OBJECTIVES: Hearing implants (HIs) are used in patients with hearing loss that cannot benefit from hearing aids or reconstructive surgery. Safety plays a critical role in the assessment of these class-III medical devices but because the concept of RCTs is hardly applicable, adverse event (AE) data is typically generated from prospective or retrospective cohort studies. Summarizing AE data for HIs is therefore often complicated by low data quality. This study aims to assess quantitative methods that summarize AEs from low-level evidence, with respect to different stakeholder perspectives.

METHODS: AE data for two different HIs were collected via systematic literature review. Incidence rates (IR) were calculated as average 6-monthly rate using person-time as denominator, or as time-specific rate if the number of events and the number at risk were given at 6-monthly time intervals. IRs were summarized across publications by pooling raw data or via meta-analysis. Survival functions were calculated based on a subset of available data. A visualization for semi-quantitative assessment was developed.

RESULTS: Overall, simpler analyses allowed for inclusion of more real-world-evidence, while in-depth analyses required parameters less frequently reported in the literature. Pooling raw data for an average yearly IR was easiest but ignored the non-linearity of AE-occurrence over time. Time-specific IR required either patient-level data or relaxed assumptions about time-to-event and number at risk. Meta-analysis assumed linearity of events and struggled to deal with zero values. Survival analysis offered the most sophisticated analysis, but few datapoints were available for input.

CONCLUSIONS: Estimating IRs of AEs from real-world-evidence can be accomplished via various methodological routes. Each way has benefits, limitations and makes assumptions that stakeholders will need to weight cautiously. Eventually, enforcing standards for AE-reporting in research on HIs will have the biggest effect on more robust safety summaries. For some stakeholders, well-designed visualizations may be more informative than accurate rate estimates.