Smoking in primary headaches – a systematic review and meta-analysis

Abstract

Background

Primary headaches mainly consist of headaches such as migraine, tension-type headache (TTH), and cluster headache (CH). There is contradictory data concerning the association between smoking cigarettes and headaches. The objective of this study was to evaluate the prevalence of an association between smoking and primary headaches.

Methods

A systematic review was conducted using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines. PubMed and Scopus were screened up until February 2025. Next, those studies with eligible criteria related to this topic were included in a meta-analysis. To assess risk of bias, the Joanna Briggs Institute tool (JBI) was used. The random effects model was used to estimate odds ratios (OR) and smoking prevalence.

Results

There were 2,713 records out of which 37 studies were included in the meta-analysis. 22 out of the 37 included studies had a low risk of bias. The prevalence of smoking in migraine was 20% (95% CI: 16–24) and in migraine with aura (MwA) 27% (95% CI%: 17–27). For TTH, the prevalence was 19% (95% CI: 7–34) and in CH, it was 65% (95% CI: 55–76). The overall smoking prevalence in primary headaches was 32% (95% CI: 8–62). Current smoking was associated with an increased risk of migraine (OR = 1.29, 95% CI: 1.02–1.62, p = 0.034) compared to the group without any headache. Meta-regression revealed that neither age, sex, nor year of publication influenced this result. Current smoking was associated with a decreased risk of TTH (OR = 0.78, 95% CI: 0.68–0.89, p < 0.001). No association was observed between current smoking and CH and MwA; additionally, no association was found between former smoking and migraine. Meta-regression could not be conducted for TTH and CH.

Conclusions

Current smoking was associated with an increased risk of migraine and a decreased risk of TTH. There was no association between current smoking and CH despite the fact that over half of CH patients smoked. The influence of electronic cigarettes commonly used these days is not yet explored in primary headaches, despite the wide-spread use of such cigarettes among the young population who are vulnerable to primary headache onset.

Supplementary Information

The online version contains supplementary material available at 10.1186/s10194-025-02076-2.

Introduction

Primary headaches are the most common neurological condition characterized by having a headache as the main symptom without being caused by other diseases. The main types of primary headaches are migraine, tension-type headache (TTH), and trigeminal autonomic cephalalgias (TACs) with the main representative being cluster headache (CH, [1]). The global prevalence of active headaches is significant and may affect up to approximately 52% of subjects worldwide [2]. It is estimated that about 15.8% of people experience headaches every day [3]. Primary headaches and their comorbidities constitute a serious public health burden, generating direct and indirect costs for healthcare systems around the world [4, 5]. Although the exact pathomechanisms responsible for the development of primary headaches are not fully understood, the importance of environmental and lifestyle factors – including tobacco smoking – is postulated [6–8].

There is evidence to suggest that both active and passive smoking may increase the frequency and severity of headaches, especially migraine [9, 10] and cluster headaches [11, 12]. However, the exact relationship between smoking and migraine is not entirely clear due to minor discrepancies in the results of existing studies and the limited available evidence [6]. A similar situation occurs in the case of CH, where many patients with CH are heavy smokers, suggesting a possible link between smoking and the occurrence of CH episodes [13, 14]. In the case of TTH, the role of smoking is even less clear. There are no studies in the literature exploring the association between tobacco use and TTH; however, the overall impact of smoking on vascular health and neurological function suggests that avoiding nicotine may be beneficial in this case as well [15, 16].

Therefore, to synthesize the current gap in the literature, we are performing a systematic review with a meta-analysis. The main objective is to verify the association between smoking and the most common primary headache types, i.e. migraine, TTH, and CH. The secondary aim is to examine the prevalence of smoking in the primary headache population.

Methods

Search strategy and collection process

This systematic review was conducted using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines. Medical databases such as PubMed and Medline (Scopus) were searched until the end of February 2025. Three authors (BB, PN, JP) formulated the following key terms in order to find articles related to the main objective of the present study: (“migraine” OR “tension-type headache” OR “cluster headache” OR “primary headache” OR “aura” OR “MwoA” OR “MwA” OR “TTH”) AND (“tobacco” OR “smoking” OR “nicotine” OR “cigarette” OR “smoke” OR “vaping”). Then, the above databases were screened without a timeframe. Initially, abstracts of all identified articles were read independently by the three authors (BB, PN, JP). To reach the final results, the PRISMA 2020 guidelines and selection criteria from “Selection criteria” section were used. Only full-text available studies with eligible criteria were considered for data extraction. In the event of any disagreement between the authors regarding inclusion, a fourth author (MWP) helped resolve the issue via discussion. No automated tools were used in the searching process.

