Is the way to headache through the stomach?

Headache in general and acid-related diseases, such as gastroesophageal reflux disease and peptic ulcers, are very common disorders at the population level (1,2). Both conditions have been linked and the association between headache and gastrointestinal complaints increased markedly with increasing headache frequency (3). However, linking two common conditions in population-based studies is challenging as statistically significant associations in large databases are easily achieved and even small biases can have substantial influences on risk assessment. On the other hand, even small to moderate associations between two common conditions or their treatments can have an important impact on public health.

In this issue of Cephalalgia, Liang and colleagues report an association between prescription of proton pump inhibitors and acute headache among patients who were at least 18 years old and who had an incident headache coding (migraine, tension-type, or general headache) in the Taiwan National Health Insurance program (4). The authors used a case-crossover design to estimate the odds of having been exposed to a proton pump inhibitor in a specific time window prior to the incident headache events (index day). The authors used three time windows: seven, 14, and 28 days and controlled for a large number of other medications and comorbid conditions.

During 12 years of observation, 279,120 patients with headache in general, 24,713 patients with migraines, and 17,130 patients with tension-type headaches were identified. An increased risk of headache was noted for all three time windows in patients who have received proton pump inhibitor prescriptions. The highest risk occurred during the seven-day window (odds ratio (OR) 1.41; 95% confidence interval (CI) = 1.14–1.74) and the lowest in the 28-day window (OR = 1.20; 95% CI = 1.07–1.35). Of the five types of proton pump inhibitors, only esomeprazole (OR = 1.78; 95% CI = 1.10–3.12) and lansoprazole (OR = 1.73; 95% CI = 1.21–2.47) were associated with headache. Regarding specific headache diagnosis codes, proton pump inhibitors were associated only with codes for headache in general but not with codes for migraine. The risk for tension-type headache was increased but this finding did not reach statistical significance. In further subgroup analyses, the observed increased risk of headache for those prescribed proton pump inhibitors was apparently only in women (4).

This study has several strengths, including its large sample, available information on many medical conditions, and the use of a case-crossover design, which is an optimal study design to evaluate intermittent or transient exposure effects of acute events in short time windows and to reduce confounding by personal characteristics (5).

Even modest associations between a commonly used drug class and headache would result in a substantial impact as both the drug class and headache are very common conditions. Should we fear an enormous increase in headache among individuals taking proton pump inhibitors and as a consequence should such drugs not be recommended for patients at risk for headache? Before such a conclusion can be drawn several aspects should be considered.

First, observational studies on drug effects are threatened by confounding by indication, that is, the underlying disease for which the drug is prescribed is the cause of the outcome and not the drug itself. Indeed, several diseases involving the gastrointestinal system have been linked with overall or specific headache (6–8). Even without gastric symptoms, eradication of Helicobacter pylori had beneficial effects in a small controlled trial of patients with migraine (9). Thus, the observed association between proton pump inhibitors and headache may be driven by the link between gastrointestinal diseases and headache and not by the consequence of the drug class. The observation that the association is strongest in the time window closest to the headache may indicate that the activity of the gastric disease may be important for headache occurrence. In a large population-based cohort study, the propensity for gastrointestinal symptoms increases with increasing headache frequency (3).

Second, as some pain medication and some migraine-associated features (such as stomach upset) may increase the likelihood of proton pump inhibitor prescription, the reported association may be an overestimate. If we assume the result is causal, then a substantial increase in headache incidence should have been observed after the introduction of proton pump inhibitors. However, there is no indication of a strong increase in headache prevalence (10).

Third, the biological mechanisms as to why proton pump inhibitors could cause headache remain unclear. While there are some differences in the metabolism and bioavailability of the various proton pump inhibitors, it further remains unclear why only specific formulations should cause headache.

Last, the data used suffer from typical limitations of administrative databases, such as lack of detail on personal information. While the case-crossover design controls for stable personal characteristics, changing factors, such as dietary information, is not accounted for. In addition, the included patients very likely had a history of headache prior to the “incident” headache event that qualified them for participation. However, the role of headache history and potential (non-proton pump inhibitor) triggering factors were not evaluated.

In summary, while this study provides some evidence linking specific proton pump inhibitors with headache in general, several aspects of this potential association remain unclear and should be evaluated in targeted future analyses.

While the assessment of gastrointestinal symptoms among patients with headache disorders (and vice versa) is re-emphasized by the findings of the study, change of clinical practice in prescribing proton pump inhibitors specifically for headache patients is, based on current evidence, not supported.