP natural history study (NHS; NCT02322255) participants untreated beyond standard of care. Patients aged ≥4 years received palovarotene once daily (chronic: 5 mg; flare‐up: 20 mg for 4 weeks, then 10 mg for ≥8 weeks; weight‐adjusted if skeletally immature). The primary endpoint was annualized change in new HO volume versus NHS participants (by low‐dose whole‐body computed tomography [WBCT]), analyzed using a Bayesian compound Poisson model (BcPM) with square‐root transformation. Twelve‐month interim analyses met futility criteria; dosing was paused. An independent Data Monitoring Committee recommended trial continuation. Post hoc 18‐month interim analyses utilized BcPM with square‐root transformation and HO data collapsed to equalize MOVE and NHS visit schedules, BcPM without transformation, and weighted linear mixed‐effects (wLME) models, alongside prespecified analysis. Safety was assessed throughout. Eighteen‐month interim analyses included 97 MOVE and 101 NHS individuals with post‐baseline WBCT. BcPM analyses without transformation showed 99.4% probability of any reduction in new HO with palovarotene versus NHS participants (with transformation: 65.4%). Mean annualized new HO volume was 60% lower in MOVE versus the NHS. wLME results were similar (54% reduction fitted; nominal p = 0.039). All palovarotene‐treated patients reported ≥1 adverse event (AE); 97.0% reported ≥1 retinoid‐associated AE; 29.3% reported ≥1 serious AE, including premature physeal closure (PPC)/epiphyseal disorder in 21/57 (36.8%) patients aged <14 years. Post hoc computational analyses using WBCT showed decreased vertebral bone mineral density, content, and strength, and increased vertebral fracture risk in palovarotene‐treated patients. Thus, post hoc analyses showed evidence for efficacy of palovarotene in reducing new HO in FOP, but high risk of PPC in skeletally immature patients. © 2022 The Authors. Journal of Bone and Mineral Research published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research (ASBMR).


Introduction
I n 2006, the genetic basis of the ultra-rare, severely disabling disorder fibrodysplasia ossificans progressiva (FOP; OMIM#135100) was described, revealing potential therapeutic targets in the bone morphogenetic protein (BMP) signaling pathway. (1) The estimated point prevalence of FOP is up to 1.4 per million individuals. (2) About 97% of individuals carry the same specific pathogenic variant in the glycine-serine activation domain of the Activin A Receptor Type 1 gene (ACVR1/Activin Receptor-Like Kinase 2 [ALK2]). (3,4) This classic FOP pathogenic variant (ACVR1 R206H ) leads to continually enhanced BMP pathway signaling. (5) Individuals with ACVR1 R206H typically present with classical features of bilateral malformations of the great toes, intermittent swelling and masses, and progressive heterotopic ossification (HO) in specific anatomic patterns that often form with intermittent painful episodes of connective tissue swelling, known as flare-ups. (6)(7)(8)(9) Although some flare-ups regress spontaneously, many lead to HO, transforming the affected muscles, tendons, and ligaments into heterotopic bone. (10) FOP is associated with considerably shortened life expectancy, mainly caused by cardiorespiratory failure resulting from thoracic insufficiency syndrome. (11) Progressive accumulation of HO leads to decreased mobility and increased disability. Most individuals with FOP become confined to a wheelchair by their 30s, requiring lifelong assistance with activities of daily living (7,12,13) and reducing quality of life. (14) In a 3-year, longitudinal, non-interventional natural history study (NHS) of individuals with FOP receiving standard of care (NCT02322255), correlations >0.5 were observed between HO volume and functional disability at baseline, as assessed by the Cumulative Analogue Joint Involvement Scale (CAJIS) and by the FOP Physical Function Questionnaire (FOP-PFQ), suggesting that new HO volume represents a clinically meaningful endpoint for use in clinical studies in FOP. (15,16) Current standard of care is mainly palliative and limited to symptom management and flare-up prevention. (6,7,9,17,18) Palovarotene is an orally bioavailable, selective retinoic acid receptor gamma (RARγ) agonist. (19) RARγ is a nuclear hormone receptor expressed in chondrogenic cells and chondrocytes, where it functions as a transcriptional repressor. (20,21) RARγ agonists potently downregulate the BMP signaling pathway (22) and activate the retinoid signaling pathway, inhibiting chondrogenesis and HO. (23)(24)(25) Palovarotene reduced both trauma-induced and spontaneous HO in a conditional knock-in mouse model of FOP. (26) Subsequently, two phase 2 studies (NCT02190747 (27) ; NCT02279095 (28) ) assessed the safety and efficacy of palovarotene in FOP, with the results supporting further clinical evaluation of palovarotene for the reduction of new HO. (15) Here, we present data from MOVE (NCT03312634), (29) a phase 3 trial assessing the efficacy and safety of palovarotene in reducing new HO in adult and pediatric patients with FOP.

Materials and Methods
Patients and study design MOVE (NCT03312634) is a multicenter, open-label, phase 3 trial assessing palovarotene in patients with FOP. Part A (the main part of the trial) proceeded for 24 months, and Part B was a 24-month extension. Three interim analyses were planned in Part A: interim analysis 1 (IA1) when 35 patients completed the month 12 assessment, IA2 when all patients completed month 12, and IA3 when all patients completed month 18 assessments. A final analysis was planned for when all patients completed month 24 of Part A. In Part B, all patients could receive palovarotene for an additional 24 months until commercial availability, allowing longer-term safety and efficacy data to be obtained. We present results from IA3 of Part A, using the most complete data set with the longest total follow-up available from MOVE.
