Sage Journals: Discover world-class research

Abstract

To propose minimal important differences (MID) for the Migraine-Specific Quality of Life Questionnaire version 2.1 (MSQ v2.1). To our knowledge (to date), no published MID values exist for the MSQ v2.1 in any population. Analyses were performed on data from two pivotal clinical trials of topiramate for migraine prevention (n = 916), as well as from the QualityMetric National Headache Survey (n = 1016). Analyses included both distribution- and anchor- based MID techniques as well as group- and individual-level MID values. Group-level anchor-based MID values ranged from 3.2 [Role Restrictive domain (RR)] to 7.5 [Emotional Functioning domain (EF)], setting the minimum level of appropriate MID (which can also aid with power analysis). Individual-level distribution-based MID values resulted in highly similar estimates from two large databases: median MID of 8.5 for RR, 9.2 for Role Preventive (RP) and 12.0 for EF. Finally, individual-level anchor-based MID values ranged from 5.0 (RR and RP domains) to 10.6 (EF). For group-level purposes of calculating power for future studies, an MID of 3.2, 4.6 and 7.5 for RR, RP and EF is recommended. For within-group analyses for analysing clinical trial efficacy of each patient's change with responder analyses, 5 points is necessary for RR. For RP and EF, ranges are recommended: 5.0 to 7.9 for RP and 8.0 to 10.6 for EF. These latter two domains tend to have more error in the MID, and thus a sensitivity analysis with both ends of the range should be used to confirm significant differences in responder analyses.

Keywords

MSQ migraine prophylaxis MID MCID

Introduction

The concept of minimal important difference (MID) has received increasing attention in clinical trials research. A popular definition of MID is ‘the smallest difference in score in the domain of interest which patients perceive as beneficial and which would mandate, in the absence of troublesome side effects and excessive cost, a change in the patient's (health care) management’(1). Other investigators have modified this definition to include deterioration in quality of life (2).

MID provides interpretability in the evaluation of treatment effectiveness, linking clinical indicators with patient-reported outcomes (PROs) in health-related quality of life (HRQoL) measures. The draft Food and Drug Administration (FDA) guidance on PROs (3) has an entire section dedicated to the use of MID for PRO instruments in clinical trials. Indeed, the draft guidance proposed that responsiveness and interpretation guidelines (i.e. MID) must be demonstrated and documented for particular PRO instruments and study populations (3).

Given the importance of MID as a meaningful interpretation guideline, we have undertaken the current task of examining the MID of the Migraine-Specific Quality of Life Questionnaire (MSQ). The MSQ has been the most frequently used migraine-specific instrument in existing research on HRQoL. It has been demonstrated to have good psychometric properties in specific patient populations (4), including migraine prophylaxis patients (5).

Despite the MSQ's wide use in migraine studies, we were not able to find any research on its MID assessment. Our search yielded only one study that estimated MID for any patient-reported measure of migraine impact. The Headache Impact Test, another frequently used instrument in headache and migraine studies, was examined by Coeytaux et al. (6). According to Coeytaux, a change score of 2.3 units in Headache Impact Test reflected meaningful clinical change over time among patients with chronic daily headache.

The primary goal of the MID examination is to have clear guidelines for interpreting the meaning of differences and changes in MSQ scores. As differences of a certain magnitude may be clinically meaningless if observed for a single patient but statistically significant if observed as a mean group difference (2), we followed draft guidelines by the FDA by examining MID in the context of both mean group differences and individual change differences (i.e. responder criteria).

MID can be established with distributional and anchor-based techniques. Distribution-based methods examine the underlying distribution of results in order to garner understanding of the minimal effect needed for change. Investigators may infer effects in terms of between-person standard deviation (s.d.) units, within-person s.d. units and the standard error of measurement (s.e.m.) (2). Whereas distributional MID approaches provide ease of calculation and moderate generalization, they can be difficult to interpret with the absence of other clinical markers (2). Advantages of the anchor-based approach include reduced ambiguity of interpretations (uses patient's perspective of disease improvement, physician's perspective and associated clinical markers, and accepted population benchmarks), and the process emphasizes differences between trivial and ‘small but meaningful change’. However, different anchors can yield different MID values.

