What health authorities criticise in periodic benefit-risk evaluation reports
Health authorities assess the periodic benefit-risk evaluation reports (PBRERs) submitted to them and return their assessment as comments. The comments fall into recurring kinds of deficiency, which occur at different rates, arise in how a report is prepared, and are the target of changes to that preparation.
Why the assessment comments matter
The quality and comprehensiveness of a PBRER directly affect regulatory decision-making and, ultimately, patient safety. When authorities review a report, their assessment can lead to significant safety measures: updates to contraindications, warnings and the adverse reaction profile, changes to risk mitigation strategies, requests for additional safety data, and refinements in post-marketing surveillance plans. Health authority feedback on a PBRER enables the marketing authorisation holder, supported by its pharmacovigilance partners, to continuously refine the benefit-risk assessment and align with evolving standards, strengthening report quality and patient safety. Understanding the patterns in these assessments is an untapped opportunity for systematic quality improvement.
What was studied
One analysis covered health authority assessment reports on PBRERs submitted worldwide over four years, from 1 January 2021 to 31 December 2024, for six marketing authorisation holders; the PBRERs had been written by the holders or by their pharmacovigilance partners and vendors. There were 698 assessment reports spanning 350 distinct products. The reports were classed as simple (315), complex (246) or medium (137) complexity. Most products were established ones with a long market presence and a well-characterised safety profile (265), followed by lifecycle products in active post-marketing phases with label extensions, new indications or formulation changes (81), and only four were new products approved within the previous two years. The five most represented therapeutic areas were oncology, infectious diseases, cardiovascular diseases, neuroscience and endocrinology.
The complexity classes were defined by the numbers of cases, literature references and safety topics. A simple report had fewer than 200 cases, fewer than 300 literature references and fewer than five safety topics; a medium one 200 to 500 cases, 300 to 600 references and five to ten topics; a complex one more than 500 cases, more than 600 references and more than eleven topics.
Every comment, whether a suggestion, an additional request or an identified issue, was sorted into one of five groups: safety assessment, missing or incomplete information, data inconsistencies or quality issues, structural or template non-compliance, and non-quality-related observations.
Kinds of deficiency and how often they occur
The 698 assessments carried 1635 comments, of which 879 were non-quality observations, mainly requests for complementary or additional information before the next report, leaving 756 that bore on quality. Of those, 310 (41%) concerned the safety assessment, 215 (28%) missing or incomplete information, 175 (23%) data inconsistencies and quality issues, and 56 (8%) structural or template non-compliance, which is 0.4, 0.3, 0.2 and 0.1 comments per assessment report respectively. Submissions were spread evenly over the four years, while comments were most numerous in the first year and least numerous in the second. The year-on-year variability in assessment reports and comments may reflect underlying regulatory and operational dynamics, including shifts in product lifecycle phases and changes to report frequency.
Safety assessment was the largest category of quality-related comments. Inappropriate or incomplete scientific evaluation (93 comments, 12%) covered assessments the authorities found insufficiently detailed or lacking adequate justification, such as comprehensive data and assessment not presented for fatal and life-threatening cases, or cases excluded from analysis without an adequate scientific rationale. Rejected proposals to stop monitoring a safety topic in subsequent reports (74, 10%) typically arose when the rationale for removing an important topic was judged scientifically inadequate. In one case with nine post-marketing reports of bowel obstruction, an authority found causality unconfirmed for want of clinical detail, so it did not endorse the holder's proposal for routine monitoring and required close monitoring and discussion of new evidence in future reports. In another, despite no new safety issue being identified, an authority mandated continued monitoring of osteoporosis because of a plausible mechanism by which the drug could contribute to it, angiogenesis being important for bone turnover. The category also counted insufficient documentation of cases (61, 8%) and inadequate search criteria (19, 3%). The underlying factors in these safety assessment deficiencies appear to be insufficient scientific justification and a lack of clarity in safety evaluations.
Label change recommendations (63, 8%) are counted in the same category, most often as an update to section 4.4 of the summary of product characteristics, on warnings and precautions. Most followed signal assessment processes independent of the report, with the signal assessment report then appended to it. Signal detection and assessment frequently run on dedicated pathways parallel to, but distinct from, periodic benefit-risk evaluation. A label change recommended after assessment is therefore to be read as an outcome of the assessment process rather than an indicator of report quality, unless it is explicitly linked to a gap or omission in the report.
