Results
There were no statistically significant differences in key demographic variables between the 3 groups (P > .05), except for gender (see Table 1). The headache groups did not differ in terms of key clinical variables including age of onset, disease duration, or side of CH (P > .05) (see Table 1).
Cognitive Function
Three MANOVAs were performed to compare the 3 groups in terms of intelligence, executive functioning, and memory. The mean and standard deviations for each group and the results of the analyses are presented in Table 2.
Measures of Intelligence.—Both the current IQ and estimates of premorbid IQ for the 2 headache groups were in the "average range," with no discrepancy between the 2 measures. For the HCs, current IQ was in the superior range and estimated premorbid IQ was in the high average range. There were no statistically significant differences between the 3 groups in measures of current IQ (WASI-FSIQ: F[2,29] = 3.130, P = .059), estimated premorbid IQ (NART-FSIQ: F[2,29] = 0.968, P = .392), or global cognitive functioning (MMSE: F[2,29] = 1.660, P = .208) (see Table 2).
Measures of Executive Functioning.—Executive functioning was measured using subtests of the TMT, VF, and Stroop. The MANOVA results yielded no statistically significant differences between the 3 groups on any of the measures (see Table 2), except on the letter sequencing and number-letter switching subtests of the TMT (F[2,27] = 4.492, P = .021 and F[2,27] = 3.416, P = .048 respectively). Following a Bonferroni correction, setting alpha at 0.001, these statistical significances remained.
Post hoc comparisons using Tukey showed that these significant results were due to the difference in the means of the HCs and ECH patients with the latter group performing worse (P < .05 on both tests) (see Fig. 1).
(Enlarge Image)
Figure 1.
Means and standard errors for patients with episodic and chronic cluster headache and healthy controls on 2 subtests of the Trail Making Test. HCs = healthy controls; ECH = episodic cluster headache; CCH = chronic cluster headache.
Interestingly, the ECH patients performed poorer on most subtests of TMT than CCH patients, although the differences did not reach significance (see Table 2).
Measures of Memory Function.—Memory function was assessed using the CVLT-II for verbal episodic memory, WAIS-III subtests for working memory, Warrington SRMF for non-verbal episodic memory, and CFQ for subjective memory. There were no statistically significant differences between the 3 groups on the CVLT-II and the arithmetic subtest of the WAIS-III WMI (see Table 2). There were group differences on the other 2 subtests of WAIS-III, the Warrington SRMF, and the CFQ (see Table 2 and Fig. 2). These statistical differences remained following a Bonferroni correction to account for the multiple comparisons.
(Enlarge Image)
Figure 2.
Mean and standard error for patients with episodic and chronic cluster headache and healthy controls on tests assessing memory. (a) WAIS-III Working Memory Index subtests; (b) Warrington Short Recognition Memory for Faces; (c) Cognitive Failures Questionnaire. HCs = healthy controls; ECH = episodic cluster headache; CCH = chronic cluster headache; WAIS-III = Wechsler Adult Intelligence Scale III; WMI = Working Memory Index.
Post hoc analyses of the WAIS-III subtests using Tukey's test showed a statistically significant difference in the mean scores for HCs and both ECH and CCH patients (P < .05) (see Fig. 2a). On the letter-number sequencing, only the CCH group scored significantly lower than HCs (P < .05), and the difference between HCs and the ECH group was not (P = .07). On the WMI subtest, the ECH group scored significantly lower than the HCs (P < .05), and the difference between HC and CCH groups just missed significance (P = .06). For both of these subtests, the headache groups did not differ from each other (P > .05) (see Fig. 2a).
On the Warrington SRMF, post-hoc tests showed that the differences between the 3 groups were not statistically significant (P < .05). The 3 means were homogenous (see Fig. 2b). On the CFQ, the HC means were significantly lower (less cognitive failures) than CCH (P < .05) but not the ECH patients (P > .05) (see Fig. 2c). The CCH group exhibited higher scores showing more cognitive failures, followed by ECH patients and then HCs (see Table 2).
Psychosocial Function
Two separate MANOVAs were conducted on measures of mood, disability, and QoL. The first one examined differences in mood between the 3 groups, and the second analysis evaluated differences in disease impact and QoL. The descriptive results and their significant values are provided in Table 3.
