Pith. sign in

REVIEW 4 major objections 5 minor 43 references

Mammographic density masks breast cancer risk: a latent-state model separates risk from detectability and shows that ignoring masking underestimates risk by up to 70%.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 19:55 UTC pith:62TQY7TP

load-bearing objection A clever latent-Markov masking framework undermined by treating hand-set sensitivity parameters as an empirical finding. the 4 major comments →

arxiv 2607.16793 v1 pith:62TQY7TP submitted 2026-07-18 stat.ME stat.AP

Decoupling risk and masking in mammographic density under irregular follow up using a latent Markov progression detection framework

Furkan Danisman , Zarina Oflaz , Zeynep Kalaylioglu , Mahmut Onur Kulturoglu , Lutfi Dogan This is my paper
classification stat.ME stat.AP MSC 62P1062M05
keywords mammographic densitybreast cancer riskmasking effectlatent Markov modelirregular longitudinal datadetectabilityBMI stratificationscreening sensitivity
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper argues that mammographic density does not directly cause breast cancer, but is a visible trace of an unobserved disease-progression process. By fitting a latent Markov model to irregular longitudinal density histories, the authors extract a hidden risk state; the observed density then acts only as a detection-sensitivity factor. They show that standard analyses that ignore masking are systematically biased: moving from perfect detection to a realistic masking scenario raises the estimated risk of a cancer diagnosis by about 70% for post-menopausal overweight patients and about 40% for obese patients. If true, this means that clinical decisions based on density alone substantially underestimate risk in the very groups where supplemental imaging is most debated.

Core claim

The paper's central claim is that the risk carried by mammographic density can be decoupled from its masking effect. The authors formalize this through a two-phase model: a latent Markov chain describes how an unobserved progression state evolves over time; the observed density category is an emission of that state. In the second phase, cancer diagnosis is modeled as the product of cancer development (driven by cumulative time spent in latent states) and detection at the final visit (driven only by the final density category). This yields the factorization Pr(diagnosis) = p_density × logistic(latent-state occupancy), where p_density is a detection probability. The authors show empirically th

What carries the argument

The central object is a latent Markov model over three ordered density categories (D, C, B), with an underlying four-state risk process (critical, established, emergent, low emergence). Irregular visit times are first regularized to a monthly grid under a monotone no-reversal assumption, filling gaps with a Markov bridge that conditions on observed endpoints. The key identity is Equation (7): Pr(Z_i=1 | c_i, occupancy, covariates) = p_{c_i} ∙ logit⁻¹(β0 + Σ β_j n_j + βᵀx_i). Here p_{c_i} is the detection probability of the final density category, and n_j are monthly occupancy counts of latent state j. This factorization is what separates risk (the logistic term) from detectability (p).

Load-bearing premise

Density never undergoes clinically meaningful reversals and unobserved monthly density values are missing at random, so the regularly imputed histories used to infer the latent risk states are unbiased.

