{"id":"71555365-2b58-4498-8f9d-351582f8a466","arxiv_id":"2506.05992","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"A stochastic-dynamic model with a moving immune extinction threshold replicates age-specific cumulative cancer risk curves and predicts a 1.75-fold higher breast cancer risk for women with short menstrual cycles.","lead":"This paper proposes a population-dynamics model of cancer where the immune system acts as a moving extinction threshold, and shows the model can mimic cumulative risk curves for several cancers. The authors use it to argue that menstrual-cycle and hormone-replacement effects on immune response explain the unusual shape of breast cancer risk by age.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Breast-cancer fit (Fig 7d) may be an artifact of screening and cohort effects; the menopause immune switch in Eq (12) is not independently identified.","rationale":"The reader's weakest assumption identifies the menopause immune switch as load-bearing, and I agree. The breast-cancer data are cross-sectional, from a single census year, in a population with organized biennial screening, so the slope change near age 50 can be produced by screening detection, overdiagnosis, or birth-cohort differences in HRT use rather than by an immune-system switch. Eq. (12) is inserted in exactly the age range where the constant-response model under-predicts (Cases 1 and 2 of Fig. 7), which is why the improved fit alone cannot validate the mechanism. The proposed pre-screening or true-cohort re-fit is a clean discriminator: if the transition persists in unscreened cohort data, the screening confound is ruled out; if it disappears, the central breast-cancer explanation is not supported. The 1.75-fold cycle-length prediction is also worth a robustness check on bin boundaries, but it is secondary to this validation concern. I therefore keep the reader's conditional verdict unchanged.","tokens_in":35137,"tokens_out":9125,"duration_ms":92207,"concrete_test":"Replace the cross-sectional 2016 Irish data with a true birth cohort or pre-screening incidence series, e.g., compute R(t) from SEER 1973-1980 or from a cohort with known screening history, and re-run the fitting protocol for Fig. 7 using the same three cases (constant immune response, periodic + smax after M, plus HRT). If the polynomial-to-linear transition disappears or is already captured by a constant immune response, the menopause immune switch in Eq. (12) is not necessary and the claimed mechanism is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reproduction of the unusual breast-cancer cumulative risk is the paper's key empirical success, but in Sec. 3.2.2 the fit only succeeds after inserting the time-varying immune response in Eq. (12): sm(t) drops to smin in luteal phases before menopause, rises to smax after menopause, and M is delayed for 26% of women by HRT (Fig. 7b-d). The breast-cancer data are single-year (2016) cross-sectional Irish incidence from a population with organized biennial mammography. The observed polynomial-to-linear transition near ages 44-50 could therefore be produced by screening-related overdiagnosis and lead-time, by birth-cohort differences in HRT use and reproductive history, or by the proposed menopause immune switch. The paper does not test or control for these established epidemiological confounders. Because the abstract claims 'new insights' into HRT and menstrual-cycle effects, this confound is load-bearing: if the flattening is a screening or cohort artifact rather than an immune-system switch, the central explanation for the breast-cancer curve is not established, even though the model may still be mathematically consistent.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an ODE/stochastic model of carcinogenesis in which a cluster of mutated cells grows under a modified Richards' equation with an additional immune-elimination term (Eq. 5), random mutations are drawn from a binomial distribution whose mutation probability increases linearly with age (Eqs. 8-9), and a time-varying immune response s(t) produces a moving extinction threshold A(t) approximated by Eq. (11). The authors first show that the classical Volterra/Allee model (Model 1) cannot produce the growth-versus-saturation asymmetry typical of cancer, while their Model 2 can (Sec. 2.2, Supplement Prop. 1.1). They then fit Model 2 to two mouse breast-tumour data sets (Sec. 3.1), reproduce the exponential cumulative risk of colorectal cancer with a constant immune response (Fig. 6), and reproduce the polynomial-then-linear cumulative