Skip navigation

Prophylactic removal of impacted third molars


What do you do when there is no evidence? Carry on with what you are doing because you have no evidence to stop, or stop what you are doing because there is no evidence to carry on? This is not a hypothetical question, because in this new 'evidence-based world', much of what we do is still often based on what is time-honoured rather than trial-honoured.

The obvious thing to do, and one that Bandolier tries to encourage, is to search out a systematic review, and preferably one based on randomised trials. This should give the highest level of evidence. But what if reviews themselves differ in the advice they give?

Prophylactic removal of impacted third molars

Opinions about prophylactic removal of impacted third molars vary widely. Some people seem to think it is a good thing, because it prevents later pain and suffering. Others can't see the reason for doing an unnecessary procedure unless the teeth are causing problems.

A systematic review of reviews [1] examined the pathologies associated with impacted third molars, and outcomes following surgical removal of third molars. A thorough literature search found 12 published reviews fit for inclusion, and five which were not.

For inclusion a review had to :

  1. be a review of research literature
  2. address either pathology and/or symptoms associated with impacted third molars or outcomes following surgical removal
  3. be published after 1985 and be written in English.

Results of reviewing the reviews

Only one review gave its search strategy and criteria for including a research paper. The number of references ranged from 9 to 149 (median 43 references) - with wide variations also in the number of references addressing different aspects of the problem.

And what was their conclusion?

Two reviews concluded that prophylactic removal of third molars is a good thing. The other ten concluded that there was a lack of evidence to support prophylactic removal of impacted third molars.

And the review of reviews itself concluded that in the absence of good evidence to support prophylactic removal there appears to be little justification for removing pathology-free impacted third molars.

What to do when systematic reviews disagree

Fortunately we can now find a guide through this particular minefield [2]. Ask some simple questions:

  • Do the reviews ask the same question?
  • Do the reviews include the same trials?
  • Is the same endpoint used, and are the methods of analysis comparable?

None of this is difficult stuff, but the authors of this useful contribution take us step by step through the important signposts to resolving discrepancies. Of course, there may always be times when reviewers disagree, but then that's all part of the fun. If nobody disagreed we wouldn't make much progress.

Never mind the quality?

No, with systematic reviews you must take quality into account. That the outcome of a systematic review is likely to be influenced by its methodological quality has been thoroughly examined in systematic reviews in pain [3].

This paper looked for pain meta-analyses. To be included in this study, reports had to meet the following criteria:

  1. They had to be described as meta-analyses or, if not, they had to include pooled analysis of the results of several independent primary studies. Studies in which statistical synthesis had been planned but was deemed to be inappropriate were also included.
  2. They had to incorporate trials in which pain was an outcome measure or in which analgesic interventions were compared for outcomes other than pain within the context of a painful condition (e. g. a study looking at the validity of grip strength to assess the effectiveness of NSAIDs in rheumatoid arthritis).
  3. They had to be published or accepted for publication.

Each study was evaluated independently (and blinded) by both the investigators using Oxman and Guyatt's index [4]. The details of this are given on the next page, together with information about its use, so Bandolier readers can use it themselves. Only items scored as 'yes' or 'not applicable' were regarded as present and those scored as 'no', 'cannot tell' or 'partially' were regarded as absent. The maximum possible overall score for a given study was 7.

Did quality make a difference?

Of the reports found, the conclusions were positive in 60 meta-analyses (75%), negative in 7 (9%) and uncertain in 12 (15%). One meta-analysis (1%) did not reach any conclusion.

The distribution of the overall scores between meta-analyses with positive conclusions and those with negative or uncertain results was different (chi-squared 18.2, p = 0.006). Meta-analyses of high quality were less likely to produce positive results. Sixteen out of 19 meta-analyses with negative or uncertain results had overall quality scores above the median value, while that was the case for only 20 of the 60 with positive results.

Meta-analyses which included only randomised trials were less likely to produce positive conclusions (19 of 31, 61%) than those which included other study designs as well as or instead of RCTs (41 of 49, 84%). The difference was statistically significant (chi-square = 5.07; p = 0.024).


Reviews which are of high quality tended to be less enthusiastic about an intervention. Those which used only randomised trials were also less enthusiastic. This serves to drive home the lesson about using high quality evidence in making choices.

So if we are faced with reviews, systematic or not, which disagree, go for the highest quality. And if reviews are not systematic, or contain no randomised trials, like those for the prophylactic removal of wisdom teeth, regard them with a cold and fishy eye!


  1. F Song, DP Landes, A-M Glenny, TA Sheldon. Prophylactic removal of impacted third molars: an assessment of published reviews. Oral Surgery 1997 182: 339-46.
  2. AR Jadad, DJ Cook, GP Browman. A guide to interpreting discordant systematic reviews. Canadian Medical Association Journal 1997 156: 1411-6.
  3. AR Jadad, HJ McQuay. Meta-analyses to evaluate analgesic interventions: a systematic qualitative review of their methodology. Journal of Clinical Epidemiology 1996 49: 235-43.
  4. Oxman AD, Guyatt GH. Validation of an index of the quality of review articles. Journal of Clinical Epidemiology 1991 44: 1271-8.

previous or next story in this issue