What would settle it

In a screening cohort where interval cancers are recorded (cancers detected between scheduled screenings), fit the same latent Markov model but with detection status observed for interval cases. If final density category retains predictive value for interval cancer after conditioning on latent-state occupancy, then Assumption 1—that the latent history captures all risk information—is contradicted and the decoupling claim collapses.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Standard risk analyses that ignore masking systematically underestimate the predictive effect of dense tissue, by roughly 70% in post-menopausal overweight patients and 40% in obese patients.
  • Cumulative time spent in high-risk latent states, not single-visit density, is the actionable risk signal—two patients with the same final density can have very different risk histories.
  • The same observed density category maps to different latent-risk structures across BMI groups, so density should be interpreted within BMI strata, not on a single scale.
  • If the underestimation is as large as estimated, the case for supplemental imaging in post-menopausal overweight and obese women is stronger than currently believed.
  • Higher parity accelerates movement toward lower-risk states while family history slows it, offering modifiable and non-modifiable prognostic markers visible in screening histories.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same latent-progression-plus-detectability architecture applies to any biomarker that both predicts disease and degrades test sensitivity—for example, lung nodule conspicuity in CT screening or lesion visibility in ultrasound—so the decoupling idea is portable beyond mammography.
  • A direct external test would apply the framework to a cohort with interval cancers (cancers diagnosed between screenings), where detection status is known rather than latent; the estimated masking correction should match the empirical sensitivity difference between dense and non-dense breasts.
  • If hormonal replacement therapy or other causes produce genuine reverse density transitions, the monotone Markov bridge will systematically mis-attribute those histories; allowing reversible transitions would reveal how sensitive the 70% figure is to this assumption.
  • The sensitivity parameters (p_B, p_C, p_D) could be pinned down, rather than scanned, by linking mammography records to interval-cancer rates, which would turn the paper's bracketed risk estimates into point estimates.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes a two-phase framework for mammographic density histories observed at irregular visits. Phase 1 regularizes the histories onto a monthly grid by assuming a monotone D→C→B density process and fits a latent Markov model within BMI groups. Phase 2 links the inferred latent-state occupancy counts and the final density category to the observed binary diagnosis through the factorization Pr(Z=1) = p_{c_i} logit^{-1}(β0 + Σβ_j n_j + β^T x), where p_c are detection probabilities treated as sensitivity parameters. The paper reports that accounting for masking increases estimated breast cancer risk by about 70% in post-menopausal overweight patients and about 40% in obese patients, and concludes that the framework decouples risk from detectability.

Significance. The question addressed — how mammographic masking affects risk estimates — is clinically and methodologically important. The manuscript is transparent in some respects: it explicitly states that the detection probabilities are not identifiable, propagates uncertainty through random completion, posterior path sampling, and the bootstrap, and reports sensitivity analyses for K=3 and K=5. However, the central quantitative claim (the 70%/40% masking-induced increases) is a deterministic consequence of the hand-set detection probabilities, and the 'empirically supported' scenario is not anchored to absolute sensitivity data. As a conditional sensitivity analysis the framework is coherent; as an empirical quantification of masking it is not identified. The score test for the central assumption tests only a weak implication and cannot validate the completed histories or the assumed p vector.

major comments (4)
  1. [§4.2, Eq. (7), Table 7] The abstract and conclusions report that accounting for masking 'raised estimated breast cancer risk by 70% for post-menopausal overweight patients and about 40% for obese patients'. These numbers are not identified from the data: the detection probabilities (p_B,p_C,p_D) are imposed sensitivity parameters, and the estimates in Table 7 are refits of β for each chosen vector. In particular, scenario (iii) (0.8,0.3,0.2) is labeled 'empirically supported', but the cited reference (Mandelson et al., JNCI 2000) reports odds ratios for interval versus screen-detected cancer by density, not absolute screening sensitivities. A detection probability of 0.2–0.3 for dense categories is at the low end of published estimates, so the 70%/40% inflation is a sensitivity result conditional on a plausible-but-unverified assumption, not an empirical estimate. The conclusion should be reframed as a conditio
  2. [§3.1, 'Irregular Missing Observations'] The monthly histories used in every subsequent analysis are constructed under a hard monotone D→C→B assumption. The patient(s) with reverse transitions are excluded, boundary months are carried forward/backward, and gaps between different endpoint categories are filled from a Markov bridge estimated after removing reverse transitions. This is load-bearing because the occupancy counts n_j^i entering Eq. (7) are computed from these completed histories. If visit times are informative (as the paper's own literature review in §1 acknowledges) or if reversals are clinically meaningful (e.g., HRT), the completed histories and all Phase 2 coefficients are biased. The paper does not report sensitivity analyses under alternative missingness mechanisms, informative visit processes, or relaxed monotonicity. The score test in Table 8 only uses the final density and cannot diagnose bias in the imputed
  3. [§3.3, Assumption 1, Appendix Table 8] The 'decoupling' of risk from masking is imposed by Assumption 1, which states that risk depends on the latent-state occupancy and not on the observed density trajectory. The empirical justification is an efficient score test for H0: δ_B=δ_C=δ_D=0 in a model that adds the final density category to the risk component. This test has two limitations. First, it checks only the final density, not the entire observed trajectory, so it does not validate the full conditional-independence assumption in Assumption 1. Second, the latent states are themselves estimated from the density histories, so the test is partly circular: the inferred occupancy counts are functions of the observed densities, and conditioning on them is expected to absorb much of the density effect by construction. The score test also depends on the p vector, which is fixed rather than estimated. The paper should state clearly
  4. [§2 and §4, outcome definition] The outcome Z_i is diagnosis at the final screening visit, not incident cancer, and the cohort is a retrospective, highly enriched sample (40% cases). The model in Eq. (7) attempts to recover undiagnosed cancers through the p_c factor, but with no follow-up past the final visit and no external calibration, 'cancer development probability' in Assumption 1 is not directly observed. The reported odds ratios therefore describe diagnosis conditional on reaching the final visit under the assumed detection mechanism, not a validated risk of cancer development. This strengthens the need to frame the results as conditional sensitivity estimates rather than empirical risk quantification.
minor comments (5)
  1. [Abstract vs. Discussion] The abstract says the masking-induced increase is about 70% for post-menopausal overweight patients; the Discussion says 'roughly 75%'. The numbers should be consistent or reconciled.
  2. [§3.1] 'Only two of 2464 observed visit-to-visit transitions move in the reverse direction. We therefore exclude the patient exhibiting a reverse transition' — if there are two reverse transitions, clarify whether one patient contributed both or whether two patients were excluded.
  3. [Table 5] Some initial-state odds ratios are effectively unbounded (e.g., 77305.195), indicating near-complete separation. These should be reported with a note about separation, rather than as stable point estimates.
  4. [Tables 3–4] Several state-by-visit cells have very small counts (e.g., overweight last-visit CR, n=1), making the reported percentages unstable. The paper should flag these cells or suppress them.
  5. [§4.2] The phrase 'empirically supported detectability scenarios' is too strong. The cited evidence does not determine absolute detection probabilities, and the phrase should be tempered throughout.

