{"id":"1af35399-60d1-485d-aca2-d3a545052f5c","arxiv_id":"2505.05534","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A dynamic network simulation shows that early vaccination and reductions in one-time partnerships, even only among the highest-risk quarter of men, cut mpox infections by roughly 30 percent.","lead":"This paper simulates how mpox spreads through a modeled network of 10,000 American men who have sex with men, testing behavior changes and vaccination. It finds that cutting one-time sexual partnerships and vaccinating early, especially in the highest-risk group, substantially reduces infections.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No calibration to the observed 2022 US mpox epidemic: the model's no-intervention baseline (16%) is roughly ten times the observed attack rate (~1.5%), so the quantitative 30% and 5.5% claims are not tied to the outbreak being modeled.","rationale":"The reader identifies assortativity as the weakest assumption and also lists the absence of validation against the observed US mpox epidemic curve among the reasons for CONDITIONAL. I partially agree: the assortativity concern is plausible, but the more load-bearing issue is the model's failure to reproduce the observed scale of the 2022 outbreak. The paper's no-intervention baseline of 15.98% and its main intervention outcome of about 11-12% are an order of magnitude larger than the observed US attack rate of roughly 1.5% among the at-risk MSM population the authors themselves cite in Appendix A.1. Because the central quantitative claims are stated as absolute percentages and reductions, a model that does not match the empirical outbreak curve cannot support those specific numbers without a calibration step. The qualitative direction of the results is likely robust, and the paper has real strengths: public code, clear algorithms, and extensive sensitivity analyses. For that reason I would not move the verdict to REJECT; the existing CONDITIONAL verdict is appropriate, but the conditions should explicitly require a calibration or validation exercise against the observed epidemic curve before the quantitative results are presented as forecasts.","tokens_in":41036,"tokens_out":5326,"duration_ms":64126,"concrete_test":"Using the published code, recalibrate the per-contact transmission probability beta (Table 1, currently 0.9 with no citation) so that the main intervention scenario reproduces the observed US 2022 mpox burden (about 30,000 cases among 1,998,039 at-risk MSM, roughly 1.5% attack rate), then recompute the relative reduction and the intervention-timing heatmap in Figure 6. If the 30% reduction or the 5.5% pre-vaccination estimate changes materially, or if no value of beta reproduces the observed epidemic scale without breaking other assumptions, the quantitative central claim should be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing concern is that the model is never calibrated or compared against the observed 2022 US mpox epidemic, and its baseline scale is far too large. In Section 3.1, the no-intervention scenario infects 15.98% of the 10,000-node network, and the main targeted intervention still leaves 11.97% infected. The paper's own at-risk population denominator (Appendix A.1) is 1,998,039, while the US reported roughly 30,000 cases in the 2022 outbreak, about 1.5%. The model therefore predicts, even with interventions, roughly an order of magnitude more infections than actually occurred. Sensitivity analyses (Figures A5-A9) vary infection parameters and transmission probability, but none varies the model to match the observed epidemic curve or final size, so the quantitative headline '30% reduction' and '5.5% infected with pre-vaccination' are not tied to the empirical target. This does not undermine the qualitative conclusion that earlier behavior change and vaccination reduce transmission; it does mean the specific numeric reductions should not be read as validated estimates for the 2022 US outbreak until shown robust under calibration to observed incidence.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper develops a dynamic agent-based network model of mpox transmission among 10,000 US MSM, with main, casual, and one-time partnerships evolving over time. Using parameters from ARTnet and Atlanta surveys, it simulates SEIR dynamics and compares interventions: universal versus targeted behavior change (reducing one-time partnership formation by 50%) and vaccination with CDC-derived availability. The central findings are that the targeted intervention reduces cumulative infections by about 30% relative to no intervention, earlier vaccination (one year pre-outbreak) reduces final size to 5.5%, and one-time partnerships become the dominant transmission route after the first weeks, with median Rt rising from 0.6 to 1.48 by day 28. The paper includes extensive sensitivity analyses and public code.","tokens_in":41265,"tokens_out":6438,"duration_ms":63480,"significance":"The strength of the manuscript is its transparent and mechanistic framework: the code is public, the parameters are tabulated with sources, each scenario is run for 100 simulations with percentile intervals, and sensitivity