risk of breast cancer in Irish women by adding a menstrual-cycle immune switch, distributed menopause, and HRT-delayed menopause (Eq. 12, Fig. 7d). Finally, they report a model-predicted 1.75-fold increase in breast-cancer risk for 22-25-day cycles relative to 25-31-day cycles, compared with the 1.86-fold ratio observed by Yuan et al. 1988 (Sec. 3.2.3).","tokens_in":35554,"tokens_out":7099,"duration_ms":76868,"significance":"If established, the paper would make a useful contribution: a single process-based model with an extinction threshold generates several qualitatively different incidence patterns, and the moving-threshold concept is transferable to other time-varying immune contexts. The formal asymmetry result for Model 1 versus Model 2 is a clean and permanent mathematical contribution; the mice-data fits in Sec. 3.1 are genuine empirical tests with plausible estimated carrying capacities; and the colorectal exponential fit is a nontrivial qualitative success. The breast-cancer application, however, is at present an in-sample proof-of-concept rather than an established explanation: the key mechanism in Eq. (12) is not independently identified, the data are a single cross-section with known screening and cohort confounders, and the cycle-length and HRT 'predictions' are made with the same mechanism used to fit the curve. With a strengthened statistical and identifiability analysis, and with claims scaled to the evidence, this could become a solid paper.","major_comments":[{"comment":"The abstract and Conclusions claim that the model 'accurately reproduces' the breast-cancer cumulative risk and 'predicts' HRT and cycle-length effects, but every component of the breast-cancer fit is calibrated to the same 2016 Irish cumulative-risk curve: p0, pT, smin, and smax in Table 2, the menopause distribution, and the 26% HRT fraction and gamma-distribution parameters in Table 3 are all chosen so that Fig. 7(d) visually matches the black curve in Fig. 7. There is no holdout age band, no second population, no goodness-of-fit statistic, and no confidence interval on either R(t) or R-hat(t). With at least seven free parameters available for a 16-point curve, a visual match cannot support the strength of the reported claim. Please provide an out-of-sample check (for example, fit on ages 0-45 and predict the 50-85 age bands, or fit the NCRI 1994-2021 averaged data and predict the 2016 census-year curve) and quantify uncertainty with bootstrap or profile-likelihood methods.","section":"Abstract, §3.2, Tables 2-3, Fig. 7"},{"comment":"The post-menopausal flattening of the cumulative risk is produced by the discrete switch sm(t)=smax for t>M together with the HRT-delayed effective menopause M; without this switch (Cases 1 and 2 in Fig. 7b-c) the model under-predicts risk above age 50. This switch is not measured or independently estimated; it is supported only by qualitative progesterone-immunology citations and by the improved fit in Fig. 7(d). The observed polynomial-to-linear transition at ages 44-50 could instead arise from the organized biennial screening programme (overdiagnosis and lead-time), from birth-cohort differences in HRT use and reproductive history in the single-year 2016 cross-section, or from the proposed immune switch. Because the abstract claims 'new insights' into HRT, this identifiability problem is load-bearing. A concrete test would be to fit the model to screening-adjusted or age-period-cohort-decomposed incidence, or to a population with a different screening policy, and to compare the implied smin/smax switch with independent evidence on progesterone-related immune modulation in women.","section":"§3.2.2, Eq. (12), Fig. 7"},{"comment":"The 1.75-fold cycle-length ratio is not a free prediction: the menstrual-cycle mechanism in Eq. (12) is already the mechanism used to fit the overall breast-cancer curve in Fig. 7(d), so the cycle-length comparison in Fig. 9 is a consistency check of the same fitted mechanism rather than an independent test of the model. The quoted ratio also depends on modeling choices whose sensitivity is not reported: the lower bound of 22 days is set as approximately two standard deviations below the mean (Supplement §2.2.6), the luteal phase is fixed at 14 days, and the comparison target is a 1988 Shanghai study with a different birth cohort and screening context than the 2016 Irish data. Please report the ratio as a function of the chosen cycle-length bounds, give Monte Carlo uncertainty estimates for R-hat(51, tf), and state the population mismatch explicitly.","section":"§3.2.3 and Supplement §2.2.6"},{"comment":"The parameter values used in the complete model (Table 3) are obtained by fitting Model 2 to two measurements from a single untreated human breast cancer (Fornvik et al. 2016), with A and K fixed a priori. The uncertainties from this two-point boundary-value problem are not propagated into Fig. 7 or Fig. 9, and the text does not report how strongly the final breast-cancer curves depend on r, mu, s, and nu=10. Because these progression parameters are then held fixed while p0, pT, smin, and smax are tuned, the effective number of degrees of freedom in the breast-cancer fit is larger than a casual reading of Table 2 suggests. Please add a sensitivity analysis (for example, doubling or halving r, mu, and s and refitting the remaining parameters) or provide a joint uncertainty propagation.","section":"Supplement §3 and Table 3"}],"minor_comments":[{"comment":"The binomial distribution is printed as n!/(m!(n-m!)) in Eq. (8); the denominator should be m!(n-m)!.","section":"Eq. (8)"},{"comment":"The variable a in A=10^a is used in Eqs. (7) and (11) but is not defined in the main text; please define it explicitly near the first occurrence.","section":"Eqs. (7) and (11)"},{"comment":"The caption says 'black triangles in Fig. 5(a)' for Dataset 2, but Dataset 2 is plotted in Fig. 5(b).","section":"Fig. 5 caption"},{"comment":"There is a duplicated article in the sentence 'falls again below the the (blue/red) moving extinction threshold'; please correct.","section":"Supplement §1.5.2"},{"comment":"The text describes the pre-menopausal breast-cancer curve as a '7th order polynomial', but the supporting analysis in Supplement Fig. 5 fits a power law R(t)=a t^b with b≈7; these are different functional forms, and the evidence for 'linear thereafter' is based on a cubic fit without tests of coefficient significance. Please clarify the terminology and report confidence intervals for the fitted exponents.","section":"§3.2 and Supplement §4"}],"recommendation":"major_revision","confidential_remarks":"The authors themselves call Sec. 3 a 'proof of concept' at its start, and this is in tension with the stronger language in the abstract and conclusions ('accurately reproduces', 'predicts'). The core mathematical contribution is solid and the revisions I request are within the scope of a major revision: scale the claims, add out-of-sample or screening-adjusted checks, propagate uncertainty, and test sensitivity of the cycle-length and HRT conclusions. The supplementary material is essential to the paper; please ensure it is published in accessible form."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know: this paper's basic dynamical-systems move is solid and worth stealing. Model 2 — Richards growth plus an additive immune death term -s(t) — gives the desired property that small mutated clones grow faster than large tumours saturate for ν≥1, which classical Volterra/Allee models cannot do. That is a real result, and the mouse-tumour fits are convincing. The stochastic mutation layer with a moving extinction threshold is a natural way to connect mutation accumulation to incidence, and the cycle-length prediction (relative risk 1.75 vs observed 1.86) is a nice check.\n\nBut the headline claim about reproducing breast-cancer incidence is not as strong as the abstract says. The post-menopause flattening is only reproduced after inserting the menopause-dependent immune switch in Eq. (12), and without it the model underpredicts old-age risk. That switch is not independently identified; it is a fitting device. The data are single-year cross-sectional Irish incidence from a screening population, and the paper never addresses the obvious alternative that the flattening is a screening/cohort artifact rather than a progesterone-driven immune change. So the 'new insights' into HRT and menstrual cycle are conditional on a premise that has no direct evidence.\n\nTwo further issues. The key proof that Model 2 achieves |λ_grow/λ_sat| ≥ 1 is cited to the supplement, but the supplement does not contain it—only the approximation for the extinction threshold. That needs to be supplied. And the fit relies on several tuned parameters (p0, pT, smin, smax) with no uncertainty quantification, so 'accurately reproduces' is overclaiming.\n\nThe paper is clear and honest about being a proof of concept, but the abstract oversells. Still, the framework is genuinely novel and the dynamics are correct, so it deserves a serious referee. I would send it out, asking for the missing proof, an uncertainty analysis, and a comparison against a simple screening/cohort alternative.","headline":"A genuinely novel modeling framework with a clean fix to Allee-type growth, but the breast-cancer story is oversold: the key immune-switch mechanism is fitted to the data and confounding explanations are left untested.","tokens_in":35900,"tokens_out":3368,"would_cite":false,"duration_ms":34667,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92C50","92D25"],"pacs":[],"model":"deepseek-v4-flash","headline":"One simple model with a moving immune extinction threshold reproduces exponential and breast-cancer risk curves and predicts a cycle-length risk ratio close to observation.","keywords":["cancer modelling","extinction threshold","immune system","Allee effect","breast cancer","menstrual cycle","hormone replacement therapy","cumulative risk"],"falsifier":"Measure age-specific cumulative breast cancer risk in a large cohort of women who never used HRT and were never screened: the model with distributed menopause but no HRT under-predicts risk after age 50, so if such a cohort still shows the same linear postmenopausal rise as the general population, the menopause/HRT switch cannot be the cause.","tokens_in":34899,"feed_emoji":"🧬","tokens_out":9714,"duration_ms":92201,"temperature":0.7,"pith_summary":"The paper claims that a single simple model—random mutations plus an immune response that acts as an extinction threshold—can reproduce several qualitatively different real cancer datasets. With a fixed immune response, the model yields the exponential age-specific cumulative risk seen in most cancers, illustrated by colorectal cancer in women. With an immune response that dips in the luteal phase and stops dipping after menopause, it reproduces the unusual polynomial-then-linear breast cancer risk curve, including the effect of hormone replacement therapy. The same mechanism predicts a 1.75-fold higher breast cancer risk for women with 22–25 day cycles than for women with 25–31 day cycles, close to the observed 1.86-fold difference. If correct, the paper establishes that a moving extinction threshold can explain these incidence patterns and produces testable predictions about cycle length and HRT.","feed_headline":"Cancer model with moving immune threshold reproduces real risk curves","feed_subtitle":"One model also predicts a 1.75-fold higher breast cancer risk for 22–25 day cycles, matching epidemiology.","key_machinery":"The load-bearing object is the moving extinction threshold, an Allee-type critical cluster size separating immune elimination from cancer establishment. In the model it is approximated by \\(\\bar{A}(t)\\approx(s_m(t)-\\overline{m}(t))/(\\nu r-\\mu(\\nu+a\\ln 10))\\), where \\(s_m(t)\\) is the daily immune-elimination rate, \\(\\overline{m}(t)\\) is the mean daily mutation count, and \\(\\nu,r,\\mu\\) are growth parameters. The threshold is what makes most mutated clusters die out and only threshold-crossing clusters become cancer; when \\(s_m(t)\\) varies with menstrual cycle phase and menopause, the threshold moves and changes the shape of the age-specific risk curve.","core_discovery":"The paper's central claim is that a single daily-update model, \\(x(t+1)=x(t)+\\nu r x(t)-\\nu\\mu x(t)^{(\\nu+1)/\\nu}-s_m(t)+m(t)\\) clipped at zero, captures both carcinogenesis and tumour progression: \\(m(t)\\) is a binomially distributed number of new mutations whose per-cell probability rises linearly with age, and \\(s_m(t)\\) is the number of mutated cells the immune system removes each day. The immune term creates a critical cluster size, the extinction threshold \\(A(t)\\), below which the immune system eradicates the cluster and above which it grows into cancer. With a constant immune response the model produces the exponentially rising cumulative risk typical of most cancers and fits colorectal cancer data; with an immune response that weakens during the luteal phase of the menstrual cycle and returns to full strength after an effective menopause, it reproduces the polynomial-then-linear breast cancer risk curve once distributed menopause and HRT use are included. The same simulations fit breast cancer growth in two mouse datasets, reproduce a 1.75-fold relative risk for short versus longer cycles against the observed 1.86-fold value, and place most transitions to cancer in the luteal phase.","pith_inferences":["Beyond the paper: if the extinction threshold is the true control point, strengthening immune clearance during the luteal phase should measurably shift the age of first threshold crossing in animal models.","Beyond the paper: the rescue-event mechanism implies that temporary immunosuppression need not be fatal—a cluster can cross the threshold and fall back below it—so timing-based immune interventions may deserve study.","Beyond the paper: the model offers a dynamical alternative to screening-artifact explanations of the postmenopausal breast cancer plateau; comparing never-screened, never-HRT cohorts with the model's predictions would separate the explanations."],"forward_implications":["A constant immune response is sufficient to explain why most cancers show exponentially rising age-specific cumulative risk.","The postmenopausal flattening of breast cancer risk can be explained dynamically by the end of cyclic progesterone-related immune weakening, without invoking separate biological mechanisms for young and old ages.","Shorter menstrual cycles raise risk because more days are spent with a lowered extinction threshold; the model quantifies this as a 1.75-fold effect for 22–25 versus 25–31 day cycles.","Most modelled transitions to breast cancer occur in the luteal phase (63–76%), consistent with progesterone as the main cycle-associated driver.","The same moving-threshold mechanism can be applied to other cancers and to time-varying factors such as infections, immune deficiencies, and immunotherapy."],"supporting_citations":[{"why":"The paper's supplementary material; supplies all derivations, parameter tables, and the Monte-Carlo and curve-fitting procedures every fit relies on.","marker":"[44]"},{"why":"Mouse breast cancer progression dataset 1, used to show Model 2 fits the growth data while Model 1 fails.","marker":"[41]"},{"why":"Mouse breast cancer progression dataset 2, used to compare Model 1 and Model 2 on immunocompetent mice.","marker":"[51]"},{"why":"Registry incidence data for colorectal and breast cancer in women, from which the observed cumulative risk curves are computed.","marker":"[57]"},{"why":"Epidemiological observation of a 1.86-fold higher breast cancer risk for short cycles; the model's 1.75-fold prediction is tested against it.","marker":"[79]"},{"why":"Provides prevalence and duration data for postmenopausal HRT use used to set the 26% proportion and the gamma duration distribution.","marker":"[77]"},{"why":"Provides HRT duration data used to model the extension of effective menopause in Case 3.","marker":"[78]"},{"why":"Two measurements of untreated human breast cancer growth used to fix parameters of the complete model.","marker":"[20]"}],"fun_headline_variants":["Moving immune threshold matches real cancer risk curves","Cycle-driven immune shifts fit breast cancer epidemiology","One dynamic model reproduces human and mouse cancer data","Immune threshold that moves explains cancer risk patterns","Moving threshold model fits human and mouse cancer data"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central explanation rests on the assumption that the postmenopausal flattening of breast cancer risk is caused by a switch in immune response at menopause; if the flattening instead comes from screening overdiagnosis, cohort changes, or another biological pathway, the breast-cancer explanation collapses even though the mathematics may remain consistent.","fun_headline_variants_meta":{"raw":{"variants":["Moving immune threshold matches real cancer risk curves","Cycle-driven immune shifts fit breast cancer epidemiology","One dynamic model reproduces human and mouse cancer data","Immune threshold that moves explains cancer risk patterns","Moving threshold model fits human and mouse cancer data"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000588,"raw_usage":{"total_tokens":2783,"prompt_tokens":991,"completion_tokens":1792,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":607,"completion_tokens_details":{"reasoning_tokens":1722}},"tokens_in":607,"tokens_out":1792,"duration_ms":13111,"temperature":1.0,"reasoning_tokens":1722,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T06:03:04.865830+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure age-specific cumulative breast cancer risk in a large cohort of women who never used HRT and were never screened: the model with distributed menopause but no HRT under-predicts risk after age 50, so if such a cohort still shows the same linear postmenopausal rise as the general population, the menopause/HRT switch cannot be the cause.","supporting_citations":[{"cited_title":"F¨ ornvik, K","cited_arxiv_id":null,"evidence_quote":"Two measurements of untreated human breast cancer growth used to fix parameters of the complete model."}],"review_version":1}