1 INTRODUCTION

Hemophilia A is characterized by a deficiency of clotting factor (F) VIII. The hallmark of this lifelong bleeding disorder, especially in people with moderate/severe hemophilia, is recurrent bleeding into index joints (ankles, knees and elbows) that may lead to progressive, irreversible joint destruction (arthropathy), impaired health-related quality of life, and the requirement of more intensive treatment.1 This intra-articular bleeding can precipitate a biological cascade with adverse effects that may persist despite the clearance of blood from the joint.2

Primary prophylaxis is the preventive administration of clotting factor on a regular basis, before the onset of joint damage in people with hemophilia.3 Primary prophylaxis is recommended as the standard of care for the treatment of boys with moderate/severe hemophilia and a severe bleeding phenotype by the World Federation of Hemophilia (WFH), with initiation ideally before age 3 years.4

Dose- and frequency-escalated primary prophylaxis is an escalation regimen in which young boys with moderate/severe hemophilia A initially receive weekly intravenous factor VIII (FVIII) infusions, with an increase in dose/infusion frequency based on clinically significant breakthrough bleeding into joints.5-7 This management approach is less costly than standard full-dose primary prophylaxis which, for hemophilia A, involves intravenous infusion of a standard half-life FVIII concentrate, ideally on alternate days, a minimum of 3 times per week.8 Dose/frequency-escalation programs, such as the Dutch intermediate prophylaxis regimen, aim to reduce the number of intravenous FVIII infusions at a young age, when vascular access can be very difficult.7

Evaluation of joint structure and function on any prophylaxis regimen is essential to determine its effectiveness. Magnetic resonance imaging (MRI) is the most sensitive imaging test because of its capacity to detect both early soft-tissue changes and osteochondral abnormalities.9, 10

Most reports of long-term imaging studies compare cross-sectional joint outcomes, often comparing various joint measures, such as physical examination and/or measures of joint function, with MRI. As a result, the clinical relevance of joint changes on MRI is still not fully understood, but it is considered the gold standard, as it is the best tool available to assess early soft-tissue and osteochondral abnormalities.11 Few studies have reported interval disease progression with serial MRI over a prolonged time interval or the potential of specific MRI findings to predict subsequent joint progression.

This study describes end-of-study and interval MRI changes in boys enrolled in the dose- and frequency-escalated Canadian Hemophilia Prophylaxis Study (CHPS), and assesses whether specific MRI findings and self-reported joint bleeding episodes may be predictive of subsequent joint deterioration.

2 MATERIALS AND METHODS 2.1 Study design and participants

CHPS was a longitudinal, single-arm study with 11 participating hemophilia treatment centers (HTCs) located across Canada. Boys with severe hemophilia A, between the ages of 1 and 2.5 years, were consecutively enrolled over a 10-year period from 1997 to 2007. Data were collected until December 31, 2012, or the end-of-study MRI assessments, whichever was later.

Detailed descriptions of the study design have been previously published.5, 6 In brief, boys were eligible for enrollment in the study if they had normal index joints (ankles, knees, and elbows) as determined by plain radiography and joint assessment based on physical examination. Exclusion criteria included a history of three or more bleeds into any index joint, present or past history of a circulating inhibitor to FVIII (level ≥0.6 BU), and competing risk factors such as hepatitis C.

The study was approved by the research ethics boards at all participating sites. Parents or guardians gave written informed consent.

2.2 Procedures

Participants attended study visits every 3 months for the first 5 years, and thereafter every 6 months. All boys were treated with dose- and frequency-escalated primary prophylaxis according to an a priori approved protocol based on bleeding criteria,5, 6 using standard half-life recombinant FVIII. The prophylaxis regimen started with once weekly infusions of 50 IU/kg of FVIII (step 1), and escalated to twice weekly infusions of 30 IU/kg of FVIII (step 2) and three times/week or alternate-day infusions of 25 IU/kg of FVIII (step 3).

Bleeding and treatments were recorded by parents, guardians, or the participants themselves, in diaries and confirmed by study personnel at each study visit.

2.3 Imaging acquisition

Standard joint view radiographs (X-rays) and MRI studies of the six index joints were planned at two time points during the study: ages 6 and 12 years (±2 years), referred to as interval and end-of-study time points, respectively. Due to constraints on access to research MRI machines, several subjects were unable to have their imaging studies within this period; a number of images were acquired outside of the planned window. This also accounted for some missing data at both time points. To minimize this, some participants were flown to the central site (The Hospital for Sick Children, Toronto) for MRI studies when possible. Repeat X-rays were not required if participants had clinical X-rays within the previous 3 years.

MRI acquisition was performed on 1.5 T MRI scanners in all participating HTCs and included T2* gradient echo images (repetition time, 600 milliseconds; echo time, 20 ms; flip angle, 20°; bandwidth, 15.63; matrix, 256×192; number of excitations, 2; average field-of-view, 12 cm [to be adjusted according to the subject’s joint size; slice thickness, 4 mm; gap, 0 mm]). The acquisition included coronal and sagittal planes for ankles and knees and axial, sagittal, and coronal planes for elbows. No contrast material was used, except for one subject who had intravenous administration of gadolinium for another clinical indication. Elbows were imaged separately with surface coils; both knees and both ankles were imaged simultaneously with extremity or head coils, respectively, according to joint size. The images from all participating HTCs were collected in a TeraRecon database at the coordinating center.

2.4 Image interpretation

Plain X-rays were scored using the Pettersson score (range, 0-13 per index joint where 0 represents no evident joint damage and 13 the maximum possible score).12 MRIs were scored using the International Prophylaxis Study Group (IPSG) MRI score, a scale that allocates 9 points for the soft-tissue domain (effusion/hemarthrosis, synovial hypertrophy, and hemosiderin) and 8 points for the osteochondral domain (surface erosions, subchondral cysts, and cartilage degradation). A score of 0 represents no joint changes and 17 the maximum possible score for an individual index joint.13

X-rays and MRIs were independently read by two experienced radiologists with >10 and 20 years of experience (JS and PB), blinded to subject identity and clinical data, and unblinded to the order of the MRI examinations—as previously recommended for paired readings in chronologic order.14 A tutorial calibration session between the readers was conducted before the individual review of the MRI examinations.14 Discrepant readings were discussed and both radiologists agreed upon a consensus score.

2.5 Clinical information

Descriptive clinical information included the number and location of index joint hemorrhages, before study entry and up to the time of the MRI examination, age at study entry, the number of days on study, and age at time of the MRI studies.

2.6 Outcome measures and statistical analysis

The primary outcome measure for this study was evidence of end-of-study osteochondral changes on MRI. Secondary outcomes for this analysis were evidence of index-joint bleeds on MRI, and the predictive ability of bleeding and interval MRI findings on subsequent joint deterioration.

The association between the number of prior lifetime index joint bleeds and clinical information (age at time of MRI, age at start of prophylaxis) was investigated with Spearman correlation coefficients (rs), with random intercepts for subjects to account for within-subject clustering. Strength of the correlations were interpreted according to the following definitions: ≤0.40 indicated poor, >0.40 to ≤0.6 moderate, >0.60 to ≤0.80 strong, and >0.80 excellent agreement/correlation.15

Descriptive statistics were used (median, interquartile range, and range of values) to characterize the outcome measures at each time point.

The relationships between bleeding and change in MRI scores (i.e., the difference in score between the interval and end-of-study MRI), and interval MRI scores and end-of-study MRI and Pettersson scores were determined using a generalized linear mixed model with random intercepts for subjects to account for within-subject clustering. Results are expressed as odds ratios and 95% confidence intervals were estimated using standard errors.

The interreader reliability of interpretation of the IPSG MRI scale in this study was tested using intraclass correlation coefficients (ICCs).16, 17 ICC and r ≤ 0.40 indicated poor, >0.40 and ≤0.60 moderate, >0.60 and ≤0.80 substantial, and >0.80 excellent agreement/correlation.15, 17

Unlike in previous reports,6 for those subjects who enrolled in the study toward the end of the recruitment period and thus had only one imaging study, we considered that their end-of-study imaging, regardless of their age. This resulted in some subjects being outside the a priori defined age range for their end-of-study imaging assessments.