The purpose of the present study was to provide estimates of MID, with the goal of facilitating the interpretability of commonly used HRQoL indicators in migraine treatment and prevention, the three domains of the MSQ v2.1: Role Restriction (RR), Role Prevention (RP) and Emotional Function (EF). Both distribution- and anchor-based techniques were employed in order to provide breadth and robustness to the MID results. Additionally, two different datasets were employed to capture both clinical trial participants (who forego medication prior to trial commencement—wash-out—and thereby have more migraine-related impairment) and participants from a longitudinal national survey on migraines (for whom no wash-out period was undertaken given that no intervention was conducted as part of the trial).

Methods

Population characteristics

The analyses were performed on two datasets: one database was the pooled data from two topiramate clinical trials (7, 8), and another database was the QualityMetric National Headache Survey (NHS) (9). The two trials of topiramate for migraine prevention were randomized, double-blind, placebo-controlled clinical trials of identical design, one conducted in the USA and Canada (7) and the other in the US only (8). This combined sample of 916 patients were aged 12–65 years, had a minimum 6-month history of migraine (International Headache Society criteria) and experienced 3–12 migraines per month but no more than 15 headache days a month during the 28-day prospective baseline period. Patients were randomized to placebo or topiramate 50, 100 or 200 mg/day, and titrated over 8 weeks. The maintenance therapy continued for 18 weeks.

The QualityMetric NHS data comprised responses from 1016 adults residing in the contiguous 48 states of the USA (9). Participants were required to: be 18–65 years old, inclusive; reside permanently in the household called; have had a headache at least once in the past 4 weeks (prior to interview); be able to converse in English; and be physically and mentally competent to participate in the interview. Phone interviews were conducted by an independent research organization with trained telephone operators. It should be noted that these participants were not subjected to any particular intervention as a part of the QualityMetric NHS study.

Statistical analyses

Missing data handling

To minimize the impact of missing data on results and power, a Bayesian multivariate imputation method was used to impute values for any missing value (10, 11). Proper handling of missing data in MID studies can provide greater accuracy in distributional-based MID analyses (as Bayesian multivariate imputation will provide efficient and accurate standard errors compared with techniques such as casewise deletion or mean replacement) and allows for analysis of all observed data without overinflating significance in relationship for anchor-based analyses (given proper controls of significance calculations with Bayesian multivariate imputation) (10).

MID calculations: distribution techniques

Three analyses were used to examine the distribution-based MID of the MSQ. First, between-patient variability was assessed on each of the MSQ domains at baseline, based on s.d. Many studies have found that a ½ s.d. is a reasonable MID criterion (2). Second, the s.e.m. was examined. s.e.m. is often used in distribution-based MID analyses because it is sample independent (12). Therefore, we examined the s.e.m. associated with sufficient change per the recommendations of Wyrwhich et al. (12). Within-patient variability was assessed on each of the MSQ domains. Finally, Cohen's d (13, 14) was examined for changes over time (15) throughout the course of the study, as Cohen's d effect size is a measure of the differences between scores, based on their s.d. We examined the effect sizes through change scores, as is commonly conducted with MID analyses (16).

MID calculations: anchor-based techniques

Unlike many medical conditions, migraine has no gold standard for objective and observable clinical measures. Nevertheless, the anchors used in this study are similar to those that have been accepted in many clinical trials as criteria in determining MID (1, 6, 15, 17–23)—these anchors represent the decrease in frequency of an undesirable event (i.e. migraine attack) or the patient's perception that change has occurred, as measured by another scale.