Circularity Check

2 steps flagged

The 70%/40% masking-driven risk increases are deterministic consequences of the hand-set detection probabilities in Eq. (7), and the risk/detectability 'decoupling' is built into the model by defining latent states as a re-encoding of the density history.

specific steps
  1. self definitional [Section 3.3, Eq. (7); Section 4.2, detectability scenarios; Abstract Results]
    "because pB, pC, pD are not identifiable from the observed diagnoses alone, we do not estimate them; instead we treat them as sensitivity parameters and refit (7) over clinically admissible values satisfying pD < pC < pB... Moving from the no-masking baseline—the assumption implicit in the current literature—to an empirically supported detectability scenario, the increase in the estimated risks is not marginal."

    Equation (7) sets Pr(Z_i=1 | c_i, {n_j}, x) = p_{c_i} logit^-1(beta0 + sum_j beta_j n_j + beta^T x). The observed diagnosis Z_i is fixed; choosing p_{c_i}<1 for dense categories forces the fitted logistic term (and hence the beta coefficients) upward for patients with dense final categories. The reported 70%/40% increases are therefore a deterministic function of the selected p vector, especially scenario (0.8, 0.3, 0.2), and not an empirical estimate of masking from the data. The paper openly states p is not identifiable, so the 'quantification' is a restatement of the assumed sensitivity input rather than a data-derived prediction.

  2. self definitional [Section 3.2 'Phase 1: Latent Markov Model of Density Progression'; Section 3.3, Assumption 1; Table 1]
    "Our central assumption is that mammographic density itself does not carry breast cancer risk; rather, the risk is carried by an unobserved progression state, of which density is only an observed manifestation... Assumption 1: Conditional on the cumulative latent-state occupancy and covariates, the observed density trajectory carries no additional information about cancer development."

