Effectiveness of Exercise and Manual Therapy as Treatment for Patients with Migraine, Tension-Type Headache or Cervicogenic Headache: An Umbrella and Mapping Review with Meta-Meta-Analysis

: The aim of the study was to perform a mapping and umbrella review with meta-meta-analysis (MMA) to synthesise and critically evaluate the effectiveness of manual therapy (MT) and aerobic exercise (AE) in relation to pain intensity, frequency, disability and quality of life in patients with migraines, tension-type headaches (TTH) and cervicogenic headaches (CGH). A systematic search was conducted in PubMed, PEDro, Scielo and Google Scholar up to December 2020. A total of 18 articles met the inclusion criteria, and only 8 were included in the quantitative analysis. The MMA showed results in favour of the interventions in terms of pain intensity and quality of life in migraine, TTH and CCH. Data were also in favour of the intervention in terms of pain frequency in migraine and in terms of disability in TTH. However, there were no signiﬁcant effects on pain frequency in TTH and CGH. The results showed moderate evidence to suggest that AE reduces pain intensity in patients with migraine. In addition, the evidence in favour of MT or a mixed intervention (including therapeutic exercise) was also moderate in terms of reducing pain intensity in patients with TTH.

Headaches lead to important deteriorations in patients' quality of life and involve significant economic repercussions, medical expenses, work incapacity and social and familiar impact. Headache disorders were the third-leading cause of disability in 2016 [2].
Regarding the therapeutic approach for headaches, there are both pharmacological and nonpharmacological interventions. Pharmacological treatment appears to be efficient for some acute cases and for prophylaxis [3]. However, this treatment is not effective in all cases, and they can become chronic disorders. Chronic headaches generate an increase in the number of medical consultations at the cost of the health system and can even generate medication-overuse headache [4][5][6].
The nonpharmacological approach provides therapeutic options to be assessed for the treatment of headaches, such as therapeutic exercise and manual therapy. Some

Materials and Methods
This umbrella and mapping review was performed according to the Preferred Reporting Items for Overviews of Systematic Reviews including the harms checklist (PRIO-harms). The PRIO-harms tool is composed of 27 items and 56 sub-items [11]. The protocol of this systematic review and meta-analysis was registered in an international register prior to starting the review (PROSPERO, CRD42020222573).

Review Inclusion Criteria
The inclusion criteria for this review were based on methodological and clinical factors, including population, intervention, control, outcomes and study type [12].

Population
The individuals selected for the articles were patients older than 18 years, diagnosed with migraine, TTH and CGH.

Intervention and Control
The interventions were any type of therapeutic exercise and/or manual therapy performed by a physical therapist or health professional. Studies in which the intervention was performed by chiropractors or osteopaths were excluded. The intervention could be provided as an independent treatment or combined with other types of intervention. The control group could include any type of intervention, when it was possible to isolate and evaluate the effectiveness of manual therapy and/or therapeutic exercise.

Outcomes
The measures employed to assess the results and effects were pain intensity, frequency of symptoms, disability and/or quality of life.

Study Design
We included systematic reviews (with or without meta-analysis) of randomised controlled trials (RCTs) or controlled clinical trials (CCTs). No language restrictions were applied, as recommended by international criteria [13].

Search Strategy
We conducted a search for articles on PubMed, PEDro, Scielo and Google Scholar. The last search was run on December 2020.
The following PubMed search strategy was employed and was adapted to the rest of the databases: [Title/Abstract]) OR ("physical therapy" [Title/Abstract]) OR ("musculoskeletal manipulations" [MeSH Terms]) OR ("manual therapy" [MeSH Terms]) OR ("manual therapy") OR ("manipulation spinal"). We also used the following search filters: "systematic review" and "meta-analysis".
Two independent reviewers conducted the search using the same methodology, and differences during this phase were resolved by consensus. The reference sections of the original studies were screened manually, and the authors were contacted for further information if necessary.

