Beyond significance: a 10-year analysis of negative findings in ECNP Congress abstracts (2016–2025)

BACKGROUND Publication bias and the underreporting of negative findings remain major challenges in biomedical research. Initiatives such as registered reports, the EQUATOR Network and the TOP guidelines aim to improve transparency and the visibility of null results. Within neuropsychopharmacology, the ECNP Best Negative Data Prizes -introduced in 2018 (preclinical) and 2021 (clinical)- represent a field-specific effort to incentivise the reporting of negative findings. Whether these efforts are reflected in poster abstracts presented at the ECNP Congress remains unclear.

AIMS To examine temporal trends in the reporting of positive, negative and mixed findings in ECNP poster abstracts (2016–2025), and to compare periods before and after the 2018 introduction of the ECNP Best Negative Data Prize.

METHODS Poster abstracts from ten consecutive ECNP Congresses (2016–2025) were retrieved from Neuroscience Applied conference supplements (ScienceDirect; n=9,606). A-priori exclusions (narrative reviews, protocols, case reports/series, descriptive studies without a comparative hypothesis, and insufficient methodology) yielded 7,439 included abstracts. Each abstract was assigned to one of four categories based on its reported results: positive (statistically significant in the hypothesised direction), negative (non-significant or not supporting the hypothesis), mixed (coexistence of both), or non-classifiable; outcome proportions were computed within the classifiable subset. Screening and classification were assisted by a large language model under pre-specified criteria with manual validation of a random subset. Temporal trends were analysed using logistic regression with year as a continuous variable. Pre/post-2018 comparisons used the introduction of the preclinical prize as the marker year. Sensitivity analyses recoded non-classifiable cases as positive (most conservative assumption against the observed trend).

RESULTS Of 7,439 included abstracts, 7,179 (96.5%) were classifiable. Among these, the proportion of positive findings declined from 85.8% in 2016 to 67.8% in 2025 (per-year OR=0.89, 95%CI 0.87-0.90, p<0.001). Negative findings rose from 2.6% in 2016 to a peak of 17.5% in 2023, before moderating to 7.9% in 2025 (per-year OR=1.15, 95%CI 1.12-1.18, p<0.001). Mixed findings increased from 11.6% to 24.3% (per-year OR=1.08, 95%CI 1.06-1.10, p<0.001). Pre/post-2018 comparison showed a near-tripling of negative findings (3.7% vs 12%; OR=3.5, p<0.001) and a doubling of the combined non-positive category (negative plus mixed: 16.0% vs 32.9%; OR=2.6, p<0.001). Sensitivity analyses confirmed these patterns.

CONCLUSIONS Over a decade, ECNP poster abstracts show a substantial increase in the reporting of negative and mixed findings. This trend aligns with growing attention to null results across biomedical research, to which initiatives such as the ECNP Best Negative Data Prizes may have contributed. However, the proportion of negative findings remains below estimates from prospectively registered studies, suggesting persistent underrepresentation. The attenuation observed in 2025 highlights the need for continued monitoring to determine whether this represents a transient fluctuation or a genuine reversal of progress. Limitations include reliance on AI-assisted classification, restriction to accepted abstracts, an observational design that precludes formal causal attribution, and the absence of a comparator congress.

References

No references