analyses cover transmission probability, infection parameters, isolation compliance, and population size. If the results are interpreted as relative scenario comparisons, the model provides useful qualitative insights into the value of early and targeted interventions. However, the lack of any calibration to the observed 2022 US mpox epidemic means the quantitative estimates (16%, 30%, 5.5%) are not empirically validated; the paper currently presents them as though they describe the 2022 outbreak, which overstates their status.","major_comments":[{"comment":"The model's no-intervention baseline infects 15.98% of 10,000 nodes (Section 3.1), and the main targeted intervention still leaves 11.97% infected, whereas the observed US 2022 mpox attack rate was roughly 1.5% (about 30,000 cases against the 1,998,039 at-risk population in Appendix A.1). The model is never calibrated to any observed epidemic curve, and the sensitivity analyses in Figures A5–A9 do not include a scenario reproducing the observed final size. Consequently, the headline estimates of a 30% reduction and a 5.5% final size are not validated estimates for the 2022 US outbreak; they should be reframed as relative scenario results or supplemented with a calibration analysis.","section":"§3.1 and Appendix A.1"},{"comment":"The one-time partnership rate is parameterized inconsistently. The text calls πo,k the daily probability of forming a one-time partnership and says that π=0.286 corresponds to about 8 one-time partners per month, but Algorithm 3 samples the daily number of one-time partners as n_o ~ Geometric(1-πo,k), whose mean is πo,k/(1-πo,k). For stratum 6 this gives 0.4 partners per day, roughly 12 per month, not 8. This discrepancy changes the effective one-time partnership contact rate throughout the simulation and should be corrected or explicitly justified.","section":"§2.1.2, Table 1, and Algorithm 3"},{"comment":"The claim that Rt at t=0 equals R0 is not supportable because the initially infected nodes are deliberately seeded in the top two one-time-partnership strata (Section 2.2). The reported per-relationship Rt values at t=0 are conditional on these high-activity seeds and should be labeled as such; otherwise readers may misinterpret them as population-average basic reproduction numbers.","section":"§3.3"}],"minor_comments":[{"comment":"The caption labels both lower panels as 'Panel C'; the second should be 'Panel D'.","section":"Figure A5 caption"},{"comment":"The text contains the typographical error 'N = 5,0000' for the 5,000-node network.","section":"Appendix A.2.3"},{"comment":"Reference [8] is incomplete ('617; 2024'); please provide the full citation.","section":"References"},{"comment":"The abstract states that behavior change and vaccination 'reduce cumulative infections by 30%', but the results show that the reduction is driven almost entirely by behavior change, with vaccination adding no meaningful effect; the abstract should specify that the combined intervention's effect is dominated by behavior change.","section":"§3.1 and Abstract"}],"recommendation":"major_revision","confidential_remarks":"The calibration issue is the main barrier. If the authors reframe the manuscript as a mechanistic scenario analysis rather than a model of the 2022 US epidemic, the paper may be publishable after addressing the parameter inconsistency and the Rt labeling issue. I would not reject on qualitative grounds."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this is a solid, well-documented simulation paper with a real methodological kernel, but its quantitative headline claims are not tied to the outbreak it claims to model. The no-intervention baseline infects 16% of the network, roughly ten times the observed US attack rate of about 1.5% among at-risk MSM, and this discrepancy is never addressed.\n\nThe genuinely new piece is a dynamic agent-based network with three tie types—main, casual, and one-time partnerships—with data-informed rewiring and per-relationship-type reproductive numbers estimated by tracking infection sources. That is a modest extension of existing temporal configuration-model work, but it is a useful one: the finding that sustained partnerships seed the outbreak while one-time partnerships sustain it is mechanistically plausible and clearly presented. The execution is careful: public code, 100 independent runs per scenario, percentile bands, and an unusually wide sensitivity sweep covering transmission probability, isolation compliance, infection parameters, and population size. The authors also state their limitations honestly.\n\nThe soft spot is the one the stress-test flagged, and it holds up on reading. The model is never calibrated or compared with the observed 2022 US epidemic curve or final size. Sensitivity analyses vary parameters but never ask which settings reproduce the empirical outbreak, so the absolute reductions—30% fewer infections, 5.5% infected with early vaccination—are not validated as estimates for the 2022 outbreak. The qualitative direction is robust and consistent with earlier modeling, but the quantitative scale should not be quoted as realistic. Two minor gaps: beta = 0.9 per-contact transmission probability is asserted without a direct source, and input-parameter uncertainty is not propagated into the reported ranges.