Diversity when interpreting evidence in network meta-analyses (NMAs) on similar topics: an example case of NMAs on diabetic macular oedema
Systematic Reviews volume 12, Article number: 189 (2023)
Different network meta-analyses (NMAs) on the same topic result in differences in findings. In this review, we investigated NMAs comparing aflibercept with ranibizumab for diabetic macular oedema (DME) in the hope of illuminating why the differences in findings occurred.
Studies were searched for in English and Chinese electronic databases (PubMed, Embase, Cochrane Library, Web of Science, CNKI, Wanfang, VIP; see detailed search strategy in the main body). Two independent reviewers systematically screened to identify target NMAs that included a comparison of aflibercept and ranibizumab in patients with DME. The key outcome of interest in this review is the change in best-corrected visual acuity (BCVA), including various ways of reporting (such as the proportion of participants who gain ≥ 10 ETDRS letters at 12 months; average change in BCVA at 12 months).
For the binary outcome of BCVA, different NMAs all agreed that there is no clear difference between the two treatments, while continuous outcomes all favour aflibercept over ranibizumab. We discussed four points of particular concern that are illustrated by five similar NMAs, including network differences, PICO (participants, interventions, comparators, outcomes) differences, different data from the same measures of effect, and differences in what is truly significant.
A closer inspection of each of these trials shows how the methods, including the searches and analyses, all differ, but the findings, although presented differently and sometimes interpreted differently, were similar.
With the rapid growth of biomedical evidence, systematic reviews provide an opportunity to make healthcare decisions based on comprehensive syntheses of the best available evidence on a topic [1,2,3]. Current knowledge may be imperfect, but decisions should be better informed when made in the light of the best, most up-to-date knowledge. It is essential that the systematic review itself is both clear and accurate for local interpretation by healthcare decision-makers (healthcare practitioners and policy-makers) [3, 4]. A problem arises when different research teams use similar approaches to synthesise evidence but report conflicting results. In this review, we investigate one such example in the hope of shedding light on the reasons for the differences in findings.
Our example comes from network meta-analyses (NMAs) comparing aflibercept (a vascular endothelial growth factor inhibitor) with ranibizumab (a monoclonal antibody) for diabetic macular oedema (DME). Diabetic retinopathy (DR) is a common complication of diabetes and the leading cause of blindness in the population of working age , with DME present in 4–7% of the population with DR . DME impairs vision-related functioning and quality of life (QoL) . It is the leading cause of moderate to severe visual impairment in the population with diabetes. DME is a significant economic burden for patients and public health systems [8, 9]. The therapeutic goal for people with DME is to improve visual function and vision-related QoL . Anti-vascular endothelial growth factor (VEGF) is recommended as a first-line treatment in several clinical guidelines [11, 12]. Aflibercept and ranibizumab are commonly used in clinical practice, but direct comparisons of these two drugs are limited. NMA then becomes an attractive option, as NMAs use previous studies of the two drugs directly compared with other controls to create statistically indirect comparisons of the two drugs of current interest.
We searched English and Chinese electronic databases (PubMed, Embase, Cochrane Library, Web of Science, CNKI, Wanfang, VIP (with search strategies described in Additional file 1)). Two authors (XNH and FQ) independently screened the search results to identify target NMAs that included a comparison of aflibercept and ranibizumab in patients with DME. For each included NMA, two authors (XNH and FQ) independently extracted the general study characteristics (such as search databases, number of studies, PICO [participants, interventions, comparators, outcomes] information) and assessed the quality of the included NMAs using the AMSTAR-2 tool . Disagreements at each stage were resolved by author team discussion and consensus. Tables and figures were used to summarise and present descriptive information.