    The latent states U_i are estimated exclusively from the regularized density sequence via the latent Markov model, and Table 1 shows essentially deterministic emissions (each density category maps to latent states with probabilities 1 or 0). Thus the 'risk-bearing' occupancy counts n_j are a re-encoding of the density history. Claiming that density carries no risk once this latent re-encoding is included, and then interpreting the remaining final-density effect as 'masking,' is a definitional separation imposed by the model, not an empirical decoupling. The score test in Appendix Table 8 only adds the final density category to the same fitted risk model, so it cannot validate the assumed p vector or the causal direction; it merely checks one component of a trajectory that is already summar

full rationale

The paper is transparent that p=(pB,pC,pD) is not identifiable and is treated as a sensitivity parameter, but the headline conclusion converts this sensitivity analysis into an empirical claim: 'accounting for masking ... raised the estimated breast cancer risk by 70% ... and about 40%.' That increase is an algebraic consequence of Eq. (7): with observed diagnoses fixed, lowering p for dense categories forces the risk logit upward, so the reported masking effect is a restatement of the assumed p vector, not a quantity the study identifies from the data. The 'decoupling' of risk from detectability is likewise built into the modeling assumptions: latent states are fitted to the density trajectories, risk is assigned to those latent states, and density is then said to act only through detectability. The score test for Assumption 1 is not load-bearing against this concern, because it merely tests whether final density adds to a risk model that already contains latent occupancy derived from the same density data. The self-citation to the authors' prior dataset (ref. 11) is not the main circular mechanism; the circularity is in the phase-2 sensitivity quantification and the definitional construction of latent risk. Excluding this central issue, the Phase 1 progression findings (parity, family history, menopausal associations) are ordinary empirical model outputs and are not themselves circular.

Axiom & Free-Parameter Ledger

4 free parameters · 7 axioms · 2 invented entities

The central result rests on three layers the reader funds: (1) a monotone no-reversal density model used to impute all missing months; (2) a latent Markov structure whose states are labeled post hoc; and (3) unidentifiable detection probabilities set by hand. The first two are domain assumptions, the last is the main load-bearing choice because the 70%/40% estimate is a function of it.

free parameters (4)
  • Detection probabilities (pB, pC, pD) = Scenarios: (1,1,1), (0.8,0.6,0.4), (0.8,0.3,0.2)
    Not identifiable from observed diagnoses; chosen by hand as sensitivity inputs. The 70%/40% risk-increase result is a direct function of these values.
  • Number of latent states K = 4
    Selected as second-lowest BIC in both BMI groups and aligned with four clinical cancer stages; K=5 had the lowest BIC; sensitivity checks at K=3 and K=5 are reported but not used for the main conclusions.
  • Latent Markov model parameters (initial-state logits, transition logits, emission probabilities) = Estimated via EM using LMest
    These define latent-state occupancy counts entering the Phase 2 risk model; they are estimated from the regularized density data, so all downstream odds ratios inherit their uncertainty.
  • Imputation transition probabilities pDC, pCB in the monotone Markov model = Estimated from observed intervals
    Used to stochastically fill gaps between differing observed categories; if the monotone model is wrong, imputed histories and latent states are affected.
axioms (7)
  • domain assumption Mammographic density evolves monotonically D to C to B with no reversals (one reverse-transition patient excluded).
    Section 3.1; 2 of 2464 transitions reverse, and one patient is excluded; the imputation of all missing density months depends on this.
  • domain assumption Missingness of density at unobserved months is ignorable; visit times carry no additional information.
    Section 3.1 frames irregular visits as a missing-data problem; no model for an informative visit process is fit, despite citing literature on informative visiting.
  • ad hoc to paper Assumption 1: risk of cancer development depends on latent-state occupancy only, not on the observed density trajectory.
    Section 3.3; this is the central decoupling assumption; tested only indirectly by adding final density category to the risk model.
  • ad hoc to paper Assumption 2: detection depends on final density category only, with Pr(D=1|C=0,.)=0 and ordering pD<pC<pB.
    Section 3.3; pB,pC,pD are set by hand, and this assumption defines the masking effect.
  • standard math Diagnosis is the product Zi = Ci Di (cancer present and detected).
    Section 3.3; product rule used to factor the observed diagnosis probability.
  • domain assumption Latent Markov structure with multinomial-logit initial/transition probabilities on a common 80-month grid.
    Section 3.2; standard latent Markov model but imposes functional form for covariates and homogeneous monthly transitions.
  • ad hoc to paper Four latent states can be ordered and labeled LR/EmR/EsR/CR based on emission and transition matrices.
    Section 4.1.2; labels are post hoc and used to interpret risk coefficients; ordering is data-driven but not externally validated.
invented entities (2)
  • Latent disease progression state U_i(t) (K=4 states) no independent evidence
    purpose: Unobserved risk-bearing process inferred from density; density is treated as a manifestation rather than a causal risk factor.
    No external handle (biomarker, interval cancer, pathology) is used to validate the latent states; the score test only checks whether final density adds risk information, not whether latent states are causally meaningful.
  • Latent risk labels (Low, Emergent, Established, Critical) no independent evidence
    purpose: Interpretive labels for the four fitted states.
    Assigned post hoc from emission and transition patterns; labels are not identified by data alone.

