One-hundred seventy-two participants aged 6 to 17 years (M = 13.09, SD = 3.02) completed the study procedure. Of these, 80 were patients with migraine (59 girls), 23 were patients with tension-type headache (TTH, 18 girls), and 69 were healthy controls (HC, 43 girls). Regarding the time when headaches started, 6 participants declared the onset within the last few months, another 6 approximately half a year before the study, 14 a year before the study, and 76 participants declared that the headaches started more than one year ago. All three groups were balanced in terms of age (χ22 = 4.31, p = .116) and gender distribution (χ22 = 3.20, p = .202). The PedMIDAS scores were significantly higher in migraine and TTH patients than in HC (χ22 = 122.82, p < .001).

Spearman correlation coefficients for all the variables are presented in Table 2. Due to the high correlation between (1) Mechanical Pain Sensitivity in trigeminal and control areas, (2) TENS detection thresholds on the left and right side, and (3) TENS pain thresholds on the left and right side, these pairs of variables have been averaged for the binomial logistic regression models.

Groups did not differ in MDT, either for the control area (χ22 = 0.98, p = .613) or the trigeminal area (χ22 = 4.96, p = .084). There was also no difference in MPT for the trigeminal area (χ22 = 3.47, p = .177), but the groups differed in MPT for the control area (χ22 = 11.64, p = .003). Specifically, migraine patients had significantly lower MPT thresholds than TTH patients (p = .025) or HC (p = .011). There was no difference between TTH patients and HC (p > .99).

We found significant differences in MPS in both control and trigeminal areas (χ22 = 18.63, p < .001; χ22 = 8.26, p = .016; respectively). In both areas, the sensitivity was increased in migraine patients as compared with HC (p < .001, p = .011, respectively). TTH patients did not differ from HC or from migraine patients in any of the areas (all p > .350). All the results from QST are presented in Fig. 1.

Quantitative Sensory Testing scores between groups. Mechanical Detection Threshold for the control and trigeminal areas is presented in Panels A and B. Mechanical Pain Threshold for the control and trigeminal areas is presented in Panels C and D. Mechanical Pain Sensitivity for the control and trigeminal areas is presented in Panels E and F

Note. * - p < .05; *** - p < .001. C – Control area forearm, T – Test area trigeminal V2. In the Panel A one participant with score 4.59 is not presented in the healthy group to ensure plot’s readability

We found significant differences in TENS detection thresholds on both left and right sides (χ22 = 51.86, p < .001; χ22 = 49.15, p < .001; respectively). Post hoc analyses showed a similar pattern of results, i.e., both migraine and TTH patients had lower detection thresholds on both sides as compared to HC (all p < .001). The patient groups did not differ (p = .680, p > .99, for the left and right sides respectively).

Groups did not differ in TENS pain thresholds for the left and right sides (χ22 = 5.04, p = .081; χ22 = 3.20, p = .202; respectively). TENS scores across groups are presented in Fig. 2.

Transcutaneous Electrical Nerve Stimulation scores between groups. Detection thresholds for the left and right sides are presented in Panels A and B. Pain thresholds for the left and right sides are presented in Panels C and D. Note. *** - p < .001

Patients with migraine and TTH showed higher scores in the olfactory threshold test, indicating increased olfactory sensitivity, χ22 = 19.33, p < .001. Both groups scored higher than HC, (p = .007 for migraine patients, p < .001 for TTH patients). Olfactory sensitivity was not different between the two patient groups (p = .121).

For the trigeminal threshold, we found an opposite effect, χ22 = 10.50, p = .005. HC showed greater trigeminal sensitivity as compared to migraine patients (p = .036) and TTH patients (p = .019). Patients did not differ in the trigeminal sensitivity levels (p = .582).

Groups did not differ in odor identification ability, χ22 = 0.02, p = .991. All the results related to the chemosensory function are presented in Fig. 3.

Between-groups comparisons of chemosensory abilities: olfactory threshold (Panel A), trigeminal threshold (Panel B), and odor identification ability (Panel C). Note. * - p < .05; ** - p < .01; *** - p < .001; higher scores in Sniffin’ Sticks threshold tests indicate lower detection thresholds

The first binomial logistic regression model classifying participants into healthy and patient groups demonstrated that age, MPS, olfactory threshold, trigeminal threshold, and TENS detection and pain threshold are significant or trend-level predictors of the group classification (see Table 3 for regression coefficients and p-values). Therefore, all these variables were included as predictors in the final model. The final model showed that MPS (z = 3.51, p < .001), olfactory threshold (z = 2.88, p = .004), trigeminal threshold (z=-2.07, p = .038), TENS detection (z=-5.66, p < .001) and pain (z = 3.00, p = .003) thresholds were all significant predictors. These results indicate that variability in these sensory measures aided classification of participants into healthy and patient groups, with higher MPS, olfactory sensitivity, and TENS pain threshold increasing the likelihood of being classified as patient, and higher trigeminal sensitivity and TENS detection threshold decreasing this likelihood. AIC values decreased from 148.6 for the first model to 143.8 for the final model, suggesting a better fit for the latter one. Overall, scores in MPS, olfactory and trigeminal thresholds, TENS detection and pain thresholds classified participants into healthy or patient groups with sensitivity of 87% and specificity of 75%.

The initial model classifying patients into TTH and migraine groups showed that only olfactory and trigeminal thresholds are significant predictors of the classification (z=-2.66, p = .008; z = 2.46, p = .014), with MPT for the trigeminal area approaching a significance level (z=-1.88, p = .060; full model presented in Table 4). These three predictors were the ones included in the final model and all of them became statistically significant. Trigeminal threshold (z = 2.48, p = .013) was a positive predictor, indicating that higher trigeminal sensitivity increased likelihood of being a migraine patient. Contrary, olfactory threshold (z=-2.58, p = .010) and MPT for the trigeminal (z=-2.73, p = .006) area were negative predictors, indicating that greater olfactory sensitivity and higher MPT decreased the likelihood of classification as a migraine patient. AIC values decreased from 110.98 for the initial model to 101.59 for the final model, suggesting a better fit of the latter one. Overall, scores in MPT of the trigeminal area, olfactory threshold, and trigeminal threshold classified patients into migraine and TTH groups with sensitivity of 95% and specificity of 26%. When the classification threshold was adjusted from 0.50 to 0.75, the classifier had sensitivity of 79% and specificity of 57%.

The time since the first presence of headaches did not correlate with any of the sensory measures from QST (r ≤ .15, p ≥ .131), TENS (r ≤ .09, p ≥ .361) or olfactory tests (r ≤ .10, p ≥ .315). Only for the trigeminal threshold we found a weak, negative, significant correlation (r=-.21, p = .31) demonstrating that trigeminal sensitivity was higher in the participants whose headaches started recently and decreased in patients who experienced symptoms for several years.