Selection Criteria and Data Extraction
Initially, analyses were performed by two independent reviewers who assessed the relevance of the systematic reviews (with and without a meta-analysis) regarding the study questions and objectives. The first analysis was performed based on each study's title information, abstract and keywords. If there was no consensus or the abstract did not contain enough information, the full text was reviewed.
In the second phase of the analysis, the full text was assessed if the studies met all of the inclusion criteria. Differences between the reviewers were resolved by a process of discussion, and consensus was moderated by a third reviewer [14]. Data described in the Results section were extracted by means of a structured protocol that ensured that the most relevant information was obtained from each study [15].

Methodology Quality Assessment
The two independent reviewers assessed the methodological quality of the selected systematic reviews based on the Modified Quality Assessment Scale for Systematic Reviews (AMSTAR) developed by Barton et al. This scale presents 13 items, each worth 2 points (with "yes" scoring 2; "in part" scoring 1; "no" scoring 0), and the maximum possible score is 26. A high-quality cutoff of 20 or more points was provided by the developers of the scale [16].
The two independent reviewers assessed the quality of the studies employing the same methods, and disagreements on the final quality assessment score were resolved by consensus with a third independent reviewer. The inter-rater reliability was estimated using the kappa coefficient (κ): κ > 0.7 indicated a high level of agreement between the reviewers; κ of 0.5-0.7 indicated a moderate level of agreement, and κ < 0.5 indicated a low level of agreement [17].

Risk of Bias Assessment
We assessed the risk of bias using the Risk of Bias in Systematic Reviews tool (ROBIS), which evaluates the quality across 3 phases: (1) relevance assessment; (2) identification of concerns with the review process through 4 domains related to study eligibility, study identification and selection, data collection and study appraisal, and synthesis and findings; and (3) judgment on the risk of bias. The ROBIS tool includes signalling questions to evaluate specific domains by answering "yes", "probably yes", "probably no", "no" or "no information". The risk of bias is therefore judged as "low", "high" or "unclear" [18].
The two independent reviewers evaluated the risk of bias in the selected studies using the same methodology; disagreements were resolved through consensus and mediation by a third reviewer. The interrater reliability was estimated using the same κ cutoffs described in the section Methodology quality assessment.

Evidence Map
We presented the scientific evidence of each systematic review with meta-analysis through a visual map. We created 2 different maps, one descriptive and the other using Appl. Sci. 2021, 11, 6856 4 of 44 effect size data, and the information from each review was provided using the following criteria: 1.
Number of studies (figure size): The size of each figure is directly proportional to the number of original studies included in each of the meta-analyses.

2.
Study Population (bubble colour) and type of intervention (symbol): The type of population evaluated in each study is represented by a colour (green: migraine; blue: TTH; yellow: CGH). The type of therapeutic intervention determined each symbol inside the bubble (x: manual therapy; −: therapeutic exercise; +: mixed intervention). In addition, in the second mapping, the plot of the figure represents the study variable. 3.
x-Axis: In the descriptive mapping, each of the study variables is represented on the x-axis. In the second mapping, each of the reviews was classified according to the size effect as described by Hopkins (Hopkins et al., 2009). The categorisation of the effect size is described in the section Data synthesis and analysis. 4.
y-Axis: The descriptive mapping represents the quality of each of the reviews on the y-axis according to the AMSTAR scale. In the second mapping, the reviews were sorted into the following 4 categories according to the Physical Activity Guidelines Advisory Committee (PAGAC): strong, moderate, limited or not assignable.

Qualitative Analysis
As described earlier, we relied on the assessments of each systematic review with meta-analysis for the methodological quality of the primary studies, using AMSTAR for the included reviews and PAGAC for assessing the evidence across reviews.
For the PAGAC analysis, the findings were evaluated according to 5 criteria: (1) the applicability of the study sample, exposures and outcomes to the research question, (2) generalisability to the population of interest, (3) the risk of bias or study limitations, (4) the quantity and consistency of findings across studies and (5) the magnitude and precision of the effect. The strength of the evidence was classified as strong, moderate, limited or not assignable [19].