\n\nThis paper is for readers who want a transparent, flexible mechanistic STI network model and who care about the relative timing and targeting of interventions. The per-tie Rt analysis alone is worth a look. It deserves a serious referee, with the expectation that calibration to the observed epidemic, or a clear reframing of the claims as exploratory rather than predictive, is needed before acceptance.","headline":"A transparent, reproducible mechanistic network model whose qualitative intervention findings hold up, but whose headline numbers are not calibrated to the observed 2022 US mpox epidemic.","tokens_in":41805,"tokens_out":2085,"would_cite":true,"duration_ms":25137,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Halving one-time partnerships and vaccinating the top-risk quarter of men cuts mpox infections by about 30 percent, a dynamic network model suggests.","keywords":["mpox","agent-based model","sexual network","men who have sex with men","behavior change","vaccination","effective reproductive number","intervention timing"],"falsifier":"Contact-tracing data from the 2022 outbreak showing that most late-phase infections spread through repeated main or casual partnerships rather than one-time encounters would contradict the central claim. A second check is comparing the model's predicted weekly incidence decline after the day-70 behavior change with observed mpox case counts.","tokens_in":40766,"feed_emoji":"🦠","tokens_out":7719,"duration_ms":70142,"temperature":0.7,"pith_summary":"This paper claims that the 2022 mpox outbreak among men who have sex with men can be reproduced by a dynamic sexual-network model, and that halving one-time partnerships plus vaccination—even when applied only to the 25% of men most likely to form such partnerships—cuts cumulative infections by about 30%. It further claims that earlier intervention matters more: beginning vaccination a year before an outbreak leaves only 5.5% of men infected, averting nearly 10% of the population's infections. The model also shows that sustained main and casual partnerships drive the first weeks of transmission, while one-time partnerships become the dominant transmission route thereafter. If these claims hold, outbreak planners should prioritize early, targeted outreach to the highest-risk groups rather than waiting for universal coverage.","feed_headline":"Targeted sex-network interventions cut mpox cases 30%","feed_subtitle":"A 10,000-person sexual network model shows cutting one-time partnerships and vaccinating early are the levers.","key_machinery":"The central object is a dynamic configuration-model sexual network: 10,000 nodes with fixed counts of main and casual partners drawn from observed relationship-type proportions, main and casual edges rewired with geometrically distributed durations, and daily one-time partnerships formed by shuffling and pairing stubs. A discrete-time stochastic SEIR model runs on this network, with per-partnership contact probabilities and a transmission probability of 0.9 per sexual contact. The mechanism that carries the argument is infection-source attribution: every infection records which relationship type transmitted it, allowing the authors to compute relationship-specific effective reproductive numbers $R_t^*$ and the proportion of infections attributable to one-time, casual, and main partnerships over time.","core_discovery":"The central discovery is that one-time sexual partnerships, rather than sustained ones, determine the long-run size of a mpox outbreak, and that interventions aimed specifically at the people who form such partnerships are almost as effective as universal interventions. In a simulated population of 10,000 MSM with dynamic main, casual, and one-time partnerships, the no-intervention scenario infects about 16% of the population over 250 days. Adding a 50% reduction in one-time partnership formation and vaccination among only the top two sexual-activity strata (the 25% of men most likely to have a one-time partner) brings cumulative infections down to around 11% of the population, a reduction of about 30% and roughly 500 averted infections. Infection-source tracking shows that the effective reproductive number for one-time partnerships rises from 0.6 at day 0 to 1.48 at day 28, while it falls for casual and main partnerships, meaning one-time partnerships sustain the outbreak after the first weeks.","pith_inferences":["If real MSM networks are assortative by risk rather than randomly mixed, targeting the top 25% could be even more efficient, though the quantitative 30% reduction would shift; this is an extrapolation from the model's no-assortativity assumption.","The relationship-specific $R_t^*$ trajectories imply that monitoring one-time partnership rates, not just case counts, could serve as an early warning indicator for mpox-like STI outbreaks.","A natural testable extension is to fit the same network architecture to other short-infectious-period STIs and check whether one-time partnerships dominate late transmission, which would generalize the