What do you do if you want to know what is the best treatment for a condition, and treatment A and treatment B have both been compared to placebo, but have never been compared to each other in RCTs? If there are multiple options, rather than just 2, then it is usual that several have not been directly compared to each other.

One way to approach this is to do what is called a network meta-analysis, where you try and calculate the relative efficacy of each treatment, and then estimate whether A is better or worse than B. Is this valid? I am not sure… (and I am not alone Li T, Puhan MA, Vedula SS, Singh S, Dickersin K: Network meta-analysis-highly attractive but more methodological research is needed. BMC Med 2011, 9:79.) If the samples were very homogeneous and many details of the research protocols, such as timing and duration of intervention, timing and nature of outcome measures etc etc were identical, then a network meta-analysis might be enticing. Unfortunately that is not often the case.

A new network meta-analysis has been published (Haas DM, Caldwell DM, Kirkpatrick P, McIntosh JJ, Welton NJ: Tocolytic therapy for preterm delivery: systematic review and network meta-analysis. BMJ 2012, 345) examining the efficacy of tocolytics to suppress preterm labour and prevent preterm delivery. The analysis is interesting, but rather complex.  They first did systematic reviews of all of the pairwise comparisons for which there was trial data, then performed the network thing. The final evaluation gives a probability that a particular agent is the best, 2nd best, 3rd best and so on, for delaying delivery for a least 48 hours, for maternal side effects etc. They indicate that antiprostaglandin agents (mostly indomethacin, but 4 others have been studied and are lumped together as one category in this meta-analysis) are probably the most effective for delaying delivery for at least 48 hours, and oxytocin receptor antagonists are probably the safest (although the antiprostaglandin agents came close). The authors also confirmed that there is no clear evidence of clinical benefit, in terms of reduction in mortality or the risk of acute lung disease in the baby. A skeptical editorial accompanies the article (skeptical of the benefits of tocolytics, not of the systematic review). Having said that the evidence that these agents are no better or worse than placebo is fairly weak. The quality of the studies is generally good, as Dr Haas and co-workers showed in another recent publication, with the major exception that too many of them are too small, especially those comparing to placebo.

There has been a network meta-analysis in neonatology, only one that I am aware of, (Jones LJ, Craven PD, Attia J, Thakkinstian A, Wright I: Network meta-analysis of indomethacin versus ibuprofen versus placebo for PDA in preterm infants. Archives of Disease in Childhood – Fetal and Neonatal Edition 2011, 96(1):F45-F52). Published last year, and with a much simpler network, indomethacin and ibuprofen versus placebo. They seemed to demonstrate that both drugs are equally effective at closing the PDA, but that ibuprofen may lead to more BPD.

** Update ** I knew there was a recent article published about how to read and use a network meta-analysis, but my RAM failed me (my real actual memory) and my database search was fruitless, partly because the authors call this a multiple treatment comparison meta-analysis rather than ‘network’. Anyhow an addition to the JAMA series ‘A User’s Guide to the Medical Literature’ takes you through the topic in a very useful, usable, way.  (Mills EJ, Ioannidis JP, Thorlund K, Schunemann HJ, Puhan MA, Guyatt GH: How to use an article reporting a multiple treatment comparison meta-analysis. JAMA : the journal of the American Medical Association 2012, 308(12):1246-1253.)