Most missing or incomplete information comments concerned missing data elements within the report (169 comments, 22%), such as absent clinical exposure data, for instance exposure in managed access or compassionate use programmes, and the omission of significant new information on cases received in earlier periods. Incomplete responses (20, 3%) misinterpreted or insufficiently addressed questions from an authority's earlier assessment, as with tabular summaries of cases that omitted alternative explanations, or safety assessments that drew only on safety outputs and omitted clinical database outputs. Questions left unanswered (20, 3%) showed no specific trend and were attributed to systemic gaps, requests misclassified by safety teams, or procedural oversights, such as deadlines missed through technical problems with regulatory affairs communication systems, questions the holder's safety team deprioritised as non-critical, and the provision of outdated assessment reports that omitted relevant requests. Information listed in the local label but not specified in the reference safety information (3, 0.4%) and unspecified labelling discrepancies (3, 0.4%) completed the category. These issues were driven by gaps in presenting or adequately discussing relevant safety data, including data the authorities had requested being omitted or insufficiently addressed.
Inconsistencies in a PBRER usually stem from fragmented authoring or misaligned data management. Inconsistency across sections (45 comments, 6%) covered data or counts that differed between sections, appendices or the executive summary, most commonly between the main sections and the appendices; examples were safety concerns listed in section 16.3 but not reflected in section 16.4, and numbers of specific adverse reactions that differed between sections and appendices. Inconsistency in numbers (34, 4%) was attributed to errors in sales data and in the numbers of events or cases presented for safety issues or in appendices relative to previous reports. Most often the numbers differed from the previously submitted report, as with cumulative case counts lower than those reported before, or different numbers of cumulative events presented for identified and potential risks. The remaining subcategories were inconsistency with other submitted documents (33, 4%), calculation errors (25, 3%), comments on the overall quality of the report (19, 3%) and elements the authorities considered uninterpretable (19, 3%).
Most structural non-compliance comments called for closer alignment with GVP Module VII (43 comments, 6%) and showed no specific trend, as with updating section 3 on actions taken in the interval for safety reasons, or describing post-authorisation use in special populations in section 5.2.2, in line with the module. Formatting errors in tables and text (13, 2%) remained minor, but their presence suggested that authorities continue to attend to template presentation standards.
Recommended changes to report preparation
Recommendations drawn from the analysis address the most critical gaps among the recurring issues judged significant for their impact on patient safety and their volume, and focus on systemic improvements in processes, tools and competencies that marketing authorisation holders and their pharmacovigilance partners can implement.
For questions from earlier assessments, which concentrate in sections 15 and 16, the recommendations are a centralised repository of assessment reports within the holder's safety unit, which only one of the six holders studied had; clear ownership of each request, whether safety lead, coordinator or writer, and its deadlines, defined at the kick-off call; sharing the full assessment report rather than only the comments with the writer; and using AI to flag open requests and track them to completion. For rejected requests to stop monitoring, which suggest inadequate scientific justification for the proposed change, the recommendations are collaboration between the holder's safety and regulatory teams at the start of the reporting period to understand what the authority expects, and requests substantiated by comprehensive evaluations drawing on multiple data sources, including literature, clinical and epidemiological data.
For insufficient scientific evaluation, evident in sections 9.2, 15 and 16, the recommendations are a clear analysis strategy covering data sources, search terms and presentation, flexible enough to take in emerging information, documented and approved by the safety lead, with every update discussed and documented; a subject-matter-expert framework that maps report complexity to the skills and experience of the expert, which in one team's experience has helped facilitate rigorous evaluations; mandatory medical review for complex topics; and training that builds writers' and reviewers' scientific acumen, such as workshops, bootcamps and quarterly or biannual trend analysis of authority assessments and quality-check findings.
Data omissions or inconsistencies within PBRER sections can constitute a procedural deficiency. For missing data, in sections 5, 7 and 11 and the appendices, the recommendations are a first-draft self-check and a peer check focused on data completeness, and, where several writers share a report, a lead writer who makes sure the data are complete and consistent across sections.
Addressing the gaps in risk evaluations by complementing the assessment reports directly enhances patient safety by ensuring comprehensive surveillance of emerging safety topics. Improved scientific evaluations and complete responses to authority queries enable timely identification of potential risks, leading to appropriate risk minimisation measures. Structured processes for data management and quality control reduce the likelihood of overlooking critical safety information and safeguard the benefit-risk profiles of medicinal products. These improvements create a more robust pharmacovigilance system that can better detect, assess and mitigate safety issues.