Selection criteria

Before the search process, the inclusion criteria were set: full-text articles in English; original type of studies; headache diagnosis based on the International Classification of Headache Disorders (ICHD) at any version or International Headache Society (IHS) criteria; control group in the included study for association calculation had to be without any other headache diagnosis; patients with primary headache had to be marked in current smokers/former smokers/never smokers or included study had to describe methods used to define smoking such as smoking duration expressed in time range, in pack-years or amount of cigarettes per day; data from the study had to be presented as number of patients in particular groups. Additionally, only cigarette-smoking was considered; other forms of nicotine addiction were excluded. Exclusion criteria included articles such as reviews, letters to the editor, animal studies, conference abstracts, book chapters, duplicates and non-English papers; studies that did not contain precise data about smoking with division into active/past smoking or number of cigarettes per day etc.; studies without exact diagnosis of migraine based on validated criteria – probable migraine was not considered for inclusion; studies where the same register databases of patients were used to create different analyses; studies with missing data; studies where the main focus was on a serious medical condition such as stroke, heart attack, or multiple sclerosis, because smoking has a negative impact on these conditions [17, 18].

Methods to assess risk bias

Risk of bias was assessed independently by the three authors (BB, PN and JP) using a critical appraisal tool, the Joanna Briggs Institute tool (JBI CAT, [19]). After assessment of all publications included in this systematic review by each of the above-mentioned authors, the results were discussed thoroughly and final decisions regarding the assessment of risk of bias were made. In the event of disagreement, a fourth author (HM) resolved problems regarding risk of bias assessment via discussion. Due to the large methodological diversity of the included studies, appropriate JBI checklists assigned to each research design such as case-control, case-series, cohort, cross-sectional, and randomised control trial (RCTs) were used. For cross-sectional studies, eight questions measured the risk of bias; for case-control studies, there were ten questions; in cohort studies, there were 11 questions; case-series studies contained ten questions; and in RCTs, there were 13 control questions. The possible answers on risk of bias questions were “Yes”, “No”, “Unclear”, “Not applicable”. A “low” risk of bias was considered if a cross-sectional study received ≥ 6 times “yes” answers, ≥ 7 times “yes” in case-control and case-series articles, 8 ≥ times “yes” in cohort studies, and ≥ 9 times “yes” in RCTs. A “high” risk of bias was noted if cross-sectional studies received ≤ 3 times “yes” answers, case-control and case-series studies received ≤ 4 times “yes” answers, cohort studies ≤ 5 times “yes” answers, and there were ≤ 6 times “yes” answers in RCTs. A “moderate” assessment was given if studies scored between the described ranges.

Meta-analysis

The data from the included studies was extracted to a Microsoft Excel spreadsheet (Microsoft Corporation, 2010 version, USA). Results aggregation was performed in the JASP program (Jeffreys’s Amazing Statistics Program, JASP Team, version 0.19.3), and the ’metafor’ script of the R language environment (R version 4.3.1) was used to create a forest plot of prevalence. Heterogeneity was assessed by calculating the Cochran Q statistic and using an I² test. Considering the diversity of the collected studies, random effect models were selected in the meta-analyses. The adopted measure of final effect was the odds ratio (OR) corresponding confidence intervals (CI). In order to detect a possible publication bias, the symmetry of funnel plots was analyzed using the trim-and-fill method, and the Egger test was also used. In order to assess the extent to which the assumptions of the meta-analysis and the included studies influenced the obtained total results, a sensitivity analysis was also performed. In all statistical tests, a p < 0.05 value was considered significant. Meta-regression was used to assess the extent to which various moderating variables (age of the study participants, percentage of females in the study group and year of publication) impacted on the results.

Results

Study inclusion

There were 2,713 search results in PubMed and Scopus. Of these 2,713 papers, there were 659 duplicates. After the duplicates were removed, the remaining 2,054 were screened by abstract and title. Next, there were 81 non-English studies, and 872 were articles not relevant to the subject of our study. Among the remaining 1,101 articles, 88 were not available to be retrieved. Of the potentiall eligible reports, there were 327 reviews, two animal studies, 4 book chapters, 69 letters to editors, 135 case reports, ten conference abstracts, and 62 editorials. In the rest of the articles, there were other reasons for exclusion such as inappropriate criteria for headache diagnosis (125 articles); in 42 papers, headache diagnosis and smoking data were for serious medical conditions; in ten, there was missing data; in 12, probable migraine diagnosis was conflated with migraine diagnosis; in five, control groups contained other types of headaches; in 102, smoking was not described according to inclusion criteria; 71 articles presented data incorrectly. Finally, 37 studies were included in the presented systematic review. Details of the steps in the selection process are shown in Fig. 1.

Fig. 1.

PRISMA 2020 flow diagram

Study characteristics

Thirty-seven studies were included, and these came from many countries; that is, one from Egypt [12], two from the Netherlands [20, 21], eight from the USA [22–29], two from Brazil [30, 31], two from South Korea [32, 33], four from Italy [34–37], three from Taiwan [38–40], two from Turkey [41, 42], two from China [43, 44], two from Germany [45, 46], and individual reports was from Iran [47], France [48], Norway [49], Spain [50], Syria [51], Greece [52], Ukraine [53], Denmark [54], and India [55]. The studies involved 24,502 participants. The most common methods to diagnose headache were ICHD criteria. Smoking was assessed in different ways, such as interview, questionnaire, or both methods combined. Different categories of smoking were utilized in the studies such as active smokers, former smokers, and never-smokers; current smokers, former smokers, and never-smokers; all smokers, current smokers, and past smokers. Some studies provided additional smoking group characteristics including amount (defined as cigarettes per day or packets or packyears), duration (years of smoking or packyears) or smoking group categorization such as mild, moderate, or heavy smoking. Twenty-seven of the 37 studies described migraine [20–23, 25–31, 33, 35–39, 41, 42, 44, 45, 47, 49, 50, 53, 55], five studies showed MwoA and MwoA [21, 35, 39, 42, 49], four studies examined smoking in TTH [30, 40, 46, 52], and nine studies were about smoking in CH [12, 24, 32, 34, 36, 40, 43, 48, 54].

Twenty-two out of 37 studies contained comparison groups marked as “control group” besides groups with primary headaches [20, 21, 26–32, 34, 35, 39, 41, 44, 45, 47, 49–52, 54, 55]. However, in seven out of these 22 studies, comparison group was not clearly described as patients without any headache or there was just presented as the control group involving patients with other types of headaches (primary or secondary), therefore in order to avoid potential bias risk of mixing patients with other headaches and patients without any headaches, only prevalence of smoking in primary headaches was extracted from these 7 articles [20, 26, 29, 31, 47, 50, 51]. Additionally, if covariates such as age, sex, year of publications needed for meta-regression in calculation of association between smoking and primary headaches could not be simply extracted, this study was only involved in the prevalence estimation [42, 45]. Finally, 14 studies were used to estimate the association between primary headaches and smoking [21, 27, 28, 30, 32, 34, 35, 39, 41, 44, 49, 52, 54, 55]. From all 37 studies [12, 20–33, 35–50, 52–55], the prevalence of smoking was calculated. The described study characteristics are given in Table 1.

Table 1.

The main characteristics of the studies included in this paper

No.	Author	Study type	Country	Year of publication	Criteria used to assess headache	Age of study population	Type of headache	Headache subjects			Method used to assess smoking	Smoking group characteristic in patients with primary headaches (e.g. current smokers/former smokers/never smokers) and smoking measurement method used in included studies (e.g. smoking duration expressed in time range or in pack-years, amount of cigarettes per day)
								Smokers	Nonsmokers	Total number of subjects
1.	Hamdy et al. [12]	Cohort	Egypt	2024	ICHD-3	Mean age: 42.08 ± 10.93 (20–66)	Cluster	144	28	172	Interview	Smoking group characteristic: -Smokers -Non-smokers Smoking measurement: Smoking was definied as -Smoking index -Smoking duration -Average cigarettes/day -Age at onset of smoking -History of exposure to smoking during childhood
2.	Al-Hassany et al. [20]	Crss-sectional	Netherlands	2024	ICHD-2	Median age (IQR): 62.2 (55.5 − 72.8)	Migraine	159	926	1085	Interview	Smoking group characteristic: -Current smokers -Former smokers Smoking measurement: -Smoking was definied as pack-years
3.	True et al. [22]	Cohort	USA	2024	ICHD − 3 beta	Mean age (SD): 43.1 (12.15)	Migraine	188	1612	1800	Interview	Smoking group characteristic: -Current smokers Smoking measurement: -None
4.	Belitardo de Oliveira et al. [30]	Cohort	Brazil	2024	ICHD − 2	Mean age (95%CI): TTH: 48.6 (48.3–48.8) Migraine: 47.3 (46.9–47.7)	Migraine	142	843	985	Questionnaire	Smoking group characteristic: -Current smokers Smoking measurement: -None
							TTH	453	3239	3692
5.	Kim et al. [32]	Cross-sectional	South Korea	2024	ICHD − 3 and ICHD − 3 beta	Mean age ± SD: 37.8 ± 9.6	Cluster	186	237	423	Questionnaire	Smoking group characteristic: -Current smokers Smoking measurement: -None
6.	Cho et al.[33]	Cross-sectional	South Korea	2024	ICHD-3	Age at baseline, n (% of whole included population): 20–29 y.o.: 34 (5.1); 30–39 y.o.: 44 (6.4); 40–49 y.o.: 60 (7.3); 50–59 y.o.: 32 (3.8)	Migraine	36	134	170	Questionnaire	Smoking group characteristic: -Current smokers Smoking measurement: -None
7.	Silvestro et al. [34]	Cross-sectional	Italy	2021	ICHD − 3	Median age (IQR): 40 (36–43.5)	Cluster	15	5	20	Interview	Smoking group characteristic: -Smoker Smoking measurement: -Smoking was definied as > 15 cigarettes/day
8.	Wang et al. [38]	Cross-sectional	Taiwan	2023	ICHD − 3	Mean age ± SD: 41.7 ± 13.9	Chronic migraine	56	564	620	Interview	Smoking group characteristic: -Current tobacco use Smoking measurement: -None
9.	Liampas et al. [52]	Case-control	Greece	2021	ICHD − 3	Mean age ± SD: 46.0 ± 11.6	TTH	14	16	30	Interview	Smoking group characteristic: -Smoker -Never smoker Smoking measurement: -Smoking was definied as active smoker or quit less than 1 year ago