Examination of MID for a migraine instrument should involve the examination of various anchors in order to determine (i) which anchors are related to changes in the MSQ scale, and (ii) the range of plausible MID values from the various related anchors. Ideally, correlations should be both significant and have an effect size near the medium range (which is at r = 0.3) (13). The within-group analyses involved examination between which MSQ change scores were related to no change on the anchor and which were related to minimal change, similar to the process used by Juniper and colleagues (21). To conduct these analyses, we used logistic regression to examine the relationship between each of the three MSQ subscales and the anchors. Regression coefficients were used to derive odds ratios (ORs) of each anchor as a function of 1-, 2-, 5- and 10-point baseline MSQ scale score differences (24). The critical assumption of linearity for such an analysis was evaluated by inspecting separate regression coefficients for each 5-point increment of the MSQ scale scores and by performing the linear-by-linear association test for linearity (25). The key for interpreting MID with the ORs is how many points the MSQ subscales need to change to move a minimal but important difference away from an OR of 1.0. Moving about 25% above the OR (or an OR of 1.25) is often considered to be a meaningful minimal movement (26).

Average monthly migraine rate

This anchor came from the topiramate clinical trials data (7, 8). The average monthly migraine rate was calculated by dividing the total number of migraine periods for each phase (e.g. baseline, or double-blind phase) by the total duration (in days) for that particular phase, times 28 days. The mean average monthly migraine rate was 4.3 (ranging from 0 to 15.4) for the baseline, and 2.8 (ranging from 0 to 21.5) for the double-blind phase. For the anchors in the topiramate databases, no universally agreed upon cut-off for minimally important reduction in frequency exists in the change in monthly migraine rate or number of severe migraines per month. Some have argued that a 30% reduction is sufficient, whereas others have suggested that no less than 50% reduction should be obtained. Therefore, we examined patient-level MID analysis with 30, 40 and 50% reduction on each of these scales.

Migraine status

Also based on the QualityMetric data, this dichotomous variable asked participants to indicate if they had ever been told by a doctor that the participant suffers from migraine headaches. A score of 1 was given for those answering yes, whereas a score of 0 was given for those answering no.

Four-grade migraine disability

The four-stage migraine disability anchor came from the QualityMetric data. This disability score was developed by Stewart et al. (27) and is a four-grade severity scale from the Migraine Disability Assessment (MIDAS) questionnaire. Based on the MIDAS score, patients were categorized into one of the following four categories: MIDAS 0–5 was minimal or infrequent disability, MIDAS 6–10 was mild or infrequent disability, MIDAS 11–20 was moderate disability, and MIDAS 21 or more was severe disability. Because 80% of the respondents in the QualityMetric data were categorized as a disability of 1 (minimal or infrequent disability), we combined the other three disability categories (2, 3 and 4) in our analysis, yielding a dichotomous variable.

Migraine frequency improvement

Also based on the QualityMetric data, this dichotomous variable asked participants to indicate if they were having more or less headaches compared with 3 months ago. A score of 1 was given for those having about the same or fewer headaches, whereas a score of 0 was given if they reported having more headaches compared with 3 months ago.

Migraine intensity improvement

Based on the QualityMetric data, this dichotomous variable asked participants if their headaches bothered them more now compared with 3 months ago. A score of 1 was given for those with the same or less bother currently, whereas a score of 0 was given if they were more bothered compared with 3 months ago.

Migraine impact improvement

A few different indicators were used to measure how reductions in migraine impact one's life. Specifically, we looked at response differences on the following questions: limited in everyday physical activities, feeling frustrated or irritable because of headaches, or limited in daily activities (including housework, work and/or social activities). Participants were also asked to assess their overall quality of life (single item). All questions were asked based on their current state. However, scores for the current study were calculated as a 1 for those with the same or less negative impact and a score of 0 was given if they had greater negative impact compared with their scores 3 months ago.

Results

Distribution-based analyses

For group-level data in the topiramate clinical trial datasets, the s.d. s for RR, RP and EF were 16.6, 19.8 and 24.8, respectively. Therefore, the ½ s.d. MID would suggest that MIDs of 8.3 points, 9.9 points and 12.4 points are required for RR, RP and EF, respectively. In the non-trial data from the QualityMetric database, s.d. s were 22.7 (½ s.d. = 11.6), 21.7 (½ s.d. = 10.8) and 28.0 (½ s.d. = 14.0) for RR, RP and EF, respectively. Although these scores are slightly higher, MSQ data from the QualityMetric database were scored on a non-standard 5-point scale (28). Thus, when we attenuate for this difference we end up with ½ s.d. s of 8.6, 8.5 and 11.5 for RR, RP and EF, respectively—similar to the scores from the topiramate clinical trials data.

