Interpreting Changes in Quality of Life in Atrial Fibrillation
Among the 214 enrolled patients, 196 had no change or improved on the Patient Global Change form, and 189 had no change or improved on the Physician Global Change form. Baseline demographic and clinical characteristics of the 210 patients with Patient Global Change data are presented in Table I. Most of the patients were males with paroxysmal AF and moderate symptoms. Patients scheduled for ablation tended to have lower baseline AFEQT scores (51.2 ± 20.2) than those with no change (69.4 ± 22) or planned pharmacological change (58.7 ± 21) in treatment (P < .01).
Anchor-based Approach.Table II shows the mean improvement in AFEQT scores using the Patient Global Change forms as the anchor. Patients who rated themselves as having "no or unimportant improvement" (0 and ±1) had a mean increase in AFEQT score of 6.9 ± 17 units, whereas those who rated themselves as having had "small improvements" had a mean change in AFEQT score of 20.9 ± 18.2 points. The change in AFEQT scores for those with a moderate improvement was 18.9 ± 21 units.
We also calculated the change in AFEQT score for patients who worsened. There was only 1 patient in each of "moderately" and "markedly" worse by global change; there were 12 patients with "slightly worse" (small decrease) global score, and they had a small decrease in the AFEQT score (–2.2 ± 12.3). The difference in AFEQT score change between patients with a slightly worse global score and "unimportant changes" in global score was 9.1 AFEQT unit, similar to the difference between the "slightly better" (small increase) in global score and unimportant changes group of 14 units.
The correlation between change in AFSS and change in AFEQT was – 0.72 (P < .0001). Correlations for the association between AFEQT ratings and global patient transition ratings at month 3 were also calculated. Baseline AFEQT score was not significantly correlated with the Patient Global Change assessment with r = – 0.11 (P = .13), indicating that there does not need to be an adjustment of change in AFEQT score for different baseline scores. The baseline AFEQT score also negatively correlated with the change in AFEQT score (r = – 0.52, P < .01), so that for each unit of higher baseline AFEQT score, the average change in AFEQT score was 0.5 units smaller. This suggests that there was a ceiling effect so that subjects with a higher AFEQT baseline score did not have as much potential to improve compared with those that may have a lower baseline AFEQT score. Month 3 AFEQT versus Patient Global Change r = 0.35 (P < .01) and change in AFEQT versus Patient Global Change r = 0.46 (P < .01) (Figure), suggesting that the global change assessment may be a better anchor for change in AFEQT but did not explain all variances.
(Enlarge Image)
Figure.
Correlation between Patient Global Change and change in AFEQT.
The patients and physicians both reported the identical category of change in global AFEQT for individual patients in less than half of cases (n = 101, 48%). In 51 cases, the physician global rating of change indicated more improvement than the patient rating, and in only 18 cases did the physician rating indicate less improvement than the patient rating.
The Physicians' Global ratings of change were not able to distinguish those with small clinical improvement from those with no clinical change, as assessed by the similarity in mean 3-month changes in AFEQT scores (Table III). However, the patients deemed by their physicians to have improved by a moderate amount had a mean change in AFEQT scores of 21.3 ± 20 points.
Table IV describes the differences in mean scores by category of change assessed by the AFSS. Of the 210 enrolled patients, there were 172 patients who did not deteriorate on the AFSS. Those who had an unimportant improvement in AFSS had a mean increase in AFEQT score of 5.8 ± 13 units. Those who had a small improvements in the AFSS (3- to 5-point decrease ) had an increase in AFEQT scores of 12.0 ± 15.4 units, and those with a moderate improvements in symptoms (6- to 8-point decrease in AFSS) had a mean increase in AFEQT scores of 17.9 ± 11.8 points (P < .01).
From the anchor-based estimation, we assessed the sensitivity and specificity of AFEQT scores to detect a change of >+3 units of improvements (defining this cutoff as a moderate improvement per Juniper et al and Crosby et al) on the Patient or Physician Global Change forms. An improvement in 19-point change in AFEQT score was associated with a sensitivity of 52% and specificity of 79% to detect at least a 3-unit increase in Patient Global Change. Similarly, the estimated improvement by 21 points using the Physician Global Change as anchor corresponded to a sensitivity of 55% and a specificity of 87%. Table V provides a summary of physician ratings and AFSS as anchors.
The Cohen effect size method. The calculated "moderate improvement" in AFEQT of 18.9 from Patient Global Change and 21.3 from Physician Global Change corresponded to a 0.9 and 1.1 SD units in effect sizes, respectively, which are considered to be large.