behavior-change messaging.","The model's prediction of 5.5% infection with year-early vaccination could be compared with observed outcomes in settings that vaccinated MSM communities before sustained local transmission."],"forward_implications":["A campaign aimed at the 25% of men most likely to form one-time partnerships can avert roughly 30% of infections, nearly matching the effect of universal intervention.","Vaccination begun a year before an outbreak could reduce cumulative infections to about 5.5%, showing that pre-outbreak preparedness can largely substitute for later behavior change.","Because one-time partnerships become the dominant transmission route after the first weeks, interventions that reduce their frequency will have their largest effect later in the outbreak.","The near-equivalence of targeted and universal intervention suggests limited vaccine supply can be routed to high-risk strata without much loss of population-level benefit."],"supporting_citations":[{"why":"Supplies the empirical relationship-type proportions, contact probabilities, and one-time partnership probabilities that parameterize the network.","marker":"[21]"},{"why":"Supplements the network parameters with additional MSM survey data used for the same distributions.","marker":"[23]"},{"why":"Provides the agent-based modeling approach for HIV spread that the paper extends to three partnership types.","marker":"[6]"},{"why":"Introduces the temporal configuration model that underlies the dynamic rewiring of partnerships.","marker":"[20]"},{"why":"Is the statistical-network modeling study whose case-reduction estimates the paper compares with its own.","marker":"[15]"},{"why":"Supplies the survey evidence that about 50% of MSM reduced one-time partners, motivating the behavior-change intervention.","marker":"[16]"},{"why":"Provides the vaccine efficacy estimates used in the vaccination model.","marker":"[34]"},{"why":"Provides the diagnosis-delay and isolation assumptions used in the epidemic model.","marker":"[17]"}],"fun_headline_variants":["Targeting one-time partnerships cuts mpox by 30%","One-time partners drive mpox spread; targeted cuts work","Early vaccination plus fewer casual links cut mpox 30%","High-risk vaccination plus fewer one-time partners cuts mpox 30%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the real sexual network of men who have sex with men can be represented by random pairing of partnership stubs with no assortativity by risk or demographics, so if actual partner choice is strongly assortative, the estimated benefits of targeting the top 25% could change.","fun_headline_variants_meta":{"raw":{"variants":["Targeting one-time partnerships cuts mpox by 30%","One-time partners drive mpox spread; targeted cuts work","Early vaccination plus fewer casual links cut mpox 30%","High-risk vaccination plus fewer one-time partners cuts mpox 30%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001187,"raw_usage":{"total_tokens":4972,"prompt_tokens":1092,"completion_tokens":3880,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":708,"completion_tokens_details":{"reasoning_tokens":3808}},"tokens_in":708,"tokens_out":3880,"duration_ms":25662,"temperature":1.0,"reasoning_tokens":3808,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:05:42.344665+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Contact-tracing data from the 2022 outbreak showing that most late-phase infections spread through repeated main or casual partnerships rather than one-time encounters would contradict the central claim. A second check is comparing the model's predicted weekly incidence decline after the day-70 behavior change with observed mpox case counts.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplements the network parameters with additional MSM survey data used for the same distributions."},{"cited_title":"Temporal Configuration Model: Statistical Inference and Spreading Processes","cited_arxiv_id":null,"evidence_quote":"Introduces the temporal configuration model that underlies the dynamic rewiring of partnerships."},{"cited_title":"Modelling the impact of vaccination and sexual behavior adaptations on mpox cases in the USA during the 2022 outbreak","cited_arxiv_id":null,"evidence_quote":"Is the statistical-network modeling study whose case-reduction estimates the paper compares with its own."},{"cited_title":"Morbidity and Mortality Weekly Report Strategies Adopted by Gay, Bisexual, and Other Men Who Have Sex with Men to Prevent Monkeypox virus Transmission-United States","cited_arxiv_id":null,"evidence_quote":"Supplies the survey evidence that about 50% of MSM reduced one-time partners, motivating the behavior-change intervention."},{"cited_title":"Vaccine Effectiveness of JYNNEOS against Mpox Disease in the United States","cited_arxiv_id":null,"evidence_quote":"Provides the vaccine efficacy estimates used in the vaccination model."},{"cited_title":"The Roles of Unrecognized Mpox Cases, Contact Isolation and Vaccination in Determining Epidemic Size in Belgium: A Modeling Study","cited_arxiv_id":null,"evidence_quote":"Provides the diagnosis-delay and isolation assumptions used in the epidemic model."}],"review_version":1}