Next, we examined the s.e.m. of each scale based on the topiramate clinical trials data. The s.e.m. for the three domain scales was found to be 4.8, 7.9 and 10.6 for RR, RP and EF, respectively.

Finally, within-patient differences were examined. The effect sizes for change score differences by treatment group are presented in Table 1. For RR, effect sizes ranged from 0.19 (for 50 mg) to 0.44 (for 100 mg). For RP, effect sizes ranged from 0.23 (for 50 mg) to 0.35 (for 100 mg). Finally, for EF effect sizes ranged from 0.05 (for 50 mg) to 0.30 (for 100 mg). These resulted in between-group MIDs of 3.2, 4.6 and 7.5 for RR, RP and EF, respectively.

Table 1

Between-group difference in mean Migraine-Specific Quality of Life Questionnaire (MSQ) change scores from baseline to end-point (day 183) by pair-wise comparison between treatment groups

MSQ scales	TPM 50 mg vs. placebo		TPM 100 mg vs. placebo		TPM 200 mg vs. placebo
MSQ scales	P	ES	P	ES	P	ES
RR	0.0389	0.19	< 0.0001	0.44	0.0004	0.34
RP	0.0148	0.23	0.0002	0.35	0.0090	0.25
EF	0.6090	0.05	0.0016	0.30	0.0536	0.18

TPM, Toperimate; ES, effect size; RR, Role Restrictive; RP, Role Preventive; EF, Emotional Functioning.

Anchor-based approach

Table 2 provides correlations between the RR, RP and EF scales of the MSQ. Correlations with the anchors range from small to medium effects, with most correlations achieving significance (partially because of the large power from the topiramate trials).

Table 2

Pearson correlations and descriptive statistics for all anchors

	Average monthly migraine rate (base)†	Average monthly rate of severe migraine (base)†	MIDAS 4 grades (base)‡	Fewer headaches‡	Bother less‡	Limit less‡ (physical activities)	Frustrated less‡	Limit less (daily activities)‡	Overall better QOL‡
RR-Base	−0.20∗	−0.19∗	−0.51∗	0.10	0.18∗	0.21∗	0.25∗	0.28∗	0.00
RP-Base	−0.11∗	−0.19∗	−0.49∗	0.11	0.19∗	0.23∗	0.28∗	0.28∗	0.00
EF-Base	−0.14∗	−0.08∗	−0.40∗	0.11	0.15∗	0.19∗	0.29∗	0.33∗	0.01
RR-Change	0.02	0.05	0.13∗	0.24∗	0.24∗	0.27∗	0.24∗	0.22∗	0.18∗
RP-Change	0.00	0.08∗	0.07	0.20∗	0.19∗	0.29∗	0.17∗	0.24∗	0.17∗
EF-Change	0.03	0.03	0.07	0.22∗	0.23∗	0.26∗	0.24∗	0.16∗	0.17∗
Mean	4.32	2.30
s.d.	2.32	0.42
Range	0–15.4	0.3–3	1–4	0–1	0–1	0–1	0–1	0–1	0–1

∗

P < 0.05.

†

N = 916.

‡

N = 300.

MIDAS, Migraine Disability Assessment; RR, Role Restrictive; RP, Role Preventive; EF, Emotional Functioning.

Because the MSQ used in the QualityMetric data was based on a truncated 5-point MSQ ordinal scale for each item (29), we examined how truncating the MSQ data from the topiramate trials to a 5-point scale would influence the anchor-based interpretations in order to give and adjustment for the use of the QualityMetric MSQ data. Results of these comparisons showed negligible differences.

Analysis of the relationships between the nine migraine frequency, intensity and impact improvement anchors and change scores with the MSQ subscales indicated that all three of the MSQ subscale change scores were significantly associated with two of the anchors: having fewer headaches compared with 3 months ago, and less bothered by headaches compared with 3 months ago. On the other four anchors, changes in EF were significantly related to less limitation in everyday physical activities, and RR and RP changes were significantly related to changes in overall QoL improvement.

