Health-related quality of life and a cost-utility simulation of adults in the UK with osteogenesis imperfecta, X-linked hypophosphatemia and fibrous dysplasia

This study used cross-sectional data from an ongoing UK-based multi-centre prospective cohort study: RUDY (Rare and Undiagnosed Diseases Study). RUDY is a novel web-based registry and patient-driven research platform designed to improve the understanding of all aspects of rare musculoskeletal diseases [13]. Ethical approval was obtained from the UK’s South Central Research Ethics Committee (LREC 14/SC/0126).

Potential participants were informed of the study and invited to visit the study website (www.​rudystudy.​org) during clinic visits at participating sites, through relevant patient groups, Facebook, Twitter and the web searches [13]. Potential participants registered on the website and provided basic demographic information. Inclusion criteria included individuals aged 18 and above with a clinical diagnosis of OI, FD or XLH. Once registered, eligible individuals were contacted via email to arrange a telephone consent appointment. Once informed consent was received, participants were granted full access to a secure personalised homepage on the study website. Demographic information collected via the study website included age, sex and patient-reported comorbidities. A recent clinic letter was requested from the participant’s doctor to verify their diagnosis. Participants were invited to complete the EQ-5D-5 L questionnaire [14] electronically via their study website homepage or using the paper version which was sent via post. Non-responders were sent reminders via email and social media during the data collection period. Participants were excluded if they had not fully completed the EQ-5D-5 L questionnaire by the final day of the data collection period. Data was collected from September 2014 to March 2016.

The EQ-5D-5 L is a widely used generic preference-based tool developed by the Euroqol group for measuring health-related quality of life [14]. The EQ-5D-5 L divides health into five dimensions (mobility, self-care, usual activity, pain/discomfort and anxiety/depression), each assessed with a 5-response-level questions [15]. Health states can then be weighted using country-specific ‘value sets’, which reflect the preferences of the general population. This results in a health utility score for each 3125 health states on a scale ranging from −0.28 (negative scores denoting health states worse than death) through 0 (dead) to 1 (full health). The recently published value set for England [16] was used in this study as a value set for the whole United Kingdom remains in development and most participants came from England.

The EQ-5D-5 L also includes a visual analogue scale (VAS) from 0 (worst health) to 100 (best health), which participants use to self-rate their health status on the day of completing the questionnaire.

Descriptive statistics were used to present participants’ baseline characteristics and EQ-5D-5 L responses. Health states were converted into health utility scores using the England value set using all five response levels [16]. Mean and standard deviation of the VAS score and health utility score were calculated for each disease group. The one-way analysis of variance (ANOVA) test was used to assess statistical significant differences between mean scores for each disease group. To aid interpretation, response levels for each dimension were divided into three categories: ‘no and slight problems’ (level 1 and 2), ‘moderate; (level 3) and, ‘severe and extreme problems’ (level 4 and 5). Significant differences between the three disease groups for the proportion of participants’ responses in each category were assessed using the Fisher’s exact test. Statistical significance was set as p ≤ 0.05.

For the economic simulation, we chose OI as the case study because it reported the largest number of individuals. We split the group into tertiles to identify those individuals reporting the lowest health utility scores and therefore the greater potential improvement (Additional file 1: Figure S1). For our simulation, we considered a hypothetical treatment that would be applied to those in the lower tertile of health utility scores over a period of 10 years. Patients under treatment would attain 75% of their individual potential improvement (the difference between their reported score and 0.745, the mean of the middle tertile). We conducted sensitivity analysis on the percentage improvement producing results for 55%, 65%, 75%, 85% and 95%.

The treatment was considered to have a constant nominal cost per year. Where total costs are reported, they are discounted at the standard rate of 3.5% per year, as are health utilities according to NICE’s guidance for technology assessments [17]. Full treatment effect was assumed to be achieved linearly within the first 12 months. We further assumed no effect on mortality in either arm as the simulation was run for 10 years only. Health utility was however decreased every year by the coefficient found associated with age for all OI patients (−0.005 per year) in a linear regression on reported health utility. Quality-adjusted life years were estimated for each individual in each arm and added over the 10 year period.

All data analysis was carried out using STATA IC version 14.

A total of 174 participants were recruited to the study. 109 participants fully completed the EQ-5D-5 L questionnaire (response rate 63%). The baseline characteristics of respondents are shown in Table 1 and there were no statistically significant difference between baseline characteristics of the responders and non-responders (p > 0.05) for age, gender and number of comorbidities. Amongst those diagnosed with OI and using Sillence’s classification, 18 participants had type I, five had type III, five had type IV and 15 reported an unknown type.