General study characteristics
Everything seemed varied in almost every NMA (Table 1). Although the question under investigation was consistent, the searches, the numbers of trials used, and the definitions for eligible participants, comparisons from which to source data and acceptable outcomes mostly lacked rigid consistency. The work for these NMAs spanned at least 5 years based on the search dates (2013 to 2017). All included NMAs searched three main databases (MEDLINE/PubMed, EMBASE, and Cochrane Library). Korobelnik et al.  and Régnier et al.  additionally searched abstracts and unpublished data, and Muston et al.  and Virgili et al.  additionally searched other clinical trial registration platforms. The main difference in the eligibility criteria is the inclusion of interventions (such as bevacizumab, triamcinolone acetonide or pegaptanib was not in all NMAs). Therefore, the inconsistent inclusion of trials in the NMAs was due to number of reasons, the most important being the scope of the search and the inclusion criteria (such as the dosage of interventions, the regimen [PRN (pro re nata) or T&E (treat and extend)]). Details of the included trials in these NMAs and the situation of overlap are shown in Additional file 2. Despite lacking a rigid consistency of definitions, study identification and sampling across the five NMAs, the participants, studies, and outcomes were, generally speaking, recognisable to those with clinical and academic experience in this field.
Methodological quality of the included NMAs
The overall quality of the AMSTAR-2 assessment (Table 2) showed that only the Cochrane NMA  was of high quality, and the remaining NMAs had major methodological limitations. The percentage of NMAs that satisfied AMSTAR-2 seven critical domains were as follows: protocol registered before commencement of the review (item 2) ~ 40% (2/5), adequacy of the literature search (item 4) ~ 80% (4/5), justification for excluding individual studies (item 7) ~ 40% (2/5), risk of bias from individual studies being included in the review (item 9) ~ 80% (4/5), appropriateness of meta-analytical methods (item 11) ~ 20% (1/5), consideration of risk of bias when interpreting the results of the review (item 13) ~ 40% (2/5), and assessment of presence and likely impact of publication bias (item 15) ~ 20% (1/5). Regarding the potential sources of conflict of interest (item 16), Zhang et al.  and Virgili et al.  did not involve authors from the industry, the author teams of Korobelnik et al.  and Régnier et al.  were from the industry, and Muston et al.  were funded by the industry. In addition, funding and conflict of interest information was well reported in all five NMAs. Full details of the assessment and supporting information are provided in Additional file 3.
Key results of the included NMAs
Best-corrected visual acuity (BCVA) is an important outcome to assess the effect of interventions in people with DME. It is usually measured using Early Treatment Diabetic Retinopathy Study (ETDRS) letters. Different base trials may report this outcome in different ways, and then, in turn, different NMAs can choose different trials and varied ways of reporting. Table 3 shows, how, for the identical binary outcome (the proportion of participants who gain ≥ 10 ETDRS letters at 12 months), different NMAs collected data from different trials, and, partly due to this, arrived at slightly different point estimates—although they all agreed that there was no clear difference between the two treatments (all 95% credible intervals [CrI] straddled ‘one’). Mostly, different decisions for the trial choices according to PICO information are all considered to be clinically meaningful in different NMAs. These differences then lead to results that are not identical. For example, one reviewer may feel that a systematic difference in participants in a particular trial may make it inappropriate to network with data from other trials (e.g. Ishibashi et al. , only included in the sensitivity analysis in Régnier et al. , included in the main analysis of Korobelnik et al. ). Variations in the dosage of treatments may, in the view of one review team, make a study ineligible but be acceptable to other researchers (e.g. Massin et al. , included in Régnier et al. , excluded from Korobelnik et al. /Muston et al. ). Time point of outcome assessment can lead to further differences (e.g. Nguyen et al. , included in Régnier et al. , excluded from Korobelnik et al. /Muston et al. ). These decisions, all made with the best of intentions, lead to the inclusion of different trials contributing to the final—slightly different—results as illustrated in Table 3.
In addition, the BCVA measure can be reported as a continuous score of average change (see Table 3—average change in BCVA at 12 months, 95% CrI straddled ‘zero’ indicates no clear differences). We reproduced the results from each NMA and illustrated how the same measure is reported in different ways and different combinations across the five NMAs and found that a pattern does arise. Continuous outcomes all favour aflibercept over ranibizumab. Only Zhang et al.  did not reach conventional levels of statistical significance for this outcome (its 95% CrI included a negative value), but all the mean scores do favour the aflibercept group. The binary scores, however, tell a different story—and in no case, when a binary cut-off of ≥ 10 was used—was a statistically significant difference seen (these 95% CrIs straddled ‘one’). In the two later NMAs—including the one considered to be of the highest quality by AMSTAR-2—a binary cut-off of ≥ 15 was employed. For this particular outcome, both NMAs suggest that there exists a statistically significant difference in favour of people allocated to aflibercept (the 95% CrIs were > 1).