Table 3 presents the ORs for migraine status and migraine disability based on different change scores on MSQ, controlling for age and gender. Compared with our criterion of an OR of at least 1.25, a 5-point difference on RR and RP and an 8-point difference on EF are recommended.

Table 3

ORs for anchors based on different change scores for Migraine-Specific Quality of Life Questionnaire subscales

Change	Headache-RR	Disability-RR	Headache-RP	Disability-RP	Headache-EF	Disability-EF
−1	1.05	1.07	1.04	1.06	1.03	1.03
−3	1.15	1.22	1.13	1.18	1.10	1.10
−5	1.27	1.39	1.22	1.32	1.17	1.18
−8	1.46	1.70	1.38	1.55	1.28	1.30
−10	1.60	1.94	1.50	1.73	1.36	1.39

An odds ratio (OR) of 1.25 was used to delineated the minimum value for a meaningful OR.

RR, Role Restrictive; RP, Role Preventive; EF, Emotional Functioning.

Finally, for changes on the anchors from the topiramate clinical trial databases, between- and within-group MIDs were used to examine the amount of change associated with a single s.e.m. for each of the percent change groups (30, 40 and 50). An s.e.m. of 4.85 (RR), 7.90 (RP) and 10.63 (EF) indicated marked differences in the drop in average monthly migraine [all better–same–worse (BSW) analyses were significantly different at the P < 0.0001 level; detailed tables are available from the first author]. As noted previously, between-group differences were not expected for baseline scores given that the randomization should have resulted in comparable patient groups assigned to each of the treatment arms. It was not until treatment began that differences should have occurred, thus favouring importance on the within-group differences over time.

Discussion

Results from the current study are summarized in Table 4. Between-group MIDs ranged from 3.2 (RR) to 7.5 (EF)—lower than the within-group MIDs, as is typically the case (6). Distribution-based within-group MIDs were larger for the ½ s.d. approach than for the s.e.m. approach. This is to be expected with scales that have higher internal consistency, such as the MSQ scales, as strong internal consistency shrinks the s.e.m. (30). Indeed, the largest discrepancy between the ½ s.d. approach and the s.e.m. approach was for RR, the scale with the highest internal consistency (9 points vs. 5 points). The s.e.m. approach is likely to be more accurate, as the ½ s.d. approach does not presume to know anything about the internal consistency of the measure, thereby not appropriately rewarding reliable measures with a drop in MID.

Table 4

MID summary of results, segregated by distribution- and anchor-based approaches, as well as by group and individual approaches

Scale	Analysis	Database	Distribution-based	Anchor-based
Group-level MID
RR	Effect size	T	3.2	NA
RP	Effect size	T	4.6	NA
EF	Effect size	T	7.5	NA
Individual-level MID
RR	½s.d.	T & QM	8.3 (T), 8.6 (QM)	—
RP	½ s.d.	T & QM	9.9 (T), 8.5 (QM)	—
EF	½ s.d.	T & QM	12.4 (T), 11.5 (QM)	—
RR	s.e.m.	T	4.8	—
RP	s.e.m.	T	87.9	—
EF	s.e.m.	T	10.6	—
RR	Logistic	Q	—	5.0
RP	Logistic	Q	—	5.0
EF	Logistic	Q	—	8.0
RR	BSW	T	—	4.9
RP	BSW	T	—	7.9
EF	BSW	T	—	10.6

MID, minimal important difference; RR, Role Restrictive; RP, Role Preventive; EF, Emotional Functioning; BSW, Better–Same–Worse analysis.

Anchor-based approaches for the within-group designs were rather consistent between the two techniques employed. For RR, a 5-point difference is suggested by both logistic regression and BSW analyses. RP obtained MIDs of 5 and 8 for logistic and BSW analyses, respectively. Although there is no compelling rationale to select one analysis over the other, the difference is expected: clinical trial data tend to show larger necessary MIDs than general population data given the large gains frequently experienced during a clinical trial because of careful inclusion and exclusion criteria. For now, it may be best to examine both of these values with clinical trial data, and some research in this area suggests that very little difference should be found when examining the efficacy of clinical trial data using small differences on the MID. Finally, EF was found to have an 8- and 10-point MID using logistic regression and BSW analyses, respectively. Again, these differences are in the expected direction, and examining both values may be prudent for the time being.