Measures of Mood, Disability, and QoL.—The first analysis showed significant group differences on all measures of mood and autonomic dysfunction: BDI-II (F[2,25] = 18.482, P = .000), HADS-A (F[2,25] = 5.497, P = .011), HADS-D (F[2,25] = 19.140, P = .000), BHS (F[2,25] = 9.358, P = .000), and SCOPA-AUT (F[2,25] = 9.931, P = .000). These statistically significant differences revealed the same pattern: patients with CCH had the highest scores on all of the measures, followed by ECH, and then the HCs (see Fig. 3a). When alpha was adjusted (P = .001), these statistically significant differences remained.
(Enlarge Image)
Figure 3.
Means and standard errors for (a) hopelessness (BHS), depression (BDI-II and HADS-D) and anxiety (HADS-A) and (b) quality of life for the patients with chronic or episodic cluster headache and healthy controls. BHS = Beck Hopelessness Scale; BDI-II = Beck Depression Inventory II; HADS-A = Hospital Anxiety and Depression Scale-Anxiety; HADS-D = Hospital Anxiety and Depression Scale-Depression; EQ5D = EuroQoL quality of life measure; ECH = episodic cluster headache; CCH = chronic cluster headache; CH = cluster headache; HCs = healthy controls.
A post hoc Tukey test showed that means for HCs and the ECH group, and HCs and CCH patients were significantly different on BHS, BDI-II, and HADS-D, with the patients scoring worse on all measures (all P < .005). Post hoc comparisons of means on HADS-A showed a statistically significant difference only between HCs and patients with CCH (P < .05), but not the ECH group (P > .05). Similarly, post hoc analysis of means on SCOPA-AUT showed statistically significant differences between the CCH and HCs groups, and CCH and ECH groups (all P < .05).
None of the HCs scored in the moderate or severe depression range on the BDI-II. In contrast, 35% of the ECH and 44% of the CCH groups scored in the moderately depressed range, and 25% of the ECH group and 22% of the CCH patients scored in the severe depression range. This difference between the 3 groups was statistically significant, indicating that a significantly higher proportion of the patients with ECH or CCH experienced moderate or severe depression compared with the HCs (χ2[2,29] = 12.541, P < .005).
Similarly, on the HADS-D, while none of the HCs achieved "caseness" indicating definite depression, 38% of the patients with ECH and 38% of the CCH patients did so. The association between group (HC, ECH, CCH) and "caseness" (yes vs no) for depression on the HADS-D approached significance (χ2[2,27] = 5.304, P = .071). However, a significantly higher proportion of the patients with ECH or CCH achieved "caseness" for anxiety on HADS-A than the HCs. (χ2[2,28] = 9.307, P < .05). One participant in the HCs group (8%), 38% of the ECH group, and 75% of the CCH patients achieved "caseness" for anxiety as measured on HADS-A.
The second analysis revealed no differences in QoL between the 3 groups (EQ5D health state: F[1,13] = 1.049, P = .325; EQ5D thermometer: F[1,13] = 0.963, P = .344). Similarly, there were no group differences between the headache groups concerning the impact of their headache (HIT-6: F[1,13] = 0.026, P = .875) or headache-related disability (MIDAS: F[1,13] = 2.674, P = .126). We then combined the 2 headache groups, to compare QoL between HCs and CH patients. The results revealed no difference in the EQ5D health state (F[1,28] = 0.054, P = .818) but showed a statistically significant group difference in the EQ5D visual analog scale rating of overall health status (F[1,28]: 31.104, P = .001), which as expected was higher in HCs (85.75 ± 9.29) than CH patients (51.39 ± 19.86) (see Fig. 3b).
Correlational Analysis.—A number of correlational analyses were carried out using Pearson correlation coefficient, to examine the relationship between demographic and clinical variables (age, age of onset, disease duration, years of education, side of CH) and participant's mood, disease impact, QoL (as listed in Table 3) and cognitive test results listed in Table 2. There was a statistically significant negative correlation between age and MIDAS (r = −0.806, P < .005), and age and BHS (r = −0.424, P < .05). Disease duration correlated negatively and statistically significantly with WASI FSIQ (r = −0.459, P < .05) and also with 2 measures of the CVLT-II (negatively with the long delay free recall: r = −0.463, P < .05; and positively with false positives on the recognition test: r = 0.527, P < .05). Furthermore, there was a negative correlation between age at onset and MIDAS (r = −0.466, P < .05), and years of education and the category subtest of VF (r = −0.414, P < .05). None of the other correlations were statistically significant or noteworthy. It was not possible to statistically adjust the alpha to account for multiple comparisons due to the sample size. Thus these should be interpreted with caution.