pith-pipeline@v1.3.0-alltime-deepseek · 4092 in / 5632 out tokens · 196742 ms · 2026-08-01T19:55:41.121213+00:00 · methodology

0 comments
read the original abstract

Background: Mammographic density is a strong marker of breast cancer risk, yet it also reduces mammographic sensitivity through masking. Screening densities are observed at irregular times, requiring methods that accommodate irregular follow-up. Methods: We propose a two-phase framework for irregular mammographic screening data. After regularizing the irregular density histories, the first phase fits a latent Markov model within each BMI group, under the assumption that density does not itself drive risk but the underlying latent disease process does. Extracting the latent-state information leaves density to act only as a detectability factor, and the second phase links the latent process and the final density to cancer diagnosis through a detection--risk model. Detection probabilities are treated as sensitivity parameters to quantify how the estimated risk changes under different masking assumptions. Uncertainty in the inferred latent states is quantified via posterior path sampling and bootstrapping. Results: In 616 patients, higher parity was associated with faster movement toward lower-risk, while a family history of breast cancer was associated with longer persistence in higher-risk. Moving from ignoring masking to accounting for masking with empirically supported detectability scenarios raised the estimated breast cancer risk by 70\% for post-menopausal overweight patients and about 40\% for obese patients regardless of menopausal status. Conclusions: By attributing risk to latent progression while letting observed density operate through detectability, the framework decouples risk from detectability, quantifies how much masking can distort breast cancer risk, and provides interpretable risk characterization under irregular screening follow-up.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

43 extracted references · 27 canonical work pages

  1. [1]

    Risk Factors and Preventions of Breast Cancer.International Journal of Biological Sciences

    Sun YS, Zhao Z, Yang ZN, et al. Risk Factors and Preventions of Breast Cancer.International Journal of Biological Sciences. 2017;13(11):1387–

    2017

  2. [2]

    Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries

    Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA: A Cancer Journal for Clinicians. 2021;71(3):209–249. doi: 10.3322/caac.21660

    doi:10.3322/caac.21660 2020

  3. [3]

    GLOBOCAN 2020: Breast Cancer Data Visualization Tool

    World Health Organization, International Agency for Research on Cancer . GLOBOCAN 2020: Breast Cancer Data Visualization Tool. Available from the Global Cancer Observatory; 2022. Accessed June 7, 2025. 17

    2020

  4. [4]

    Comparison of supplemental breast cancer imaging techniques-interim results from the BRAID randomised controlled trial

    Gilbert FJ, Payne NR, Allajbeu I, et al. Comparison of supplemental breast cancer imaging techniques-interim results from the BRAID randomised controlled trial. Lancet (London, England). 2025;405(10493):1935–1944. doi: 10.1016/S0140-6736(25)00582-3

    work page doi:10.1016/s0140-6736(25)00582-3 2025

  5. [5]

    Using clinical factors and mammographic breast density to estimate breast cancer risk: development and validation of a new predictive model.Annals of Internal Medicine

    Tice JA, Cummings SR, Smith-Bindman R, Ichikawa L, Barlow WE, Kerlikowske K. Using clinical factors and mammographic breast density to estimate breast cancer risk: development and validation of a new predictive model.Annals of Internal Medicine. 2008;148(5):337–347. doi: 10.7326/0003-4819-148-5-200803040-00004

    doi:10.7326/0003-4819-148-5-200803040-00004 2008

  6. [6]

    High-throughput mammographic-density measurement: a tool for risk prediction of breast cancer

