This is the first nationwide, population-based survey in any country of South Asia to explore the comorbidity of headache and psychiatric disorders, which are both highly prevalent in Nepal [31, 32]. Comorbidity occurred more than was expected by chance, at least with regard to migraine and headache on ≥15 days/month on the one hand, and HADS-A and HADS-cAD on the other, the associations thereby indicated (with AORs in the range 1.7–3.2) likely to be bidirectional. There were also significant associations between headache, especially headache occurring on ≥15 days/month, and neuroticism. There were no associations between any headache and HADS-D, and none between TTH and any HADS caseness. Additive and possibly synergistic effects of comorbid headache and HADS caseness were evident on QoL, highly significantly so when migraine or TTH was the index disorder and HADS caseness the comorbid disorder.

Before considering the implications of these findings, we would compare them with others as a test of veracity; but as we have noted, there are none from this Region. Outside South Asia, most studies exploring psychiatric associations with headache disorders have focused on migraine [3–11]. A Zürich cohort study first demonstrated strong relationships between migraine and both anxiety and depression [3], which were subsequently reported in the United States (US) [4, 9] and Canada [42]. Other US studies found that people with migraine had high levels of neuroticism [43, 44]. In Nepal, we found strong associations between migraine and HADS-A, HADS-cAD and neuroticism, but not HADS-D.

There is relatively little evidence regarding psychiatric associations with episodic TTH [45]. A large Swiss epidemiological study found no associations with depressive or anxiety disorders [7], and neither did we. Episodic TTH is usually a less painful disorder than migraine, and less troublesome since it lacks the range of associated symptoms that contribute to the burden of migraine [35]. Hence people with episodic TTH may not express the same extent of subjective emotional experiences as those with migraine [46].

People with chronic TTH, on the other hand, demonstrate strong psychiatric associations [47]: in the US, almost half had either a depressive or an anxiety disorder [48] while those in a Norwegian study had high levels of neuroticism [22]. We did not analyse chronic TTH specifically because it could not reliably be diagnosed by lay interviewers using a questionnaire [35], but among our participants with other headache on ≥15 days/month would have been some with this disorder. The category of headaches occurring on ≥15 days/month corresponds with what is elsewhere referred to as “CDH” [49], which has been associated with high frequencies of both depressive and anxiety symptoms [13, 50]. We found, in the Nepalese population, that headache on ≥15 days/month was associated with HADS-A, HADS-cAD and neuroticism, but not HADS-D.

Our findings, therefore, are not entirely in accord with others regarding depression. The explanation may be methodological. We found no association between HADS-D and any headache type, in line with French [27] and Norwegian studies [51] that also used HADS to establish psychiatric caseness. The studies from Switzerland [3], the US [4] and Canada [42], which found strong associations between depression and headache, instead used structured interviews based on Diagnostic and Statistical Manual (DSM) criteria [52]. A recent study from South Korea [53] obtained similar findings using the self-administered Patient Health Questionnnaire-9 (PHQ-9) [54], as did another US study [48] using the Beck Depression Inventory (BDI) [55]. Cultural factors determining how moods are expressed [56] may contribute: in many Asian countries, depression is often manifested somatically [57, 58], while HADS does not capture the somatic domains of depression so well [38]. HADS may therefore underestimate any association of depression alone with headache.

However, as was pointed out in a recent review [20] and revealed in the French study [27], depression in migraine rarely occurs alone but is almost always comorbid with anxiety. A stronger association of migraine with combined anxiety-depression than with either independently was seen in the early Zürich study [3], and similar findings in relation to the other headache types came later in Norway [51]. In our study, HADS-cAD occurred twice as often among all headache sufferers and those with migraine, and almost three times in those with headache on ≥15 days/month. Hence the association of headache with depression in our population, though not apparent in HADS-D, may have manifested in HADS-cAD. There was, however, a gender influence in this association: for any headache, the association with HADS-cAD appeared specific to females (AOR 4.3 versus 1.0 in males).

Summarising these arguments, we believe use of HADS – a screening rather than diagnostic instrument for depression – is valid among people with headache although it may underestimate caseness prevalence. We could not establish any association between headache disorders and HADS-D in Nepal, but this limitation should be taken into account.

We also note other limitations. It is a general limitation of cross-sectional studies that, while associations can be demonstrated, causality cannot. However, public-health purposes are served in the first instance by uncovering these relationships, which have important policy implications; subsequent research can investigate causation. It is a limitation of HADS that it cannot go beyond detecting caseness of anxiety or depression, which encompass heterogeneous mental disorders. This renders it difficult to identify and make provision in health-service planning for issues that might be related to particular types of these disorders. By a similar token, our headache diagnostic questionnaire captured episodic migraine and TTH, while all chronic cases were subsumed under headache on ≥15 days/month [35]. Comorbidities associated with these subtypes were likely to be different in view of their high-frequency, long-duration, intractable symptoms coupled with problems in management, but the prevalence of each would have been about 1 % or less. Categorically assessing these multiple types and subtypes would have been a cumbersome exercise requiring much greater investment. We believe that our approach, with simple validated interviewer-administered culturally-validated instruments, was a more practical option in a population with high illiteracy, and it best served our public-health purpose. Furthermore, our study had notable strengths: tried and tested methodology [33–35, 59], large sample size, high participation rate and good representativeness of the diverse population of Nepal [33, 34].