Response level frequencies for each domain by disease is shown in Table 2. The response groups by no/slight vs. moderate vs. severe/extreme are shown in Fig. 1. Pain/discomfort was the most problematic domain for participants within all three diseases. Overall 94% of respondents reported problems with pain or discomfort. Thirty one percent of adults with FD reported severe/extreme problems with pain, with a quarter of those with XLH and 16% of those with OI reporting similar levels of problem.

Mobility also caused problems to a significant proportion of respondents. Overall 73% reported some level of problem with mobility and severe/extreme problems were reported by 21% of total respondents. FD respondents compared with the other condition groups reported mobility as least problematic, 43% reporting no problems compared with 19% in the OI group and 13% in the XLH group.

Self-care was least problematic overall for participants (82% reported no or slight problems). Less than 10% of all respondents reported severe or extreme problems with self-care.

Overall the majority of respondents (68%) reported problems with usual activity to varying extent. However most reported slight problems (30% OI, 31% FD, 42% XLH) and the percentage of respondents reporting problems decreased as the severity of the response option increased.

Low levels of severe or extreme anxiety and depression were reported in all three groups (less than 10% in each group). Overall most respondents reported no and slight problems in this domain (78%). The highest percentage of FD respondents reported moderate problems (24%) compared to the other groups.

There was no statistically significant difference, using the Fisher’s exact test, in responses between disease groups for all dimensions of the EQ-5 L-5D (Mobility p = 0.775, Self-care p = 0.322, Usual activities p = 1.000, Pain/discomfort p = 0.627, Anxiety/depression p = 0.100). Among OI respondents linear regression showed a statistically significant positive correlation between age and response level (p = 0.01, Pearson’s r =0.39) in the Usual Activities domain, meaning that older age is associated to more problems in performing daily activities. For all other domains no significant correlation between age and response levels was found across all three disease groups.

The mean health utility scores generated using the England value set [16] were high for all disease groups (OI 0.656, SD 0.283; FD 0.656, SD 0.288; XLH 0.648, SD 0.290; Overall 0.654, SD 0.284), with wide distributions (Fig. 2). For all three conditions the distribution of health utility was bimodal, this pattern was most evident for XLH. No statistically significant difference between the three groups for the health utility score was found (p = 0.993).

The reported VAS scores were 69.4 (SD 21.4) for OI, 64.1 (SD 23.0) for FD and 60.8 (SD 26.9) for XLH adult patients. For all diseases the distribution was skewed to high VAS scores (Fig. 3). The graph shows a higher proportion of respondents with OI reporting high VAS scores compared with the other diseases. A statistically significant negative correlation between VAS score and age was found in the OI group using linear regression (p = 0.005 Pearson’s r = −0.42), showing that older age is associated with worse perception of self-rated health.

For our base case, where 75% of the individual potential improvement would be achieved during the first 12 months, the simulated treatment group would have accumulated 78 QALYs over the following 10 years (mean of 5.2 per person). The no-treatment group would have accrued 41 QALYs in total (2.8 per person), leaving the simulated treatment with an associated impact of +2.5 QALYs gained per person during the 10-year period. Given the small sample size, the mean QALY gain reported a high standard deviation (1.66), thus producing a wide confidence interval (−0.78 to 5.72). Figure 4 shows the progression over time of the simulated mean health utility index for each group. The QALY gain is related to the individual potential improvement at one year. Expected QALY gain per person over a 10 year period for other potential improvements were +1.83 (55% improvement during the first year), +2.15 (65%), +2.79 (85%) and +3.11 (95%).

Based on these results, we calculated the corresponding maximum annual discounted cost that a health care system would be willing to pay for the treatment at various thresholds of cost-effectiveness over the 10 years (Table 3).

For example, a health care system willing to pay up to £50,000 for each additional QALY gained by an intervention for OI patients, would be willing to pay £14,355 annually for such treatment achieving 75% of the potential improvement during the first year. After the appropriate discounting has been applied, this annual cost amounts to £123,561 over the 10 years of treatment which, expected to improve patient outcomes by an average 2.47 QALYs over the period, would equate to an incremental cost-effectiveness ratio of £50,000 per QALY. Higher treatment costs would not be considered value for money for a health care system with this willingness to pay, whereas a lower cost would be likely to represent value for money for this health care system.