BMJ meta-analysis challenges ATS pneumonia guidelines

In the June issue of Cleveland Clinic Journal of Medicine, (vol 72 number 6 june 2005, pg464 ) in their Patient Oriented Evidence that Matters section we find their summary of a recent BMJ meta-analysis of 18 studies that included 6,7049 patients. The conclusion is in the headline of the POEM section "Antibiotic choice makes little difference in community-acquired pneumonia".
The referenced article is by Mill GD, et al BMJ 2005:330 456-460. Since POEMs are said to be evidence that matters, there must be a take home lesson in each of them. Here it would appear to be "go with the [cheaper] beta lactams for community acquired pneumonia (CAP)". Yet this advice would be counter to the recommendations of both the ATS and IDS in regard to CAP. Both have said a beta lactam alone is not adequate. So when the cards of the EBM are dealt and one hand is a Meta-analysis and one is a treatment guideline from a national speciality organization, which is the trump? From a defensive medicine point of view, you probably have to go with the guidelines. The response of the ATS CAP committee members will be eagerly awaited as clearly the gaunlet is thrown down by these investigators from New Zealand. They state clearly these data support the BTS recommendation to use beta lactams and point out this is contrary to the ATS guidelines. To date none of the rapid responses seem to be from ATS committee members.
One caveat is that these patients were mild or moderately severe cases not severe cases requiring IV drugs.
Subgroup analysis showed no difference between drugs active against atypical pathogens (Mycoplasma and Chamydia) and beta lactams in cases in which those pathogens were the causative agent.A difference was shown in regard to Legionella. I would be interested to hear what the pulmonary docs at Pulmonary Roundtable think of this article. Should we abandon the "U.S. approach" i.e include coverage for the atypical pathogens based on this Meta-analysis?
I think not. Meta-analysis certaintly have their limitations which have been discussed here before. We have seen MAs reach opposite conclusions on the same issue and we have seen instances on a large RCT trumping an earlier MA.