Patients were enrolled at one of 16 international study sites in Argentina, Australia, Brazil, Canada, France, Italy, Japan, Spain, Sweden, the United Kingdom, and the United States of America (USA), with the first patient enrolled on November 30, 2017. One Australian site closed; patients were transferred to another Australian site. Seven study centers were shared by MOVE and the FOP NHS; additional centers participating in MOVE had expertise in FOP management and training for study procedures. Patients with FOP aged ≥4 years who had not experienced flareup symptoms in the 4 weeks before enrollment were eligible to participate and were required to provide written informed patient/parent consent, and age-appropriate assent where required by local guidelines. Sexually active patients were required to agree to remain abstinent or use two highly effective forms of birth control during and for 1 month after treatment. Pregnancy testing was performed both before treatment of females of child-bearing potential (defined as a premenopausal female patient ≥12 years old or at onset of menses, whichever was earlier, capable of becoming pregnant) and monthly during the trial. Patients receiving concurrent tetracycline, tetracycline derivatives, vitamin A or beta carotene, and patients treated with synthetic oral retinoids other than palovarotene within 4 weeks before screening were excluded. This trial was approved by local institutional review boards and carried out in compliance with the Declaration of Helsinki and International Council for Harmonisation Good Clinical Practice Guidelines.
The Principal Enrolled Population included all patients with the ACVR1 R206H pathogenic variant with no prior treatment with palovarotene. Patients with FOP who carried a different ACVR1 pathogenic variant comprised the Supplementary Enrolled Population and were considered in bone safety measures and analyses here.
Patients in MOVE received palovarotene orally at a dose of 5 mg daily (chronic treatment) and were instructed to take palovarotene after a full meal at a consistent time each day. At the onset of any eligible flare-up, patients received palovarotene dosed at 20 mg daily for 4 weeks, followed by 10 mg daily for 8 weeks (flare-up treatment), with the option to extend in 4-week intervals at the discretion of the investigator if the flareup was ongoing, until flare-up resolution. Eligible flare-ups were defined by the presence of at least one symptom (including, but not limited to, pain, swelling, redness, decreased range of motion, stiffness, and warmth) consistent with a patient's previous flare-ups, with patient-reported onset date and confirmed by the investigator. Flare-up treatment could also be initiated if the investigator confirmed a substantial high-risk traumatic event likely to lead to a flare-up. If the investigator confirmed a new flare-up location, marked worsening of the original flareup, or new substantial traumatic event likely to lead to a flareup at any time during treatment, the 12 weeks of palovarotene flare-up treatment (flare-up cycle) were restarted. Chronic and flare-up doses of palovarotene were weight-adjusted in patients <18 years old with <90% skeletal maturity (bone age <12 years in females and <14 years in males). (30) Palovarotene dose could also be reduced to the next lower dose in patients who experienced side effects, or discontinued if they were already receiving the lowest possible dose (Supplemental Table S1).

Efficacy outcomes
Efficacy data from patients enrolled in MOVE were compared with data from FOP NHS participants untreated beyond standard of care; individuals ≤65 years of age with clinically diagnosed FOP and a verified ACVR1 R206H pathogenic variant were eligible for inclusion in the NHS. (15) The data cut-off date was February 28, 2020.
FOP-specific clinical trial considerations developed by the Clinical Trials Committee of the International Clinical Council on FOP recommend the use of common study endpoints to allow direct comparison of outcomes. (31) These include measurement of the extent of established and new HO and functional assessments such as CAJIS. The MOVE primary endpoint was the annualized change in new HO volume as assessed by low-dose whole-body computed tomography (WBCT; excluding the head) compared with results from NHS participants. Annualized change in new HO volume was defined as the sum of the increases in HO volume per 12-month period across all body regions in which new HO occurred between WBCT scans. WBCT images were obtained at baseline and every 6 months in Part A of MOVE to document HO across different body regions and volume of total body HO; volume of new HO was quantified per region if new HO was judged to be present at post-baseline visits. WBCT image acquisition parameters were standardized across trial sites, including tube voltage, distance between slices, and slice thickness (Supplemental Table S2). HO volumes were measured by radiologists using a segmentation tool in the WBCT image reading software (MIM) with a lower limit of 150 Hounsfield units. Voxels that met this threshold were captured, so that emerging HO with calcification was identified (body fluids and non-calcified extraskeletal tissues generally have Hounsfield units <100). WBCT scans were obtained every 12 months in the NHS. Equivalent image acquisition protocols were used in MOVE and the NHS, and all post-baseline images were interpreted by two independent radiologists with adjudication by a third radiologist if needed, using standardized procedures at a central imaging core laboratory by readers who were blinded to which study an individual was enrolled in and the time point at which the image was taken. The two radiologists were unblinded to the time point for baseline images but remained blinded to study; baseline images were interpreted by consensus adjudication. Subsequent intra-and inter-read variability analyses revealed strong agreement for quantitative volumetric measurement of new HO: 98% of the assessed regions from 91% of the time points evaluated (intra-read) and 93% of regions from 89% of the time points (inter-read) did not have differences in HO volumes that would have exceeded pre-defined volume discrepancy thresholds. Lower agreement was found for qualitative assessment of whether new HO was present (90% of assessed regions from 45% of the time points [intra-read] and 93% of regions from 62% of the time points [inter-read]). Thus, the use of two independent readers and adjudication for WBCT reads was important.
Change from baseline to month 12 in the number of body regions with new HO (any body region in which the overall volume of HO increased) was assessed by WBCT, based on nine anatomical regions: right chest, neck, shoulder through midhumerus; left chest, neck, shoulder through mid-humerus; mid torso (torso below shoulder regions and above iliac crests); right arm (mid-humerus through hand); left arm (mid-humerus through hand); right hip (iliac crest, hip through mid-femur); left hip (iliac crest, hip through mid-femur); right lower leg (midfemur through foot); and left lower leg (mid-femur through foot).
Catastrophic HO was assessed in MOVE as an exploratory endpoint at month 12 and at the last time point assessed. Catastrophic HO categories were defined post hoc as annualized new HO >50 Â 10 3 mm 3 and >30 Â 10 3 mm 3 , based on 12-month observations. An additional cut-off of annualized new HO >100 Â 10 3 mm 3 was also explored.
Range of motion was assessed using the CAJIS by trained investigators at baseline and every 6 months in MOVE, and at baseline and every 12 months in the NHS. CAJIS measures range of motion across 12 joints (shoulder, elbow, wrist, hip, knee, and ankle on both right and left sides) and three body regions (jaw, cervical spine [neck], and thoraco-lumbar spine). Each joint or region is assigned a score (0: >90% range of motion; 1: 10%-90% range of motion; 2: <10% range of motion), with a maximum total score of 30. (32) Physical function was assessed using age-appropriate forms of the FOP-PFQ at baseline and every 6 months. (33) The FOP-PFQ includes questions about the activities of daily living and physical functioning. Age-appropriate forms were selfcompleted by adults (aged ≥15 years), with the option for selfcompletion for individuals aged 8-14 years or completion by a parent acting in proxy for individuals aged ≤14 years. Results were expressed as a percentage of the worst possible score (0-100%).
Physical and mental health (for individuals aged ≥15 years) or global health (for individuals aged <15 years) were assessed at baseline and every 6 months using the Patient-Reported Outcome Measures Information System (PROMIS) Global Health Scale and PROMIS Pediatric Global Health Scale, respectively. (34,35) These scales measure what individuals are able to do and how they feel. Scores were converted to T-scores for adult and pediatric scales; a value of 50 (with a standard deviation [SD] of 10) represents the average for the general population in the USA.

Safety outcomes
The Principal Safety Set included all enrolled patients with a documented ACVR1 R206H pathogenic variant who received at least one dose of palovarotene. The Supplemental Safety Set included patients in the Supplementary Enrolled Population. Adverse events (AEs) were monitored throughout the trial for all patients. The reporting period constituted the time from informed consent through to the data cut-off date of February 28, 2020.
Physical examination of patients was also carried out at baseline and every 6 months to further monitor for possible AEs. Severity was determined as mild, moderate, or severe. Serious AEs (SAEs) were those that resulted in death or risk of death, hospitalization or prolongation of hospitalization, persistent or significant disability or incapacity, congenital abnormality, or any other medically important event. AEs were coded using the Medical Dictionary for Regulatory Activities (MedDRA; version 21.0). AEs known to be associated with retinoids were further graded according to Common Terminology Criteria for AEs (CTCAE; version 4.03). AEs occurring during the trial were considered treatment-emergent (TEAEs). TEAEs for chronic and flare-up treatment were defined as any AE with an onset date during a period of chronic treatment or flare-up treatment, respectively.
Twelve-lead electrocardiogram and blood/urine samples for clinical laboratory tests (hematology, biochemistry, and urinalysis) were carried out at baseline and every 6 months. Vital signs, body weight, and suicidal ideation (individuals ≥8 years old only; assessed by the Columbia Suicide Severity Rating Scale [pediatric form used for those aged 8-11 years]) (36) were assessed at baseline and every 3 months. Additional safety monitoring was carried out on day 1 of flare-up palovarotene treatment and every 12 weeks thereafter until completion of treatment for the last flare-up or traumatic event. Safety evaluations for flare-up treatment included assessment of body weight, vital signs, laboratory tests, and suicidal ideation. Concomitant medications were assessed at every site and remote visit during both chronic and flare-up treatment.
For bone safety comparisons to the NHS, the Principal Safety Set included all patients enrolled in the NHS with available post-baseline follow-up. In the NHS, AEs were only captured if they resulted from procedures performed during the study; therefore, direct comparison with TEAEs reported in MOVE is not possible. Electrocardiograms and blood/urine samples were carried out on day 1 and at months 12, 24, and 36.

