Wednesday, July 9, 2014

Validation of the fatigue severity scale in chronic hepatitis C

Validation of the fatigue severity scale in chronic hepatitis C

The complete text article is available as a provisional PDF. The fully formatted PDF and HTML versions are in production.

Kathleen Rosa, Min Fu, Leen Gilles, Karin Cerri, Monika Peeters, Jeffrey Bubb and Jane Scott

Health and Quality of Life Outcomes 2014, 12:90 doi:10.1186/1477-7525-12-90
Published: 11 June 2014

Abstract (provisional)
Fatigue is a common symptom of chronic hepatitis C virus (cHCV) infection and a common side effect of interferon-based treatment for cHCV. This study provides confirmatory evidence of the reliability and validity of the Fatigue Severity Scale (FSS) to document fatigue in cHCV research and identifies values that indicate clinically important differences in FSS to aid in interpreting fatigue in cHCV clinical trials.

The study used data from two double-blind, randomized, placebo-controlled, Phase IIb trials evaluating the efficacy and safety of simeprevir plus peginterferon-alpha/ribavirin in treatment-naive (PILLAR, n = 386) and treatment-experienced patients (ASPIRE, n = 462) with cHCV infection. Patients completed the FSS and EuroQoL 5 dimension questionnaire (EQ-5D) at baseline and at regular intervals throughout both trials. Reliability was assessed using Cronbach's coefficient alpha at Week 24 (internal consistency reliability) and intraclass correlation (ICC) between FSS at Weeks 12 and 24 in stable patients (<0.5 g/dL hemoglobin [Hb] change between Weeks 12/24). Correlation with the EQ-5D visual analog scale (VAS) and Usual Activity domain score was used to assess concurrent validity. Clinical validity was evaluated using a case-control method to link spontaneously reported fatigue and anemia adverse events (AEs) during the study to FSS scores.

FSS total scores demonstrated good reliability (Cronbach's alpha: 0.95, 0.96; ICC: 0.74, 0.86 for PILLAR and ASPIRE, respectively) and concurrent validity (correlation with EQ-5D VAS: -0.63, -0.66) with a monotonic relationship between the EQ-5D "usual activities" item response and FSS. Clinical validity was confirmed by a significant difference between cases and controls for fatigue AEs (p < 0.05); however, anemia defined by AE or Hb abnormalities was only weakly related to FSS score. Analyses indicate that a change of 0.33-0.82 in mean FSS scores represents a meaningful improvement in fatigue, and a one-point change is a conservative indicator of an important change in individual FSS scores.

A difference of >=0.7 in mean FSS scores can be considered a clinically important difference within groups over time or between groups. A 1-point change or less in individual FSS scores indicates a clinically relevant change in fatigue.

Abbreviations AE, Adverse event; cHCV, Chronic hepatitis C virus; EOS, End of study; EOT, End of treatment; EQ-5D, EuroQoL 5 dimension questionnaire; EuroQoL, European Quality of Life; FSS, Fatigue Severity Scale; Hb, Hemoglobin; ICC, Intraclass corre lation; ITT, Intent to treat; MID, Minimal important difference; PegIFN- α , Peginterferon- α ; PRO, Patient-reported outcome; RBV, Ribavirin; RGT, Response-guided therapy; SD, Standard deviation; SE M, Standard error of measurement; SVR, Sustained virologic response; SVR24, SVR rate 24 weeks post treatment; VAS, Visual analog scale 

Discussion And Conclusion Only
View Full Text, Here
This analysis used data from two randomized, double-blind, placebo-control led studies conducted in treatment-naïve and -experienced patients with Chronic hepatitis C virus (cHCV) infection to reconfirm that the Fatigue Severity Scale (FSS) is a valid, reliable, and responsive measure of fatigue for use in HCV treatment trials. An important and unique aspect of this study was the identification of the degree of change in the FSS that can be considered clinically meaningful.

At baseline, a floor effect was observed for 6 of the 7 FSS items . Although the PILLAR and ASPIRE studies enrolled patients who potentially experienced some fatigue at baseline as a result of cHCV infection, much of the fatigue observed in these trials was attributable to side effects of treatment.

Therefore, it was not unreasonable for large floor effects, indicative of low levels of fatigue, to exist at baseline. Large floor effects would not interfere with the measurement goal of evaluating increased fatigue but could make it difficult to evaluate improvements in fatigue, potentially attenuating any observed benefit of SVR on fatigue.

Floor-ceiling effects for the FSS total score were minimal. Internal consistency reliability of the FSS was very good. The item level deletion results suggested that the first FSS item (“my motivation is lower when I am fatigued”) did not add consistent information and could potentially be removed from the scale f or future work. This finding was not unexpected given that the wording of item 1 is different from that of t he other items, being more general rather than directly evaluating the patient’s current state of fatigue. Test-retest reliability surpassed the prespecified criteria of 0.70 for both studies. As predicted, the FSS had moderate negative correlations with the EQ-5D VAS and showed a monotonic relationship with the EQ-5D “usual activities” item, with results consistent between the two studies.