Data Synthesis and Analysis
The statistical analysis was performed using meta-analyses with MetaXL software [20,21]. The same inclusion criteria for the systematic review were employed, but 2 criteria were added: (1) The Results section contained detailed information on the comparative statistical data (mean, standard deviation and/or 95% confidence interval [CI]) of the main variables and (2) data for the analysed variables were represented in at least 2 studies. The summary statistics are presented in the form of forest plots [22], which consist of a weighted compilation of all standardised mean differences (SMDs) and corresponding 95% CIs reported by each study. They provide an indication of heterogeneity among the studies. The statistical significance of the pooled SMDs were examined using Hedges' g to account for a possible overestimation of the true population effect size in small studies [23]. We interpreted the statistical significance of the pooled SMDs as described by Hopkins [24]; that is, we considered an SMD of 4.0 an extremely large clinical effect, 2.0-4.0 a very large effect, 1.2-2.0 a large effect, 0.6-1.2 a moderate effect, 0.2-0.6 a small effect and 0.0-0.2 a trivial effect. When the statistical significance of the pooled data was presented as mean difference, the meta-analysis was replicated using Meta-Essentials (ERIM, Erasmus University Rotterdam, Netherlands) with Microsoft Excel to obtain the SMD values [25]. The degree of heterogeneity among the studies was estimated by employing Cochran's Q statistic test (p < 0.1 was considered significant) and the inconsistency index (I 2 ) [26]. An I 2 > 25% is considered to represent low heterogeneity, while an I 2 > 50% is considered medium and an I 2 > 75% is considered to represent large heterogeneity [27]. The I 2 index is complementary to the Q test, although it has a similar problem with power as does the Q test with a small number of studies [27]. A study was therefore considered heterogeneous when it fulfilled one or both of the following conditions: (1) the Q-test was significant (p < 0.1) and (2) the result of I 2 was >75%. To obtain a pooled estimate of the effect in the meta-analysis of the heterogeneous studies, we performed a random-effects model, as described by DerSimonian and Laird (1986) [28]. Publication bias through the funnel plot and the sensitivity exclusion analysis was not evaluated due to the impossibility of performing it when the MMA includes fewer than 2 studies.
Eight of the studies were included in eight independent MMAs. The MMAs assessed pain intensity and the frequency of symptoms in patients with migraine, TTH and CGH independently, as well as disability in patients with TTH. heterogeneous when it fulfilled one or both of the following conditions: (1) the Q-test was significant (p < 0.1) and (2) the result of I 2 was >75%. To obtain a pooled estimate of the effect in the meta-analysis of the heterogeneous studies, we performed a random-effects model, as described by DerSimonian and Laird (1986) [28]. Publication bias through the funnel plot and the sensitivity exclusion analysis was not evaluated due to the impossibility of performing it when the MMA includes fewer than 2 studies.
Eight of the studies were included in eight independent MMAs. The MMAs assessed pain intensity and the frequency of symptoms in patients with migraine, TTH and CGH independently, as well as disability in patients with TTH.
Frequency of symptoms 4 (n = 178) Self-report headache diaries The meta-analysis showed that manual therapy was more effective than pharmacological medical care (WMD = −0.80; 95% CI -1.66 to −0.44) immediately after the intervention.
Chaibi The Intervention group had significantly more reduction in their headache intensity than the Comparison group after manual therapy intervention.
Frequency of symptoms -Self-report headache diaries There was a reduction of more than half in the frequency of symptoms post-treatment, and the results were maintained for more than six months.. The exercise intervention displayed statistically significant improvements at 7 weeks. The combination of exercise and manipulative therapy displayed statistically significant improvements. Those improvements persisted at the 12-month-follow up period.
Frequency of symptoms -Self-report headache diaries The exercise intervention displayed statistically significant improvements at 7 weeks. The combination of exercise and manipulative therapy displayed statistically significant improvements. Those improvements persisted at the 12-month-follow up period.

Disability -MVK disability scale
With the exception of one study, all reported reduction in disability but there was conflicting evidence regarding the effects of manipulative therapy. There were no significant differences between groups at the end of the treatment. However, at 4 weeks follow up, SMT showed greater improvement than medication. In patients with TTH, it appears to be an advantage for SMT at 4 weeks post-treatment, but 6 weeks later, medication was better. In patients with MH, SMT group showed an advantage after 8 weeks of treatment. The combination of amitriptyline and SMT did not provide any advantage. Patients with CGH, who received SMT, reported approximately twice the reduction in headache intensity per episode than the massage group.
Frequency of symptoms -Self-report headache diaries In patients with CGH, the SMT group showed a decrease of 69% compared with a 47% decrease in the Comparison group at 8 weeks of treatment.

