Illustrating the Sense of a Network Meta-Analysis by Means of Dedicated Plots: A Way for Making It Conceptually Easier and More Immediately Understandable.

When there is a need to compare three or more treatments for the same clinical indication, the network meta-analysis (NeMa) offers the important advantage of “incorporating” in a single analysis all the evidence available, allowing therefore to better manage any condition of multiple comparisons. Therefore NeMa overcomes the main limit of traditional meta-analysis; the latter, in fact, is able to compare two treatments against each other, but is not able to analyze the cases in which the treatment regimens to be compared are ≥ 3. Considering on the one hand two innovative treatments T1 and T2 and on the other hand the corresponding standard treatment S, it is frequent the case in which there are randomized controlled trials (RCTs) comparing T1 vs. S and also T2 vs. S, but there is a lack of controlled head-to-head comparison studies between the two innovative drugs, i.e., studies based on direct comparison of T1 with T2. Within a NeMa, the comparison between two treatments that have been the subject of a specific comparative RTC is defined “direct” (in this example, the comparisons T1 vs. S and T2 vs. S), while we usually term “indirect” the comparison between two treatments for which it would be interesting make a comparative assessment, but for which a specific RCT does not exist yet (e.g., T1 vs. T2).

According to personalized way of representing the NeMa graphs adopted by our team, the direct comparisons are depicted by means of continuous arrows, while the indirect ones are represented by dashed arrows (Fig. 1).

Figure 1

Network meta-analysis graph concerning comparisons between dabigatran, rivaroxaban and warfarin in patients with atrial fibrillation.

The statistical calculation techniques concerning NeMa are complex. Therefore, on this point we prefer to refer to the specialized literature about the subject [1-3]. Nevertheless, as for the interpretation of the results, our modality of drawing the NeMa graphs might become a major strength of this new technique, due to communicative efficacy of the instrument and its ability to synthesize the experimental evidence.

What can the reader receive from this network of arrows and numbers? First of all, the distinction between continuous arrows and dashed arrows seems to be very functional to us in order to identify direct and indirect comparisons. Moreover, the symbols +, - and =, adopted to indicate the data with statistical significance of each comparison, represent a useful notation in our opinion.

Despite the graphical effectiveness of this innovative tool, the issue of what is the real meaning of the NeMa is still alive, that is, it is open to more than one interpretation. In fact, there are more conservative positions, which recognize for the NeMa only the merit of proposing in quantitative terms what is already evident in narrative terms, and more favorable positions according to which all these statistical comparisons, both direct and indirect, deserve in full to be presented because they provide useful knowledge to the problem of multiple comparisons between subjects, e.g. between multiple different drugs. To illustrate the practical application of NeMa, two examples are given below.

Example 1 is the first example which takes into consideration two widely known anticoagulant drugs (dabigatran and rivaroxaban) studied in the treatment of patients with atrial fibrillation at risk of cardio-embolic stroke or non-central nervous system (CNS) systemic embolism (renal, splenic, mesenteric, peripheral arterial embolism, etc.). No RCT has been reported in the literature to directly compare dabigatran with rivaroxaban in the clinical setting of non-valvular atrial fibrillation (NVAF). In fact, the three currently available studies of direct comparison of dabigatran with rivaroxaban [4-6] are observational studies, more exactly retrospective cohort studies, and therefore, as such, burdened by the threat of possible biases, for example confounding by indication. On the other hand, for the prevention of cardio-embolic stroke in NVAF there are two large RCTs, the former of which includes dabigatran [7] whereas the latter is centered around rivaroxaban [8, 9], both having the warfarin as comparator drug, administered according to international normalized ratio (INR).

More exactly, in the trial by Connolly et al [7] dabigatran was administered at a dose of 110 mg twice daily to a first group of patients, at a dose of 150 mg twice daily to a second group with the aim of comparing these two groups with a third group of patients receiving warfarin according to their INR. Conversely, in the ROCKET-AF trial [8, 9], rivaroxaban was given at a dose of 20 mg per day. The two studies differ for the selection modality of patient population: for the ROCKET-AF trial the patient population had at least two risk factors and a CHADS2 score ≥ 3; for the RE-LY trial the patients had at least one risk factor and a CHADS2 score ≥ 0. In both studies, the primary end-point was the occurrence of stroke or non-CNS systemic embolism.

The results of the dabigatran trial show that, compared to warfarin, the new anticoagulant at a dose of 150 mg twice daily has greater efficacy and equal adverse reactions (major bleeding), while at a dose of 110 mg twice daily it is equally effective and shows less adverse reactions, in particular less cases of major bleeding. On the other hand, the rivaroxaban trial shows that the new anticoagulant has a substantially overlapping efficacy compared to warfarin, albeit the trend in favor of the new drug does not reach the threshold of statistical significance. In the absence of a direct comparison trial between dabigatran and rivaroxaban, the network meta-analysis summarizes all the comparative efficacy data concerning these three drugs (Fig. 1) [7-9].

Example 2 is the second example which concerns the first-line treatment of chronic myeloid leukemia in Philadelphia chromosome-positive patients. Since the time of the first trial (which dates back to 2003) [10], imatinib has been proved to be a safe and effective standard of care in first-line treatment of chronic myeloid leukemia. In 2010, a study of dasatinib as first-line [11] showed significantly better results than imatinib, and at the same exact time a study centered on nilotinib [12] demonstrated the superiority of this second drug compared to imatinib. Finally, in December 2010, a study was published that showed the superiority of the combination therapy imatinib + pegylated interferon compared to imatinib alone [13]. Head-to-head comparison trials among the three most innovative options, namely direct comparisons concerning nilotinib vs. dasatinib vs. imatinib + interferon are missing. Choosing the endpoint of the “major molecular response” at 12 months, the network meta-analysis generates the scheme illustrated in Figure 2 [11-13]. Based on the data shown, no significant difference appears to emerge between the three new treatments.

Figure 2

Network meta-analysis graph concerning comparisons between imatinib, dasatinib, nilotinib, and association imatinib plus peginterferon in chronic myeloid leukemia with positivity of chromosome Philadelphia. End-point: major molecular response.

In conclusion, NeMa might play a role of paramount importance in the evidence-based medicine and our network plots may be useful as an innovative iconographic complement for data presentation.