We conducted intent-to-treat analyses, which included all available data, including participants who were lost to follow-up. We censored subjects at withdrawal and lost to follow-up. All analyses were performed using R version 3.5.2 (R Foundation for Statistical Computing, Vienna, Austria).18

3 RESULTS 3.1 Subjects

Fifty-six boys with severe hemophilia A were followed in CHPS for a median of 10.2 (range, 0.2-16.1) years; 6 subjects were lost to follow-up (including the subject followed for 0.2 years). Forty-six of the 50 (92%) remaining subjects, followed for a median of 9.6 years (range, 4.8-16 years) had end-of-study MRI assessments. A few subjects had one or more missing joint examinations due to time constraints or personal preference, yielding a total of 89 ankles, 90 elbows, and 91 knees for analysis, representing 98% of potentially available joints. The median biologic age of the 46 boys at the time of the end-of-study imaging assessments was 11.4 years (range, 6.2-18.5 years). A timeline of study events is shown in Figure 1. Twenty-seven of the 46 subjects (59%) had interval and end-of-study MRIs. Clinical characteristics of the study cohort are presented in Table 1. Of the subjects who had at least one MRI, at the end of study, 2 of 46 (4%) remained on treatment step 1, 17 of 46 (37%) had escalated to step 2, and 27 of 46 (59%) had escalated to step 3. During the first 5 years of the study, there were 53 factor adjustments in 36 subjects, with 18 adjustments occurring in the last 5 years of the study. In the time between the interval and end-of-study MRI, 11 of 27 (41%) of subjects had dosing adjustments, including four from step 1 to step 2, one from step 1 to step 3, and six from step 2 to step 3.

Flow diagram of key study events: enrollment, interval magnetic resonance imaging (MRI), and end-of-study MRI. Each milestone includes the number of patients (N), and their median (range) age and time on study in years. The time period over which enrollment occurred is also indicated

TABLE 1. Clinical characteristics of study subjects with end-of-study MRI examinations (n = 46) showing a breakdown of the number of joints evaluated that had either 0 or at least 1 reported joint bleed at the time of the MRI and a summary of the lifetime number of reported index joint bleeds Joint Lifetime number of joints with bleeds at end-of-study MRI Lifetime number of joint bleeds at end-of-study MRI, median (range of values) 0 bleeds ≥1 bleed(s) Per joint (%)* Per joint (%)* Ankles 28/89 (31) 61/89 (69) 2 (1–17) Elbows 49/90 (54) 41/90 (46) 1 (1–46) Knees 40/91 (44) 51/91 (56) 2 (1–12) Total (all joints) 117/270 (43) 153/270 (57) 2 (1–46) * Number of cases/total number of joints evaluated. Abbreviation: MRI, magnetic resonance imaging. 3.2 End-of-study findings

Of the 46 subjects who completed the end-of-study MRI, 25 (54%) had detectable soft-tissue changes in at least one index joint, indicated by a score on the IPSG 17-point MRI scale of >0 in at least one item in the soft-tissue domain (Table 2). Soft tissue changes were detected in the ankles of 18 (39%) subjects, in the elbows of 14 (30%) subjects, and in the knees of 4 (9%) subjects. Of the 25 subjects with soft-tissue findings, 17 (68%) had evidence of soft-tissue changes in more than one joint. Synovial hypertrophy and hemosiderin deposition were noted in 25 of 46 (54%) of subjects, with effusion or hemarthrosis seen in 6 of 46 (13%).

TABLE 2. End-of-study frequency of soft-tissue findings per subject detected by MRI assessed with the International Prophylaxis Study Group 17-point MRI scale Types of MRI findings Soft-tissue domain item Severity

Number of affected individuals (n = 46)*

n (%)

Number of affected joints: n (%) Ankles Elbows Knees

(n = 89)+

n (%)

(n = 90)+

n (%)

(n = 91)+

n (%)

Effusion/hemarthrosis Any effusion/hemarthrosis 6 (13) Mild 3 (3) 3 (3) 3 (3) Moderate 0 (0) 0 (0) 0 (0) Severe 0 (0) 0 (0) 0 (0) Synovial hypertrophy Any synovial hypertrophy 25 (54) Mild 18 (20) 12 (13) 1 (1) Moderate 6 (7) 5 (6) 2 (2) Severe 3 (3) 4 (4) 0 (0) Hemosiderin Any hemosiderin 25 (54) Mild 17 (19) 12 (13) 1 (1) Moderate 7 (8) 5 (6) 2 (2) Severe 3 (3) 4 (4) 0 (0) Abbreviation: MRI, magnetic resonance imaging. * Number of individuals studied. + Number of joints studied.

