Sage Journals: Discover world-class research

Abstract

Objective

Vagal nerve activity—indexed by heart rate variability (HRV)—has been linked to altered pain processing and inflammation, both of which may underpin headache disorders and lead to cardiovascular disease (CVD). Here we examined the evidence for differences in parasympathetic (vagal) activity indexed by time- and frequency-domain measures of HRV in patients with headache disorders compared to healthy controls (HCs).

Methods

A systematic review and meta-analysis was conducted on studies investigating group differences in vagally mediated HRV (vmHRV) including time- (root-mean-square of successive R-R-interval differences (RMSSD)) and frequency- (high-frequency HRV) domain measures. Studies eligible for inclusion were identified by a systematic search of the literature, based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement.

Results

Seven studies reporting a total of 10 comparisons of patients with headache disorders (HF-HRV n = 67, RMSSD n = 122) and HCs (HF-HRV n = 64, RMSSD n = 125) were eligible for inclusion. Random-effects meta-analysis revealed a significant main effect on RMSSD (Z = 2.03, p = 0.04; Hedges’ g = −0.63; 95% CI (−1.24, –0.02); k = 6) and similar pooled effect size estimates for HF-HRV when breathing was controlled (g = −0.30; 95% CI (−0.69; 0.10)) but not when breathing was not controlled (g = 0.02; 95% CI (−0.69; 0.74)). Controlling for breathing had no effect on RMSSD.

Conclusion

vmHRV is reduced in patients with headache disorders, findings associated with a medium effect size. Suggestions for future research in this area are provided, emphasizing a need to investigate the impact of headache disorders and commonly comorbid conditions—including mental disorders—as well as the investigation of the risk for CVD in migraine in particular. We further emphasize the need for large-scale studies to investigate HRV as a mechanism mediating the association of migraine and CVD.

Keywords

Primary headache disorders migraine tension-type headache heart rate variability parasympathetic activity meta-analysis

Introduction

Autonomic nervous system (ANS) dysfunction has been linked to common headache disorders including migraine (1 –3) and tension-type headache (TTH, 4,5). The association between ANS dysfunction and migraine, in particular, is well documented within the literature. Heart rate variability (HRV) is the beat-to-beat variation in the heart rate. Parasympathetic modulation of the heart rate is fast (timescale in the order of milliseconds) while sympathetic effects are much slower (6,7). Therefore, high-frequency (HF) HRV and time-domain measures reflecting these fast changes (i.e. the root-mean-square of successive R-R-interval differences, RMSSD) provide a readily available, surrogate measure of vagal activity. While most studies have emphasized a role for sympathetic hyperactivity in headache disorders, examining a potential role for vagal parasympathetic activity may provide additional insights into the mechanism underlying headache disorders.

Importantly, there is evidence that migraine with aura (MwA) is associated with increased risk of major cardiovascular disease (CVD), myocardial infarction, ischemic stroke, and death due to ischemic CVD (8). Other studies found that both MwA and migraine without aura (MwoA) are associated with CVD and risk factors for CVD (9,10). A recent systematic review found that migraine is associated with a twofold-increased risk of ischemic stroke, which is apparent only among people who have MwA (11). Therefore, understanding the role of HRV in headache disorders might promote a deeper understanding of the increased risk for CVD in headache patients.

Vagal impairment—as indexed by HRV—may further lead to heightened inflammatory processes, which have also been associated with migraine (12). In particular, impaired vagal efferent activity may lead to excessive inflammation via failure to inhibit release of pro-inflammatory cytokines, reflecting a poorly functioning anti-inflammatory reflex (13 –18). These inflammatory processes may cause prolonged, ongoing excitation of primary nociceptive neurons leading to chronic painful conditions. Recently, we were able to show that vagally mediated HRV (vmHRV) predicts increased levels of inflammation four years later (19), providing evidence for the anti-inflammatory reflex in humans. Vagal-nociceptive networks are also involved in pain processing (20), and decreased vagal activity leads to greater somatic and visceral input via the spinothalamic track. Such loss of sensory integration due to decreased vagal activity may result in greater affective processing of nociceptive information that results in overstraining adaptive capabilities (21).

Lower vmHRV is reported in a variety of chronic painful conditions, such as chronic neck pain (22), chronic pelvic pain (23), complex regional pain (24), fibromyalgia (25,26) and irritable bowel syndrome (27). In light of this work, the vagus nerve is widely perceived as a promising target for therapeutic interventions in the treatment of chronic pain. An important area involved in descending inhibitory modulation of pain is the periaqueductal gray (PAG). It has been shown that stimulation of the ventral PAG increases HRV and decreases pain in humans with chronic pain (28). Given that this pathway is distinct from dorsal PAG stimulation, it has been suggested that analgesia with deep brain stimulation in chronic pain is associated with increased vagal parasympathetic activity (28). Considering these anatomical connections, the prominent role of the vagus nerve in pain processing, and promising findings on vagus nerve stimulation in people with chronic pain (29,30)—including migraine (31) and cluster headaches (32)—is slowly gaining recognition. Vagal activity in headache patients might be more than a simple marker of ANS dysfunction; it may actually be a major contributor to the condition itself. Recently, we were able to show (21), a negative correlation of vmHRV and pain in multivariate adjusted analysis. Most interestingly—and deserving special notice—while individuals with chronic pain reported lower vagal activity, the correlation of pain severity and vmHRV was present only in respondents without chronic pain, suggesting that the descending inhibitory pathway via the vagus nerve is disturbed in patients with chronic pain (21).

Chronic reductions in HRV are also associated with a variety of comorbidities that are themselves associated with headache disorders. While decreased HRV increases risk for morbidity, especially CVD, and is an independent risk factor for mortality (33,34), headache disorders are also associated with comorbid conditions such as depression and anxiety (35,36), which are also associated with reductions in HRV (37,38). The lifetime prevalence of major depression is approximately three times higher in patients with migraine and in individuals with severe headache disorders (39). Depression without CVD is associated with reduced HRV (37,40). While the underlying mechanisms in the association of migraine and psychiatric comorbidity remain unclear (41), vmHRV may provide a valuable tool to further investigate the association of migraine, CVD, and comorbid depression. The present systematic review and meta-analysis reviews the current evidence on differences in vmHRV addressing if vmHRV is reduced in headache patients in comparison to healthy controls (HCs).