    Li J, Szekely L, Eriksson L, et al. High-throughput mammographic-density measurement: a tool for risk prediction of breast cancer. Breast Cancer Research. 2012;14(4):R114. doi: 10.1186/bcr3238

    work page doi:10.1186/bcr3238 2012

  7. [7]

    Mammographic breast density: effect on imaging and breast cancer risk.Journal of the National Comprehensive Cancer Network

    Pinsky RW, Helvie MA. Mammographic breast density: effect on imaging and breast cancer risk.Journal of the National Comprehensive Cancer Network. 2010;8(10):1157–1165. doi: 10.6004/jnccn.2010.0085

    arXiv 2010

  8. [8]

    Mammographic density and breast cancer risk: current understanding and future prospects

    Boyd NF, Martin LJ, Yaffe MJ, Minkin S. Mammographic density and breast cancer risk: current understanding and future prospects. Breast Cancer Research. 2011;13(6):223. doi: 10.1186/bcr2942

    work page doi:10.1186/bcr2942 2011

  9. [9]

    Mammographic density and the risk and detection of breast cancer.The New England Journal of Medicine

    Boyd NF, Guo H, Martin LJ, et al. Mammographic density and the risk and detection of breast cancer.The New England Journal of Medicine. 2007;356(3):227–236. doi: 10.1056/NEJMoa062790

    doi:10.1056/nejmoa062790 2007

  10. [10]

    An overview of mammographic density and its association with breast cancer

    Nazari SS, Mukherjee P. An overview of mammographic density and its association with breast cancer. Breast Cancer (Tokyo, Japan). 2018;25(3):259–267. doi: 10.1007/s12282-018-0857-5

    doi:10.1007/s12282-018-0857-5 2018

  11. [11]

    The impact of changes in breast density over time on breast cancer risk

    Kulturoglu MO, Aydin F, Sagdic MF, et al. The impact of changes in breast density over time on breast cancer risk. Scientific Reports. 2025;15(1):23900. doi: 10.1038/s41598-025-09315-1

    work page doi:10.1038/s41598-025-09315-1 2025

  12. [12]

    Comparison of the performance of screening mammography, physical examination, and breast US and evaluation of factors that influence them: an analysis of 27,825 patient evaluations

    Kolb TM, Lichy J, Newhouse JH. Comparison of the performance of screening mammography, physical examination, and breast US and evaluation of factors that influence them: an analysis of 27,825 patient evaluations. Radiology. 2002;225(1):165–175. doi: 10.1148/radiol.2251011667

    work page doi:10.1148/radiol.2251011667 2002

  13. [13]

    Derived mammographic masking measures based on simulated lesions predict the risk of interval cancer after controlling for known risk factors: a case-case analysis

    Hinton B, Ma L, Mahmoudzadeh AP, et al. Derived mammographic masking measures based on simulated lesions predict the risk of interval cancer after controlling for known risk factors: a case-case analysis. Medical Physics. 2019;46(3):1309–1316. doi: 10.1002/mp.13410

    work page doi:10.1002/mp.13410 2019

  14. [14]

    Breast density effect on the sensitivity of digital screening mammography in a UK cohort

    Payne NR, Hickman SE, Black R, Priest AN, Hudson S, Gilbert FJ. Breast density effect on the sensitivity of digital screening mammography in a UK cohort. European Radiology. 2025;35(1):177–187. doi: 10.1007/s00330-024-10951-w

    work page doi:10.1007/s00330-024-10951-w 2025

  15. [15]

    Breast Cancer in Dense Breasts: Detection Challenges and Supplemental Screening Opportunities

    Brown AL, Vijapura C, Patel M, De La Cruz A, Wahab R. Breast Cancer in Dense Breasts: Detection Challenges and Supplemental Screening Opportunities. RadioGraphics. 2023;43(10):e230024. doi: 10.1148/rg.230024

    work page doi:10.1148/rg.230024 2023

  16. [16]

    Breast density as a predictor of mammographic detection: comparison of interval- and screen- detected cancers

    Mandelson MT, Oestreicher N, Porter PL, et al. Breast density as a predictor of mammographic detection: comparison of interval- and screen- detected cancers. Journal of the National Cancer Institute. 2000;92(13):1081–1087. doi: 10.1093/jnci/92.13.1081

    work page doi:10.1093/jnci/92.13.1081 2000

  17. [17]

