Discriminative ability of quality of life measures in multiple sclerosis

Comorbidity is increasingly recognized to be common in multiple sclerosis (MS), including physical and psychiatric comorbidity [1]; over 40% of the MS population may suffer from comorbidities depending on the age and sex of the population [2]. Some studies suggest that the presence of certain comorbidities is associated with delayed diagnosis, accelerated disability progression, and an increased risk of gadolinium-enhancing lesions on MRI [3, 4].

Several studies also report that physical and psychiatric comorbidity adversely affect health-related quality of life (HRQOL), suggesting that improved treatment of comorbidity may be one means of improving HRQOL in MS [5‐7]. However, prior studies of the association between comorbidity and HRQOL differ with respect to the instruments used to assess HRQOL [8‐11]. Instruments used in prior comorbidity studies in MS include the Short Form (SF)-12 [12], SF-36 [9], the Multiple Sclerosis Quality of Life-54 (MSQOL-54) [8] and the Health Utilities Index (HUI) [13], among others.

Previous research suggests that the HUI-III performs better than other health utility measures in discriminating between persons with MS with differing degrees of disability, being able to distinguish mild, moderate, severe, and very severely disabled persons [14]. It is unknown which instrument is most sensitive to the effect of comorbidity on HRQOL. This is relevant to future studies evaluating the impact of interventions directed at mitigating the effects of comorbidity on MS, and to studies aimed at evaluating the effects of comorbidity on MS. We aimed to compare the discriminative ability of HRQOL instruments, including the MSQOL-54, SF-36, SF-6D, and the HUI-III, in a prevalent MS population with and without comorbidity. We also examined whether the included measures performed similarly in persons with and without disability. Specifically, we hypothesized that: (1) generic measures would discriminate better between those with and without comorbidities than disease-specific measures; and (2) the HUI-III would discriminate best between persons with differing severities of disability.

Ultimately, 868 participants enrolled in this study, of whom 563 (64.9%) provided complete data for the HRQOL measures, disability, and comorbidity (Fig. 1) and were the subject of further analyses (Table 1). Their characteristics were similar to those expected for a prevalent MS population [30]. The distribution of comorbidities was as follows: none (n = 197, 35.0%); one (n = 163, 29.0%); two (n = 107, 19.0%), and three or more (n = 96, 17.1%) comorbidities. Physical comorbidities were present in 54.4% (n = 306) while psychiatric comorbidities were present in 29.1% (n = 164). The most common comorbidities were depression, hypertension, and hyperlipidemia (Table 2).

The mean (SD) score was 38.4 (12.3) for the PCS-36, 48.5 (11.9) for the MCS-36, 0.56 (0.32) for the HUI-III, 0.68 (0.13) for the SF-6D, 57.4 (22.2) for the PCS-54 and 63.2 (19.1) for the MCS-54.

In this study, we evaluated the discriminative ability of multiple HRQOL measures to distinguish between persons with and without comorbidities in MS. We also examined the ability of these HRQOL scales to discriminate between varying levels of patient self-reported disability. The examined scales were only able to distinguish between persons with and without some comorbidities, particularly depression, and differed in their ability to distinguish individuals at higher levels of disability. The MCS-36 was not a good tool to distinguish between disability levels or number of comorbidities. The PCS-54 was slightly better than the generic tools at distinguishing between differing numbers of comorbidities, but it was comparable to the generic tools for levels of disability. In general, the MCS measures performed worse than the PCS measures for both the SF-36 and MSQOL-54, with the MCS-54 outperforming the MCS-36 for both comorbidity and disability discrimination. Of the utility measures, the SF-6D and HUI-III performed comparably for discriminating comorbidity and disability statuses.

In our study no measure was able to distinguish between those with no comorbidities and those with only one comorbidity, but the PCS-54 was very good at distinguishing between persons with different numbers of comorbidities, and the PCS-36, MCS-36, SF-6D, HUI-III, and MCS-54 were also adequate. All HRQOL scales were very good at distinguishing persons with depression. Some HRQOL scales were also good at distinguishing those with hyperlipidemia and anxiety disorders. Overall the PCS-54 was slightly better at distinguishing between differing numbers of comorbidities than the generic tools. Of the utility measures, the SF-6D and HUI-III performed comparably for discriminating comorbidity statuses, although the SF-6D distinguished between persons with and without any psychiatric comorbidity and between persons with and without migraine; whereas the HUI-III did not. It is possible to achieve more precision with the PCS-54 compared with the PCS-36, SF-6D or the HUI-III but this comes at the cost of additional questions (54 vs. 36, 11, 16). However, the length of the HRQOL tool does not appear to affect a subject’s perception of the tool’s utility [26]. Prior work has not identified clear differences in reliability and validity between disease-specific and general instruments when used cross-sectionally [26].

HRQOL scores were generally lowest in those persons with severe disability in the present study consistent with previous work [14]. There was a dose-response relationship between disability level and scores on all HRQOL measures, except the MSC-36. A prior study examining the utility of generic HRQOL measures across varying levels of MS-disability found that the HUI-III performed best to discriminate between all levels of disability [14]. As correlations between the MCS-54 and MCS-36 were very high, there may not be additional value in the extra 18 items on the MS-specific questionnaire (especially considering the amount of missing items on the optional sexual functioning subscale). We used a modified MSQOL-54 scoring algorithm to account for a large amount of missing data on the sexual function subscale. Missing data on this subscale is not uncommon in MS research [26], and other studies have employed modified algorithms to account for this issue [31]. We found the HUI-III was best able to discriminate between higher levels of disability (moderate and severe) as compared to the other generic and disease-specific measures. This is especially relevant in the context of progressive MS research and trials, where appropriate patient reported outcome measures must be carefully selected to ensure adequate sensitivity to change [32].

We examined generic and disease-specific health profile and utility measures of HRQOL in a large, well-described sample of persons with MS [33]. We explored the use of these tools across disability levels and comorbidity status, which has not been previously reported in the MS population. All tools except the SF-36 and MSQOL-54 are utility scales designed to put the effects of diverse illnesses onto a single scale; this is valuable when assessing patients with comorbidities. Limitations of the present study include the online nature of the survey, which may bias toward the inclusion of participants who are more educated, and with a higher income [33]. As noted previously, participants in this study were more likely to have a university education and to live in a neighborhood with a higher median income, compared to a reference MS population. The administration of the scales was not entirely independent, though the SF-36 was designed to be administered in this manner. This study was cross-sectional and therefore we did not examine the use of these tools over time; future studies should explore the responsiveness of these tools to change and individual response shift. Due to the online nature of the survey it was not possible for respondents to skip questions; as such scales were completed in their entirety or not at all, which may have resulted in more non-response on the HRQOL questionnaires. Whereas this study provided a broad-based comparison of two types of instruments, future studies seeking to compare specific instruments should consider adopting a priori criteria for assessing instrument performance.

Our findings suggest that future studies aimed at evaluating the effects of comorbidity or of comorbidity treatment on HRQOL in MS, may reasonably choose to employ generic or disease-specific measures depending on the MS population of interest. Disease-specific measures would discriminate better between those with and without comorbidities, and between those with differing numbers of comorbidities than generic measures. In contrast, the HUI-III would discriminate best between persons with moderate and severe disability. In studies in which the participants enrolled may have high levels of disability, the HUI-III appears best able to discriminate differences between participants, suggesting this may be a good measure for trials in progressive MS, although further evaluation is needed.