Methods

Systematic literature search

A systematic search of the literature, according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (42), was employed. PubMed, PsycNET/PsycINFO, CINAHL Plus, and Web of Science (WOS) databases were searched using the search terms “heart rate variability” and “headache” OR “migraine” (see Appendix A for search strategy by database). The number of initial hits was recorded and additionally a hand search (i.e. Google, Google Scholar and other sources) was performed. Reference lists of included studies were also checked for additional studies eligible for inclusion. After removing duplicates, abstracts of all articles were screened based on predefined inclusion criteria. Studies were included if they reported (i) an empirical investigation in (ii) humans and (iii) recorded HRV in (iv) headache patients compared to HCs. All titles meeting the inclusion criteria were retrieved and reviewed in full text. Empirical investigations were defined as studies involving active data collection in humans. Reviews, meta-analysis, comments, or single-case reports were excluded. Animal studies and studies using a computational modeling approach (i.e. virtual data) were also excluded. The number of studies meeting the prespecified inclusion criteria, number of studies excluded, and reasons for exclusion were recorded (Figure 1).

Figure 1.

Systematic literature search flowchart.

Data extraction

Guidelines from the Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology (43) were used to define the HRV measurements included for analysis. Only components reflecting primarily vagal cardiac modulation were included. These included the root-mean-square of successive R-R-interval differences (RMSSD) or any spectral measure in the high-frequency (HF) range of 0.15–0.4 Hz (natural log transformed (lnHF), normalized (HFnu) or expressed as absolute power in ms² (HFP)). Authors who reported HRV but who did not report RMSSD or HF-HRV, or did not provide sufficient quantitative data (e.g. only a graphical display), were contacted to request the necessary information to derive effect size estimates and confidence intervals (CI) on the selected indices of HRV. Studies that reported multiple groups of headache patients were included as long as they reported findings against HCs, while studies that compared different groups of headache patients only were excluded. When multiple groups of headache patients (e.g. MwoA and MwA) were reported, each group was compared to the same group of HCs.

Meta-analysis

Descriptive statistics (means and standard deviations (SD)) of time- (RMSSD) and frequency- (HF-HRV) domain measures of vmHRV from HRV recordings were extracted for each group. In case descriptive statistics were reported other than as mean and SD, data were imputed if possible (44,45). Where longitudinal or pre-post data were reported, only baseline HRV was included to minimize confounding effects by experimental manipulation and conflation of effect size estimates. True effect estimates were computed as adjusted standardized mean differences (Hedge’s g). We undertook meta-analyses using both fixed-effect and random-effects models. When results of both analyses are consistent (with CI of fixed-effect analyses being included within that of the random-effects analysis), the results from random-effects models are reported, as it better conveys the variability of data. However, when fixed-effect and random-effect models give different results, results of fixed-effects analyses are reported, as they provide a more reliable estimate of the true effect (46). Possible sources of heterogeneity or inconsistency among trials in the magnitude or direction of effects were investigated. Heterogeneity was assessed using the standard I² index, chi-square, and Tau² tests (47). Bias was examined using a funnel plot of effect size against standard error for asymmetry. Meta-analytic computations were performed using RevMan (version 5.3.4, Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). Risk of bias was also assessed in regards to reporting of comorbid conditions, medication intake at the time of testing, other restrictions prior to the recording of HRV (i.e., smoking, alcohol), and sample homogeneity including age, gender, headache diagnosis, headache history and frequency reported. Furthermore, meta-regression on potential covariates was conducted if sufficient data were available. These covariates include headache diagnosis, headache history (duration of disorder), symptom/attack frequency, age differences, gender differences, differences in the length of HRV recording and control for respiration.

Results

Selection of studies

The systematic search of the literature revealed a total of 77 abstracts (after removing duplicates) that were screened for eligibility of inclusion within the present meta-analysis. A total of 18 studies were considered eligible for inclusion and retrieved in full-text (if possible). Of the 77 abstracts evaluated, seven studies (48 –54) yielding a total of 10 comparisons (HF-HRV k = 4, n = 67 headache patients and n = 64 HCs; RMSSD: k = 6, n = 122 headache patients and n = 125 HCs) were included in the meta-analysis. The systematic literature search and reasons for exclusion of studies are illustrated in Figure 1.

Included studies by clinical etiology

Sample characteristics of included studies are summarized in Table 1. The majority of studies investigated patients with migraine. Nilsen and colleagues (49) investigated differences in HRV and spontaneous baroreflex sensitivity in female migraine patients (n = 16) compared to HCs (n = 14). They recorded HRV for 24 hours and 10 minutes of paced breathing. To avoid effect size inflation by including multiple measures from the same participants, only HRV indices derived from the 24-hour measurement period was used. Long-term recordings were also obtained by Tabata and colleagues (51), who investigated differences in the circadian rhythm (48-hour recordings of HRV) in 27 patients with migraine and 24 HCs. The authors report only values derived from the Midline Estimating Statistic of Rhythms (MESOR) variable, a measure derived from cosinor analysis of long-term recordings. There were no significant differences between headache patients and HCs. We included mean and SD of RMSSD from MESOR analysis for meta-analysis. The study by Vollono et al. (52) reported HRV data in different sleep stages and wakefulness and found a significant reduction of RMSSD, HF total power, and NN50 in patients with migraine compared to HCs during waking rest. The authors compare headache patients to a larger sample of HCs and age- and sex-matched HCs, of whom the latter comparison was used for meta-analysis. Sanya and colleagues (53) assessed the autonomic regulation of the heart and peripheral blood vessels in migraine patients (n = 30) and HCs (n = 30) during baroreflex stimulation by oscillatory neck suction. Analysis of five-minute baseline recording (paced breathing) revealed no significant difference in measures of HF-HRV between patients and HCs. Thomsen and colleagues (54) compared cardiovascular response in 50 participants (n = 27 MwoA and n = 23 MwA) and HCs (n = 30), undergoing a head-up tilt test, a cold-pressor test, and a Valsalva maneuver. Data reported from the normal breathing baseline recording (150 consecutive heart beats, outside a migraine attack) expressed as mean successive difference of the numeric values of the R-R intervals (MSD index) were extracted and subjected to meta-analysis. The authors reported no significant difference between headache patients and HCs within this condition.

Table 1.

Sample characteristics of included studies.