In general, it is clear that the reader must continue to think ‘cleanly’ amidst the data which may not be so clean. Below, we discuss some points of particular concern illustrated by these five similar NMAs.
Network meta-analysis is an exciting but still evolving tool employing data from [in these cases] randomised trials in ways by which comparisons of interest can be constructed by using somewhat assumption-heavy observational methods. For example, aflibercept versus ranibizumab is the comparison of interest (referred to as the decision set). Aflibercept may only have been compared with sham injection in randomised trials. Ranibizumab, likewise, may only have been compared with sham injection. We use the term ‘supplementary set’ to refer to interventions, such as sham injection, that are included in the network meta-analysis for the purpose of improving inference among interventions in the decision set . As different selections for supplementary set, different network structures will be conducted for the same clinical problem. For example, aflibercept may have been compared with sham injection in some trials and laser in others. Ranibizumab also may have some sham-controlled trials and some where it has competed against laser. One review may choose the sham-controlled trials as the supplementary set, and another review may choose the laser-controlled trials. In selecting which competing interventions to include in the decision set, researchers should ensure that the transitivity assumption is likely to hold [4, 27]. The choice of supplementary set is mostly based on clinical considerations . Including more interventions in the network may provide more information, which leads to more precise results. Meanwhile, it brings the risk of employment of non-valid indirect comparisons and of varied and ill-understood assumptions. When theoretical assumptions are guaranteed (transitivity and consistency), there is no absolute right or wrong in the construction of a network structure. The reader of the review should carefully consider if she/he feels the network indirect comparisons are sensible and making best use of available data.
From within even a few trials, the multiple choices of what data to select begin to become obvious. Of course, these choices should be based on sound clinical rationale and taken when as blind as possible to the base trials and their results—hence the value of a peer-reviewed protocol for the conduct of the trial. It should not be a surprise that clinicians and researchers evolve their ideas and differ even at the same time. Scientific questions merit replication and, powerful and reassuring in this illustration, is that although the point estimates differ across the trials, the findings are, for the outcomes in Table 3, essentially the same—no matter what way the data are used. Readers need to consider and understand what participants are included—and excluded—in the trial; what treatments are its focus and if there are omissions; and what outcomes are being reported and why those choices were taken. On the other hand, some studies [28, 29] have shown that while some independent replication of meta-analyses (or overviews of reviews ) by different teams may be useful, there is considerable overlap and potential redundancy in published NMAs. Replication can add value or be wasteful, depending on the topic and quality of research.
Different data from the same measures of effect
Continuous data from a scale or measure provide more detailed information than dichotomous data from the same scale or measure. The dichotomous or binary is often a crude and even arbitrary cut-off within an ostensibly continuous measure. Continuous measures are, however, often a research fabrication and not truly continuous. In the real world, a 10-point decline in a score from 90 to 80 may have quite a different meaning than a 10-point decline on the same measure from 20 to 10. In addition, clinicians and patients first tend to seek if the treatment will help them, for example, ‘get better’ (a question that merits a binary answer) and then, only as second preference, seek information on the degree of improvement (meriting an answer on a continuous scale). In the examples used in this paper, there is reasonable consistency that aflibercept does help shift the mean scores in a positive direction (continuous) when compared with ranibizumab. Crude though it may be, however, the binary is easier to interpret clinically and, if great enough is likely to mean ‘getting better’ or ‘making a substantial difference’ despite shortcomings of the so-called continuous measure. In the examples in Table 3, the average change improvements seem relatively consistently to be a matter of around 4 points. It is problematic to really understand what this may mean for any one patient’s life. In averaging across the groups, something may be lost, however, that is revealed in the binary, and Table 3 gives good evidence for speculation. What trials that report a ≥ 10-point gain consistently are reviewed to show no clear difference between aflibercept and ranibizumab, but the two latest trials have a new binary to report (≥ 15-point gain) and both show an advantage for those allocated to aflibercept. Perhaps in the averaging across all people in the trials, there has been a masking of an important group of people who respond better to aflibercept. But these are clinical and research points of debate. Overall, the five NMAs have reported results that are complicated, thought-provoking, but not truly inconsistent with each other. The reader needs to consider the value of the outcome for their need. The researchers may favour the continuous measure of function, the clinician or patient, the binary cut-off for better/not better, and the policy maker the economics.
