Jan 10

Tricyclic antidepressants and headaches

Journal reference: Jackson JL, Shimeall W, Sessums L, et al.  Tricyclic antidepressants and headaches: systematic review and meta-analysis. BMJ 2010;341:c5222 [1]

Link: http://dx.doi.org/10.1136/bmj.c5222

Evidence cookie says...

Tricyclic antidepressants are possibly/probably effective in reducing the frequency and intensity of migraine and tension-type headaches.

The magnitude of effect appears moderate to large when compared to placebo [1]:

  • migraine headaches:
    • ↓ 1.4 headaches/month
    • 80% more likely to be at least half as intense
  • tension-type headaches:
    • ↓ 6.9 headaches/month
    • 41% more likely to be at least half as intense

There are substantial limitations with the reliability of the evidence base; these results should not be considered authoritative.

More details:

Article details

Study design:

systematic review and meta-analysis

Study aim:

to evaluate the efficacy and relative adverse effects of tricyclic antidepressants (TCA) in the treatment of migraine, tension-type and mixed headaches

Methods summary:

Trial eligibility:

  • Inclusion criteria:
    • randomised trials that evaluated efficacy of TCA in reducing frequency or severity of migraine or tension-type headaches
    • treatment groups required to receive a TCA regularly as a single intervention for at least four weeks (any dosing schedule)
    • comparison groups could receive placebo or a specified alternative drug or non-drug treatment
    • participants were over 18 years of age, and had diagnosed migraine or tension-type headache (frequent episodic or chronic) that could reasonably be defined on the basis of diagnostic criteria described in 1988 by the International Headache Society.
  • Excluded:
    • trials where there was additional treatment of any other drugs with possible prophylactic benefit or effect augmentation
    • secondary headaches eg. due to drug overuse, concussion or lumbar puncture


  • Medline, Embase, the Cochrane trials, PsycINFO, PsycLIT (1966 until 25 March 2010)
  • also search of unpublished literature (CRISP and FEDRIP databases)


  • articles selected by inclusion in two stages
    • initial stage:
      • independent review by two researchers of titles and abstracts to select for full text retrieval
      • only modest inter-rater agreement (κ = 0.66)
      • all articles included at this stage if recommended by either reviewer
    • second stage:
      • two other researchers independently applied inclusion and exclusion criteria using standardised forms
      • good agreement (κ = 0.86)
      • disagreements resolved by consensus
  • data independently abstracted by two authors independently onto standardised forms
    • studies reporting dichotomous outcomes with less than 50% improvement were not abstracted
    • article quality formally assessed with good agreement (κ = 0.83)
  • trials assessed for heterogeneity
  • individual patient data pooled and expressed as confidence intervals
  • comparisons were made between:
  • the following were also assessed:
    • small study effects (publication bias)
    • estimated number of unpublished trials
    • effect of study sponsorship

Results summary:

  • 1471 potential articles identified → 443 articles were retrieved and assessed for eligibility
  • 37 studies were included:
    • 13  studies on migraine headaches → all meet current criteria
    • 17 studies on tension-type headaches
    • 6 chronic mixed headaches → these were reclassified by the researchers for analysis
      • better met criteria for migraine headache than for tension-type headache
    • 1 on psychogenic headaches

Analysis of all trials:

  • 3176 participants
  • average duration of treatment 10 weeks (range 4-26 weeks)
  • drugs:
  • study quality:
    • 13 studies had dropout rates of over 30%
    • losses to follow up could not be determined in 3 studies
    • intention to treat analyses were done in only 7 trials
    • majority of trials had limitations on design and reporting
      • few studies reported method of randomisation or allocation concealment
      • blinding success was only reported in one trial, and was questionable in all

TCA vs placebo:

  • migraine headache frequency:
    • baseline: 4.7 headaches per month (95% CI 4.3-5.1)
    • effect size (average standardised mean difference): -1.00 (95% CI -1.52 to -0.48) [-1.00  is considered a large difference but note the wide confidence interval]
    • heterogeneity: I2 = 89.4% [large heterogeneity between studies; see section on methodological weaknesses]
  • migraine headache intensity, likelihood of at least 50% improvement:
    • relative risk: 1.80 (95% CI 1.24-2.62), four studies, I2 = 0.0%
  • tension-type headache frequency:
    • baseline: 16.9 headaches per month (95% CI 15.8-18.0)
    • effect size: -0.99 (95% CI -1.66 to -0.32) [-0.99  is considered a large difference but note the wide confidence interval; an effect size larger than -0.30 is considered insignificant]
    • heterogeneity: I2 = 93.8% [large heterogeneity between studies; see section on methodological weaknesses]
    • effect did not differ among the different categories of tension-type headaches
  • tension-type headache intensity, likelihood of at least 50% improvement:
    • relative risk: 1.41 (95% CI 1.02-1.89), four studies, I2 = 48.8% [note wide confidence interval and moderate level of heterogeneity; see section on methodological weaknesses]
  • side effects:
    • dry mouth, RR 2.34 (1.63-3.35)
    • drowsiness, RR 1.87 (1.25-2.70)
    • no other side-effects reached statistical significance when comparing TCA to placebo