Authors	Year	Comparison	Criteria	Duration	Frequency	HCs	N (female), HAP/HC	Age mean (SD) HAP/HC
Gass and Glaros	2013	Mixed^a vs. HCs	IHS criteria (ICHD-II)	> 6 mths	n.r.	Sex and age matched	21 (19)/19 (17)	32.86 (11.74)/30.37 (11.2)
Nilsen et al.^b	2009	MwoA vs. HCs	IHS criteria (ICHD-1)	>1 yr	2.3 (1.1) last mth	Sex and age-matched controls	16 (16)/14 (14)	23.4 (3.2)/22.8 (1.5)
Sanya et al.	2005	MwoA/MwA vs. HCs	IHS criteria	7.6 (4.5) yrs	n.r.	Sex- and age-matched	30 (25)/30 (22)	34 (2)/34 (2)
Tabata et al.	2000	MwoA/MwA vs. HCs	IHS criteria	10.5 (10.4) yrs	5.9 (5.7) p mth	Healthy volunteers	27 (n.r.)^c/24 (n.r.)^d	30.2 (12.7)/31.1 (12.1)
Thomsen et al.	1995 (a)	MwoA vs. HCs	IHS criteria	20 (1 –38) yrs	24 (0.1–72) p yr	Sex- and age-matched controls (±5 yrs)	27 (n.r.)^e/30 (n.r.)^f	n.r. (n.r.)^g/42.0 (n.r.)^h
Thomsen et al.	1995 (b)	MwA vs. HCs	IHS criteria	23 (3 –49) yrs	32 (2 –75) p yr	Sex- and age-matched controls (±5 yrs)	23 (n.r.)ⁱ/30 (n.r.)^j	n.r. (n.r.)^k/42.0 (n.r.)^l
Tubani et al.	2003	CH vs. HCs	IHS criteria	n.r.	n.r.	Cross-matched for sex and age	8 (0)/7 (0)	n.r. (n.r.)^m/n.r. (n.r.)
Vollono et al.	2013	MwoAⁿ vs. HCs	IHS criteria (ICHD-II)	n.r.^o	8.5 (1.4) p mth	Sex and age matched	8 (6)/8 (6)	48.1 (9.3)/46.7 (10.7)

HCs: healthy controls; HAP: headache patients; MwoA: migraine without aura; MwA: migraine with aura; CH: cluster headaches; IHS: International Headache Society; ICHD: International Classification of Headache Disorders; mth: month; yr: year; n.r.: not reported; TTH: tension-type headache.

Eight TTH, seven migraine, six migraine and TTH.

Two comparisons reported for this study one on long-term HRV recordings and one on short-term.

Ratio reported, men to women: 5:19.

Ratio reported, men to women: 6:21.

Ratio reported for the entire sample, men to women: 7:43.

Ratio reported for the entire sample, men to women: 5:25.

Mean age and range reported for the entire sample of headache patients, mean 42.4 (range 19–66).

Range reported: 20–66.

Ratio reported for the entire sample, men to women: 7:43.

Ratio reported for the entire sample, men to women: 5:25.

Mean age and range reported for the entire sample of headache patients, mean 42.4 (range 19–66) used for meta-regression.

Range reported: 20–66.

Age range of headache patients: 26 to 42 years, age of 34 used for meta-regression.

Comorbid sleep-related headache.

Duration reported for every participant but no sample mean.

Gass and Glaros (48) investigated whether HRV measures taken at rest from patients with headache disorders (n = 21; TTH, migraine, and mixed) differed significantly from those without headaches (n = 19). They reported that headache patients had less variability in all HRV time domain measures, increased sympathetic nervous system (SNS) activity and decreased parasympathetic nervous system (PNS) activity compared to HCs. However, significant differences between groups were reported only for pNN50 and NN50 indices of HRV—both measures of vmHRV. RMMSD and HF-HRV derived from five-minute recordings of HRV were subjected to the meta-analysis. Furthermore, one study by Tubani and colleagues (50) investigated differences in 24-hour recordings of HRV in patients with cluster headache (n = 8) compared to HCs (n = 8). Results revealed significant differences in low-frequency HRV (LF-HRV) but not HF-HRV. Reported HF-HRV values for the entire length of the recording (24-hour) were subjected to meta-analysis. Details on extracted HRV values and conditions are provided in Table 2.

Table 2.

Details on HRV recording and data extracted from included studies.

					Mean HF (SD)		RMSSD mean (SD)
Authors	Year	Comparison	Condition extracted	vmHRV finding	HAP	HCs	HAP	HCs
Gass and Glaros	2013	Mixed vs. HCs	Five min, resting (breathing recorded)	Ø	49.50 (19.44)	59.74 (18.77)	46.20 (38.14)	63.79 (33.19)
Nilsen et al.	2009	MwoA vs. HCs	24 hours	Ø	n.r. (n.r.)	n.r. (n.r.)	85.2 (42.8)	68.9 (23.8)
Sanya et al.	2005	MwoA/MwA vs. HCs	Five min baseline (paced breathing)	Ø	6.21 (0.26)ⁱ	6.45 (0.31)ⁱ	n.r. (n.r.)	n.r. (n.r.)
Tabata et al.	2000	MwoA/MwA vs. HCs	48-hour recordings, averaged to 24 hours (MESOR)	Ø	n.r. (n.r.)	n.r. (n.r.)	41.6 (0.2)^a	43.6 (0.2)^a
Thomsen et al.	1995 (a)	MwoA vs. HCs	150 consecutive heart beats, rest supine (normal breathing)	Ø	n.r. (n.r.)	n.r. (n.r.)	3.6 (0.3)^b	4.9 (0.7)^b
Thomsen et al.	1995 (b)	MwA vs. HCs	150 consecutive heart beats, rest supine (normal breathing)	Ø	n.r. (n.r.)	n.r. (n.r.)	3.6 (0.5)^b	4.9 (0.7)^b
Tubani et al.	2003	CH vs. HCs	24 hours	Ø	9.043 (3.796)	7.654 (3.002)	n.r. (n.r.)	n.r. (n.r.)
Vollono et al.	2013	MwoA^c vs. HCs	Five min, quiet wakefulness	HAP ↓RMSSD and HF^d	33.7 (n.r.)^e	40.7 (n.r.)^d,e	24.6 (n.r.)	35.4 (n.r.)

HCs: healthy controls; HAP: headache patients; HF: high-frequency; RMSSD: root-mean-square of successive R-R-interval differences; MwoA: migraine without aura; MwA: migraine with aura; CH: cluster headaches; n.r.: not reported; n.a.: not applicable; h: hour; min: minute; vmHRV: vagally mediated heart rate variability; Ø: no significant differences comparing HAP and HCs.

SD imputed from standard error.

Reported as MSD in percentage of mean R-R. SD imputed from standard error.

Comorbid sleep-related headache.

Significant lower total HF power in headache patients compared to matched controls reported. Normalized units used for analysis (no significant effect).

SD imputed from p values of Mann-Whitney U test.

Meta-analysis main effect

First, we aimed to identify an overall effect of headache disorder on vmHRV, followed by subsequent analysis that aimed to investigate differences between clinical etiologies (diagnosis). Random-effect meta-analysis, on RMSSD, revealed that headache patients (n = 122) displayed significantly lower vmHRV compared to HCs (n = 125) (Z = 2.03, p = 0.04) (g = −0.63; 95% CI (−1.24, −0.02); k = 6) indicating lower vagal activity in headache patients. CIs of fixed-effect models (Z = 4.35, p < 0.00001; Hedges’ g = −0.58; 95% CI (−minus0.85, −0.32); k = 6) were included in those of random-effect models, thus random-effects are illustrated in Figure 2 (negative effect estimates reflect lower RMSSD in headache patients).

Figure 2.

Random-effect meta-analysis main effect Forrest plot for root-mean-square of successive R-R-interval differences (RMSSD).

Random-effect meta-analyses on HF-HRV revealed no significant difference (Z = 1.26, p = 0.21) between headache patients (n = 67) and HCs (n = 64) (Hedges’ g = −0.22; 95% CI (−0.57, 0.12); k = 4). As the CI of the fixed-effects model (Z = 1.26, p = 0.21; Hedges’ g = −0.22; 95% CI (−0.57, 0.12); k = 4) was included in that of the random-effects model, results of random-effects are illustrated (Figure 3; negative effect estimates reflect lower HF in headache patients).

Figure 3.

Random-effect meta-analysis main effect Forrest plot for headache disorders-heart rate variability (HF-HRV).

Significant heterogeneity was observed for RMSSD (see test results in Figure 2), and visual examination of funnel plots for HF (Figure 4(b)) and RMSSD (Figure 4(a)) revealed considerable asymmetry for RMSSD, indicating publication bias. Substantial heterogeneity was assumed if I² was greater than 50%, indicating that 50% of the variability in the outcome cannot be explained by sampling variation.

Figure 4.

Funnel plot for (a) RMSSD and (b) HF-HRV.

As we assumed that the significant effect on RMSSD might relate to the considerable asymmetry of included studies, in particular the unlikely large effect reported by Tabata et al. (51), and given that this effect estimate was derived from a different analytical approach (MESOR analysis) we excluded the study in a secondary analysis (leave-one-out meta-analysis). When excluding the study (see outlier in Figure 4), the significant main effect on RMSSD remained in fixed-effect meta-analysis (Z = 2.33, p = 0.02; Hedges’ g = −0.34; 95% CI (−0.63, −0.05); k = 5) (CIs of fixed-effect not included in those of random-effect models). Thus, we conclude that the main effect reported for RMSSD is likely driven by study heterogeneity, but remains significant when controlling for such bias.

Risk of bias and meta-regression

Risk of bias assessment (Table 3), revealed large risk of bias regarding the reporting of comorbid conditions (i.e. depression, anxiety), medication intake, and other restrictions prior to the examination (i.e. smoking, alcohol) that have substantial effects on HRV. Furthermore, reported samples from studies showed large heterogeneity in age, gender, headache history and symptom frequency.

Table 3.

Risk of bias assessment of included studies.

Participants taking medications for cardiovascular control, such as beta-blockers, were excluded, no analgesics for any reason within 12 hours, no permanent medication reported.

void exercise or use of any caffeine, tobacco, or alcohol products 30 minutes prior.

Large SD in age (> 10 years).

Dominantly female.

Mixed sample of migraine and tension-type headache (TTH).

Any duration longer than six months.

No prophylactic drugs.

No headache or use of alcohol three days before the investigation.

Dominantly female.

Not reported in detail but similar age of onset.

Asked to refrain from consuming alcohol or caffeine for 24 hours prior to the tests.

Dominantly female.

Migraine without aura (MwoA) and migraine with aura (MwA).

Mean migraine duration of 7.6 (4.5) years with similar intensity.

Ceased taking prophylactic medication for at least one month.

Large SD in age (> 10 years).

Dominantly female.

MwoA and MwA, majority MwoA.

0.5 to 12 attacks per month with a duration of 2 to 72 hours.

One to 35 years.

Not allowed to take ergotamine or morphinomimetics 48 hours prior to examinations, sumatriptan 24 hours prior and weak analgesics six hours prior to examinations.

Headache medication but no restrictions on smoking etc.

19 to 66 years.

Dominantly female.

0.1 to 75 attacks per year.

One to 49 years.

Allowed to use.

26 to 42 years.

Large SD in age (> 10 years).

Dominantly female.

MwoA but with comorbid sleep-related headaches.

We performed a series of meta-regressions and subgroup analysis to identify potential covariates of the observed main effect found. There were not enough studies to address differences between studies recording HRV during uncontrolled (free) breathing and paced breathing using meta-regression on HF but RMSSD (Table 2). In a sub-group analysis on HF, studies that controlled or recorded breathing showed a significant greater pooled effect estimate for HF (g = −0.30; 95% CI −0.69; 0.10)), compared to studies using uncontrolled conditions (g = 0.02; 95% CI (−0.69; 0.74)). Controlled breathing was not a significant covariate of RMSSD (β = 0.822, SE = 0.637; 95% CI (−0.427; 0.637); p = 0.197). Further subgroup analysis on etiology indicated lower RMSSD in migraine patients (g = −0.93; 95% CI (−1.69; –0.18)) compared to a sample of patients with a variety of headache disorders, including migraine and TTH (g = −0.48; 95% CI (−0.48; 0.15)). Furthermore, patients with cluster headache showed increased HF (g = −0.48; 95% CI (−0.65; 1.41)) and patients with migraine decreased HF (g = −0.18; 95% CI (−0.63; 0.27)). No other covariate yielded a significant influence on the effects reported.

Discussion

Research on differences in vagal activity in headache patients may have important implications for better understanding the elevated risk for CVD in these individuals. The present meta-analysis is the first to summarize the current evidence on vagal activity as indexed by vmHRV in headache patients. Meta-analysis revealed a significant and sizeable main effect on RMSSD (Hedges’ g = −0.63) and similar effect-size estimates for HF-HRV when breathing was controlled (Hedges’ g = −0.30). The effect for both selected indices of vmHRV revealed similar findings, indicating decreased vagal activity in headache patients.

While our initial analysis indicated significant differences between headache patients and HCs on RMSSD only, subgroup analysis showed an effect of respiratory control on the reported effects for HF-HRV, but not RMSSD. Our findings demonstrate that studies controlling for respiration reported larger effect sizes than those that did not control for respiration. Thus, the results for RMSSD and HF-HRV are consistent when respiration is controlled. The fact that respiration had a greater effect on HF-HRV than on RMSSD is not surprising. Our group (55) as well others have shown that RMSSD is less affected by respiration than other indices of HRV including spectrally derived HF-HRV (56). HRV indices derived from frequency domain analysis provide information of different quality and detail compared to time-domain analysis (57). While RMSSD and HF are highly correlated (58), it has been suggested that time-domain parameters can be estimated with less bias and considerably smaller variability as compared with frequency-domain parameters (59).

While we selected the two most prominent and frequently used parameters of vmHRV (RMSSD and HF) for the present analysis, other indices reflecting vagal activity such as the NN50 or pNN50 component are of interest. Three studies by Gass and Glaros (48), Tabata et al. (51) and Vollono et al. (52) reported these and found significantly lower NN50 and pNN50 in headache patients compared to HCs in line with our findings of lower vagal activity indexed by RMSSD and HF.

Of the four studies reporting RMSSD, only two studies initially reported significant differences between headache patients and HCs (49,52), while three found no significant differences in the selected indices of vagal activity (48,51,54). One study that reported significant differences found RMSSD to be greater in headache patients (49), and one found RMSSD to be lower in headache patients (52). Except for one study (52), none of the studies on HF-HRV initially reported significant differences between headache patients and HCs (48,50,53). This further highlights the utility of meta-analysis to provide a more objective assessment of the extant literature.

Results from subgroup analysis on diagnosis of headache disorder revealed lower vagal activity in migraineurs compared to mixed samples of patients with a variety of headache disorders, including migraine and TTH, and patients with cluster headache. These findings provide first the evidence for a unique association of migraine and vagal activity, lending further support for our hypothesis that reduced vagal activity may provide a physiological pathway linking migraine and CVD. However, it is important to note that there are insufficient data to determine robust differences in vagal activity comparing different diagnosis of primary headache disorders. Regarding a potential pathway linking HRV and CVD in headache patients, the distinction between different headache disorders is crucial, as different diagnoses are associated with different risk factors for CVD (60), and it is proposed that different mechanism underlie the elevated CVD risk in MwA in comparison to other headache types (61,62). In general, symptom history and frequency—both important diagnostic qualities—are inadequately reported in most of the included studies, and as a result we did not have sufficient data to perform meta-regression on symptom/attack frequency. In fact only two studies provided adequate reporting of headache frequency (49,52) and headache history (52,53). We analyzed the reported mean duration of headache symptoms as a potential covariate on RMSSD but found no significant differences. Given that, in most cases, studies on migraine patients did not distinguish between MwA and MwoA symptoms, we were not able to address the potential impact of migraine aura. Critically, past research demonstrates that the association of migraine and CVD might depend on the presence of aura symptoms (63). One of the included studies explicitly addressed differences in MwA and MwoA patients (54), but found no significant differences. Given that we were not able to address differences in vagal activity between MwoA and MwA using meta-analysis, future studies on differences between the two etiologies are encouraged. Thus, future studies are needed to (i) compare different diagnosis of primary headache disorders, (ii) investigate the association of symptom severity (i.e. frequency, intensity, duration of attacks) and vmHRV, as well as (iii) the (longitudinal) association of headache history and vmHRV, and (iv) address differences in MwA and MwoA.

While we found age not to be a significant covariate of HF-HRV, interpretations should be drawn with caution given the relatively small mean differences in age between samples (Table 1) included in the present analysis. HRV decreases with age (64 –67) and future studies should investigate age as an important covariate—in particular linking the age of onset of symptoms and age at time of HRV recording. We were unable to explore potential gender differences in HRV that have recently been reported (68 –72) as included studies reported only mixed (women and men) samples. Lastly, we compared short- (<1 hour) with long-term recordings (24 or 48 hours) using meta-regression, but found no significant differences. It is noted that guidelines for the measurement of HRV (45) suggest that spectral analysis of 24-hour long-term HRV (where spectral estimates are calculated over long data epochs that are not likely to be stationary) may not accurately reflect autonomic modulation, which may be better captured by estimates based on shorter data epochs. Future studies should take this into account. Furthermore, no study recording HRV for at least 24 hours (49 –51) addressed potential differences between awake and sleep periods. The data by Vollono et al. (52) show that such analyses may be important to address the involvement of the vagus nerve, as during nighttime, movement artifacts are minimized and the contribution of sympathetic influences is low, in particular when further exploring sleep-related headache.

The present systematic review and meta-analysis faces several limitations that need to be addressed in future research, especially given the availability of primary research on this topic. Overall, this analysis was performed on a relatively small data set, including only seven studies, all with high risk of bias (Table 3). We were not able to retrieve one full-text of a study conducted in Poland (73) that compared 40 headache patients to 62 HCs. One study conducted in children and adolescents (74) with potential data of interest, that also reviewed the existing literature without taking a meta-analytical approach, did not report the respective indices of interest (HF-HRV or RMSSD) and we were not able to obtain the data. Two studies published in 1993 by the same group of authors, investigating ANS function in patients with TTH (75) and migraine (76), did not report sufficient baseline data (i.e. a graphical display only instead of reporting exact means and SDs) on vagal indices of HRV, and again, we were not able to obtain the data. Another three studies, two by Zigelman et al. (77,78) and one by Appel et al. (79) were insufficient in terms of reporting data of interest, and we were unable to contact the authors. Pierangeli and colleagues (80) explored differences in cardiovascular responses to the tilt test and Valsalva maneuver in HCs compared to patients with MwoA and MwA, but provide data on HF as a graphical display only and we were unable to contact the authors. Within a commentary to a paper by Kurth et al. (10), Perciaccante et al. (81) report some preliminary findings on HRV differences in headache patients compared to HCs. Again, however, we were not able to obtain the necessary data to include the study in the present meta-analysis. The study by Shechter et al. (82) reported RR variation but no indices of vmHRV, and we were not able to obtain the data. We also requested details on the HRV measures taken in the study by Aygül et al. (83) that reported only RR variation but not values for HF and RMSSD, but did not receive a response from the authors. Studies that illustrate but insufficiently report findings on vagal indices of HRV are inconsistent, reporting lower parasympathetic activity in headache patients compared to HCs (73) or no differences (74 –76,79). Given the exclusion of these 11 studies (73 –83) with potential data of interest, the present analysis is limited and not completely reflective of the existing evidence.

