Not often enough for the trials to separate it from placebo. Zepbound’s label pools two placebo-controlled trials in which 2,519 adults took the drug for up to 72 weeks, and its side-effect table lists every reaction that reached at least 2 percent and ran above placebo [1]. Nausea made that table at 25 to 29 percent against 8 percent on placebo, and dizziness made it at 4 to 5 percent against 2 percent, but headache is not on it. Across twelve randomized trials of tirzepatide in 10,615 people with type 2 diabetes, the risk ratio for headache was 1.00, with a 95% confidence interval from 0.65 to 1.55 [2].
The question is easy to picture: a few weeks in, the day after a dose step, when lunch was half a sandwich and the water bottle stayed full in the bag. Whether the pen caused the headache or the day did is exactly what a placebo arm exists to settle. That is why the answer here leans on trial tables first and on reports second, and why the reports get their own section rather than a vote.
What the label table can and cannot say
The table has a threshold with two parts, and headache could have failed either one. Its absence means headache was either under 2 percent on Zepbound or no more common than on placebo, and the label does not print a headache row that would say which [1]. The table does show where the drug’s burden sits, and it sits in the gut: vomiting at 8 to 13 percent against 2 percent, constipation at 11 to 17 percent against 5 percent.
The same kind of table does catch a headache signal when one exists, which is the useful comparison. Wegovy’s label lists headache at 14 percent of 2,116 people on semaglutide 2.4 mg against 10 percent of 1,261 on placebo [6]. That is four extra headaches per hundred people, and it also shows how common headache is without any drug, since one person in ten on placebo reported one. These were separate trials with different people, so the pair is a contrast between two labels rather than a head-to-head result, and the full comparison across every outcome sits in which GLP-1 has the worst side effects.
The randomized figure for tirzepatide itself comes from diabetes trials, not the obesity program. The meta-analysis pooled twelve trials, grouped in its subgroups by comparator type, and headache landed at a risk ratio of exactly 1.00 [2]. Dizziness did not: it rose significantly, to a risk ratio of 1.64 (95% CI 1.04 to 2.57), and the authors report the higher risk in patients aged 60 or younger. A reader who feels light-headed rather than head-sore is describing the side effect the trials did detect.
Why the reports keep coming anyway
Spontaneous reporting systems tell a louder story. In the FDA’s adverse event database, a study of 25,110 neuropsychiatric reports for the GLP-1 class found headache reported more often than expected, at a reporting odds ratio of 1.74 (95% CI 1.65 to 1.84) [3]. The median time from starting the drug to the reported event was 16 days, with an interquartile range of 3 to 66 days, across those neuropsychiatric reports as a group rather than headache alone.
The European database shows the same pattern for tirzepatide specifically. Headache made up 2.51 percent of tirzepatide’s reports there, a reaction its European product summary does not list, and the authors ask for cautious interpretation of reporting biases [4]. On Facebook, headache was the adverse event mentioned most often in posts about tirzepatide, in 78 of 4,202 posts, or 1.86 percent [5].
None of those numbers is a rate among people taking the drug. A reporting odds ratio compares how often headache shows up in reports for one drug against reports for all the others, and it can rise with attention as well as with harm. The gap between what trials count and what reports count has come up before on this site, in the psychiatric-risk pooling and in what the adverse event reports actually list. Reports are where a signal is first noticed, and a placebo-controlled trial is where it is either confirmed or not.
The two routes the label does name
The first is low blood sugar, and it belongs mostly to people with type 2 diabetes. In the trial of adults with diabetes, blood glucose below 54 mg/dL was reported in 4.2 percent of people on Zepbound against 1.3 percent on placebo, and in 10.3 percent of those also taking a sulfonylurea [1]. The patient information lists headache among the signs of low blood sugar, alongside sweating, shakiness and confusion. In the trial without diabetes the same reading appeared in 0.3 percent of people on the drug and in none on placebo.
The second is fluid loss. The label ties nausea, vomiting and diarrhea to dehydration, and reports low blood pressure in 1.6 percent of people on Zepbound against 0.1 percent on placebo, higher in those also taking blood-pressure medicine [1]. It says that hypotension also occurred alongside gastrointestinal side effects and dehydration. The stomach side of that story, and how long it lasts through each dose step, is covered in how long nausea lasts on a GLP-1.
A headache is not the same question as migraine
Someone who already lives with migraine is asking something different, which is whether the drug changes attacks they were having before it. That question has its own evidence, mostly on semaglutide and triptan prescriptions, and it is laid out in whether GLP-1 drugs help migraine. The FDA database adds one awkward note to it, a small migraine signal for the class at a reporting odds ratio of 1.28 (95% CI 1.06 to 1.55) [3]. Headache is also the main symptom of raised pressure around the brain, a condition these drugs are being studied to treat, as the intracranial hypertension evidence explains.
What remains open is the detail a person actually wants in week three. No trial behind the label published headache by dose step, by severity or by how many days it lasted, so there is nothing to say about whether a headache after the move to 10 mg fades by the next pen. Tiredness sits in the same gap, and the fatigue evidence shows how little the trials were built to catch.