This is often a controversial subject because there hasn't been much, if any scientific research focusing on dogs being able to alert to migraines or how exactly they do it. Many trainers often compare it to the same training as Diabetic Alert Dogs. (DAD's) I've been told that research can't scientifically be done because there is no way to test when the migraine actually starts and stops so there is no way to pinpoint the exact start and end of the trail. Therefore any studies currently out are based on gathering date from personal annotates and self reporting surveys.
The migraine alert/response is often a mix of natural alert and taught or reinforced behavior. Behaviors that are reinforced get repeated.
I've taught 2 personal SD's how to learn to alert to my migraines. It's a struggle trying to teach something that can be so vastly different for different people! But if neurologists struggle with figuring out all of the individualized issues in migraine sufferers, how are we supposed to figure out what works for us without trial and error? I'm convinced my migraines have a smell the dogs can pick up on. I do fun scent work with all my dogs to teach them how to use their nose to discriminate between smells. The dog's nose is their strongest sense and dogs naturally know how to sniff. But scentwork helps them learn that certain smells are more important than others. My first SD always smelled my ears & fingers, then would either alert or walk away if an alert wasn't needed. Azul will air sniff, stare at me, and air sniff again before giving me an alert or deciding I'm fine. The last time we taught this class I decided to try and teach Azul the scent that happens with a bad auto-immune flare of rapid inflammation. He picked that up really fast and now also alerts me when I'm about to swell up by licking the area that is going to swell or licking my hand if he can't reach that area.
Cardiac Alerts are another place where there is no scientific data as of yet to determine if dogs are actually alerting to the medical issue or not. There are also various types of cardiac medical issues. It's also very challenging to tell if the dog is alerting to an actual smell that changes in the person or if the dog is hearing the heartrate change or breathing patterns change, or some other body language change.
Personal stories of dogs alerting & responding to cardiac issues are widely spread. There are now some small programs working on training this task, but it's still not largely accepted in the Service Dog Community as a task that can be trained in a specific way.
Diabetic alerts on the other hand have been the topic of multiple research studies. Diabetics can measure their blood glucose levels practically any time, therefore there is an exact start time with proof of need and often an exact end time. Now high & low numbers vary from person to person slightly, but there is a documentable need for when it's time to take action.
Still with any alert it's difficult to know if the dog is picking up a smell or seeing a behavior. But most programs that train DAD's use scent training and also insist that a dog be trained response tasks like getting a glucose tester, getting a food/drink that will help the person, etc. This way the person can document and demonstrate if needed an actual response task. You can't fake an alert when needed, but you can cue the response behavior and get a response if you need to making that a more reliable task.
This is a sample from the Scent-based Medical Alert Workgroup that is currently running. The Crazy2Calm Canine Coaches offer this class twice a year.