Differences in what is truly significant
As for individual studies, results for meta-analyses are reported with a point estimate with an associated confidence interval. The confidence interval describes the uncertainty inherent in any estimate and describes a range of values within which we can be reasonably confident that the true effect lies . If the confidence interval is relatively narrow (e.g. 2.5 to 5.5 in Virgili et al. , Table 3), the effect size on the mean change in BCVA is known reasonably precisely. If the interval is wider (e.g. 1.5 to 7 in Régnier et al. , Table 3), the uncertainty is greater, although there may still be enough precision to make decisions about the utility of the finding. Intervals that are very wide indicate that we have little knowledge about where the true effect actually is and our certainty is eroded.
When the synthesis of continuous data produces confidence intervals of the data’s point estimates that are both on the positive side (e.g. Muston et al.’s  1.90 to 8.52, Table 3), or both on the negative, then the finding has reached a pre-stated level of statistical significance. When the synthesis of binary data produces confidence intervals of the data’s point estimates that are both greater than one (e.g. Muston et al.’s  1.12–4.20, Table 3), or both less than one, then the finding has reached a pre-stated level of statistical significance. However, as has been suggested above, the statistically significant findings of an outcome measure may not have a great clinical impact. An outcome may be statistically significant, yet clinically insignificant. It is easy for confusion to arise when the same data are commented upon by one set of reviewers discussing findings from the statistical perspective and another set of reviewers considering the clinical meaning of the data. Careful consideration is required from the reader to understand the assessment of the reviewers—are they reporting the clinical or statistical perspective—or a mixture of both.
A further danger of differing interpretations of the same findings lies in when confidence intervals straddle 0 (for continuous data) or 1 (for binary data—as for all the ≥ 10-point gain findings in Table 3). The findings are certainly not statistically significant, but it is easy for reviewers and readers of the trials to confuse ‘no evidence of an effect’ with ‘evidence of no effect’. When confidence intervals are wide, for example, the 0.63 to 4.06 of Muston et al.  in Table 3, they straddle 1 or unity. In this case, it is wrong to claim that aflibercept has ‘no effect’ or is ‘no different’ from ranibizumab—both statements carry too much certainty. As the confidence interval for the estimate of the difference in the effects of the treatments overlaps with no effect (in this binary example, 1), the analysis is compatible with both a true beneficial effect and a true harmful effect. It is true that there is no clear difference, but one drug is not clearly different from the other. If a true beneficial effect is mentioned in the conclusion, a true harmful effect should also be mentioned and discussed. It is so easy for reviewers and readers of those trials to take one side of a message from the same data and leave another half of the message less considered. Again, the reader has to be vigilant that balance has been achieved by the reviewers and if it has not, then be balanced themselves.
As always, really thinking about the meaning of findings is key. Together, the point estimate and confidence interval provide information to assess the effects of the intervention on the outcome. For example, in the evaluation of these drugs on BCVA, it could have been decided that it would be clinically useful if the medication increased BCVA from baseline by 5 letters—and at the very least 2 letters. Virgili et al.  report an effect estimate of an increase from a baseline of four letters with a 95% confidence interval from 2.5 to 5.5 letters. If this finding is based on good methods (see above), this allows the conclusion that aflibercept was useful since both the point estimate and the entire range of the interval exceed the criterion of an increase of two letters. The Régnier et al.  trial reported a similar point estimate (4.5 letters) but with a wider interval from 1.5 to 7 letters. In this case, although it could still be concluded that the best estimate of the aflibercept effect is that it provides net benefit, the reader could not be so confident as the possibility still has to be entertained that the effect could be between 1.5 and 2 letters—a low range that had been pre-specified to be of little clinical value. So, in this example, the latter, higher-quality trial had the confidence intervals that were reassuring of a net benefit for one compound. But the contrast of Régnier et al.  and Virgili et al.  serves well to illustrate how very similar findings may justify subtly different implications. The reviewers carry a responsibility to help the reader through clear reporting and thoughtful inclusive explanations—but where this has not happened, the readers may have to do this for themselves.