Disability
-HDI There appears to be an advantage for chiropractic SMT after 8 weeks.  After the intervention, all articles showed significant improvements over the Comparison group, and better improvements were found when the intervention combined different types of manual therapy.
Frequency of symptoms -Self-report headache diaries The results showed significant improvements for the Intervention group following the intervention. Regarding this group, better outcomes were found when different types of manual therapy were combined.

Disability -HDI
All treatment groups showed significant improvements in the overall HDI count, revealing better improvements for the combination of cervical manipulation and suboccipital inhibition.
Quality of life -SF-12, HIT-6, SF-36 The studies that evaluated quality of life outcome showed inconclusive results. There was evidence that therapeutic exercise on the craneocervical and shoulder region significantly improved, in the medium term, the intensity of pain.
Frequency of symptoms -Self-report headache diaries The results showed that therapeutic exercise on the craneocervical and shoulder region significantly improved the frequency of symptoms in the medium term, with a strong level of evidence.

Disability -HDI
There was improvement in the disability outcome at medium term regarding therapeutic exercise on the craniocervical and shoulder region, with a strong level of evidence. The results showed greater reduction for the Intervention group in the short-and mid-term.
Frequency of symptoms -Self-report headache diaries The Intervention group performed significantly better in terms of headache frequency at 4 and 8 weeks. In addition, the results showed significantly more reduction in headache frequency at 12 months.

Disability -HDI, NDI
Significantly fewer patients in the Intervention group experienced a 50% reduction in headache index score after 4 weeks. In addition, there was a significant difference in favor of the Intervention group after 3 weeks.

Characteristics of the Included Systematic Reviews
Our study included 18 systematic reviews (with or without a meta-analysis), comprising 95 original studies, 90 RCTs, 4 CCTs and 1 non-controlled clinical trial (included by [29]), with a total of 9188 participants. Several original studies appeared in different systematic reviews, but none of the reviews included the same studies (Table S1).
In terms of the populations in the systematic reviews, 9 studies (n = 3693) included patients diagnosed with migraine [7,8,29,33,35,37,[39][40][41] In three of the systematic reviews, the Intervention group used AE with other forms of exercise therapy in patients with migraine [8,39] or TTH [29]. Ten systematic reviews had an Intervention group that used various types of manual therapy in patients with TTH [30,32,34,38], CGH [44] or populations with various types of headaches included in this study [33,35,37,40,43]. In the remaining five studies, the intervention used a combination of both treatments in patients with TTH [42], CGH [31,36] or several types of headache [7,41].

Results of the Methodology Quality Analysis
The scores ranged from 7 to 24 out of a possible 26 points, with a mean score of 17.28 ± 4.75 points. Only 6 (33.33%) of the 18 studies were considered high quality, with a score above 20 points [8,37,[40][41][42]44] (Table 3). Table 3. Quality assessment scores. 1. Explicitly described to allow replication; 2. Adequate number and range of databases; 3. Alternative searches; 4. Adequate range of key words; 5. Non-English-language papers included in the search; 6. Inclusion criteria explicitly described to allow replication; 7. Excludes reviews which do not adequately address inclusion and exclusion criteria; 8. Two independent reviewers assessing selection bias; 9. Quality assessment explicitly described to allow replication; 10. Meta-analysis conducted on only homogeneous data or limitations to homogeneity discussed; 11. Confidence intervals/effect sizes reported where possible; 12. Conclusions supported by the meta-analysis or other data analysis findings 13. Conclusions address levels of evidence for each intervention/comparison.
The items with the highest scores were those related to the assessment of selection bias and the adequate description of the quality assessment. The items with the lowest scores were those related to language restrictions and the reporting of confidence intervals and effect sizes. The inter-rater reliability of the methodological quality assessment was high (κ = 0.790).
The domain for "synthesis and findings" had the highest risk of bias, with 10 (55.55%) studies scoring a high risk of bias. In domains for "study eligibility criteria", "identification and selection of studies" and "data collection and study appraisal", 9 (50%) of the 18 studies had a low risk of bias. The inter-rater reliability for the risk of bias assessment was high (κ = 0.849). Table 4. Risk of bias assessment in systematic reviews through the ROBIS scale.

