How common headaches are and their links to demographics in adults across Peru

The study was conducted in accordance with the Declaration of Helsinki [11], with the protocol and study instruments approved by the Institutional Committee on Research Ethics of the Peruvian University Cayetano Heredia. All participants gave verbal consent to their inclusion in conformity with this approval.

Data were collected anonymously, and managed in compliance with data protection laws.

This was a cross-sectional study of the general population of Peru aged 18–65 years. We followed the Global Campaign’s standardized methodology, which has been published in detail [2] and employed in many previous studies [1].

Accordingly, we used cluster-sampling to generate a sample representative of the population of Peru, with random selection of participants from communities in the five geographical regions of Cajamarca, Lima, Piura, Puno and San Martín. Trained health workers made unannounced door-to-door visits to randomly chosen households in each, during the period May to November 2019. From each household, the interviewers randomly selected and interviewed one adult member, following a structured questionnaire. Since only 22% of Peruvians lived outside urban areas [3], we deliberately oversampled rural areas to ensure sufficient statistical power for association analysis.

We aimed for a sample of N > 2,000 in accordance with guideline recommendations [2].

Interviews were performed using modules from the Headache-Attributed Restriction, Disability, Social Handicap and Impaired Participation (HARDSHIP) questionnaire developed by the Global Campaign [12], translated into South American Spanish in accordance with its translation protocol for hybrid documents [13].

The demographic module included age, gender, marital status, educational level, household income and employment. The headache modules began with neutral screening questions (“Have you ever had headache?” and “Have you had headache in the last year?”) and continued, when appropriate, with diagnostic questions based on ICHD-3 criteria [14] (directed towards the most bothersome headache when more than one type was reported). Specific questions on headache on the day preceding interview (“headache yesterday” [HY]) made it possible to estimate point (1-day) prevalence. Subsequent modules addressed attributable burden, but these are not reported here.

Diagnoses were derived algorithmically, following methods previously described [2, 12]. Participants reporting headache on ≥ 15 days/month (H15+) were first identified. Those also reporting a monthly consumption of acute medication on ≥ 15 days were diagnosed as probable MOH (pMOH); those not, were classified as “other H15+”, without attempt at further diagnosis. (The threshold for medication overuse was set at ≥ 15 days rather than ≥ 10 days [14] because the vast majority of the population in Peru would have access only to over-the-counter analgesics.) Remaining participants (with headache on < 15 days/month) were diagnosed, according to the algorithm, in the following order: definite migraine, definite TTH, probable migraine, probable TTH [14]. Definite and probable diagnoses were combined in adjusted prevalence estimates and association analyses.

Demographic characteristics of the sample were analysed descriptively and reported as proportions (%) with 95% confidence intervals (CIs) or using means and standard deviations (SDs). Gender was recorded as either male or female, age as a continuous variable, habitation as urban or rural. The male-female and urban-rural ratios were compared with those of the national population using chi-squared tests, and mean age using one-sample t-test.

Observed prevalences of headache overall, and of each type, were reported as proportions (%) with 95% CIs, along with age-, gender- and habitation-adjusted estimates taking account of distributions of these variables among Peru’s general population. Observed point prevalence was reported as a proportion (%). Predicted point prevalence was calculated from observed 1-year prevalence and mean headache frequency reported in days/month.

In association analyses for each headache type, caseness (positive for the type) was compared with non-caseness (all others in the sample). We used bivariate and multivariate analyses, calculating unadjusted odds ratios (ORs) and adjusted ORs (aORs), with 95% CIs, and included gender, age, habitation, altitude of dwelling, marital status, education level and household income. For these analyses, age was categorized into five groups (18–25, 26–35, 36–45, 46–55 or 56–65 years). Since two of the five sampled regions were at high altitude (Cajamarca: 2,750 m; Puno: 3,800 m) and three at or near sea level (Lima: 0 m; Piura: 55 m; San Martín: 340 m), we categorized altitude (≤ 350, 2,500–3000 and > 3,500 m). Marital status was categorized as single, married, or separated/divorced/widowed; education level as none, school, or college/university; and household income as < 600, 600–899, 900-1,299, 1,300-2,199 or ≥ 2,200 Peruvian nuevo sol (PEN) approximating to national quintiles (in June 2019, USD 1.00 = PEN 3.32).

We considered p < 0.05 as significant. SPSS version 28 was used for all analyses except for adjusted prevalence estimates, for which Microsoft Excel version 16 was used.

Gender distribution in the sample (1,065 males [49.6%], 1,084 females [50.4%]) matched that of the country’s population (chi-squared [1, N = 2,149] = 0.07; p = 0.80). Mean age in the sample (42.0 years ± SD 13.7) was somewhat higher than in the population aged 18–65 years (37.9 years [15]; t(df = 2148) = 13.9; p < 0.001). Because we deliberately oversampled rural participants (35.6%: see Methods), the urban-rural ratio in the sample differed from the national ratio (22% rural [15]; chi-squared [1, N = 2,149] = 231.1; p < 0.001).