Bone safety
Given the previously known class effects of retinoids related to the potential for impaired bone growth, (37,38) careful assessments of epiphyseal abnormalities and growth were included in MOVE and the rest of the palovarotene clinical program. For patients aged <18 years with open epiphyseal plates, knee and hand/wrist radiographs for assessment of epiphyseal plate and distal femoral angle, and linear height by stadiometer and knee height (in triplicate) for assessment of linear growth were collected at baseline and every 6 months thereafter. WBCT scans from patients <18 years were also reviewed by two blinded independent radiologists to assess growth plate morphology of the bilateral hand/wrist and knee, tibial and femoral long bone lengths, and hip morphology to monitor for signs of avascular necrosis. Hip morphology was similarly monitored in adult patients. Differences in interpretation were adjudicated by a third independent radiologist. During MOVE, premature physeal closure (PPC) was identified as a risk of palovarotene treatment, leading to the Data Monitoring Committee (DMC) recommending an increase in frequency of bone safety radiograph monitoring from every 6 months to every 3 months for patients who had not reached 100% closure of the growth plates and were receiving flare-up dosing. Patients in MOVE from both the Principal Enrolled Population and the Supplementary Enrolled Population were included in bone safety analyses.
In the NHS, a protocol amendment required participants aged <18 years to undergo knee and hand/wrist radiographs at months 12, 24, and 36; once skeletally mature, participants no longer required knee and hand/wrist radiographs. This protocol amendment occurred late in the program but before PPC risk identification in MOVE. Some radiographs obtained in the NHS also served as baseline radiographs for those who entered MOVE. Although WBCT was primarily used to assess change in HO volume during the NHS, it was also used in safety reads assessing bilateral hand/wrist and knee epiphyseal plates, femur and tibia lengths, and hip morphology.
Post hoc biomechanical computed tomography analysis and vertebral fracture assessment Systemic retinoids have effects on normal skeletal development and repair. (37)(38)(39) However, routine assessment of bone mineral density (BMD) is not possible in patients with FOP using standard techniques, such as dual-energy X-ray absorptiometry (DXA), which are confounded by adjacent HO. To investigate potential risks of reduced BMD in patients treated with palovarotene, WBCT scans from patients in MOVE were retrospectively evaluated and compared with those from NHS participants using a new computational method (biomechanical computed tomography; VirtuOst ® software, Version 2.3, O.N. Diagnostics, Berkeley, CA, USA) for calculating vertebral BMD, bone mineral content (BMC) and strength (at L1, L2, or adjacent vertebrae). (40) Additionally, vertebral fracture assessment was performed using a semiquantitative morphometric shape analysis (VirtuOst ® software, Version 2.0; O.N. Diagnostics, Berkeley, CA, USA; assessment was from T4 to L4) for patients in MOVE versus NHS participants. This was a retrospective analysis and not based on clinical symptoms; the analysis measured vertebral deformities and then classified these as mild, moderate, or severe fractures based on the Genant grading criteria using percentage deformity. (41) Both analyses were performed for baseline and 12-month WBCT scans, with adjustments made for baseline age.

Statistics
The principal Full Analysis Set (FAS) was defined as all patients in the Principal Enrolled Population in MOVE with a baseline and at least one post-baseline HO volume measurement. For efficacy comparisons with NHS participants, the principal FAS also included all NHS participants with available baseline and at least one post-baseline HO volume measurement.
The sample size was determined via simulation based on available WBCT HO volumes from the NHS and the observed efficacy of palovarotene treatment in phase 2 studies in FOP. (27,28) With the NHS contributing data from approximately 90 participants untreated beyond standard of care and a planned sample size of 80 patients in the MOVE principal FAS, the probability of declaring significance at IA3 was approximately 0.89. This was based on the assumption that palovarotene reduces the overall volume of new HO by 65% compared with an untreated population. The overall power of the trial was projected to be 0.92.
The primary efficacy analysis for annualized change in new HO volume used a Bayesian compound Poisson model (BcPM) with square-root transformation of new HO volume to compare data from the principal FAS from MOVE with those from the NHS. (42)(43)(44)(45)(46) The Bayesian model updated prior assumptions about new HO (number of body regions with new HO, new HO volume, and the rate of change in new HO volume) with data from MOVE and the NHS, giving a probability distribution for the ratio of annual mean change in HO volume for MOVE versus NHS participants. Square-root transformation of new HO volumes was utilized to mitigate the high variability in new HO volumes observed in participants with FOP in the NHS by shrinking very large volumes of new HO. A compound Poisson distribution was used as it appropriately fits these data by modeling the number of body regions with new HO (ie, the number of events) and then conditionally modeling the volume of new HO in each region (ie, the magnitude of an event when it occurs). Zero growth was imputed for regions in scans with missing or negative values (eg, reductions in the volume of HO over time due to bone remodeling or measurement variability). Annualized change in HO volume was compared by modeling the ratio of the annual mean change in HO volume in palovarotene-treated versus NHS participants as the product of the modeled percent reduction in the annual event rate and the percent reduction in the growth rate per event. Age (<18 versus ≥18 years) and sex were included as covariates. However, it was later established that there were two key limitations of this prespecified Bayesian primary analysis with square-root transformation: a bias was introduced through the difference in WBCT frequency between the NHS (every 12 months) and MOVE (every 6 months), and negative changes in annualized HO volume were incompatible with both square-root transformation and the Bayesian analysis.