We have summarised the methods and findings of five NMAs of the same topic which produced what seemed like somewhat different findings from similar data sets. Closer inspection of each of these trials shows how the methods, including the searches and analyses, all differ, but the findings, although presented differently and sometimes interpreted differently, were similar. That five different NMAs—all using different datasets, networks, and review teams broadly producing similar findings—must be reassuring that there is some consistency in the results of the aflibercept/ranibizumab comparison.
As always, the critical reader of a review should think about the review in detail. This is helped by long-established checklists . Furthermore, Grading of Recommendations Assessment, Development, and Evaluation (GRADE) offers a transparent and structured process for developing and presenting summaries of evidence, including its quality, for systematic reviews and recommendations in health care .
As is common in different trials and reviews, outcomes—even the same measures—can be legitimately reported in several different ways. There is no avoiding the need to think through what the numbers really mean in terms of people, services, and policies. This may necessitate careful, subtle, humane, and expert consideration.
Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.
Best-corrected visual acuity
Early Treatment Diabetic Retinopathy Study
Diabetic macular oedema
Grading of Recommendations Assessment, Development, and Evaluation
Participants, interventions, comparators, outcomes
Pro re nata
Quality of life
Treat and extend
Vascular endothelial growth factor
Bastian H, Glasziou P, Chalmers I. Seventy-five trials and eleven systematic reviews a day: how will we ever keep up? Plos Medicine. 2010;7(9):e1000326. https://doi.org/10.1371/journal.pmed.1000326.
Mulrow CD. Rationale for systematic reviews. BMJ. 1994;309(6954):597–9. https://doi.org/10.1136/bmj.309.6954.597.
Shea BJ, Reeves BC, Wells G, Thuku M, Hamel C, Moran J, Moher D, Tugwell P, Welch V, Kristjansson E. AMSTAR 2: a critical appraisal tool for systematic reviews that include randomised or non-randomised studies of healthcare interventions, or both. Bmj. 2017;358:j4008. https://doi.org/10.1136/bmj.j4008.
Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, Welch VA, editors. Cochrane Handbook for Systematic Reviews of Interventions version 6.4 (updated August 2023). Cochrane; 2023. Available from www.training.cochrane.org/handbook.
Ciulla TA, Amador AG, Zinman B. Diabetic retinopathy and diabetic macular edema: pathophysiology, screening, and novel therapies. Diabetes Care. 2003;26(9):2653–64. https://doi.org/10.2337/diacare.26.9.2653.
Fenwick EK, Xie J, Ratcliffe J, Pesudovs K, Finger RP, Wong TY, Lamoureux EL. The impact of diabetic retinopathy and diabetic macular edema on health-related quality of life in type 1 and type 2 diabetes. Invest Ophthalmol Vis Sci. 2012;53(2):677–84. https://doi.org/10.1167/iovs.11-8992.
Hariprasad SM, Mieler WF, Grassi M, Green JL, Jager RD, Miller L. Vision-related quality of life in patients with diabetic macular oedema. Br J Ophthalmol. 2008;92(1):89–92. https://doi.org/10.1136/bjo.2007.122416.
Chen E, Looman M, Laouri M, Gallagher M, Van Nuys K, Lakdawalla D, Fortuny J. Burden of illness of diabetic macular edema: literature review. Curr Med Res Opin. 2010;26(7):1587. https://doi.org/10.1185/03007995.2010.482503.
Scanlon PH, Martin ML, Bailey C, Johnson E, Hykin P, Keightley S. Reported symptoms and quality-of-life impacts in patients having laser treatment for sight-threatening diabetic retinopathy. Diabet Med. 2006;23(1):60–6. https://doi.org/10.1111/j.1464-5491.2005.01736.xIF:4.213Q2.