Reported lifetime prevalence (headache ever) was 83.6%, higher among females (87.5% [95% CI 85.3–89.4]) than males (79.6% [77.1–82.0]; p < 0.001). Observed 1-year prevalence of any headache was 64.6%, also higher among females (71.5%) than males (57.6%) (Table 1). TTH was the most common type (37.7%), with similar prevalence in each gender, while migraine, next most common, was almost twice as prevalent among females (28.2%) as among males (16.4%) (Table 1). H15 + was reported by 4.5%, females more than males, with about one quarter (1.1%) diagnosed as pMOH (Table 1).

In bivariate analyses we found associations of TTH and other H15 + with habitation (see below). Accordingly, we adjusted prevalence estimates with respect to population demographics to take account of habitation as well as age and gender. The effects were small: adjusted 1-year prevalence estimates were 65.5% for any headache, 22.8% for migraine, 38.9% for TTH, 1.2% for pMOH and a somewhat reduced 2.7% for H15+ (Table 1).

HY was reported by 12.1% of participants. In comparison, point prevalence of any headache predicted from the 1-year prevalence (64.6%) and reported mean headache frequency (4.2 days/month) was a rather lower 9.0%. Among those with headache, the proportion reporting HY was 18.7%: more than half (58.3%) of those with pMOH, almost half (48.6%) with other H15+, almost a quarter with migraine (23.7%), but fewer (11.6%) among those with TTH.

The gender associations apparent in Table 1 are formally analysed in Tables 2 and 3, along with associations between the headache types and other demographic variables.

In bivariate analyses, migraine was more common among females, as already noted (OR = 2.0; p < 0.001), but TTH showed no association with gender, and the other gender associations were not significant. Migraine became less prevalent after age 45 years, but this failed to reach significance even in the oldest group (OR = 0.7; p = 0.06). Other H15 + was more common in this group (OR = 3.1; p = 0.01), but no other associations between age and headache type were seen (Table 2). Again as noted, in these analyses there were associations with habitation: TTH less prevalent (OR = 0.8; p = 0.006) and other H15 + more prevalent (OR = 2.0; p = 0.005) among rural dwellers than urban. Marital status showed no associations; educational level appeared to be negatively associated with other H15+; income showed no clear associations (Table 2).

In the multivariate (adjusted) analyses (Table 3), female gender remained positively associated with migraine (aOR = 2.1; p < 0.001). The negative association of migraine with age became significant in those aged 56–65 years (aOR = 0.6, p = 0.01). The associations between habitation and TTH and other H15 + lost significance. Migraine was associated with being married rather than single (aOR = 1.5; p = 0.003). pMOH was most prevalent in those with no education (aOR = 7.1; p = 0.04), and (albeit not significantly, with small numbers) in those with mid-range income (Table 3). Those earning least had least TTH (aOR = 0.6, p = 0.03).

In bivariate analyses, migraine was more prevalent at high altitude, and significantly so at very high altitude (> 3,500 m) (OR = 1.8; p < 0.001). Other H15 + was also more prevalent in the high-altitude areas. TTH on the other hand, was less prevalent at 2,500-3,000 m (OR = 0.5; p < 0.001) (Table 2). In multivariate analyses, the association between migraine and very high altitude was confirmed (aOR = 2.5; p < 0.001), while TTH remained less prevalent at high altitude (aOR = 0.4; p < 0.001).

Most interesting was the strong association of migraine with high altitude, confirming similar indications from earlier studies in Peru [7, 8] and a more definite finding in Nepal [10], a similarly mountainous country with settlements at both high and low altitudes. There were differences, however. In Nepal, sampling occurred over the range of altitude from near sea level to > 2,500 m, but only to a limited extent (n = 216) at higher altitude [16]. Here in Peru, we missed the middle range, but captured almost 800 participants dwelling at > 2,500 m. This was pertinent because, in Nepal, there was an approximately linear association between migraine prevalence and altitude up to 2,500 m; above this level there was evidence of decline, but higher altitudes were not investigated. Therefore, the finding here in Peru of a very strong association (aOR = 2.5) with very high altitude (> 3,500 m) is intriguing.

There is no clear explanation, but it presumably lies in the physiological compensations that occur at high altitude [16]. This demands further study.

This was a carefully conducted nationally representative study using standardized and well tested methodology in a large and adequate sample (N = 2,149) with a high participating proportion (90.1%). These were considerable strengths. The limitations were those always associated with cross-sectional questionnaire-based studies: dependence on recall, and limited ability to diagnose H15+.

The study was conducted in accordance with the Declaration of Helsinki, with the protocol and study instruments approved by the Institutional Committee on Research Ethics of the Peruvian University Cayetano Heredia. All participants gave verbal consent to their inclusion in conformity with this approval.

Not applicable.

TJS is a Director and Trustee of Lifting The Burden, and associate editor of The Journal of Headache and Pain. There were no other competing interests.

The original data are held on file at the Peruvian University Cayetano Heredia, Lima, Peru, and the analytical dataset at Norwegian University of Science and Technology. Once analysis and publications are completed, they will be freely available for non-commercial purposes to any person requesting access in accordance with the general policy of the Global Campaign against Headache.