A timeline of MOVE and its planned interim analyses is shown in Supplemental Fig. S1. At IA2, futility was pre-defined as a <5% posterior probability of a >30% reduction in annualized new HO volume on the square-root transformed scale. Assuming a onesided, overall type I error rate of 2.5%, the Lan-DeMets alphaspending function with O'Brien-Fleming boundary was used to determine stopping criteria. On January 15, 2020, the prespecified analyses of efficacy met futility criteria (Supplemental Fig. S2; probability of >30% reduction in new HO volume: 4.9%) and palovarotene dosing was subsequently paused. The data were then unblinded and the raw mean annualized new HO volumes with and without palovarotene treatment were compared. Post hoc analyses, including a BcPM omitting the square-root transformation and a weighted linear mixed-effects (wLME) model, were also performed and the results reviewed by an independent DMC (with expertise across FOP, biostatistics, pediatric bone health, musculoskeletal radiology, hematology, and oncology). When using non-transformed HO volumes, the Bayesian analysis would have exceeded the prespecified boundary for early efficacy at IA2 (Supplemental Fig. S3) and thus would not have triggered futility. Because of the discordance between the results of the analysis with, versus without, square-root transformation at IA2, it was not possible to be confident in concluding treatment futility based on the square-root transformed data. The DMC recommended the study continue.
Post hoc analyses at IA3 therefore included the same Bayesian analysis without square-root transformation, in addition to the prespecified analysis. There were no formal prespecified efficacy or futility boundaries at IA3. Additional post hoc analyses of the primary endpoint included a wLME model, in which a patient-level random effect was used to account for the correlation among repeated measures on the same patient (for those individuals with post-baseline WBCT in both MOVE and the NHS, who therefore contributed results to both at IA3). Baseline HO volume divided by age (average new HO per year of life) was included as a covariate. Because of prolonged treatment interruptions in individuals <14 years of age following United States Food and Drug Administration (FDA) institution of a partial clinical hold on December 4, 2019, because of risk of PPC, and in individuals ≥14 years of age from January 24, 2020, following crossing of the futility boundary at IA2, efficacy analyses include only WBCT data collected on or before these interruptions (Supplemental Fig. S1). Further details on the statistical analyses in MOVE are provided in Supplemental Video S1.  Table S3).

Baseline demographics and characteristics
The annualized change in HO for the 97 patients in the principal FAS (mean age 14.6 years, SD: 8.7; 52.6% male) was compared with that of 101 participants from the NHS (mean age 17.3 years, SD: 9.7; 55.4% male). At IA3, 39 individuals contributed data to both MOVE and the NHS as they had transferred from the NHS to MOVE and had post-baseline WBCT scans from both studies. Baseline

Change in new HO volume
Square-root transformed annualized volume of new HO At IA3, the prespecified Bayesian analysis fitted a 5% reduction in square-root transformed annualized volume of new HO in palovarotene-treated patients in MOVE versus NHS participants untreated beyond standard of care. There was no significant difference in the square-root transformed mean volume of annualized new HO in patients treated with palovarotene compared with participants in the NHS (140.2 mm 3 versus 149.8 mm 3 ; probability of any reduction in new HO: 65.4%) when setting body region negative HO volumes to zero (Supplemental Fig. S5; wLME analysis estimate of treatment effect: p = 0.773).
It was later recognized that the difference in visit schedules between MOVE and the NHS (WBCT scans every 6 and 12 months, respectively), with the square-root transformation, introduced a bias, particularly in the Bayesian analysis because of its modeling of new HO between WBCT scans. Collapsing the new HO observed during the first 12 months in MOVE as if it were observed entirely at month 12, to mimic the visit schedule of the NHS, yielded a 16% reduction in square-root transformed new HO volume and increased the probability of any reduction in new HO to 90.7%.

Non-transformed mean annualized volume of new HO
The post hoc Bayesian analysis fitted a 36% reduction in non-transformed mean annualized volume of new HO in palovarotenetreated patients compared with NHS participants (probability of any reduction in new HO: 99.4%) (Supplemental Fig. S6). There was a 60% reduction in the non-transformed raw mean annualized volume of new HO in palovarotene-treated patients compared with NHS participants when comparing annualized new HO as reported, ie, including negative HO values (Fig. 1A). A wLME analysis yielded a consistent result, estimating the LSmean annualized reduction in new HO volume difference to be 10.9 Â 10 3 mm 3 (95% confidence interval [CI]: À21.2 Â 10 3 , À0.6 Â 10 3 ; p = 0.039), a reduction of 54% in the LSmean new HO volume in MOVE compared with the NHS (Fig. 1B).
Analysis of new HO volume data for only the 39 individuals who transferred from the NHS to MOVE, who therefore contributed results to both studies, showed a reduction in the non-transformed raw mean annualized new HO volume of 54% in MOVE compared with the same individuals in the NHS (Supplemental Fig. S7).

Variability in the annualized net change in HO volume
There was more variability in the annualized net change in HO volume among NHS participants compared with palovarotenetreated patients (Fig. 1A). Additionally, 5/101 participants (5.0%) in the NHS and 28/97 patients (28.9%) in MOVE had net negative annualized new HO, meaning that the measured volume of HO was reduced overall.

Secondary and exploratory outcomes
At month 12, the mean number of body regions with new HO since baseline was 1.3 (SD: 1.4) in palovarotene-treated patients compared with 1.5 (SD: 1.6) in NHS participants. No significant differences were observed between palovarotenetreated patients and NHS participants (Supplemental Table S4). The non-transformed annualized volume of new HO was lower in palovarotene-treated patients than in NHS participants across all body regions considered (Supplemental Table S5 Changes in functional endpoints over 18 months measured by the CAJIS (Fig. 2) and FOP-PFQ (Fig. 3) were similar between MOVE patients and those receiving standard of care in the NHS. Similar changes were observed in physical and mental health function, as assessed using the age-appropriate PROMIS Global Health Scale (Fig. 4).

Safety
A total of 99 patients were included in the MOVE Principal Safety Set. At data cut-off, 34.3% of patients had exposure to palovarotene for >12-18 months and 51.5% of patients had exposure to palovarotene for >18-24 months. Overall mean exposure to palovarotene was 3141.4 mg (SD: 1714.8); 70/99 patients received flare-up dosing (mean exposure: 2302.1 mg; SD: 1682.7) (Supplemental Table S6).
All patients in MOVE reported at least one TEAE and 97.0% reported at least one retinoid-associated TEAE ( Table 2). The maximum TEAE severity experienced was mild in 32.3% of patients, moderate in 45.5%, and severe in 22.2%. The most commonly reported TEAEs were mucocutaneous events such as dry skin (68.7%), lip dryness (46.5%), alopecia (34.3%), drug eruption (28.3%), and pruritus (26.3%) and musculoskeletal events such as arthralgia (33.3%). These events were managed with prophylactic and/or symptomatic therapy (such as analgesics, skin emollients, lip moisturizers, artificial tears, or other treatments as recommended by the investigator) in addition to dose reductions where necessary; the majority were mild in severity. Incidences of mucocutaneous events were generally similar between chronic and flare-up treatment. The majority of TEAEs that led to palovarotene dose reductions were mucocutaneous events; the incidence of dose reduction due to mucocutaneous TEAEs was higher during flare-up treatment (34.3%) than during chronic treatment (5.1%), as was the overall incidence of dose reduction due to TEAEs (flare-up treatment: 40.0%; chronic treatment: 11.1%).
Partial or complete PPC (MedDRA preferred term "epiphyses premature fusion"; n = 20) or epiphyseal disorder (frayed metaphyseal edge; n = 1) SAEs were observed in 21/57 (36.8%) patients who were <14 years of age at baseline. All events occurred in the Principal Safety Set. Two of the PPC SAEs occurred post-treatment (115 and 305 days after last dose), and 18 occurred during treatment. Nine PPC SAEs occurred in male patients and 11 occurred in female patients. Twelve of the 20 patients who experienced PPC SAEs were either females <8 years of age or males <10 years of age (based on skeletal maturity data, girls and boys achieve approximately 80% of their adult height at 8 and 10 years of age, respectively). (47) Five of the PPC SAEs occurred in the 18 patients <14 years of age who had only received chronic treatment, whereas the remaining 15 occurred in the 39 patients <14 years of age who had received chronic and flare-up treatment. PPC findings were observed as early as 6 months after initiating treatment; the majority were observed at or after 12 months. PPC in patients who received only chronic palovarotene treatment was typically observed between 12 and 18 months. In general, over the course of the trial, bone age advanced concordantly with chronological age in individuals with and without PPC; linear height, knee height, and femoral and tibial height changes from baseline and growth velocities are summarized in Supplemental Tables S7-S13 and Supplemental Figs. S8 and S9. Bone growth results are also shown for NHS participants in the Supplemental Materials.
In the post hoc biomechanical computed tomography analysis, across all ages, most adults with FOP showed low vertebral BMD, BMC, and bone strength at baseline (normative data do not allow a comparison for pediatric patients). Patients treated with palovarotene showed greater reductions from baseline to month 12 in BMD, BMC, and bone strength compared with untreated individuals from the NHS (4.0 mg/cm 3 [Supplemental Table S14], 0.17 g [Supplemental Table S15], and 156-newton [Supplemental Table S16] decreases, respectively). Additionally, in the post hoc vertebral fracture assessment, the percentage of individuals with any vertebral fractures identified at baseline was 21.5% for those enrolled in MOVE to be treated with palovarotene and 23.6% for those enrolled in the NHS; the percentage with new-onset vertebral fractures identified at month 12 was higher in the palovarotene-treated group (28.4%) versus untreated NHS participants (11.6%) (Supplemental Table S17). This equates to a 3.3 times higher risk of radiologically identified new-onset vertebral fracture with palovarotene treatment at month 12 (Supplemental Table S18). The vertebral fractures identified were mostly mild (Grade 1; 20% ≤ deformity < 25%). No severe vertebral fractures were identified in either MOVE (in the period examined from baseline to month 12) or the NHS.
In total, 29/99 (29.3%) patients in the MOVE Principal Safety Set experienced at least one serious TEAE. Additionally, 9/99 (9.1%) patients experienced a TEAE leading to permanent discontinuation of palovarotene; of these, four experienced musculoskeletal and connective tissue disorders (all PPC), two experienced skin and subcutaneous tissue disorders (pruritus and dry skin), one experienced an infection (furuncle), one experienced malnutrition and decreased mobility, and one experienced intentional self-injury. Nine patients experienced fracture AEs, two of which were SAEs. All were deemed unrelated to treatment. There were no deaths in the trial.

Discussion
HO is a characteristic feature of FOP and leads to cumulative disability and shortened life expectancy. (11) Cross-sectional results from the FOP NHS found that the volume of HO in individuals with FOP correlates with the clinical outcome measures CAJIS and FOP-PFQ. (15) This indicates that HO substantially contributes to the decreased mobility and increased disability experienced by individuals with FOP and represents an endpoint that is clinically meaningful to patients and measurable over the relatively short-term course of a clinical trial. (15) Here, data are presented from the MOVE trial, a phase 3, multicenter, open-label study to The prespecified MOVE primary efficacy analysis used squareroot transformed data due to the high variability in HO volume observed in the NHS; a square-root transformation for the analysis of HO volume data has not, to our knowledge, been used previously or since in studies in FOP. (15,27,48) Prespecified futility criteria were met using this analysis method at IA2; however, following post hoc results obtained at IA2, Bayesian analysis of nontransformed data was reconsidered at IA3 when all patients had completed 18 months of follow-up. Non-transformed results at IA3 showed substantial efficacy for palovarotene in patients with FOP (99.4% probability of reduction in new HO volume; 60% reduction in volume of new HO), whereas square-root  transformed results did not (65.4% probability of reduction in new HO volume). These results were supported by wLME analyses; however, these post hoc results without square-root transformation may be more impacted by high variability in new HO volumes. There were two key limitations of the prespecified Bayesian primary analysis with square-root transformation, which were not present in the originally planned wLME primary analysis without square-root transformation. First, a bias was introduced through the difference in WBCT frequency between the NHS (every 12 months) and MOVE (every 6 months). Second, negative changes in annualized HO volume were not anticipated and were incompatible with both square-root transformation and the Bayesian analysis. As noted in the Results, collapsing the HO data from the first 12 months in MOVE increased the evidence for efficacy, even when using square-root transformed HO volumes. Importantly, analysis was also conducted at IA3 for the 39 patients who contributed results to both MOVE and the NHS, having transferred from the NHS to MOVE. These patients served as their own controls, avoiding confounding of data by differences between individuals in MOVE and the NHS; however, these patients were, by definition, older when in MOVE than when in the NHS. Among these 39 patients who contributed data to both studies, mean annualized new HO volume was 54% lower with palovarotene compared with their prior phase receiving no treatment beyond standard of care in the NHS (Supplemental Fig. S7).
Notably, net negative annualized new HO was reported more frequently in MOVE (29%) than in the NHS (5%). Negative new HO values may have been due to measurement variability or the impact of palovarotene on bone remodeling. However, the reasons for these differences are not fully understood, and it is unclear whether palovarotene can influence maturation of HO.
Fewer patients treated with palovarotene experienced catastrophic new HO compared with participants from the NHS. However, despite this result and the lower whole-body volumes of new HO reported in patients treated with palovarotene compared with NHS participants, no significant differences were observed at month 12 in the total number of body regions with new HO since baseline between the two cohorts. Additionally, substantial changes in functional outcomes were not observed in this trial. Cross-sectional analysis of NHS data previously indicated that CAJIS and FOP-PFQ are not as sensitive as volumetric HO via WBCT over the brief periods examined in clinical studies (15) ; CAJIS total scores increase by approximately 0.5 points per year on a scale from 0 to 30. (15,32) Data from MOVE further indicate that it may not be possible to see a functional treatment effect using available tools within this timeframe, highlighting that functional outcome studies are extremely challenging in ultra-rare diseases with highly variable progression such as FOP. (18) Therefore, in addition to the correlation between mean joint HO volume and joint CAJIS score observed in the NHS and evidence that HO can have a major clinical impact even if it does not ankylose a joint, data from MOVE further support the use of new HO volume as a clinically relevant endpoint to assess the impact of treatment on future function. (15,31,49) However, one challenge of using new HO volume as the primary endpoint is the large variability in the amount of new HO, caused by differing anatomic sites of involvement as well as different rates of progression between individuals; it is difficult to determine what constitutes a clinically meaningful reduction in new HO volume over relatively short clinical trial timescales. Both location and volume of new HO impact function, such that in some regions a small amount of HO may ankylose a joint, whereas in other regions a large amount of HO may not. Longer-term follow-up of CAJIS and joint-specific changes in HO may provide further insight.
The AE profile of palovarotene was generally consistent with other known systemic retinoids, consisting mainly of mucocutaneous and musculoskeletal AEs. (50,51) Retinoid-associated TEAEs, especially related to skin and mucus membranes, were the most commonly reported AEs, but were typically mild and managed with prophylactic and/or symptomatic therapy in addition to dose reductions. Mucocutaneous events were the most frequent events leading to palovarotene dose reduction, which occurred more frequently during flare-up treatment. Nine patients discontinued palovarotene due to TEAEs in the MOVE trial, no deaths were recorded, and no serious AEs were noted affecting the kidneys, liver, or cardiovascular system.