In addition, it is critical to note that all included studies carry a potential high risk of bias (Table 3). In general, most studies failed to report medication in sufficient detail, an important issue considering well-known effects on HRV (37,84). Furthermore, most studies failed to control for frequent comorbid conditions described in the context of headache disorders (i.e. anxiety and depression) that are also linked with HRV. We also found large methodological differences between studies regarding the recording of HRV. Specifically, some studies asked participants to avoid exercise, smoking or alcohol consumption prior to the recording of HRV, while others did not control for such influences. We addressed publication bias and study heterogeneity excluding one study on RMSSD (38), reporting an unlikely large effect and found the main effect for RMSSD to hold.

Lower vagal activity, indexed by vmHRV, is associated with an increased risk for CVD. Furthermore, lower vmHRV is associated with greater inflammation and less inhibitory control in pain processing. Chronic pain patients show lower HRV, and vagus nerve stimulation in chronic pain patients (including migraine and cluster headache) may lead to pain release and a simultaneous increase in HRV. Therefore, reduced vmHRV may reflect an important mechanism underlying headache disorders and an increased risk for CVD in headache patients. Three potential models (61) may underpin the link between headache and CVD: (i) the frequent experience of headache may lead to lifestyle changes such as decreased physical activity, altered smoking habits or an unhealthy diet—all of which may reduce vmHRV—that elevate CVD risk; (ii): an unfavorable CVD risk profile is associated with reduced vmHRV, which could subsequently lead to the development of headache; and (iii) genetic or environmental causes common both to headache and an unfavorable CVD risk profile may lead to reductions in vmHRV. Here we emphasize how vmHRV may relate to headache disorders and CVD; however, future research is needed to determine whether vmHRV is actually a mediating factor of headache disorders and its relationship to CVD. Future studies are needed to: (i) carefully distinguish between different diagnosis based on established criteria, (ii) examine homogenous samples of patients of the same age, gender, with the same diagnosis, and headache characteristics (i.e. frequency, intensity, headache history) that (iii) control for comorbid conditions, and (iv) determine the impact of medication on vagal activity, in addition to primary headache disorders and comorbid conditions.

Conclusion

We demonstrate here that vmHRV is reduced in patients with primary headache disorders, findings associated with a medium effect size. Future studies are needed to further examine the association between headache disorders—and migraine in particular—vmHRV, common comorbid conditions and risk for CVD. We emphasize the necessity of future large-scale studies to further investigate these pathways.

Clinical implications

This is the first meta-analysis on vagal activity indexed by heart rate variability in headache patients compared to healthy controls.

Vagal activity is reduced in headache patients compared to healthy controls.

Lower vagal activity provides a mechanism linking headache disorders—in particular migraine—to cardio vascular disease and comorbid depression.

Footnotes

Funding

This work was supported by FAPESP, a Brazilian research funding institution in the state of São Paulo, and by Ohio State University, which has helped to initiate collaborative activities between the authors of the current manuscript.

Acknowledgment

We thank Fiorella Pepe, executive secretary, Segreteria Società Italiana di Medicina Interna, for providing us with the full text of one included paper.

Appendix A—search strategy by database

References

Peroutka

. Migraine: A chronic sympathetic nervous system disorder. Headache 2004; 44: 53–64.

Rubin

Graham

Pasker

, et al. Autonomic nervous system dysfunction in common migraine. Headache 1985; 25: 40–48.

Thomsen

Olesen

. The autonomic nervous system and the regulation of arterial tone in migraine. Clin Auton Res 1995; 5: 243–250.

Yerdelen

Acil

Goksel

, et al. Autonomic function in tension-type headache. Acta Neurol Belg 2007; 107: 108–111.

Yerdelen

Acil

Goksel

, et al. Heart rate recovery in migraine and tension-type headache. Headache 2008; 48: 221–225.

Levy

. Neural control of cardiac function. Baillieres Clin Neurol 1997; 6: 227–244.

Appelhans

Luecken

. Heart rate variability as an index of regulated emotional responding. Rev Gen Psychol 2006; 10: 229–240.

Kurth

Gaziano

Cook

, et al. Migraine and risk of cardiovascular disease in women. JAMA 2006; 296: 283–291.

Bigal

Kurth

Santanello

, et al. Migraine and cardiovascular disease: A population-based study. Neurology 2010; 74: 628–635.

10.

Kurth

Gaziano

Cook

, et al. Migraine and risk of cardiovascular disease in men. Arch Intern Med 2007; 167: 795–801.

11.

Schürks

Rist

Bigal

, et al. Migraine and cardiovascular disease: Systematic review and meta-analysis. BMJ 2009; 339: b3914–b3914.

12.

Waeber

Moskowitz

. Migraine as an inflammatory disorder. Neurology 2005; 64(10 Suppl 2): S9–S15.

13.

Thayer

. Vagal tone and the inflammatory reflex. Cleve Clin J Med 2009; 76(Suppl 2): S23–S26.

14.

Thayer

Fischer

. Heart rate variability, overnight urinary norepinephrine and C-reactive protein: Evidence for the cholinergic anti-inflammatory pathway in healthy human adults. J Intern Med 2009; 265: 439–447.

15.

von Känel

Thayer

Fischer

. Nighttime vagal cardiac control and plasma fibrinogen levels in a population of working men and women. Ann Noninvasive Electrocardiol 2009; 14: 176–184.

16.