Evidence Map
Figures 3 and 4 present the results of the evidence map for the 18 studies. In addition, Table 5 shows the levels of evidence of the meta-analyses included in the study according to the PAGAC. The domain for "synthesis and findings" had the highest risk of bias, with 10 (55.55%) studies scoring a high risk of bias. In domains for "study eligibility criteria", "identification and selection of studies" and "data collection and study appraisal", 9 (50%) of the 18 studies had a low risk of bias. The inter-rater reliability for the risk of bias assessment was high (κ = 0.849). Figures 3 and 4 present the results of the evidence map for the 18 studies. In addition, Table 5 shows the levels of evidence of the meta-analyses included in the study according to the PAGAC.
Interventions using AE with or without resistance exercise were an effective approach in patients with migraine or TTH in the short term [8,29,39]. Interventions based on manual therapy showed differences in favour of the Intervention group in patients with TTH [30,32,34,38], and in populations that included patients with migraine, TTH and CGH in the short-term [33,35,40,43] and at follow-up [35]. In patients with CGH, manual therapy showed positive results in the short-term [36,44], but not in the medium-or long-term [44]. The combination of manual therapy and therapeutic exercise showed differences favouring the Intervention group in patients with migraine, TTH or CGH in the short-term [41] and medium-term [7,31]. In one study, when comparing manual therapy with or without exercise and medication, they found significant differences in the short-term, but not in the long-term in patients with TTH [42].

Appl. Sci. 2021, 11, x FOR PEER REVIEW 24 of 34
Interventions using AE with or without resistance exercise were an effective approach in patients with migraine or TTH in the short term [8,29,39]. Interventions based on manual therapy showed differences in favour of the Intervention group in patients with TTH [30,32,34,38], and in populations that included patients with migraine, TTH and CGH in the short-term [33,35,40,43] and at follow-up [35]. In patients with CGH, manual therapy showed positive results in the short-term [36,44], but not in the medium-or longterm [44]. The combination of manual therapy and therapeutic exercise showed differences favouring the Intervention group in patients with migraine, TTH or CGH in the short-term [41] and medium-term [7,31]. In one study, when comparing manual therapy with or without exercise and medication, they found significant differences in the shortterm, but not in the long-term in patients with TTH [42].
either side of the box represent a 95% confidence interval (CI).
Interventions using aerobic exercise have shown significant differences in headache frequency reduction in patients with migraine [8,39] or migraine and TTH [29]. Treatment with manual therapy showed positive results in studies involving patients with TTH in the short-term [30,32,34]. There was also significant differences in patients with CGH in the short-term [36,44] and medium-term [44]. Studies that included populations with migraine, TTH or CGH found positive results favouring the manual therapy group in the short-term [35,40,43] and medium-term [35,43]. Studies using a combination of therapeutic exercise and manual therapy showed a decrease in headache in the days immediately following treatment in patients with migraine, TTH or CGH [33,42] and in both the shortand medium-term [7,31]. Luedtke et al. found significant differences in patients with migraine and TTH but not in the case of CGH [41].

Quality of Life
Quality of life was evaluated in five studies: two included a meta-analysis [8,37], while three performed a qualitative synthesis [32,34,40]. Two studies showed significant differences in the increase in quality of life when using aerobic exercise in patients with migraine [8] or manual therapy in patients with migraine, TTH or CGH [37]. However, the remaining three studies showed no significant differences between groups or showed inconclusive results when the intervention included different manual therapy techniques in patients with TTH [32,34] or migraine and TTH [40].
Most studies that evaluated disability associated with headache employed manual therapy, finding positive results in favour of the Intervention group in the short-term in patients with TTH [32,34,38] or GCH [44]. Studies that included patients with migraine, TTH or CGH also showed favourable results after employing manual therapy in the shortterm [33,35,37,40,43] and at follow-up [37]. Manual therapy with or without therapeutic exercise produced medium-term improvements in patients with migraine or TTH [7], with conflicting evidence because only one study evaluated disability [31].
With regard to the quantitative analysis, the MMA of disability in patients with TTH revealed a statistically significant difference in two meta-analyses through manual therapy [38,43] (SMD = −0.27; 95% CI −0.47 to −0.06; p < 0.05). Based on Cochran's Q statistical test, the heterogeneity among the reviews was not considered significant (Q = 1.37, p = 0.24). However, we found low evidence of heterogeneity according to the I 2 (I 2 = 27%) ( Figure 12).

