SYSTEMATIC LITERATURE REVIEW ON THE PERFORMANCE OF DYNAMIC TRANSMISSION MODELS IN PNEUMOCOCCAL DISEASE
Author(s)
Sajad Emamipour, PhD1, Meine Zijlstra, MSc1, Benjamin Althouse, PhD2, Cornelis Boersma, PhD1, Mark Rozenbaum, MBA, PhD3, Maarten Postma, PhD4.
1Health-Ecore, Zeist, Netherlands, 2Pfizer, New York, NY, USA, 3Pfizer, Capelle aan den IJssel, Netherlands, 4University of Groningen, Groningen, Netherlands.
1Health-Ecore, Zeist, Netherlands, 2Pfizer, New York, NY, USA, 3Pfizer, Capelle aan den IJssel, Netherlands, 4University of Groningen, Groningen, Netherlands.
OBJECTIVES: Dynamic transmission models (DTMs) are recommended by good-practice guidance when a vaccination program affects pathogen transmission or alters the distribution of circulating strains, because they structurally capture indirect (herd) effects and strain replacement that static models cannot. However, DTMs of pneumococcal conjugate vaccine (PCV) programs are complex, and their predictive accuracy beyond the calibration period has not been empirically evaluated. We conducted a systematic literature review (SLR) to assess how accurately published pneumococcal DTMs predict invasive pneumococcal disease (IPD) incidence compared with observed surveillance data.
METHODS: We systematically searched PubMed, Embase, and Cochrane were systematically searched through October 2025 following PRISMA guidelines for published DTMs of Streptococcus pneumoniae that evaluated any PCV and reported IPD incidence beyond the calibration period. Model projections were compared with national surveillance data. Only studies in which the vaccine modelled matched the vaccine implemented in the target population were eligible. The COVID-19 pandemic period (2020-2023) was excluded, unless pandemic-related assumptions were explicitly incorporated. Predictive accuracy was quantified as the mean absolute percentage error (MAPE) between projected and observed incidence over the full projection period.
RESULTS: Of 2,689 titles screened, eight met the inclusion criteria, spanning four countries (preliminary results). These studies evaluated the impact of PCV7, PCV13, PCV15 and PCV20 with projection horizons of 1 to 11 years; all used compartmental deterministic structures. Predictive accuracy varied substantially, with MAPE ranging from 9% to 131.9% (median 20.8%). Accuracy was poorest for a study with COVID-19 pandemic assumptions.
CONCLUSIONS: This is the first study to assess the predictive accuracy of pneumococcal DTMs, yet few studies allowed such evaluation. Accuracy varied widely, likely reflecting heterogeneity in projection horizon, vaccine generation, and epidemiological setting. The recommended preference for dynamic over static PCV modeling therefore rests on theory rather than demonstrated predictive performance, and current recommendations may warrant re-examination.
METHODS: We systematically searched PubMed, Embase, and Cochrane were systematically searched through October 2025 following PRISMA guidelines for published DTMs of Streptococcus pneumoniae that evaluated any PCV and reported IPD incidence beyond the calibration period. Model projections were compared with national surveillance data. Only studies in which the vaccine modelled matched the vaccine implemented in the target population were eligible. The COVID-19 pandemic period (2020-2023) was excluded, unless pandemic-related assumptions were explicitly incorporated. Predictive accuracy was quantified as the mean absolute percentage error (MAPE) between projected and observed incidence over the full projection period.
RESULTS: Of 2,689 titles screened, eight met the inclusion criteria, spanning four countries (preliminary results). These studies evaluated the impact of PCV7, PCV13, PCV15 and PCV20 with projection horizons of 1 to 11 years; all used compartmental deterministic structures. Predictive accuracy varied substantially, with MAPE ranging from 9% to 131.9% (median 20.8%). Accuracy was poorest for a study with COVID-19 pandemic assumptions.
CONCLUSIONS: This is the first study to assess the predictive accuracy of pneumococcal DTMs, yet few studies allowed such evaluation. Accuracy varied widely, likely reflecting heterogeneity in projection horizon, vaccine generation, and epidemiological setting. The recommended preference for dynamic over static PCV modeling therefore rests on theory rather than demonstrated predictive performance, and current recommendations may warrant re-examination.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
EPH207
Topic
Epidemiology & Public Health
Disease
Vaccines