10.	Delva et al. [53]	Cross-sectional	Ukraine	2022	ICHD − 3	Median age (Q1-Q3) in two groups: Fatigue 34 (24–41) No-fatigue 27 (25–37)	Migraine	19	66	85	Interview	Smoking group characteristic: -Non-smoker -Current smoker Smoking measurement: -Smoking was definied as smoking regularly for the last 1 year
11.	Yin et al. [39]	Cross-sectional	Taiwan	2021	ICHD − 3	Mean age ± SD: 36.7 ± 12.0	Migraine with aura	67	364	431	Interview	Smoking group characteristic: -Ever smoker (current/former) -Current smoker Smoking measurement: -None
							Migraine without aura	100	726	826
12.	Lipton et al. [23]	Cross-sectional	USA	2020	ICHD-3	Mean age (SD): 44.8 (13.4)	Migraine	704	3997	4701	Questionnaire	Smoking group characteristic: -Current smoker -Not current smoker Smoking measurement: -None
13.	Lund et al. [54]	Case-control	Denmark	2019	ICHD-2	Mean age (SD): 46.2 (11.5)	Cluster headache	193	207	400	Questionnaire	Smoking group characteristic: -Current smokers -Former smokers Smoking measurment: -Smoking was definied as pack-years of ever smokers
14.	Rozen [24]	Case-series	USA	2018	ICHD-3	Data not provided	Cluster headache	8	2	10	Interview	Smoking group characteristic: -Current smoker Smoking measurement: -None
15.	Chaudhuri et al. [55]	Case-control	India	2016	ICHD-3 beta	Mean age ± SD: 45.1 ± 8.9	Migraine	39	261	300	Interview	Smoking group characteristic: -Current smoker -Nonsmoker (including ex-smokers and occasional smokers) Smoking measurement: -Smoking was definied as daily smoking
16.	Monteith et al. [26]	Cohort	USA	2015	ICHD-2	Age at baseline, n (% of whole included population): <60 y.o.: 77 (29); 60–69 y.o.: 97 (37); 70–79 y.o.: 67 (26); 80+ y.o.: 21 (8)	Migraine	35	227	262	Interview	Smoking group characteristic: -Current smoker -Never smoker -Former smoker Smoking measurement: -None
17.	Gür-Özmen and Karahan-Özcan [41]	Case-control	Turkey	2016	ICHD-2	Mean age (SD): TTH: 36.5 (10); Migraine: 35.05 (9)	TTH	17	153	170	Questionnaire and interview	Smoking group characteristic: -Current smoker -Nonsmoker (including ex-smokers and never smokers) Smoking measurement: -None
							Migraine	31	139	170
18.	Jahromi et al. [47]	Cross-sectional	Iran	2013	ICHD-2	Mean ± SD Age (year): 38.4 ± 11.1	Migraine	8	261	269	Interview	Smoking group characteristic: -Active smoker Smoking measurement: -Smoking was definied as smoking in past 6 month
19.	Xie et al. [43]	Cross-sectional	China	2013	ICHD-2	Data not provided	Cluster headache	14	12	26	Interview	Smoking group characteristic: -Nonsmoker -Ex-smoker -Current smoker Smoking measurement: -Smoking was definied as years of smoking (three ranges: <10; 11–20; >20)
20.	Sacco et al. [35]	Case-control	Italy	2014	ICHD-2	Mean age ± SD: MwoA: 37.0 ± 9.3 MwA: 35.8 ± 9.8	Migraine without aura	14	36	50	Questionnaire	Smoking group characteristic: -Active smokers Smoking measurement: -None
							Migraine with aura	17	33	50
21.	Leroux et al. [48]	Cross-sectional	France	2013	ICHD-2	Mean age ± SD: 40.4 ± 11.3	Cluster headache	103	36	139	Questionnaire	Smoking group characteristic: -Current smokers -Nonsmokers -Not available Smoking measurement: -None
22.	Lambru et al. [36]	Case-control	Italy	2010	ICHD-2	Mean age ± SD: CH: 18.7 ± 11.7; MwoA: 17.9 ± 10.1	Cluster headache	150	50	200	Questionnaire (filled in by the physician)	Smoking group characteristic: -Mild tobacco intake -Moderate tobacco intake -Heavy tobacco intake -Non-smokers Smoking measurement: -Smoking was definied as amount ranges: mild (1–15 cigarettes/day), moderate (16–24 cigarettes/day) and heavy (> 25 cigarettes/day)
							Migraine	89	111	200
23.	Winsvold et al. [49]	Cohort	Norway	2013	ICHD-1	Median (quartiles): MwoA: 41 (34, 48); MwA: 40 (32, 48)	Migraine with aoura	88	210	298	Questionnaire	Smoking group characteristic: -Current smoking Smoking measurement: -Smoking was definied as current daily smoking
							migraine without aura	441	1165	1606