Jain A, Varshney N, Smith C. The evolving treatment options for diabetic macular edema. Int J Inflam. 2013;2013:689276. https://doi.org/10.1155/2013/689276.
Schmidt-Erfurth U, Garcia-Arumi J, Bandello F, Berg K, Chakravarthy U, Gerendas BS, Jonas J, Larsen M, Tadayoni R, Loewenstein A. Guidelines for the management of diabetic macular edema by the European Society of Retina Specialists (EURETINA). Ophthalmologica. 2017;237(4):185–222. https://doi.org/10.1159/000458539IF:3.757Q2.
Cheung GC, Yoon YH, Chen LJ, Chen SJ, George TM, Lai TY, H. PK, Tahija SG, Uy HS, Wong TY. Diabetic macular oedema: evidence-based treatment recommendations for Asian countries. Clin Exp Ophthalmol. 2018;46(1):75–86. https://doi.org/10.1111/ceo.12999.
Korobelnik JF, Kleijnen J, Lang SH, Birnie R, Leadley RM, Misso K, Worthy G, Muston D, Do DV. Systematic review and mixed treatment comparison of intravitreal aflibercept with other therapies for diabetic macular edema (DME). BMC Ophthalmology,15,1(2015–05–15). 2015;15(1):52. https://doi.org/10.1186/s12886-015-0035-x.
Régnier S, Malcolm W, Allen F, Wright J, Bezlyak V. Efficacy of anti-VEGF and laser photocoagulation in the treatment of visual impairment due to diabetic macular edema: a systematic review and network meta-analysis. Plos One. 2014;9(7):e102309. https://doi.org/10.1371/journal.pone.0102309.
Zhang L, Wang W, Gao Y, Lan J, Xie L. The efficacy and safety of current treatments in diabetic macular edema: a systematic review and network meta-analysis. PLoS One. 2016;11(7):e0159553. https://doi.org/10.1371/journal.pone.0159553.
Muston D, Korobelnik JF, Reason T, Hawkins N, Chatzitheofilou I, Ryan F, Kaiser PK. An efficacy comparison of anti-vascular growth factor agents and laser photocoagulation in diabetic macular edema: a network meta-analysis incorporating individual patient-level data. Bmc Ophthalmology. 2018;18(1). https://doi.org/10.1186/s12886-018-1006-9.
Virgili G, Parravano M, Evans JR, Gordon I, Lucenteforte E. Anti-vascular endothelial growth factor for diabetic macular oedema: a network meta-analysis. Cochrane Database of Systematic Reviews. 2017;6(6):CD007419. https://doi.org/10.1002/14651858.CD007419.pub5.
Ishibashi T, Li X, Koh A, Lai TY, Lee FL, Lee WK, Ma Z, Ohji M, Tan N, Cha SB, Shamsazar J, Yau CL. The REVEAL Study: ranibizumab monotherapy or combined with laser versus laser monotherapy in Asian patients with diabetic macular edema. Ophthalmology. 2015;122(7):1402–15. https://doi.org/10.1016/j.ophtha.2015.02.006. (Epub 2015 May 14).
Massin P, Bandello F, Garweg JG, Hansen LL, Harding SP, Larsen M, Mitchell P, Sharp D, Wolf-Schnurrbusch UE, Gekkieva M, Weichselberger A, Wolf S. Safety and efficacy of ranibizumab in diabetic macular edema (RESOLVE Study): a 12-month, randomized, controlled, double-masked, multicenter phase II study. Diabetes Care. 2010;33:2399–405. https://doi.org/10.2337/dc10-0493.
Nguyen QD, Shah SM, Heier JS, Do DV, Lim J, Boyer D, Abraham P, Campochiaro PA. Primary end point (six months) results of the Ranibizumab for Edema of the mAcula in diabetes (READ-2) study. Ophthalmology. 2009;116(11):2175-81.e1. https://doi.org/10.1016/j.ophtha.2009.04.023.