Huston

Tracey

. The pulse of inflammation: Heart rate variability, the cholinergic anti-inflammatory pathway and implications for therapy. J Intern Med 2011; 269: 45–53.

17.

Thayer

Loerbroks

Sternberg

. Inflammation and cardiorespiratory control: The role of the vagus nerve. Respir Physiol Neurobiol 2011; 178: 387–394.

18.

Papaioannou

Pneumatikos

Maglaveras

. Association of heart rate variability and inflammatory response in patients with cardiovascular diseases: Current strengths and limitations. Front Physiol 2013; 4: 174–174.

19.

Jarczok

Koenig

Mauss

, et al. Lower heart rate variability predicts increased level of C-reactive protein 4 years later in healthy, nonsmoking adults. J Intern Med 2014; 276: 667–671.

20.

Randich

Gebhart

. Vagal afferent modulation of nociception. Brain Res Brain Res Rev 1992; 17: 77–99.

21.

Koenig J, Loerbroks A, Jarczok MN, et al. Chronic pain and heart rate variability in a cross-sectional occupational sample: Evidence for impaired vagal control. Clin J Pain. In press.

22.

Kang

Chen

, et al. Disability in patients with chronic neck pain: Heart rate variability analysis and cluster analysis. Clin J Pain 2012; 28: 797–803.

23.

Cho

Choi

Kim

, et al. Heart rate variability in assessment of autonomic dysfunction in patients with chronic prostatitis/chronic pelvic pain syndrome. Urology 2011; 78: 1369–1372.

24.

Terkelsen

Mølgaard

Hansen

, et al. Heart rate variability in complex regional pain syndrome during rest and mental and orthostatic stress. Anesthesiology 2012; 116: 133–146.

25.

Martínez-Lavín

Hermosillo

Rosas

, et al. Circadian studies of autonomic nervous balance in patients with fibromyalgia: A heart rate variability analysis. Arthritis Rheum 1998; 41: 1966–1971.

26.

Lerma

Martinez

Ruiz

, et al. Nocturnal heart rate variability parameters as potential fibromyalgia biomarker: Correlation with symptoms severity. Arthritis Res Ther 2011; 13: R185–R185.

27.

Mazurak

Seredyuk

Sauer

, et al. Heart rate variability in the irritable bowel syndrome: A review of the literature. Neurogastroenterol Motil 2012; 24: 206–216.

28.

Pereira

Wang

, et al. Ventral periaqueductal grey stimulation alters heart rate variability in humans with chronic pain. Exp Neurol 2010; 223: 574–581.

29.

Multon

Schoenen

. Pain control by vagus nerve stimulation: From animal to man … and back. Acta Neurol Belg 2005; 105: 62–67.

30.

George

Nahas

Bohning

, et al. Vagus nerve stimulation therapy: A research update. Neurology 2002; 59(6 Suppl 4): S56–S61.

31.

Hord

Evans

Mueed

, et al. The effect of vagus nerve stimulation on migraines. J Pain 2003; 4: 530–534.

32.

Mauskop

. Vagus nerve stimulation relieves chronic refractory migraine and cluster headaches. Cephalalgia 2005; 25: 82–86.

33.

Thayer

Yamamoto

Brosschot

. The relationship of autonomic imbalance, heart rate variability and cardiovascular disease risk factors. Int J Cardiol 2010; 141: 122–131.

34.

Kemp

Quintana

. The relationship between mental and physical health: Insights from the study of heart rate variability. Int J Psychophysiol 2013; 89: 288–296.

35.

Pesa

Lage

. The medical costs of migraine and comorbid anxiety and depression. Headache 2004; 44: 562–570.

36.

Antonaci

Nappi

Galli

, et al. Migraine and psychiatric comorbidity: A review of clinical findings. J Headache Pain 2011; 12: 115–125.

37.

Kemp

Quintana

Gray

, et al. Impact of depression and antidepressant treatment on heart rate variability: A review and meta-analysis. Biol Psychiatry 2010; 67: 1067–1074.

38.

Kemp

Brunoni

Santos

, et al. Effects of depression, anxiety, comorbidity, and antidepressants on resting-state heart rate and its variability: An ELSA-Brasil Cohort Baseline Study. Am J Psychiatry 2014; 17: 1328–1334.

39.

Breslau

Schultz

Stewart

, et al.

Headache and major depression: Is the association specific to migraine?

Neurology 2000; 54: 308–313.

40.

Kemp

Quintana

Felmingham

, et al. Depression, comorbid anxiety disorders, and heart rate variability in physically healthy, unmedicated patients: Implications for cardiovascular risk. PLoS One 2012; 7: e30777–e30777.

41.

Antonaci

Nappi

Galli

, et al. Migraine and psychiatric comorbidity: A review of clinical findings. J Headache Pain 2011; 12: 115–125.

42.

Moher

Liberati

Tetzlaff

, et al. Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. BMJ 2009; 151: 264–269.

43.

Task force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. Heart rate variability: Standards of measurement, physiological interpretation, and clinical use. Eur Heart J 1996; 17: 354–381.

44.

Wiebe

Vandermeer

Platt

, et al. A systematic review identifies a lack of standardization in methods for handling missing variance data. J Clin Epidemiol 2006; 59: 342–353.

45.

Glass

McGaw

Smith

Measuring study findings. In: Glass

McGaw

Smith

(eds). Meta-analysis in social research, Beverly Hills, CA: Sage Publications, 1981, pp. 93–152.

46.

Higgins JPT and Green S. handbook for systematic reviews of interventions. Version 5.1.Cochrane 0, http://handbook.cochrane.org (accessed 22 April 2015).

47.

Higgins

JPT

Thompson

. Quantifying heterogeneity in a meta-analysis. Stat Med 2002; 21: 1539–1558.

48.

*Gass

Glaros

. Autonomic dysregulation in headache patients. Appl Psychophysiol Biofeedback 2013; 38: 257–263.

49.

*Nilsen

Tronvik

Sand

, et al. Increased baroreflex sensitivity and heart rate variability in migraine patients. Acta Neurol Scand 2009; 120: 418–423.

50.

*Tubani

Baratta

Giorgino

, et al. Heart rate variability in cluster headache. Ann Ital Med Int 2003; 18: 42–46.

51.

*Tabata

Takeshima

Burioka

, et al. Cosinor analysis of heart rate variability in ambulatory migraineurs. Headache 2000; 40: 457–463.