24.	Rozen [25]	Cross-sectional	USA	2011	ICHD-2	Not provided	Migraine	23	94	117	Interview	Smoking group characteristic: -All smokers -Current smokers -Past smokers Smoking measurement: -Smoking was definied as smoking continuously for at least 1 year during patients lifetime
25.	Lopez-Mesonero et al. [50]	Cross-sectional	Spain	2009	ICHD-2	Age rangeof participants: 19–26 years	Migraine	17	41	58	Questionnaire	Smoking group characteristic: -Current smokers Smoking measurements: -Smoking was definied as two group divided by amount per day: >10 cigarettes per day and < 10 cigarettes per day
26.	Peterlin et al. [27]	Cross-sectional	USA	2010	ICHD-2	Mean age (SD): 38.9 (12.9)	Migraine	79	172	251	Interview	Smoking group characteristic: -Current smoker -Former smoker -Never smoker Smoking measurement: -None
27.	Benseñor et al. [31]	Cross-sectional	Brazil	2010	ICHD-2	Mean age (SD): 70.6 (5.25)	Migraine	48	117	165	Questionnaire	Smoking group characteristic: -Current smoker -Former smoker -Never smoker Smoking measurement: -None
28.	Straube et al. [45]	Cross-sectional	German	2010	ICHD-2	Different mean age for each study population: -Study of Health in Pomerania (SHIP) population: EM: 50.1 CM 61.0 -Cooperative Health Research in the Region of Augsburg (KORA) study population: EM: 50.0 CM 60.8: - Dortmund Health study population: EM: 47.5	Migraine	89	320	409	Interview	Smoking group characteristic: -Active smokers Smoking measurement: -None
29.	QiHong et al. [44]	Cross-sectional	China	2009	ICHD − 2	Mean age (SD): 29.2 (8.2)	Migraine	96	233	329	Interview	Smoking group characteristic: -Current smoker Smoking measurement: -Smoking was definied as over 5 years, above 15 packs/month
30.	Tietjen et al. [28]	Cross-sectional	USA	2007	ICHD − 2	Mean age ± SD: 37.6 ± 10 Range: 18–62	Migraine	39	132	171	Questionnaire	Smoking group characteristic: -Current smoker -Ever smoker Smoking measurement: -None
31.	Scher et al. [21]	Cross-sectional	Nederland	2005	ICHD-1	Adjusted mean (Standard error): 41.9 (0.4)	Migraine	229	391	620	Questionnaire	Smoking group characteristic: -Current smoker -Never smokers -Ex-smokers Smoking measurement: -Smoking status was definied as occasional (1 cigarette/day), light (1 to 19 cigarettes/day), heavy (20 cigarettes/ day)
32.	Misakian et al. [29]	Cross-sectional	USA	2003	ICHD − 1	Mean age (SD): 55.2 (6.2)	Migraine	218	1691	1909	Questionnaire	Smoking group characteristic: -Current smoker -Past smoker -Never smoker Smoking measurement: -None
33.	Martini et al. [51]	Cross-sectional	Syria	2025	ICHD-3	Mean age ± SD: 25.941 ± 6.58	Migraine	52	467	519	Questionnaire	Smoking group characteristic: -Current smoker -Non-smoker Smoking measurement: -Smoking status was categorized as packet per day or 2 packets per day and hookah smoking
34.	Dinia et al. [37]	Cohort	Italy	2013	ICHD-2	Mean years ± SD Two groups: with white matter lesion − 49.1 ± 12.9;without white matter lesion − 40.0 ± 12.2	Migraine	10	31	41	Interview	Smoking group characteristic: -Smoker -Non-smoker Smoking measurement: -Smoking was definied as ≥ 5 cigarettes per day
35.	Celik et al. [42]	Cross-sectional	Turkey	2005	ICHD − 1	Mean age (SD): 34.59 (10.81)	Migraine	27	50	77	Interview	Smoking group characteristic: -Current smoker -Non smoker Smoking measurement: -Smoking was definied as < 20 cigarettes/day and > 20 cigarettes/day
36.	Lin et al. [40]	Cross-sectional	Taiwan	2004	ICHD − 1	Age mean ± SD (range): 39.2 ± 12.2 (18–84)	Cluster headache	61	43	104	Questionnaire	Smoking group characteristic: -Current smoker -Non-smokers -Ex-smokers Smoking measurement: -None
37.	Steiner et al. [46]	RCT	Germany	2003	ICHD − 1	Mean age (SD) in five different studied groups: Acetylsalicylic acid 500 mg: 39.9 (11.8) ASA 1000 mg: 41.0 (12.3) Paracetamol 500 mg: 39.7 (11.4) Paracetamol 1000 mg: 38.4 (11.8) placebo: 40.6 (11.4)	TTH	124	418	542	Interview	Smoking group characteristic: -Current smoker Smoking measurement: -None