No prior studies have examined the MID for the MSQ in any migraine population, either undergoing acute treatment or treated with prophylaxis. Therefore, the results of this study are novel. Although MIDs have been derived for another migraine instrument (6) and for other pain instruments (e.g. (31, 32)), MIDs are quite specific to the measure for which they were derived (2) and comparisons across instruments are not advised. Results from this study will be helpful in guiding future studies: the within-group MIDs can be used to examine a significant treatment effect by conducting a responder analysis (33), and between-group differences can help guide power analyses for future studies (6).

Conclusions

The results of various distribution- and anchor-based approaches suggest the following MIDs are appropriate. For group-level purposes of calculating power for future studies, an MID of 3.2, 4.6 and 7.5 for RR, RP and EF, respectively, is recommended. For within-group responder analyses, MID should be 5 points for RR, between 5 and 8 points for RP, and between 8 and 10 points for EF.

Acknowledgement

This work was supported by Ortho-McNeil Janssen Scientific Affairs, LLC.

References

Jaeschke

Singer

Guyatt

. Measurement of health status: ascertaining the minimally clinically important difference. Control Clin Trials 1989; 10:407–15.

Guyatt

Osoba

Wyrwhich

Norman

Clinical Significance Consensus Meeting Group . Methods to explain the clinical significance of health status measures. Mayo Clin Proc 2002; 77:371–83.

Food and Drug Administration—Department of Health and Human Services. Guidance for industry—patient-reported outcome measures: used in medical product development to support labeling claims (draft guidance). Rockville, MD: Food and Drug Administration 2006. 2006 Contract No.: Document Number.

Martin

Pathak

Sharfman

Adelman

Taylor

Kwong

Validity and reliability of the Migraine-Specific Quality of Life Questionnaire (MSQ Version 2.1). Headache 2000; 40:204–15.

Cole

Lin

Rupnow

MFT

. Psychometric efficacy of the Migraine Specific Quality of Life Questionnaire Version 2.1 (MSQ v2.1) for migraineurs on prophylaxis. Qual Life Res 2007; 16:1231–7.

Coeytaux

Kaufman

Chao

Mann

DeVellis

. Four methods of estimating the minimal important difference score were compared to establish a clinically significant change in Headache Impact Test. J Clin Epidemiol 2006; 59:374–80.

Brandes

Saper

Diamond

Couche

Lewis

Schmitt

Topiramate for migraine prevention: a randomized controlled trial. JAMA 2004; 291:965–73.

Silberstein

Neto

Schmitt

Jacobs

for the MIGR-001 Study Group . Topiramate in migraine prevention: results of a large controlled trial. Arch Neurol 2004; 61:490–5.

Ware

Jr Kosinski

Bjorner

. Practical implications of item response theory and computerized adaptive testing: a brief summary of ongoing studies of widely used headache impact scales. Med Care 2000; 38:1173–82.

10.

Cole

. How to deal with missing data: conceptual overview and details for implementing two modern methods. In: Osborne

, ed. Best practices in quantitative methods. Thousand Oaks, CA: Sage 2007:214–38.

11.

Rubin

Schenker

. Multiple imputation in health-care databases: an overview and some applications. Stat Med 1991; 10:585–98.

12.

Wyrwhich

Tierney

Wolinsky

. Further evidence supporting an SEM-based criterion for identifying meaningful intra-individual changes in health-related quality of life. J Clin Epidemiol 1999; 52:861–73.

13.

Cohen

. Statistical power analysis for the behavioral sciences, 2nd edn. Hillsdale, NJ: Lawrence Erlbaum 1988.

14.

Cohen

. A power primer. Psychol Bull 1992; 112:155–9.

15.