Elman MJ, Aiello LP, Beck RW, Bressler NM, Sun JK. Randomized trial evaluating ranibizumab plus prompt or deferred laser or triamcinolone plus prompt laser for diabetic macular edema. Ophthalmology. 2010;117(6):1064-77.e35. https://doi.org/10.1016/j.ophtha.2010.02.031.
Mitchell P, Bandello F, Schmidterfurth U, Lang GE, Massin P, Schlingemann RO, Sutter F, Simader C, Burian G, Gerstner O. The RESTORE study: ranibizumab monotherapy or combined with laser versus laser monotherapy for diabetic macular edema. Ophthalmology. 2011;118(4):615–25. https://doi.org/10.1016/j.ophtha.2011.01.031.
Korobelnik JF, Do DV, Schmidt-Erfurth U, Boyer DS, Holz FG, Heier JS, Midena E, Kaiser PK, Terasaki H, Marcus DM, Nguyen QD, Jaffe GJ, Slakter JS, Simader C, Soo Y, Schmelter T, Yancopoulos GD, Stahl N, Vitti R, Berliner AJ, Zeitz O, Metzig C, Brown DM. Intravitreal aflibercept for diabetic macular edema. Ophthalmology. 2014;121(11):2247–54. https://doi.org/10.1016/j.ophtha.2014.05.006.
Googe J, Brucker AJ, Bressler NM, Qin H, Aiello LP, Antoszyk A, Beck RW, Bressler SB, Ferris FL, Glassman AR, Marcus D, Stockdale CR. Randomized trial evaluating short-term effects of intravitreal ranibizumab or triamcinolone acetonide on macular edema after focal/grid laser for diabetic macular edema in eyes also receiving panretinal photocoagulation. Retina (Philadelphia, Pa). 2011;31(6):1009–27. https://doi.org/10.1097/IAE.0b013e318217d739.
Do DV, Quan DN, Boyer D, Schmidt-Erfurth U, Heier JS. One-year outcomes of the DA VINCI study of VEGF trap-eye in eyes with diabetic macular edema. Ophthalmology. 2012;119(8):1658–65. https://doi.org/10.1016/j.ophtha.2012.02.010.
Clinicaltrial.gov. Safety, efficacy and cost-efficacy of ranibizumab (monotherapy or combination with laser) in the treatment of diabetic macular edema (DME) (RESPOND). NCT01135914 2018 [Available from: https://www.clinicaltrials.gov/ct2/show/NCT01135914?term=RESPOND&cond=DME&rank=1.
Critical appraisal checklist for a systematic review. 2021. Available from: https://www.gla.ac.uk/media/Media_64047_smxx.PDF. Accessed 20 March 2021.
Naudet F, Schuit E, Ioannidis JPA. Overlapping network meta-analyses on the same topic: survey of published studies. Int J Epidemiol. 2017;46(6):1999–2008. https://doi.org/10.1093/ije/dyx138.
Siontis KC, Hernandez-Boussard T, Ioannidis JP. Overlapping meta-analyses on the same topic: survey of published studies. Bmj. 2013;347:f4501. https://doi.org/10.1136/bmj.f4501.
Belbasis L, Bellou V, Ioannidis JPA. Conducting umbrella reviews. BMJ medicine. 2022;1(1):e000071. https://doi.org/10.1136/bmjmed-2021-000071.
Guyatt G, Oxman AD, Akl EA, Kunz R, Vist G, Brozek J, Norris S, Falck-Ytter Y, Glasziou P, DeBeer H, Jaeschke R, Rind D, Meerpohl J, Dahm P, Schünemann HJ. GRADE guidelines: 1. Introduction-GRADE evidence profiles and summary of findings tables. Journal of Clinical Epidemiology. 2011;64(4):383–94. https://doi.org/10.1016/j.jclinepi.2010.04.026.
We thank everyone who kindly provided assistance during our preparation of this manuscript.
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Ethics approval and consent to participate
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Wu, J., Adams, C., He, X. et al. Diversity when interpreting evidence in network meta-analyses (NMAs) on similar topics: an example case of NMAs on diabetic macular oedema. Syst Rev 12, 189 (2023). https://doi.org/10.1186/s13643-023-02349-4