Abbreviations: ICHD International classification of headache disease, second and third versions, IHS International, headache society, TTH tension-type headache, CH cluster headache, EM episodic migraine, CM chronic migraine, SD standard deviations, IQR interquartile range, RCT randomized controlled trials

Smoking prevalence in primary headaches

Twenty-seven studies embracing 18,574 participants contained data about current smoking and migraine [20–23, 25–31, 33, 35–39, 41, 42, 44, 45, 47, 49, 50, 53, 55]. Due to high heterogeneity among selected articles (I² = 96%, p < 0.01), random effect model was used to estimate prevalence. The weighted-pooled prevalence of smoking in migraine was 20% (95% CI: 16–24). This result is presented in Fig. 2.

Fig. 2.

The prevalence of smoking in migraine

Five studies with 999 patients were focused on patients with MwA [21, 35, 39, 42, 49]. Due to high heterogeneity among selected articles (I² = 97%, p < 0.01), random effect model was used to estimate prevalence. The weighted-pooled prevalence of smoking in MwA was 27% (95% CI%: 17–27). This result is presented in Fig. 3.

Fig. 3.

The prevalence of smoking in MwA

Four studies embraced 4,434 patients with TTH [30, 40, 46, 52]. Due to high heterogeneity among selected articles (I² = 91%, p < 0.01), random effect model was used to estimate prevalence. The weighted-pooled prevalence of smoking in TTH was 19% (95% CI: 7–34). This result is presented in Fig. 4.

Fig. 4.

The prevalence of smoking in TTH

Nine studies was conducted in 1,494 patients with CH [12, 24, 32, 34, 36, 40, 43, 48, 54]. Due to high heterogeneity among selected articles (I² = 95%, p < 0.01), random effect model was used to estimate prevalence. The weighted-pooled prevalence of smoking in CH was 65% (95% CI: 55–76). This result is presented in Fig. 5.

Fig. 5.

The prevalence of smoking in CH

In summary, from 37 included studies there were pooled-prevalence of smoking calculated embracing 24,502 patients with headaches. Due to high heterogeneity among selected articles (I² = 100%, p = 0), a random effect model was used to estimate prevalence. The weighted-pooled prevalence of overall smoking in primary headaches was 32% (95% CI: 8–62). This result is presented in Fig. 6.

Fig. 6.

The overall prevalence of smoking in primary headaches

Association between smoking and primary headaches

Seven studies had migraine groups which involved 3,463 patients and groups of patients without any headache which were described as control groups. The test of heterogeneity was Q = 15.7, df = 6, p = 0.015, I² = 69.1%. However, the tunnel graph analysis from Figure S1 (supplementary material) as well as the results of Egger’s test (p > 0.05) indicate a lack of any statistically significant publication bias. Current smoking was associated with increased risk of migraine diagnosis in comparison to the control group (OR = 1.29, 95% CI: 1.02–1.62, p = 0.034). This result is presented in Fig. 7.

Fig. 7.

The meta-analysis results of current smoking and migraine in comparison to control patients. Abbreviations: OR, odds ratio; CI, confidence interval

Meta-regression was performed to consider the influence of three moderators on the effect of the meta-analysis. The result indicates that none of the factors (i.e. age, gender and year of publication) is a significant predictor of migraine (p > 0.05). Although the effect size decreases in more recent publications (b = -0.040), increases with more women in the migraine group (b = 0.012) and decreases with age (b = -0.008), these coefficients do not differ significantly from the zero value (p > 0.05). These results are summarized in Table S1 in the supplementary materials. The three bubble plots in Fig. 8 visualize the predicted effects of the moderators (year of publication, mean age of the study participants, and the proportion of women in the study) along with the effect size estimates and confirm that the effect of these moderators is nonsignificant.

Fig. 8.

The bubble plots of meta-regression

Only two studies considered former smoking in migraine. The test of heterogeneity was Q = 1.9, df = 1, p = 0.168, I² = 47.5%. The results of Egger’s test (p > 0.05) indicate a lack of any statistically significant publication bias. Former smoking was not associated with migraine diagnosis in comparison to the control group (OR = 0.80, 95% CI: 0.59–1.09, p = 0.152). This result is presented in Fig. 9.

Fig. 9.

The meta-analysis results of former smoking and migraine in comparison to control patients. Abbreviations: OR, odds ratio; CI, confidence interval

Three studies presented a migraine division into MwA and MwoA. The test of heterogeneity was Q = 0.722, df = 2, p = 0.697, I² = 0.00%. Current smoking was not associated with MwA diagnosis in comparison to the MwoA group (OR = 1.06, 95% CI: 0.86–1.31, p = 0.600). This result is presented in Fig. 10.

Fig. 10.

The meta-analysis results of current smoking and MwA in comparison to MwoA patients. Abbreviations: OR, odds ratio; CI, confidence interval

Three studies described smoking in TTH. The test of heterogeneity was Q = 1.087, df = 2, p = 0.581, I² = 0.00%. Due to the value of the I² statistic, a random effect model was used in the meta-analysis. Current smoking was associated with decreased risk of TTH diagnosis in comparison to the control group (OR = 0.78, 95% CI: 0.68–0.89, p < 0.001). This result is presented in Fig. 11.

Fig. 11.

The meta-analysis results of current smoking and TTH in comparison to control patients. Abbreviations: OR, odds ratio; CI, confidence interval

Three studies described smoking in CH. The test of heterogeneity was Q = 22.55, df = 2, p < 0.001, I² = 95.33%. The results of Egger’s test (p > 0.05) indicate a lack of any statistically significant publication bias. Current smoking was not associated with CH diagnosis in comparison to the control group (OR = 2.15, 95% CI: 0.41–11.27, p = 0.367). This result is presented in Fig. 12.