Of the 46 subjects who completed the end-of-study MRI, 18 (39%) had detectable osteochondral changes in at least one index joint, indicated by a score on the IPSG 17-point MRI scale of >0 in at least one item in the osteochondral domain (Table 3). Osteochondral changes were detected in the ankles of 10 (22%) subjects, in the elbows of 10 (22%) subjects, and in the knees of 2 (4%) subjects. Of the 18 subjects with osteochondral changes, 8 (44%) had evidence of changes in more than one joint, for a total of 26 joints with osteochondral changes. Surface erosions were seen in 14 of 46 (30%) subjects, subchondral cysts in 10 of 46 (22%) and cartilage loss in 18 of 46 (39%). Of the 26 joints with osteochondral changes, there was no reported prior hemarthrosis in 3 (12%).

TABLE 3. End-of-study frequency of osteochondral changes per subject detected by MRI assessed with the International Prophylaxis Study Group 17-point MRI scale Types of MRI findings Osteochondral domain item Sub-item

Number of affected individuals (n = 46)*

n (%)

Number of affected joints: n (%) Ankles Elbows Knees

(n = 89)+

n (%)

(n = 90)+

n (%)

(n = 91)+

n (%)

Subchondral bone or joint margins, n (%) Any surface erosion 14 (30) 8 (9) 9 (10) 1 (1) Half or more of the articular surface eroded in at least one bone 1 (2) 0 (0) 1 (1) 0 (0) At least one subchondral cyst 10 (22) 6 (7) 6 (7) 1 (1) Subchondral cysts in at least two bones, or cystic changes involving a third or more of the articular surface in at least one bone 5 (11) 2 (2) 4 (4) 0 (0) Cartilage loss, n (%) Any loss of joint cartilage height 18 (39) 11 (12) 12 (13) 2 (2) Loss of half or more of the total volume of joint cartilage in at least one bone 5 (11) 3 (3) 2 (2) 1 (1) Full-thickness loss of joint cartilage in at least some area of at least one bone 8 (17) 4 (4) 6 (7) 0 (0) Full-thickness loss of joint cartilage including at least one half of the joint surface in at least one bone 2 (4) 0 (0) 2 (2) 0 (0) Abbreviation: MRI, magnetic resonance imaging. * Number of individuals studied. + Number of joints studied.

At the end of study X-rays, osteochondral changes (Pettersson score >0) were present in 24/259 (9.3%) joints in 17/46 (37%) subjects. Twenty (83.3%) of the 24 joints also had corresponding osteochondral changes in MRI; the remaining 4 of 24 (16.7%) did not. Conversely, of the 26 joints with MRI osteochondral changes, 6 (23.1%) did not have damage on X-rays. The median (range) end-of-study Pettersson score for all index joints was 0 (0-10 for ankles, 0-9 for elbows, and 0-5 for knees). There was an excellent correlation between the Pettersson scores and the osteochondral subscores of the IPSG MRI scale (r = 0.91; P < .0001).

3.3 MRI and bleeds

The MRI scores and Pettersson scores are compared with the total number of reported index joint bleeds in Figure 2A-C. Some joints had abnormal MRI/X-ray scores without reported bleeds, while other joints had normal MRI/X-ray scores despite many reported bleeds. Osteochondral MRI subscores, total MRI scores, and Pettersson scores had a moderate correlation with the number of reported index joint bleeds in the elbows (r = 0.48, P < .0001; r = 0.49, P < .0001; and r = 0.53, P < .0001, respectively). In ankles, weak correlations were noted between total number of study joint bleeding episodes and osteochondral MRI subscores, total MRI scores, and Pettersson scores (r = 0.30, P = .003; r = 0.34, P = .001; and r = 0.25, P = .01, respectively). The knees did not show a correlation between osteochondral changes, total MRI scores or Pettersson scores and bleeding episodes.

