Does intravitreal bevacizumab adversely affect long term development? Two simultaneous systematic reviews say yes, or no.

A reliable answer to the above question would require a large multicentre RCT comparing intravitreal bevacizumab (IVB) to laser, powered for long term outcomes. Such a trial does not currently exist.

As a result, 2 groups have just published systematic reviews of the observational studies that have compared outcomes of non-randomized cohorts with bevacizumab compared to either no treatment or to laser therapy.

The 2 reviews come to opposite conclusions!

Kaushal M, et al. Neurodevelopmental outcomes following bevacizumab treatment for retinopathy of prematurity: a systematic review and meta-analysis. J Perinatol. 2020, state “Bevacizumab treatment for severe ROP is associated with increased risk of cognitive impairment and lower cognitive and language scores in preterm infants”.

Tsai CY, et al. Neurodevelopmental Outcomes After Bevacizumab Treatment for Retinopathy of Prematurity-A Meta-Analysis. Ophthalmology. 2020. on the other hand state “severe NeuroDevelopmental Impairment risk was not increased in ROP patients after IVB treatment. Bayley-III scores were similar in the IVB and control groups, except for a minor difference in motor performance”.

In addition to cohort comparisons there is a small amount of data (n=16) from one of the centres involved in the BEAT-RoP RCT. All of the other studies reviewed were comparisons of non-randomized cohorts.

Such studies are fraught with potential bias. In our centre, for example, we were at first only using IVB for babies with BPD who were extubated and fragile and who we really didn’t want to re-intubate for laser surgery. They were therefore higher risk than laser treated babies. We also were using IVB mostly for babies with posterior disease who are not necessarily comparable to babies with zone 2 retinopathy.

Why would 2 almost simultaneous systematic reviews produce diametrically opposite results?

The first thing I did was to look at the tables with the included studies. Kaushal et al includes 13 studies, whereas Tsai has 8, three of which are not in Kaushal.

The discrepancies seem to be because Tsai included 2 studies that compared outcomes of babies who received IVB to babies with no treatment, and in one of those cases to a second control group of babies without retinopathy. You would think that such studies would show a difference in outcomes between IVB and control but in fact they showed very little. Those studies were not eligible for the review of Kaushal et al.

Kaushal includes 2 studies only reported as abstracts, which were not in Tsai’s publication list, and included 2 studies published in 2020 which may have appeared after Tsai finished their literature review. In addition 2 of the studies in Kaushal’s review only supplied mortality data, and one other does not appear to have supplied any data used in their analyses. The main difference in data sources, therefore, seems to be that Kaushal included Zayek et al and Arima et al from 2020, whereas Tsai included the above-mentioned studies with untreated controls.

As for the results, the definitions of “Severe Neurodevelopmental Impairment” are similar in the 2 reviews, and both reviews conclude that the 95% CI include an RR of 1.0, but Tsai’s analysis includes 5 studies and an RR of 1.52 (95% CI 0.91, 2.54) whereas Kaushal includes 3 studies (only 2 of which are in the Tsai analysis) and an RR of 1.33 (95% CI 0.74, 2.39).

As for the scores on the cognitive composite of the Bayley 3 evaluation, the Kaushal review, based on 6 studies, shows that cognitive scores are 1.8 points less with IVB than laser (the figure axis title wrongly states that this result “favours IVB”) 95% CI -3.5, -0.1; whereas Tsai et al also have 6 studies of IVB vs laser (only 3 of the studies are in both reviews) and a difference in cognitive scores of 1.69, 95% CI -4.9, +1.6. The 2 studies in Tsai’s review that compared laser to no treatment are calculated separately as showing little difference with wide confidence intervals (-2.6, 95% CI -8.2, +3).

In a similar way, but with a more marked difference, the scores on the Bayley 3 language composite are lower in the Kaushal review, 5.4 points less with IVB than laser (95% CI -9.2, -1.6), but in the Tsai review the difference in scores is only 1.36, (95% CI -5.5, +2.8).

What does this all mean? Basically, I don’t think you can rely on the results of these SRs to give an answer to the question. Systematic reviews of observational studies suffer from the same problems as the observational studies they are based on. Differences in characteristics of the babies treated with either therapy are likely, and, no matter how the data are adjusted, such biases remain.