Fig. 12.

The meta-analysis results of current smoking and CH in comparison to control patients. Abbreviations: OR, odds ratio; CI, confidence interval

A meta-regression was not available for TTH, CH, and former-smoker calculations due to the limited number of included studies.

Risk of bias assessment in included studies

All thirty-seven included studies were assessed for risk of bias. Of these, thirty-six were observational studies [12, 20–40, 42–45, 47–56] and one was an experimental study [46]. Among the observational studies, twenty-three were cross-sectional studies [20, 21, 23, 25, 27–29, 31–34, 38–40, 42–45, 47, 48, 50, 51, 53], six were case-control studies [35, 36, 41, 52, 54, 55], six were cohort studies [12, 22, 26, 30, 37, 49], and one was a case-series [24] study. Within the cross-sectional studies, one study received seven “yes” answers [20] and thirteen received six “yes” answers, resulting in a low risk of bias [21, 23, 27–29, 31, 34, 38–40, 44, 47, 48]; while four studies received five “yes” answers [25, 33, 43, 45] and two received four “yes” answers [32, 51], thus receiving moderate risk of bias note; and the last three studies received three positive answers [42, 50, 53], resulting being assessed as having a high risk of bias. In the group of case-control studies, one study received nine “yes” answers [36], three studies received eight “yes” answers [35, 41, 54], one seven positive answers [55] and thus they having a low risk of bias; one study received six “yes” answers [52] and consequently it was assessed as having a moderate risk of bias. The distribution of responses in the cohort studies was as follows: two studies received eight “yes” responses, which allowed them to be classified as having a low risk of bias [12, 49]; two studies had seven positive answers [26, 37], one six positive answers [22] and were classified as moderate risk of bias; one study got four “yes” responses and it had high risk of bias [30]. The one case-series study [24] received six “yes” responses, resulting in a classification as having a moderate risk of bias. The randomized controlled study [46] received ten positive responses, and its risk of bias was assessed as low. In summary, 22 out of 37 included studies received “low” risk of bias note [12, 20, 21, 23, 27–29, 31, 34–36, 38–41, 44, 46–49, 54, 55], 11 out of 37 were identified as moderate risk of bias [22, 24–26, 32, 33, 37, 43, 45, 51, 52] and 4 studies received high risk of bias note [30, 42, 50, 53]. Additional details on the responses to individual questions for each study are provided in the Table S2 from supplementary materials.

Discussion

Smoking is a risk factor for many diseases such as stroke, cardiovascular disease and lung cancer [57]. It is responsible for over 20% of males deaths annually [58]. In 2019, there were 1,14 billion smokers worldwide, and its prevalence was estimated at approximately 15% [59]. Pooled prevalence of smokers in patients with migraine in our study was estimated as 20%. These similar values may indicate that patients with migraine slightly more often smoke than the general population. However, in a previous review of literature related to this topic, smoking prevalence in migraine varied between 25.9% and 51%, and thus the authors suggested smoking is more common in this group of patients [6]. The above difference between presented review and this study may be the result of either their not conducting meta-analysis or from including single studies with other methodologies and participant characteristics [60–63]. Tobacco smoke is often indicated as a trigger for migraine [50, 64–67]. Our study confirmed that current smoking is associated with increased risk of migraine. However, some authors deny the smoking association in migraine [39, 55, 68]. Since smoking may trigger a migraine attack, as was mentioned earlier [50, 64–67], it is interesting that the prevalence of smoking in patients with migraine was comparable with the global population or patients with TTH. Several mechanisms have been proposed which explain the association between migraine and smoking. Tobacco smoke contains nicotine and other chemicals, such as cadmium, that can affect central nervous system structures [69] or stimulate expression of pro-inflammatory molecules resulting in peripheral and central sensitization which can contribute to the development of headache pain [70–72]. The other theoretical pathways include oxidative stress, decreased NO generation, and increased production of hormones helping in trigger attack [6, 73–75]. Taking into consideration the migraine subtypes, it has been suggested that aura attacks in patient with MwA are also associated with smoking [76, 77]. Our study did not show an association with smoking habits in MwA in comparison to MwoA, but Salhofer-Polanyi et al.’s study suggested that smoking might decrease the risk of MwoA while increasing MwA risk [76]. Also, former smoking was not associated with migraine. It is difficult to discuss these results with those in the literature, because similar observations have not been conducted. There is a need to emphasize that data for the present study were extracted from studies which were not mainly focused on smoking topics; additionally, in prevalence and association calculation between smoking and migraine was noted high heterogeneity, therefore, our results need to be interpreted carefully and may be different in other populations.

