Skip navigation

Diagnosing migraine

 

Headaches are common. Everyone has a headache at some time in their lives, but for some it is more frequent, sometimes or always. Those headaches may be migraines, but they may also be "normal" headaches, brought on by tension, or whatever.

Bandolier has al ways found that a bottle or so of red wine brings on a migraine. How can we tell if it is a migraine? We have the International Headache Criteria for diagnosing migraine, but not always to hand. A new meta-analysis [1] tells us what clinical features are important, and what irrelevant, in diagnosing migraine.

Review

The review used a MEDLINE search up to May 1999 for English language papers with sensitivity and specificity of historical features in patients with primary headaches of migraine, tension-type headaches, or cluster headaches. Headaches caused by other underlying clinical conditions were not examined. Information from studies was pooled to generate overall sensitivity, specificity and likelihood ratios of migraine compared with tension-type headache. The criteria for diagnosis of headache may have been IHS criteria, or similar. Results are given for IHS criteria pooled and all. They were much the same, so the information for all criteria is given below as it represented a larger data set.

Table 1: Headache features in migraine compared with tension-type headache

Symptoms

Likelihood ratios

Positive

Negative

LR+

95% CI

LR-

95% CI

Nausea

19.2

15-25

0.20

0.19-0.21

Photophobia

5.8

5.1-6.6

0.25

0.24-0.26

Phonophobia

5.2

1.5-5.9

0.38

0.36-0.40

Activity makes it worse

3.7

3.4-4.0

0.24

0.23-0.26

Unilateral

3.7

3.4-3.9

0.43

0.41-0.44

Throbbing/pulsing

2.9

2.7-3.1

0.36

0.34-0.37

Precipiting features

The folklore is that a whole range of factors precipitate a migraine attack. These might be cheese or chocolate, or stress or red wine, or whatever. The finding of this review was that most common precipitating factors were not discriminating for migraine compared with tension-type headache. Positive likelihood ratios were about 1 for stress, alcohol, weather change, menstruation, missing a meal, lack of sleep and perfume or odour.

The only factors with positive likelihood ratios were chocolate, cheese and any food (Table 2). The negative likelihood ratios for these were not much different from 1, indicating that absence of these features was unhelpful in making a diagnosis.

Table 2: Headache precipitants in migraine compared with tension-type headache

Precipitating factor

Likelihood ratios

Positive

Negative

LR+

95% CI

LR-

95% CI

Chocolate

7.1

4.5-11.2

0.82

0.73-0.93

Cheese

4.9

1.9-12.5

0.68

0.62-0.73

Any food

3.6

2.8-4.6

0.59

0.56-0.62

Family and personal medical history

When compared with patients who had no history of headaches, a family history of migraine, childhood vomiting attacks and motion sickness all had moderate positive likelihood ratios for diagnosis of migraine (Table 3). Negative likelihood ratios were unhelpful.

Table 3: History in migraine compared with patients with no history of headaches

Factor

Likelihood ratios

Positive

Negative

LR+

95% CI

LR-

95% CI

Family history of migraine

5

4.4-5.6

0.47

0.46-0.49

Childhood vomiting attacks

2.4

1.9-2.9

0.79

0.75-0.82

Motion sickness

2.2

1.9-2.5

0.79

0.75-0.82

Comment

This is all very helpful stuff. Perfume, or a period, or a change in the weather might produce a headache, but they are no guide as to whether that headache is ma migraine or not. The likelihood ratios are very, very helpful in this.

Firstly we know those factors that are not important, and those that are. For those that are important, we know the relevant weights. But we can use the signs, symptoms, predisposing factors and history to make diagnosing a migraine much easier.

For example, suppose a young man complains of unilateral throbbing headaches, made worse by any physical activity. His mother had migraines also, and he thinks that chocolate may bring it on.

We know that the prevalence of migraine in men is about 6%. This is our pre-test probability. By sequentially using the likelihoods for unilateral headache (3.7; red in Figure 1), throbbing (2.9; blue in Figure 1), exacerbation by physical activity (3.7; green in Figure 1) and family history (5.0; black in Figure 1) we arrive at a post-test probability of about 95% that the young man has a migraine.

Figure 1: Likelihood ratio nomogram

Now we can ask some questions in the MIDAS questionnaire to discover about the level of disability he has, and then we will have a good idea of how to treat him using data from the DISC study , and if he needs a triptan which is likely to be the most effective.

Reference:

1 GW Smetana. The diagnostic value of historical features in primary headache syndromes. Archives of Internal Medicine 2000 160: 2729-2737.