52.

*Vollono

Gnoni

Testani

, et al. Heart rate variability in sleep-related migraine without aura. J Clin Sleep Med 2013; 9: 707–714.

53.

*Sanya

Brown

von Wilmowsky

, et al. Impairment of parasympathetic baroreflex responses in migraine patients. Acta Neurol Scand 2005; 111: 102–107.

54.

*Thomsen

Iversen

Boesen

, et al. Transcranial Doppler and cardiovascular responses during cardiovascular autonomic tests in migraineurs during and outside attacks. Brain 1995; 118: 1319–1327.

55.

Hill

Siebenbrock

. Are all measures created equal? Heart rate variability and respiration - biomed 2009. Biomed Sci Instrum 2009; 45: 71–76.

56.

Penttilä

Helminen

Jartti

, et al. Time domain, geometrical and frequency domain analysis of cardiac vagal outflow: Effects of various respiratory patterns. Clin Physiol 2001; 21: 365–376.

57.

Sinnreich

Kark

Friedlander

, et al. Five minute recordings of heart rate variability for population studies: Repeatability and age-sex characteristics. Heart 1998; 80: 156–162.

58.

Goedhart

van der Sluis

Houtveen

, et al. Comparison of time and frequency domain measures of RSA in ambulatory recordings. Psychophysiology 2007; 44: 203–215.

59.

Kuss

Schumann

Kluttig

, et al. Time domain parameters can be estimated with less statistical error than frequency domain parameters in the analysis of heart rate variability. J Electrocardiol 2008; 41: 287–291.

60.

Cirillo

Stellato

Lombardi

, et al. Headache and cardiovascular risk factors: Positive association with hypertension. Headache 1999; 39: 409–416.

61.

Winsvold

Hagen

Aamodt

, et al. Headache, migraine and cardiovascular risk factors: The HUNT study. Eur J Neurol 2011; 18: 504–511.

62.

Bigal

Kurth

, et al. Migraine and cardiovascular disease: Possible mechanisms of interaction. Neurology 2009; 72: 1864–1871.

63.

Kurth

Gaziano

Cook

, et al. Migraine and risk of cardiovascular disease in women. JAMA 2006; 296: 283–291.

64.

Zhang

. Effect of age and sex on heart rate variability in healthy subjects. J Manipulative Physiol Ther 2007; 30: 374–279.

65.

Yeragani

Sobolewski

Kay

, et al. Effect of age on long-term heart rate variability. Cardiovasc Res 1997; 35: 35–42.

66.

Umetani

Singer

McCraty

, et al. Twenty-four hour time domain heart rate variability and heart rate: Relations to age and gender over nine decades. J Am Coll Cardiol 1998; 31: 593–601.

67.

Reardon

Malik

. Changes in heart rate variability with age. Pacing Clin Electrophysiol 1996; 19: 1863–1866.

68.

Saleem

Hussain

Majeed

, et al. Gender differences of heart rate variability in healthy volunteers. J Pak Med Assoc 2012; 62: 422–425.

69.

Ramaekers

Ector

Aubert

, et al.

Heart rate variability and heart rate in healthy volunteers. Is the female autonomic nervous system cardioprotective?

Eur Heart J 1998; 19: 1334–1341.

70.

Tanaka

Kimura

Goyagi

, et al. Gender differences in baroreflex response and heart rate variability in anaesthetized humans. Br J Anaesth 2004; 92: 831–835.

71.

Huikuri

Pikkujämsä

Airaksinen

, et al. Sex-related differences in autonomic modulation of heart rate in middle-aged subjects. Circulation 1996; 94: 122–125.

72.

Thayer

. A meta-analysis of sex differences in heart rate variability. Psychosomatic Medicine 2014; 76: A10–A11.

73.

Prusiński

Trzos

Rozentryt

, et al. Studies of heart rhythm variability in migraine. Preliminary communication [article in Polish]. Neurol Neurochir Pol 1994; 28: 23–27.

74.

Ebinger

Kruse

Just

, et al. Cardiorespiratory regulation in migraine. Results in children and adolescents and review of the literature. Cephalalgia 2006; 26: 295–309.

75.

Pogacnik

Sěga

Mesec

, et al. Autonomic function testing in patients with tension-type headache. Headache 1993; 33: 63–68.

76.

Pogacnik

Sega

Pecnik

, et al. Autonomic function testing in patients with migraine. Headache 1993; 33: 545–550.

77.

Zigelman

Kuritzky

Appel

, et al. Propranolol in the prophylaxis of migraine—evaluation by spectral analysis of beat-to-beat heart rate fluctuations. Headache 1992; 32: 169–174.

78.

Zigelman

Appel

Davidovitch

, et al. The effect of verapamil calcium antagonist on autonomic imbalance in migraine: Evaluation by spectral analysis of beat-to-beat heart rate fluctuations. Headache 1994; 34: 569–577.

79.

Appel

Kuritzky

Zahavi

, et al. Evidence for instability of the autonomic nervous system in patients with migraine headache. Headache 1992; 32: 10–17.

80.

Pierangeli

Parchi

Barletta

, et al. Power spectral analysis of heart rate and diastolic blood pressure variability in migraine with and without aura. Cephalalgia 1997; 17: 756–760.

81.

Perciaccante

Fiorentini

Valente

, et al. Migraine and heart rate variability. Arch Intern Med 2007; 167: 2264–2264.

82.

Shechter

Stewart

Silberstein

, et al. Migraine and autonomic nervous system function: A population-based, case-control study. Neurology 2002; 58: 422–427.

83.

Aygül

Deniz

Orhan

, et al. R-R interval variation in migraine patients. East J Med 2006; 11: 1–6.

84.

Licht

de Geus

van Dyck

, et al. Longitudinal evidence for unfavorable effects of antidepressants on heart rate variability. Biol Psychiatry 2010; 68: 861–868.

Vagally mediated heart rate variability in headache patients—a systematic review and meta-analysis

Abstract

Objective

Methods

Results

Conclusion

Keywords

Introduction

Methods

Systematic literature search

Data extraction

Meta-analysis

Results

Selection of studies

Included studies by clinical etiology

Meta-analysis main effect

Risk of bias and meta-regression

Discussion

Conclusion

Clinical implications

Footnotes

Funding

Acknowledgment

Appendix A—search strategy by database

References