There was observation that patients with CH have a tendency to use different addictive substances including cigarettes [78]. Therefore, in accordance with that fact the highest prevalence of cigarette smoking in primary headaches belongs to patients with CH in our study. There are proofs in the literature confirming our results [12, 79–81]. Common smoking in CH might be partially explained by the opinion that CH is more widespread in men [82], and also male subjects are more often smokers [63]. However, the pooled-prevalence was estimated on 64% smoking in patients with CH which is extremaley high value. Lund et al. and Rozen et al. [54, 83], independently, compared smoking habits in patients with CH for both sexes, and their results showed a much higher prevalence of cigarette usage in males [54]. Besides the highest prevalence of smoking among primary headaches, there was a lack of association between CH and smoking. However, in the literature, CH is considered to be positively associated with smoking, and smoking increases CH frequency and severity [11, 84]. Additionally, nicotinism might be associated with a worse clinical course of CH [85]. Our result may, however, stem from the limited number of studies included in the meta-analysis. As well as the presented calculations included only patients with CH without any co-existing headache or severe systemic disease, therefore it may be the cause of non-significant association between CH and smoking. Observed high heterogeneity in smoking prevalence and association in this meta-analysis may limit generalizability of our findings. On the other hand, patients who have never smoked appear to have an earlier onset of CH [13, 86]. Moreover, there is no clear evidence that quitting smoking reduces the frequency or intensity of CH attacks, suggesting a more complex relationship between smoking and CH involving genetic or neurobiological factors [87, 88]. We conclude that since the highest prevalence of smoking was always in CH, and some authors have automatically associated it with CH, but smoking in CH may result from other, as yet unidentified factors.

According to our analysis, 19% of patients with TTH currently smoke. In the literature, there is little evidence regarding tobacco use in TTH. The study by Sabah et al. [89] reports cigarette smoking is present among 62.74% of patients with TTH. However, this research had specific settings, i.e. the study was conducted by electronic questionnaires, and smoking was not clearly defined. Our study showed that smoking may protect against TTH attacks, because smoking was associated with decreased risk of TTH. There are no similar results in the literature. We have to emphasize that this result is only based on three studies, and it was not possible to calculate the meta-regression. On the other hand, the pathophysiology of TTH is often connected with the serotonergic system [90]. and smoking is known for its influence on serotonin levels, which may provide a potential connection between smoking and TTH [91]. However, this observation requires further analysis, because we obtained contrasting results.

It is worth mentioning that smoking is more widespread among disadvantaged patients, such as those with low socioeconomic status or low income [92, 93]. In turn smoking is positively associated with increased all types of headache or migraine frequency among patients with low socioeconomic status [94]. The presented association may potentially have an impact on received results, but we were not able to assess the impact of low-income status and socioeconomic status as moderator factors in a meta-analysis, because not every study contains these data. Also comorbidity disease in patients with primary headache may have an impact on smoking. For example, smoking is significantly associated with increased risk of depression, which is a common disease in migraine [95–97]. Similar situation concern occurring psychiatric diseases in patients with CH [54]. By contrast lifestyle has a significant impact on headaches, especially not adequate and unhealthy lifestyle can exacerbate migraine [98, 99]. So, our results without checking the influence of comorbid disease or lifestyle to primary headache have to be interpreted carefully.

In addition, nicotine may currently also be delivered in other modern products such as electronic cigarettes. These devices have never been studied for their effect on primary headaches, but previous studies indicate that headaches, not necessarily primary headaches, may be a side effect of using electronic cigarettes [100]. Electronic cigarettes are marketed as a safer option for smoker and may help people to quit smoking; however, it is mostly young people who use these products, and they are also vulnerable to primary headaches [101, 102]. Therefore, there is a need for a modern study on this topic.

The present study had several limitations. This is the first attempt to summarize data on smoking and primary headaches utilizing a systematic review and meta-analysis. However, different authors have used other statistical methods in studies related to our subject of our interest, and only some of these consider data about confounding factors. In our meta-analysis, we performed meta-regression to minimize confounding factors, but due to the insufficient number of studies adjusting results to factors such as socio-economic status, level of education and alcohol consumption, we were only able to use age, gender and year of publication as moderators. Meta-regression could not be performed in prevalence of smoking calculation due to lack of basic population data in every included study. It is also important to emphasize that not all studies included in the prevalence analysis were included into association calculations, because most of them did not contain a control group of patients without headache. Additionally, some studies contain comparison groups consisting of patients with other types of headache, or probable migraine diagnosis, or primary headache subtypes. This type of study was only included in prevalence in order to elicit the most objective results. In our opinion, the reduced number of studies in association calculations is the biggest limitation to our study.

Conclusions

The prevalence of smoking in migraine and TTH was moderate. Current smoking was associated with increased risk of migraine and decreased risk of TTH. The direction of these associations remains unclear, so by no means is smoking advised to avoid TTH. There was no direct association between current smoking and CH, despite the fact that over half of patients with CH smoked. The influence of electronic cigarettes commonly used these days is not yet explored in primary headaches, despite their wide-spread use. Due to the potential limitations of the presented study, the results have to be interpreted carefully, and studies focused only on the topic of smoking are needed in the future.

Authors’ contributions

MWP and BB was involved in the conception and visualization of the study. BB, JP, PN, MWP collected the data and prepared a manuscript. BB was involved in data analysis. MWP, SB, HM supervised the study. MWP, HM, MS and SB revised the final version of the manuscript. All authors have read and agreed to the published version of the manuscript.

Funding

Not applicable.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.