Results from clinical validity analyses relating the FSS to hemoglobin levels were mixed. In general, total FSS scores were similar in patients with normal or mild hemoglobin abnormalities. This is consistent with current descriptions of mild anemia as symptom free [27,28]. Patients with moderate or high (at least Grade-2) hemoglobin levels differed from patients without significantly reduced hemoglobin levels, with a highly significant effect in the PILLAR study and a moderate, non-significant effect in the ASPIRE study.

Reports of anemia Adverse event (AEs) were only weakly related to FSS. There w as a non-significant trend towards a greater change in total FSS score from baseline to t he recall period for the reported event for patients with anemia AEs compared with controls in both studies. In contrast, there was a strong association between reports of fatigue AEs and FSS, with change in total FSS score from baseline to the recall period for the event significant ly higher for cases versus controls in both studies.

The FSS also had a strong relation with SVR in both studies, with pa tients achieving SVR at study end having significantly lower FSS total scores than patient s without SVR. Our findings are consistent with the recommendations provided by Sarkar et al. in their report of the fatigue results from the Virahep-C study [2].

Based on a simple Visual analog scale (VAS) measure of fatigue collected before during and after treatment for HCV, the authors reported that the presence and severity of fatigue ultimately declined in patients with sustained clearance of HCV. They commented that their findings indicated that HCV therapy can lead to significant and sustained improvement in clinical symptoms, and that the measurement of fatigue using VAS is successful in capturing these changes, with improvements in fatigue being most convincing in those patients with moderate to severe levels of fatigue at baseline.

The authors concluded that those patients with relatively nonsignificant biochemical or histologic disease, but with troublesome symptoms such as fatigue, should be considered for antiviral therapy [2]. The responsiveness analysis indicated that the FSS score measure s change where change in fatigue related concepts exists; however, responsiveness to changes in anemia produced weak results.

While we expected a relationship between self-reported fatigue and anemia AEs, other causes for fatigue have been suggested in the literature, in cluding increased HCV viremia in the brain and central nervous system [29]. However, at the time of writing no studies have definitively linked central nervous system levels of HCV viremia with patient- reported fatigue. A multivariate approach could potentially uncover the interactive relationship of anemia as a treatment-emergent AE and SVR in predicting fatigue, and may be a fruitful area of future research.

In addition to understanding whether a PRO instrument can detect change , it is also essential to understand the meaning associated with the change. The point at which a change score becomes important has been characterized as the MID. The MID is often misunderstood to represent the point difference that must be present between treatment groups in order for a difference to be considered clinically meaningful. Instead, the MID should be considered to be a within-group phenomenon [30] and any between-group effect size should actually be smaller than the MID [31].

During the design of a study, clinical knowledge of the indication and the expected effect of both treatment and control should guide est mates of effect size. For example, one might expect a very small effect between a treatment and an active comparator, and a larger effect between a placebo and treatment.

The accurate estimation of sample size to test a certain effect size takes into account the precision of the endpoint. On this basis, the MID should be used to guide decisions about change within a group that can be considered meaningful. A clinically meaningful between-groups effect size is not an artifact of the instrument in use, but instead should be determined based upon clinical and statistical considerations. As indicated in the FDA’s PRO Guidance [32], these methods are not appropriate for determining the precise point at which a score change indicates clinical significance but are useful for evaluating the meaning of an observed score change.

The MID or the meaning associated with changes in FSS scores was assessed using distributional estimates. Results from both studies suggest that an interpretable and meaningful improvement in fatigue occurs when there is an observed-group mean change in FSS total score of between 0.33 and 0.82.

Although the planned anchor-based analyses could not be performed because of insufficient sample with significantly suppressed hemoglobin levels, additional information on the importance of change scores can be gained through evaluation of the analyses involving the EQ-5D and fatigue AE events as previously discussed. Considered together, the results support a one-point change in an individual’s FSS score as a reasonable and appropriate responder definition.

These results have been confirmed in a pooled analysis of three simeprevir phase III clinical trials [33]. Together these results extend our understanding of the utility and interpretation of the FSS for research in patients with cHCV. The earlier study by Kleinmanet al. used screening and baseline visits to evaluate reliability and convergent validity of the FSS with other outcomes [18].

Our study adds to this by showing the link between FSS scores and clinicians’ ratings for fatigue and anemia, but also hemoglobin abnormalities and SVR. The current study provides guides that can be used to evaluate group means or individual scores that have been confirmed using data from over 800 patients, including treatment-naïve and treatment- experienced patients, during treatment for cHCV.

Future studies using the FSS can better estimate the statistical power required for evaluating fatigue effects given these findings. Our findings suggest that the FSS may also be used to evaluate whether fatigue levels warrant initiation of HCV treatment and to ensure treatment-emergent fatigue is detected and managed successfully.

These analyses indicate the FSS is a valid and reliable PRO tool appropriate for use in clinical trials involving patients with HCV infection. The current results support a responder definition of one point on the FSS score; however, cumulative response distributions at critical time points are recommended to aid interpretation of parametric results.

Further work is required to understand the relationship between fatigue, anemia, and hemoglobin abnormalities, particularly in treatment-experienced HCV patient s undergoing retreatment; this will likely require inclusion of a large patient population and a multivariate approach to separate out disease effects, positive response to treatment, and adverse effects of treatment.

Download provisional PDF.

No comments:

Post a Comment