Quality of Life
Quality of life was evaluated in five studies: two included a meta-analysis [8,37], while three performed a qualitative synthesis [32,34,40]. Two studies showed significant differences in the increase in quality of life when using aerobic exercise in patients with migraine [8] or manual therapy in patients with migraine, TTH or CGH [37]. However, the remaining three studies showed no significant differences between groups or showed inconclusive results when the intervention included different manual therapy techniques in patients with TTH [32,34] or migraine and TTH [40].
Most studies that evaluated disability associated with headache employed manual therapy, finding positive results in favour of the Intervention group in the short-term in patients with TTH [32,34,38] or GCH [44]. Studies that included patients with migraine, TTH or CGH also showed favourable results after employing manual therapy in the shortterm [33,35,37,40,43] and at follow-up [37]. Manual therapy with or without therapeutic exercise produced medium-term improvements in patients with migraine or TTH [7], with conflicting evidence because only one study evaluated disability [31].
With regard to the quantitative analysis, the MMA of disability in patients with TTH revealed a statistically significant difference in two meta-analyses through manual therapy [38,43] (SMD = −0.27; 95% CI −0.47 to −0.06; p < 0.05). Based on Cochran's Q statistical test, the heterogeneity among the reviews was not considered significant (Q = 1.37, p = 0.24). However, we found low evidence of heterogeneity according to the I 2 (I 2 = 27%) ( Figure 12).

Discussion
The present research aimed to synthesise and critically evaluate the evidence on the effectiveness of manual therapy and therapeutic exercise in patients with migraine, TTH or CGH. Results showed, in general, that the interventions with MT and exercise were beneficial for patients with TTH and CCH, and aerobic exercise was beneficial for patients with migraine. Most of the quantitative results favoured the interventions, with statistically significant results supporting the use of MT and exercise to reduce pain intensity in CGH and TTH but not to reduce the frequency of headaches. In addition, the MMA suggested a positive effect of AE for reducing the pain and frequency of migraine. The methodological quality of the eight meta-analyses included in our MMA is low for three of them [38,39,43] and high for the rest; and the risk of bias in general is high, except for three studies that had good methodological quality and also a low risk of bias [8,40,41].

Migraine
Currently, the management of patients with migraine is complex, given there is no specific treatment that fits all patients, and it sometimes leads to medication overuse and other adverse effects [45]. A lot of effort has been made to find a conservative non-pharmacological approach [46,47]. Therapeutic exercise prescriptions have added some hope for the treatment of this population [48,49]. This non-pharmacological intervention, which lacks adverse effects, helps minimise the need for drugs or other invasive interventions. If the headaches can be improved by reducing the frequency or intensity of the symptoms, medication overuse and adverse effects would be prevented.
Most of the included reviews on migraine have compared AE with other interventions, and the results of moderate, continuous AE appear to be positive. Along these lines, the results of the present MMA showed that AE can reduce pain intensity and frequency of headache in migraine patients, with moderate and limited evidence, respectively, and large effect sizes and low heterogeneity. The sensitivity analyses could not be performed since only two studies were included in each MMA (this also occurred in six out of the eight analyses performed).
In recent decades, the exercise intervention studied for migraine in most of the RCTs comprised moderate, continuous AE performed 2-5 days per week and for 40-60 min at 60-75% VO2max. The new approaches to therapeutic exercise for migraine use high-intensity exercise or interval exercise (of moderate or high intensity) [50,51]. There is still a lack of evidence regarding the difference between AE intensities and modalities, because

Discussion
The present research aimed to synthesise and critically evaluate the evidence on the effectiveness of manual therapy and therapeutic exercise in patients with migraine, TTH or CGH. Results showed, in general, that the interventions with MT and exercise were beneficial for patients with TTH and CCH, and aerobic exercise was beneficial for patients with migraine. Most of the quantitative results favoured the interventions, with statistically significant results supporting the use of MT and exercise to reduce pain intensity in CGH and TTH but not to reduce the frequency of headaches. In addition, the MMA suggested a positive effect of AE for reducing the pain and frequency of migraine. The methodological quality of the eight meta-analyses included in our MMA is low for three of them [38,39,43] and high for the rest; and the risk of bias in general is high, except for three studies that had good methodological quality and also a low risk of bias [8,40,41].

