The impact of severe haemophilia and the presence of target joints on health-related quality-of-life

CHESS is a retrospective, non-interventional study of severe, inherited haemophilia A and B across five European countries (France, Germany, Italy, Spain, and the UK). The purpose of the study was to generate an annualised economic burden of the condition through reporting of 12-month, haemophilia-related direct and indirect resource use [16]. One hundred and thirty-nine haematologists provided demographic and clinical information for 1285 adult (≥18 years) patients, 541 (42%) of whom completed a corresponding questionnaire covering out-of-pocket medical expenditure, HRQOL (measured using the EQ-5D-3 L) and work loss.

Study exclusion criteria was limited to patients diagnosed with an inhibitor at the time of study capture (n = 52), due to a differing risk profile for bleeds and subsequent target joint development among these patients [12, 17].

The EQ-5D is one of the most frequently used generic tools for assessing HRQOL [18]. The questionnaire consists of two sections, the first of which (health state) consists of five dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression), with responses rated on an ordinal (1–3) scale. For each of the five dimensions, the respondent indicates the statement best describing their ‘health state’ at that time, with 1 equivalent to ‘no issues’ and 3 to ‘severe issues’. A health state index ‘utility’ score based on country-specific ‘value sets’ is derived through an amalgam of the five responses, with scores generally ranging from 0 (equivalent to ‘dead’) to 1 (‘perfect health’), though scores of less than zero (states ‘worse than dead’) can also be derived [19].

The second part of the EQ-5D is a visual analogue scale (VAS). The respondent is asked to indicate how good or bad their health is on that day, based on a scale from 0 to 100, where zero represents ‘the worst health’ the respondent can imagine, and 100 ‘the best health state’ the person can imagine. Corresponding country population norms for both the EQ-5D index score and the VAS were used in this analysis, with the exception of Italy, for which the Spanish value set was used.

A ‘target joint’ as defined in the CHESS study encompasses any joint with known chronic synovitis; in contrast to previous clinical studies [20], study investigators were given discretion as to how this may be further defined with respect to bleed frequency and period of observation. In order to explore the differential impact of costs associated with lower and upper body joint deterioration, target joints were categorised into two groups based on their location. ‘Upper body’ target joints were those in the shoulders, elbows, wrists, neck, and spine; ‘lower body’ target joints consisted of hips, knees, and ankles. The target joint variable was assessed in three ways: as a binary 0/1 variable; as a binary 0/1 variable split into upper and lower body joints; and as a discrete variable.

Demographic data and EQ-5D index score and VAS responses were compared between the sample of patients with no reported target joints and those with one or more reported target joints. Means were used to describe continuous variables; categorical variables are described as frequencies and proportions. Standard t-tests were conducted in order to test for between-group differences.

The marginal effect of the presence of one or more target joints on EQ-5D-3 L index score was assessed using a generalized linear model (GLM). Health utility data on a ‘poor’-‘good’ scale is often negatively skewed with a large volume of unity values (i.e. perfect HRQOL) and a long ‘tail’ from a select group of less healthy ‘outlier’ patients. The GLM is an extension of the linear regression framework (Eq. 1) suitable for nonparametric outcome variables [21, 22]. The GLM requires a link function relating the conditional mean to the covariates, and a distribution ‘family’ to specify the relationship between the variance and the mean [23]. The log-link function (Eq. 2) in combination with a Poisson distribution (Eq. 3) is frequently used to model EQ-5D index scores and was employed for this analysis [24]. Index scores were transformed to a ‘disutility’ index score (nY = 1-Y) in order to fit the model specification [25, 26]. A confirmatory analysis of the family and link functions was conducted using the modified Park test [21, 22].

Where i = 1, …, n

Additional model covariates were country of residence, patient age, and physician-reported presence of mental illness (anxiety and/or depression) via a stepwise inclusion method. Results are presented as average mean effects (AME). All statistical analysis was conducted using Stata 13 [27].

The average age of the patient cohort was 37.7 years old, with the majority of patients (62.9%) aged between 31 and 40 years (Table 1). Almost two-thirds (61.0%) of patients were receiving therapy via a prophylaxis regimen at the time of the study.

A total of 714 target joints were recorded across the study population (mean 1.39; SD 1.44; range 0–9) (Fig. 1). Three hundred and fifty-eight patients (69.5%) were reported diagnosed with one or more target joints, with the majority (79%) diagnosed with one or two target joints. The majority (52.5%) of patients had target joints exclusively in the lower body.

The mean EQ-5D index score in the sample was 0.77 (SD 0.27) (Table 2). Age was found to have a negative impact on HRQOL: index scores were found to decrease as patients progressed into each 10-year age cohort. Patients from Germany had the highest index score (mean 0.90; SD 0.12) and the United Kingdom the lowest scores (mean 0.59, SD 0.37). Patients receiving prophylaxis had lower mean index scores compared to on-demand (mean 0.80 versus 0.75).

Patients with no recorded target joints had significantly higher utilities than those with one or more target joints (mean 0.87 versus 0.73), with scores deteriorating as the number of target joints increased (Fig. 2). Index scores among patients with an upper body target joint was broadly similar to those with a lower body target joint (mean 0.77 versus 0.73). Patients with both an upper and lower body target joint had lower index scores versus those with target joints in one location (upper or lower body) (mean index score 0.68, SD 0.31) (Fig. 3).

Patients with no target joints reported the highest VAS scores (mean 74.3, SD 0.9) (Table 3), while patients with both an upper and lower body target joint reported the lowest VAS scores (mean 64.4, SD 18.3). VAS scores among patients with an upper body target joint were identical to those with a lower body target joint (mean 67.9). Mean reported VAS scores followed a downward trend as the number of target joints increased (Fig. 4).

Across all five dimensions of the EQ-5D-3 L, the majority of patients reported no problems, with fewer than one in ten patients reporting ‘extreme’ problems in any dimension (Table 4). Across the cohort as a whole, as well as within the cohort of patients with target joints, pain/discomfort and anxiety/depression were the most common dimensions in which ‘extreme’ problems were reported. The cohort of patients with one or more target joints consistently reported a higher frequency of extreme problems compared to those with no target joints: whether that be in mobility (0.0% versus 1.4%); self-care (0.6% versus 2.0%); usual activities (0.6% versus 2.8%); pain/discomfort (1.3% versus 4.7%); and anxiety/depression (3.2% versus 5.6%).

The presence of one or more target joints was associated with lower EQ-5D index scores (AME -0.120; SD 0.026; p < 0.001) (Table 5). Patients in Germany or Italy with one or more target joints had the highest index scores (AME 0.110 and 0.121 respectively; p < 0.001); patients in the UK were found to have the lowest mean index scores (AME -0.154, SD 0.060). The AME of patient age on index scores was minimal yet significant (mean 0.005; SD 0.001; p < 0.001). Anxiety/depression was found to be associated with lower index scores (AME -0.106; SD 0.001; p < 0.001).