  • migraine headache frequency:
    • no difference detected at statistical significance (very wide confidence intervals and large heterogeneity between studies)
  • migraine headache intensity, likelihood of at least 50% improvement:
    • relative risk1.72 (95% CI 1.15-2.55), three studies, I2 = 9.2%
  • tension-type headache frequency:
    • no difference detected at statistical significance (very wide confidence intervals and large heterogeneity between studies)
  • tension-type headache intensity, likelihood of at least 50% improvement:
    • relative risk: 1.73 (95% CI 1.34-2.22), four studies, I2 = 0.0%

Study conclusion:

Tricyclic antidepressants are effective in preventing migraine and tension-type headaches and are more effective than selective serotonin reuptake inhibitors or placebo, although with greater adverse effects.  The effectiveness of tricyclics seems to increase over time.


  • 3176 participants (from the 37 trials)
  • Average duration of treatment 10 weeks (range 4-26 weeks)
  • Ranges of study size 10-554 participants (mean 70 participants)
  • 73% of participants were female.  All but one study had more female than male participants.
  • Mean age of participants was 39.6 years
  • Depression:
    • 7 studies excluded participants with depression
    • 11 studies measured depression at trial entry, ranging from 21% to 98%
    • depression not measured in the remainder

Methodological weaknesses

  • only modest initial agreement between researchers on studies selected for full-text retrieval
    • there is a possibility that the literature search is not exhaustive
  • source trials tended to be small, short and generally poor methodological quality
  • recategorisation of the participants in trials on mixed type headache into the migraine group may not be diagnostically valid for all the participants
  • substantial heterogeneity between studies in the meta-analyses and only a proportion of this is explained by the differences in duration of the trials
    • in particular, the results for migraine and tension-type headache frequency in the comparison between TCA and placebo had I2 of about 90%
    • that is, almost all of the variation in the estimates of treatment is due to study heterogeneity [2]; this is reflected in the wide confidence intervals (in the case of the effect size for TCA vs placebo for tension-type headache, the 95% confidence interval range is from large effect size to insignificant effect size)
    • the pooled estimates from meta-analysis are of questionable validity when there is such large heterogeneity between studies
  • decision to exclude data from studies reporting dichotomous results showing less than 50% improvement with treatment was not explained → likely to overestimate the effect of TCA
  • depression is an important and uncontrolled confounding factor in over half the studies included in the meta-analysis:
    • 19 (of 37) trials did not measure or report on depression
    • of the 11 studies that measured the rate depression at trial entry, the range was 21-98%
  • high dropout rates bias results towards increased effect of TCA
  • detected publication bias likely biases results
  • although the authors candidly discuss the weaknesses and unreliability of their results in their discussion, this is not reflected in their article abstract or conclusion

Methodological strengths

  • generally rigorous methodology
  • reviewed and reported study quality
  • attempted to source data from both published and unpublished trials
  • independent review for selection into study and of data by duplicate (two different researchers) at each stage

Biases and conflicts of interests

  • nil declared and none seem obvious

Clinical relevance to primary health care

Tricyclic antidepressants (TCA) are commonly used for headache prevention in Australia.  In this meta-analysis, TCA were found to be more effective than placebo and selective serotonin reuptake inhibitors (SSRI) for both migraine and tension-type headache intensity. The study authors also report that TCA substantially reduced the number of days of headache (compared to placebo) but that conclusion is much less reliable and of unclear validity due to large study heterogeneity.

Caution needs to be taken when interpreting these results.  There are substantial biases that are likely to overestimate the efficacy of TCA. It is unclear what proportion of the study population suffered from depression; there is good reason to assume that this is not insignificant as the rate of depression in the small subset of trials that measured it ranged from 21-98%. This is a very significant uncontrolled confounding factor that threatens internal validity. Unfortunately, there have been few high quality studies of TCA in headaches. The lack of long term follow up limits the conclusions that can be made for long term use of TCA in headaches despite the association found between treatment duration and efficacy in the meta-analysis.

Nevertheless, despite the methodological limitations, TCA remain clinically significant for the treatment of both migraine and tension-type headaches.  The magnitude of effects appears moderate to large. A take home message for Australian general practitioners in primary health care is that tricyclic antidepressants are possibly/probably effective for the prevention of both migraine headaches and tension-type headaches, but there is relatively little reliable evidence. It should be noted that this study does not provide evidence comparing TCA to other commonly used medications for headaches such as β-blockers.


  1. Jackson JL, Shimeall W, Sessums L, et al. Tricyclic antidepressants and headaches: systematic review and meta-analysis. BMJ 2010;341:c5222
  2. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med. 2002 Jun 15;21(11):1539-58

Co-editor: Michael Tam

Print Friendly, PDF & Email

Leave a Reply

%d bloggers like this: