Improve harms reporting in trials of microelectronic retinal implants

We read with interest Alqahtani et al. ¹ which presents a systematic review of the safety and efficacy of microelectronic retinal implants to treat degenerative diseases such as retinitis pigmentosa. Those authors reviewed studies of human implantees that reported safety-related outcomes. They considered all implant types (including epiretinal, subretinal, and suprachoroidal) and all studies with follow-up ⩾ 1 year. We commend the authors’ focus on this important topic. While much attention has been paid to implant efficacy,^2,3 relatively little has been paid to safety. ⁴ Understanding retinal implant safety is key because it enables informed decision-making by patients and clinicians. ⁴

We object to the conclusions drawn by Alqahtani et al., specifically, that “all types [of retinal implant] showed a generally acceptable safety profile” and that “adverse events [were] mostly treatable.” ¹ Those conclusions rely on the authors’ flawed analysis of the safety studies they reviewed, specifically, that, in those studies, reporting bias “was minimised. . . because the authors reported the serious and non-serious adverse events, the technical failures of implants, and any unsuccessful implantations.” ¹ This is wrong. Previously, we systematically reviewed the same literature ⁴ ; unlike Alqahtani et al., we analyzed and quantified bias by using the McMaster Tool for Assessing Quality of Harms Assessment and Reporting. ⁵ Our use of this tool revealed some high-quality harm assessment and reporting (e.g., Kitiratschky et al. ⁶ ). However, in this literature, harm assessment and reporting quality were typically poor. ⁴ Indeed, our meta-analysis of adverse events associated with implantation was not straightforward; many articles we reviewed lacked important details, few provided a clear definition of “serious adverse event,” and many did not specify whether the adverse events reported were all adverse events ascertained, or a subset thereof. Adverse event reporting that is incomplete and lacks specificity is not unique to the retinal implant literature. ⁷

Alqahtani et al. also report that ‘adverse events [were] mostly treatable and declining over time’ [emphasis added]. ¹ It is unclear what analysis techniques those authors used to draw this conclusion. We note that to reason about the timing of adverse events associated with retinal implantation after having excluded short-running (<1 year follow-up) studies from consideration is a design flaw. In our systematic review, ⁴ which included studies of any duration, we extracted the precise timing of serious adverse events where possible. Because of poor quality reporting, this undertaking was not without its frustrations. We used these timing data to populate an empirical cumulative density function, and then showed, using Monte Carlo simulation, that serious adverse events associated with implantation appear log-uniformly distributed throughout follow-up. ⁴

Systematic review with meta-analysis is the gold standard for evaluating the safety and efficacy of medical devices. ⁸ We enjoin clinical researchers to improve their conduct and reporting of clinical trials of microelectronic retinal implants ⁴ : articles should report a precise definition of ‘serious adverse event’ conforming to international standards, ⁹ specify whether all adverse events are reported, and report the precise timing of serious adverse events. The adoption of these recommendations would facilitate future meta-analyses which, in turn, would enable the development of a reliable safety profile of the various types of retinal implant. This work lacks no urgency – how else can patients and clinicians be enabled to make informed decisions?