Migraine
Currently, the management of patients with migraine is complex, given there is no specific treatment that fits all patients, and it sometimes leads to medication overuse and other adverse effects [45]. A lot of effort has been made to find a conservative nonpharmacological approach [46,47]. Therapeutic exercise prescriptions have added some hope for the treatment of this population [48,49]. This non-pharmacological intervention, which lacks adverse effects, helps minimise the need for drugs or other invasive interventions. If the headaches can be improved by reducing the frequency or intensity of the symptoms, medication overuse and adverse effects would be prevented.
Most of the included reviews on migraine have compared AE with other interventions, and the results of moderate, continuous AE appear to be positive. Along these lines, the results of the present MMA showed that AE can reduce pain intensity and frequency of headache in migraine patients, with moderate and limited evidence, respectively, and large effect sizes and low heterogeneity. The sensitivity analyses could not be performed since only two studies were included in each MMA (this also occurred in six out of the eight analyses performed).
In recent decades, the exercise intervention studied for migraine in most of the RCTs comprised moderate, continuous AE performed 2-5 days per week and for 40-60 min at 60-75% VO 2 max. The new approaches to therapeutic exercise for migraine use highintensity exercise or interval exercise (of moderate or high intensity) [50,51]. There is still a lack of evidence regarding the difference between AE intensities and modalities, because the new approaches have yet to be evaluated in depth. In this regard, contradictory results have been obtained when assessing the effects of continuous moderate versus highintensity interval training, although in general both interventions might reduce headache frequency and pain [50,51].
It is crucial to determine whether the patient with migraines also presents kinesiophobia, because it could impede successful completion of the exercise program. The first step in implementing exercise treatment for these patients should be to detect any irrational beliefs about movement, so that we can establish adequate management with a biobehavioural approach (therapeutic patient education) [52].
This section is not mandatory but can be added to the manuscript if the discussion is unusually long or complex.

Tension-Type Headache
Among the treatment options for patients with TTH, the conservative interventions applied by a physical therapist have been given a lot of weight. This type of intervention is justified based on the characteristics of patients with TTH and the involvement of the neck in TTH, which has been described and accepted worldwide [1]. Interventions such as cervical manual therapy and local exercises directed to the neck muscles are the most popular and studied approaches. These interventions are based on findings that suggest the presence of referred pain to the head coming from the cervical structures after manual provocation tests [53][54][55].
The results of the MMA showed that interventions based on MT might reduce the pain intensity (limited evidence) and disability of patients with TTH (limited evidence). The effect size is large for pain intensity with large heterogeneity. A small effect size and low heterogeneity was found for disability. No effect on frequency was found for manual therapy (limited evidence). The analyses of combined interventions (MT and/or local exercises) favoured the intervention, with moderate evidence; however the effect size was small, although with low heterogeneity.
According to the results of the present review, the physical therapy techniques included in conservative treatment for TTH were articular and soft-tissue techniques (in most cases) combined with therapeutic exercise. The articular approach included cervical and thoracic mobilisations or manipulations, mobilisations of several neck segments and tractions and soft tissue interventions applied to cervical muscle trigger points. In addition to these techniques, neck exercises were employed focused on the strength and endurance of the target muscles. In this regard, manual therapy has been largely studied; however there is a need to study other conservative exercise treatments that could also have positive effects on patients with TTH. This approach might be particularly indicated for those with chronic pain, for whom a generalised exercise intervention based on AE could have positive effects, as has been demonstrated for patients with migraine. It has been reported that patients with chronic TTH have an impairment of pain inhibition in a similar manner as has been observed in other chronic pain conditions [56].
Another interesting point is that patients with migraine as well as TTH typically have low levels of physical activity, and inactivity among pain-free individuals is a risk factor for developing a non-migraine type of headache [57]. These findings justify the implementation of interventions to increase physical activity/exercise among patients with TTH and also CGH. The combination of therapeutic patient education with exercise seems likely to lead to better adherence to exercise [58].