Results
Sociodemographic and Medical Characteristics of Participants
Among the 214 enrolled patients, 196 had no change or improved on the Patient Global Change form, and 189 had no change or improved on the Physician Global Change form. Baseline demographic and clinical characteristics of the 210 patients with Patient Global Change data are presented in Table I. Most of the patients were males with paroxysmal AF and moderate symptoms. Patients scheduled for ablation tended to have lower baseline AFEQT scores (51.2 ± 20.2) than those with no change (69.4 ± 22) or planned pharmacological change (58.7 ± 21) in treatment (P < .01).
Anchor-based Approach.Table II shows the mean improvement in AFEQT scores using the Patient Global Change forms as the anchor. Patients who rated themselves as having "no or unimportant improvement" (0 and ±1) had a mean increase in AFEQT score of 6.9 ± 17 units, whereas those who rated themselves as having had "small improvements" had a mean change in AFEQT score of 20.9 ± 18.2 points. The change in AFEQT scores for those with a moderate improvement was 18.9 ± 21 units.
We also calculated the change in AFEQT score for patients who worsened. There was only 1 patient in each of "moderately" and "markedly" worse by global change; there were 12 patients with "slightly worse" (small decrease) global score, and they had a small decrease in the AFEQT score (–2.2 ± 12.3). The difference in AFEQT score change between patients with a slightly worse global score and "unimportant changes" in global score was 9.1 AFEQT unit, similar to the difference between the "slightly better" (small increase) in global score and unimportant changes group of 14 units.
The correlation between change in AFSS and change in AFEQT was – 0.72 (P < .0001). Correlations for the association between AFEQT ratings and global patient transition ratings at month 3 were also calculated. Baseline AFEQT score was not significantly correlated with the Patient Global Change assessment with r = – 0.11 (P = .13), indicating that there does not need to be an adjustment of change in AFEQT score for different baseline scores. The baseline AFEQT score also negatively correlated with the change in AFEQT score (r = – 0.52, P < .01), so that for each unit of higher baseline AFEQT score, the average change in AFEQT score was 0.5 units smaller. This suggests that there was a ceiling effect so that subjects with a higher AFEQT baseline score did not have as much potential to improve compared with those that may have a lower baseline AFEQT score. Month 3 AFEQT versus Patient Global Change r = 0.35 (P < .01) and change in AFEQT versus Patient Global Change r = 0.46 (P < .01) (Figure), suggesting that the global change assessment may be a better anchor for change in AFEQT but did not explain all variances.
(Enlarge Image)
Figure.
Correlation between Patient Global Change and change in AFEQT.
The patients and physicians both reported the identical category of change in global AFEQT for individual patients in less than half of cases (n = 101, 48%). In 51 cases, the physician global rating of change indicated more improvement than the patient rating, and in only 18 cases did the physician rating indicate less improvement than the patient rating.
The Physicians' Global ratings of change were not able to distinguish those with small clinical improvement from those with no clinical change, as assessed by the similarity in mean 3-month changes in AFEQT scores (Table III). However, the patients deemed by their physicians to have improved by a moderate amount had a mean change in AFEQT scores of 21.3 ± 20 points.
Table IV describes the differences in mean scores by category of change assessed by the AFSS. Of the 210 enrolled patients, there were 172 patients who did not deteriorate on the AFSS. Those who had an unimportant improvement in AFSS had a mean increase in AFEQT score of 5.8 ± 13 units. Those who had a small improvements in the AFSS (3- to 5-point decrease ) had an increase in AFEQT scores of 12.0 ± 15.4 units, and those with a moderate improvements in symptoms (6- to 8-point decrease in AFSS) had a mean increase in AFEQT scores of 17.9 ± 11.8 points (P < .01).
From the anchor-based estimation, we assessed the sensitivity and specificity of AFEQT scores to detect a change of >+3 units of improvements (defining this cutoff as a moderate improvement per Juniper et al and Crosby et al) on the Patient or Physician Global Change forms. An improvement in 19-point change in AFEQT score was associated with a sensitivity of 52% and specificity of 79% to detect at least a 3-unit increase in Patient Global Change. Similarly, the estimated improvement by 21 points using the Physician Global Change as anchor corresponded to a sensitivity of 55% and a specificity of 87%. Table V provides a summary of physician ratings and AFSS as anchors.
Distribution-based Approaches
The Cohen effect size method. The calculated "moderate improvement" in AFEQT of 18.9 from Patient Global Change and 21.3 from Physician Global Change corresponded to a 0.9 and 1.1 SD units in effect sizes, respectively, which are considered to be large.
Source...