Retinoids are known regulators of growth plate chondrogenesis, (21,52,53) and may be associated with negative effects on linear growth and bone maturation in growing children. (37,38) Retinoids may also be associated with reduced BMD and increased risk of fracture, although previous findings have not been consistent, (54)(55)(56)(57)(58) possibly due to differences in retinoid pharmacokinetic and pharmacodynamic properties. As a result of the known class effects of retinoids, bone safety monitoring was put in place for MOVE.
Nine patients experienced fracture AEs in MOVE, two of which were SAEs and none of which were considered related to treatment. In the NHS, 7 participants (6.1%) experienced fracture events by month 12, and 9 (7.9%) by month 24.
A prior case report using quantitative computed tomography scanning has shown low BMD in a patient with FOP. (59) Here, the computed BMDs, vertebral body BMCs, and bone strengths as calculated in the post hoc biomechanical computed tomography analysis showed significant reductions at month 12 versus baseline in palovarotene-treated patients with FOP versus NHS participants. (59) However, at this time, the clinical significance of the results of these analyses is unclear, as biomechanical computed tomography and vertebral fracture analysis have not been validated in patients with FOP, who also have known skeletal deformities, including vertebral body malformations. (60,61) The results of these analyses suggest that patients with FOP may be at risk of low bone mass and increased bone fragility. However, the contributors are likely multifactorial, including glucocorticoid use for treating flares, immobility, and skeletal unloading due to HO. The clinical impact of these changes, both in patients with FOP overall and in those treated with palovarotene, is currently unclear, and further analyses are required. The potential risks to bone health need to be discussed with patients who may be considering palovarotene for treatment of their FOP.
PPC events observed with palovarotene in MOVE occurred sooner and more frequently than the expected class effect. (38,62) This identified risk of PPC led to increased frequency of bone safety radiograph monitoring in patients with <100% growth plate closure and to an FDA-issued partial clinical hold for individuals <14 years of age for all palovarotene trials (Supplemental Fig. S1). In MOVE, PPC or epiphyseal disorder was observed in 21/57 (36.8%) patients who were aged <14 years at baseline. This higher-than-expected incidence may be related to cumulative effect on patient growth plates, use of radiologic instead of symptomatic monitoring, or a more directed effect on RARγ in the growth plate versus other systemic retinoids. No sex-dependent differences were noted in PPC incidence. In patients who received only chronic palovarotene treatment, observed PPC events typically occurred between 12 and 18 months of treatment. All but one of the PPC SAEs in MOVE were observed first in the knee, suggesting that PPC preferentially affects the lower extremities, consistent with previous reports on retinoids. (63) In general, bone age advanced concordantly with chronological age in individuals with or without PPC, though the slowest growth velocity category (<4 cm/year) was more common in palovarotene-treated patients than NHS participants. It was not possible to exclude an effect of palovarotene, such as accelerated progression of physiologic growth plate closure, in older children.
Across the palovarotene FOP clinical program (including three phase 2 trials and MOVE), as of July 1, 2020, PPC has been reported in 24/102 (23.5%) patients aged <18 years, of whom 14 (58.3%) were aged <8 (females)/<10 (males) years; the mean age of patients who developed PPC was 8.0 years (all ages reported at time of enrollment). These data indicate that palovarotene is not suitable for use in very young children (<8 years of age for girls and <10 years of age for boys; based on skeletal maturity data, girls and boys achieve approximately 80% of their adult height at 8 and 10 years of age, respectively), (47) and that extremely careful consideration of the benefits and risks should be made when evaluating treatment of any growing child.
The MOVE trial design utilized detailed NHS data rather than a placebo arm. This was possible because the participants in the NHS were representative of the worldwide population of patients with FOP, comprising approximately 13% of all known patients. (15) To ensure comparisons between MOVE and NHS data were not confounded by unbalanced characteristics or methodology, equivalent image acquisition and assessment protocols were used for the primary outcome. There was also consistency between studies in the inclusion and exclusion criteria, standard of care, and background therapy used. (15) Furthermore, seven study centers were shared by MOVE and the NHS, and participants from the NHS were eligible for enrollment in MOVE. The 39 patients who transitioned from the NHS to MOVE help reinforce the validity of comparing NHS and MOVE data. However, the lack of a placebo arm in MOVE limited interpretation of safety data and determination of whether AEs occurred due to palovarotene treatment or disease progression.
Overall, results from IA3 of the phase 3 MOVE study showed substantial efficacy of palovarotene, as measured by reduction in new HO volume by WBCT compared with NHS participants using post hoc analytical methods. In line with expectations given the short time period, clear impacts on functional outcomes and quality of life were not observed in these interim results. Palovarotene was generally well tolerated, and the most common AEs were retinoid-associated, with the notable exceptions of the risk of PPC in a large subset of younger patients and potential risks of decreased BMD and vertebral fractures. Given the potential risk of long bone growth inhibition, particularly in very young children (<8 years of age for girls and <10 years of age for boys), extremely careful consideration of the benefits and risks should be made when deciding whether to treat growing children with palovarotene. Additionally, further monitoring for longer-term AEs with palovarotene treatment is required and is ongoing. (64) In conclusion, post hoc analyses showed evidence for efficacy in the reduction of new HO in FOP, although there was a high risk of PPC in skeletally immature patients.