Cervicogenic Headache
CGH has been included in the International Classification of Headache Disorders as a secondary headache arising from musculoskeletal disorders in the cervical spine, but not necessarily accompanied by neck pain [1,59]. Structures such as the upper cervical synovial joints, upper cervical muscles and C2-C3 intervertebral disc have been raised as possible origins of CGH [60,61]. Some findings suggest the involvement of the neck structures in CGH; for example a reduction in upper cervical rotation [62] reduced cervical flexion/extension or painful upper cervical joints as assessed by manual palpation [63]. Given the problem of relapses at the neck, conservative treatment has focused on the neck structures. Manual therapy techniques are usually employed for patients with CGH [64,65]. In the present MMA, the data suggest that MT can influence pain intensity, but not pain frequency, both with limited evidence. The effect sizes for both findings were small, with low heterogeneity.
The interventions used in the RCT included in the analysed reviews were MT and/or therapeutic local exercise or MT alone. There is evidence suggesting positive effects for therapeutic exercise on patients with neck pain [66]; thus, it would be interesting to assess the effect of therapeutic exercise alone in CGH, given that there is scarce but favourable evidence for it [67].

Clinical Implications
The present review attempts to provide clear data and conclusions to be applied in practical terms to patients with headaches. More high-quality research is needed to be able to offer more specific recommendations, but with the results of this MMA, some suggestions can be made. Patients with headache might benefit from manual therapy and exercise. Physical therapists should be able to establish a specific and appropriate treatment for patients with headache according to their type of headache. Based on the data, for example, it would be preferable to employ a general exercise intervention for patients with migraine instead of a specific MT intervention.
From a clinical point of view, even a small intervention of 2 min per day of resistance exercise can reduce headache frequency by up to 50% [68]. This result highlights the importance of frequent therapeutic exercise for patients with headaches.

Research Implications
In general terms, further research with a combination of interventions, such as exercise, education and various MT techniques, is needed for the several types of headaches. This approach could establish new models of combined treatments for patients with headache. Regarding the exercise interventions, there is a great variety of exercise options to be compared, such as high-intensity general exercise for migraine and TTH or general AE for TTH. Another pending subject to assess is the dosage of the interventions, which needs to be established through further investigation.
It would also be interesting to assess whether a high-intensity interval training intervention would be more effective than a moderate AE intervention. Finally, general strength training should be assessed in patients with migraine.

Limitations
The present review has some limitations. First, many of the included studies had low methodological quality and a high risk of bias. The results should be analysed with caution. Second, there was considerable variability between the systematic reviews in terms of the interventions used. Third, six of the MMAs performed included only two meta-analyses, and the other two MMAs included three meta-analyses. Fourth, with the small number of studies included in each of the MMAs, the sensitivity was impossible to calculate. Sixth, due to the lack of data, it was not possible to conduct an MMA for quality of life and disability. Finally, there were significant inconsistencies among the RCTs included in the meta-analyses regarding the diagnostic criteria used to classify the patients. Likewise, it could have been interesting to evaluate the effectiveness of exercise and manual therapy according to the existing subdivisions within each of the headache types, for example, in patients with migraine with or without aura; however, that information was not present in the included studies.

Conclusions
The present umbrella and mapping review with MMA provide an overview of the effects of physical therapy interventions on pain intensity, pain frequency, disability and quality of life of patients with primary headaches. The MMAs showed results in favour of the interventions in terms of pain intensity and quality of life in patients with migraine, TTH and CCH. The data are also in favour of the intervention in terms of frequency of migraine and disability in TTH. However, there were no significant effects on pain frequency in patients with TTH and CGH.
There is moderate evidence to suggest that AE reduces pain intensity in patients with migraine. In addition, the evidence in favour of MT or a mixed intervention (including therapeutic exercise) is also moderate in terms of reducing pain intensity in patients with TTH.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/app11156856/s1, Table S1: Citation matrix for the assessment of study overlap within systematic reviews.