(A) Descriptive summary of the end-of-study International Prophylaxis Study Group (IPSG) 17-point magnetic resonance imaging (MRI) total score, out of a possible 17 points, for individual index joints according to the number of self-reported index joint bleeds. (B) Descriptive summary of the end-of-study IPSG 17-point MRI soft-tissue and osteochondral subtotal scores, out of a possible 9 and 8 points, respectively, for individual index joints according to the number of self-reported index joint bleeds. (C) Descriptive summary of the end-of-study X-ray Pettersson scores for individual index joints according to the number of self-reported joint bleeds. The size of each circle corresponds to the number of cases, with the larger circles representing more than one joint. The largest number of index joints had no self-reported joint bleeds and an MRI or X-ray score of 0

3.4 Serial imaging findings

The median (range) time between the interval and end-of-study MRI for the 27 subjects with serial images was 5.8 (2.2-8.5) years.

A change in the MRI score between the two sets of imaging studies of at least 1 point was observed in 68 of 154 (44%) of index joints. Table 4 shows a breakdown of MRI items according to stability, improvement, or worsening over time. Of note, worsening of osteochondral subscores (cartilage thickness, surface erosions, and subchondral cysts) was detected in 16 index joints (7 elbows and 9 ankles in 13 subjects) with improvement in 4 index joints (4 ankles in 4 subjects); comparable figures for soft-tissue changes (effusion/hemarthrosis, synovial hypertrophy, and hemosiderin) were worsening in 25 joints (8 elbows and 17 ankles in 16 subjects) and improvement in 39 joints (13 elbows and 26 ankles in 21 subjects). Examples of subjects with serial deterioration and improvement on MRI are shown in Figures 3 and 4, respectively.

TABLE 4. Summary of interval changes in the17-point MRI scores for the subset of subjects with serial MRIs between mid- and end-of-study images (A) Improved Worsened No change

Left (N = 26)

n (%)

Right (N = 27)

n (%)

Left (N = 26)

n (%)

Right (N = 27)

n (%)

Left (N = 26)

n (%)

Right (N = 27)

n (%)

Effusion hemarthrosis 4 (15.4) 7 (25.9) 1 (3.8) 1 (3.8) 21 (80.8) 19 (70.4) Synovial hypertrophy 5 (19.2) 5 (18.5) 10 (38.5) 7 (25.9) 11 (42.3) 15 (55.6) Hemosiderin 12 (46.2) 10 (37.0) 7 (26.9) 5 (19.2) 7 (26.9) 12 (44.4) Soft-tissue subtotal 13 (50.0) 13 (48.1) 10 (38.5) 7 (25.9) 3 (11.5) 7 (25.9) Surface erosions 0 2 (7.4) 3 (11.5) 1 (3.8) 23 (88.5) 24 (88.9) Subchondral cysts 1 (3.8) 2 (7.4) 3 (11.5) 1 (3.8) 22 (84.6) 24 (88.9) Cartilage degradation 1 (3.8) 3 (11.1) 5 (19.2) 3 (11.5) 20 (76.9) 21 (77.8) Osteochondral subtotal 1 (3.8) 3 (11.1) 5 (19.2) 4 (15.4) 20 (76.9) 20 (74.1) Total score 11 (42.3) 13 (48.1) 11 (42.3) 8 (29.6) 4 (15.4) 6 (22.2) (B) Improved Worsened No change Left (N = 25) Right (N = 26) Left (N = 25) Right (N = 26) Left (N = 25) Right (N = 26) Effusion hemarthrosis 4 (16.0) 1 (3.8) 2 (8.0) 0 19 (76.0) 25 (96.2) Synovial hypertrophy 1 (4.0) 1 (3.8) 3 (12.0) 5 (19.2) 21 (84.0) 20 (76.9) Hemosiderin 6 (24.0) 3 (11.5) 3 (12.0) 4 (15.4) 16 (64.0) 19 (73.1) Soft tissue subtotal 9 (36.0) 4 (15.4) 3 (12.0) 5 (19.2) 13 (52.0) 17 (65.4) Surface erosions 0 0 4 (16.0) 3 (11.5) 21 (84.0) 23 (88.5) Subchondral cysts 0 1 (3.8) 3 (12.0) 1 (3.8) 22 (88.0) 24 (92.3) Cartilage degradation 0 0 4 (16.0) 2 (7.7) 21 (84.0) 24 (92.3) Osteochondral subtotal 0 0 5 (20.0) 2 (7.7) 20 (80.0) 24 (92.3) Total score 7 (28.0) 3 (11.5) 5 (20.0) 6 (23.1)