Long-term visual outcomes are clearly better with IVB, with much lower rates of severe myopia. I think all that you can say about long term developmental and neurological outcomes is that there remains a concern that there could be adverse impacts of IVB, but the data collected so far are conflicting. I think we should give parents a choice when retinopathy treatment is required, informing them that for aggressive or posterior disease there are advantages of IVB, and also major unknowns for the long term. Of course the ophthalmologists treating the babies have to agree to that also!

Clearly, the large multicentre RCT, powered for long term outcomes, that I mentioned at the beginning of this post, is needed. These systematic reviews suggest that such a trial should be powered to find a 5 point difference in cognitive scores on the Bayley version 3, which would need close to 150 patients per group, or alternatively a 10% difference in the proportion of children with neurological impairment or developmental delay, in this high-risk group that would need somewhere in the region of 200 babies per group, depending on what the hypothesized baseline rate is. Those sample sizes seem achievable to me without too much difficulty, and I think this should be considered a priority for our community.

Type 1 RoP with plus disease (A) and after laser surgery (B). Hwang CK, et al. Outcomes after Intravitreal Bevacizumab versus Laser Photocoagulation for Retinopathy of Prematurity: A 5-Year Retrospective Analysis. Ophthalmology. 2015;122(5):1008-15.

About keithbarrington

I am a neonatologist and clinical researcher at Sainte Justine University Health Center in Montréal
This entry was posted in Neonatal Research and tagged , , . Bookmark the permalink.

4 Responses to Does intravitreal bevacizumab adversely affect long term development? Two simultaneous systematic reviews say yes, or no.

  1. anniejanvier says:

    Totally agree that in this case parents should decide (if neonatologists and ophthalmologists let them). For those who read “breathe baby breathe”, you saw reintubating Violette for her threshold ROP could have meant terrible resp-ENT consequences and prolonged hospitalization. ENT problems are not examined either in these investigations. For some babies, taking them off the respitator was so difficult that the thought of reintubation for several hours for laser is not ideal (for neonatologists and parents). When we have asked families, presented them with the biases and pros and cons (and probably more disclosure of cons with IVB), and the 2 choices, all chose IVB.

  2. Abdul Razak says:

    Thank you Keith for summarizing the reviews. Our systematic review provided a summary of literature that is out there, which I’m sure is useful, and remind people to cautiously use Bevacizumab rather using it routinely over laser particularly when the long-term safety isn’t being established. We have provided a transparent and non-judgemental conclusion which reads

    “ In conclusion, acknowledging the limitation of confounding by indication, which can only be ruled out by prospective clinical trials, the review supports the association between bevacizumab treatment for severe ROP and adverse neurodevelopmental outcomes in preterm infants. Unfortunately, intravitreal bevacizumab injection is so easy for the ophthalmologists that it has been accepted before long-term outcomes had been rigorously studied in large cohorts. Hence, researchers should evaluate bevacizumab’s safety in large clinical trials incorporating the long-term follow-up in the study design to address this issue heavily. Till then, bevacizumab treatment for severe ROP should be cautiously used, and infants should be vigilantly monitored with long-term neurodevelopmental follow-up”

    We encourage everyone to read the review. We hope Neonatologists make informed decision making with parents on this issue. I would definitely inform parents about the ‘may be’ risks but would also tell them the issues with current data.

    • Thanks for the comment and for the systematic review, which has pointed out very clearly where we are in this messy situation. Hundreds of babies have been treated with IVB without good information about what it might do to their long term outcomes. I agree wholeheartedly with your conclusion. Asking parents to make an informed decision about their baby’s therapeutic options is essential, and right now they have to make that decision based on short term impacts, and, what I think is fairly convincing, data about long term visual outcomes.
      As Annie notes, we currently give parents the choice about the approach, and they routinely choose IVB (as would I, probably), but that choice might be different for some if we were able to accurately inform them about long term neuro or developmental impacts.

  3. Martin Keszler says:

    Thanks for that thoughtful review Keith. I agree with the conclusion that we simply don’t know the answer at this point. The confounding by indication is impossible to resolve with retrospective observational studies.
    As a point of information, there is currently an RCT of another anti-VEGF drug that is comparing aflibercept (Eylea) to laser photocoagulation Identifier NCT04101721. There will be long-term follow-up to five years (good news) but the study will be underpowered for anything but a very large effect size (bad news). It features a 3:1 randomization to the study drug vs. laser and will only enroll 112 subjects. So we may not be much smarter after that is done. An adequately powered RCT with at least 2y, preferably 5y or school age follow-up is badly needed, but given the relatively low incidence of ROP needing treatment, I don’t see that happening…
    Stay safe,

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.