Wyrwhich

Nienbar

Tierney

Wolinsky

. Linking clinical relevance and statistical significance in evaluating intraindividual changes in health-related quality of life. Med Care 1999; 37:469–78.

16.

Brozek

Guyatt

Schunemann

. How a well-grounded minimal important difference can enhance transparency of labelling claims and improve interpretation of a patient reported outcome measure. Health Qual Life Outcomes 2006; 4:69–76.

17.

Beaton

Hogg-Johnson

Bombardier

. Evaluating changes in health status: reliability and responsiveness of five generic health status measures in workers with musculoskeletal disorders. J Clin Epidemiol 1997; 50:79–93.

18.

Deyo

Inui

. Toward clinical applications of health status measures: sensitivity of scales to clinically important changes. Health Serv Res 1984; 19:275–89.

19.

Guyatt

Deyo

Charlson

Levine

Mitchell

. Responsiveness and validity in health status measurement: a clarification. J Clin Epidemiol 1989; 42:403–8.

20.

Hägg

Fritzell

Nordwall

. The clinical importance of changes in outcome scores after treatment for chronic low back pain. Eur Spine J 2003; 12:12–40.

21.

Juniper

Guyatt

Willan

Griffith

. Determining a minimal important change in a disease-specific quality of life questionnaire. J Clin Epidemiol 1994; 47:81–7.

22.

Osoba

Rodrigues

Myles

Zee

Pater

. Interpreting the significance of changes in health-related quality of life scores. J Clin Oncol 1998; 16:139–44.

23.

Wyrwhich

Wolinsky

. Identifying meaningful intra-individual change standards for health-related quality of life measures. J Eval Clin Pract 2000; 6:39–49.

24.

Bjorner

Wallenstein

Martin

Lin

Blaisdell

Piech

Interpreting the magnitude of score differences in the SF-36 vitality scale: clinical and social significance. Med Care (in press).

25.

Tabachnick

Fidell

. Using multivariate statistics, 5th edn. New York: Pearson Education 2007.

26.

Kosinski

Zhao

Osterhaus

Ware

Jr . Determining minimally important changes in generic and disease-specific health-related quality of life questionnaires of rheumatoid arthritis. Arthritis Rheum 2000; 43:1478–87.

27.

Stewart

Lipton

Kolodner

Liberman

Sawyer

. Reliability of the migraine disability assessment score in a population-based sample of headache sufferers. Cephalalgia 1999; 19:107–14.

28.

Bjorner

Kosinski

Ware

Jr . Calibration of an item pool for assessing the burden of headaches: an application of item response theory to the Headache Impact Test (HIT (TM)). Qual Life Res 2003; 12:913–33.

29.

Bjorner

Kosinski

Ware

Jr . The feasibility of applying item response theory to measures of migraine impact: a re-analysis of three clinical trials. Qual Life Res 2003; 12:887–902.

30.

Anastasi

Urbina

. Psychological testing, 7th edn. Upper Saddle River, NJ: Prentice Hall 1998.

31.

Farrar

Portenoy

Berlin

Kinman

Strom

. Defining the clinical important difference in pain outcome measures. Pain 2000; 88:287–94.

32.

Kosinski

Zhao

Dedhiya

Osterhaus

Ware

Jr . Determining minimally important changes in generic and disease-specific health-related quality of life questionnaires in clinical trials of rheumatoid arthritis. Arthritis Rheum 2000; 43:1478–87.

33.

Osoba

. Health-related quality of life outcomes in clinical trials. In: Fayers

Hays

, eds. Assessing quality of life in clinical trials, 2nd edn. New York: Oxford University Press 2005:259–74.

Minimal Important Differences in the Migraine-Specific Quality of Life Questionnaire (MSQ) Version 2.1

Abstract

Keywords

Introduction

Methods

Population characteristics

Statistical analyses

Missing data handling

MID calculations: distribution techniques

MID calculations: anchor-based techniques

Average monthly migraine rate

Migraine status

Four-grade migraine disability

Migraine frequency improvement

Migraine intensity improvement

Migraine impact improvement

Results

Distribution-based analyses

Anchor-based approach

Discussion

Conclusions

Acknowledgement

References