    Extremely dense breasts: A comprehensive review of increased cancer risk and supplementary screening methods

    Nissan N, Ochoa Albiztegui RE, Fruchtman-Brot H, et al. Extremely dense breasts: A comprehensive review of increased cancer risk and supplementary screening methods. European Journal of Radiology.2025;182:111837. doi: 10.1016/j.ejrad.2024.111837

    arXiv 2025

  18. [18]

    Longitudinal data subject to irregular observation: A review of methods with a focus on visit processes, assumptions, and study design

    Pullenayegum EM, Lim LS. Longitudinal data subject to irregular observation: A review of methods with a focus on visit processes, assumptions, and study design. Statistical Methods in Medical Research. 2016;25(6):2992–3014. doi: 10.1177/0962280214536537

    work page doi:10.1177/0962280214536537 2016

  19. [19]

    Irregular screening participation increases advanced stage breast cancer at diagnosis: A population-based study

    Ding L, Greuter MJW, Truyen I, et al. Irregular screening participation increases advanced stage breast cancer at diagnosis: A population-based study. The Breast. 2022;65:61–66. doi: 10.1016/j.breast.2022.07.004

    work page doi:10.1016/j.breast.2022.07.004 2022

  20. [20]

    Garrett RH, Patel M, Feldman BM, Pullenayegum EM. Why Recommended Visit Intervals Should Be Extracted When Conducting Longitudinal Analyses Using Electronic Health Record Data: Examining Visit Mechanism and Sensitivity to Assessment Not at Random. Statistics in Medicine. 2025;44(10-12):e70094. doi: 10.1002/sim.70094

    work page doi:10.1002/sim.70094 2025

  21. [21]

    Joint Models for Longitudinal and Time-to-Event Data: With Applications in R

    Rizopoulos D. Joint Models for Longitudinal and Time-to-Event Data: With Applications in R. Boca Raton: Chapman and Hall/CRC. 1 ed., 2012

    2012

  22. [22]

    Mixed-effects models for health care longitudinal data with an informative visiting process: A Monte Carlo simulation study

    Gasparini A, Abrams KR, Barrett JK, et al. Mixed-effects models for health care longitudinal data with an informative visiting process: A Monte Carlo simulation study. Statistica Neerlandica. 2020;74(1):5–23. doi: 10.1111/stan.12188

    work page doi:10.1111/stan.12188 2020

  23. [23]

    Doubly robust estimation, optimally truncated inverse-intensity weighting and increment-based methods for the analysis of irregularly observed longitudinal data

    Pullenayegum EM, Feldman BM. Doubly robust estimation, optimally truncated inverse-intensity weighting and increment-based methods for the analysis of irregularly observed longitudinal data. Statistics in Medicine. 2013;32(6):1054–1072. doi: 10.1002/sim.5640

    work page doi:10.1002/sim.5640 2013

  24. [24]

    Artificial intelligence in clinical research of cancers

    Shao D, Dai Y , Li N, et al. Artificial intelligence in clinical research of cancers. Briefings in Bioinformatics. 2022;23(1):bbab523. doi: 10.1093/bib/bbab523

    work page doi:10.1093/bib/bbab523 2022

  25. [25]

    Breast Cancer Prediction Using Deep Learning and Machine Learning Techniques

    Tiwari M, Bharuka R, Shah P, Lokare R. Breast Cancer Prediction Using Deep Learning and Machine Learning Techniques. 2020. SSRN Working Paper

    2020

  26. [26]

    Prediction of Breast Cancer using Machine Learning Approaches

    Rabiei R, Ayyoubzadeh SM, Sohrabei S, Esmaeili M, Atashi A. Prediction of Breast Cancer using Machine Learning Approaches. Journal of Biomedical Physics & Engineering. 2022;12(3):297–308. doi: 10.31661/jbpe.v0i0.2109-1403

    work page doi:10.31661/jbpe.v0i0.2109-1403 2022

  27. [27]

    Breast Cancer Prediction: A Comparative Study Using Machine Learning Techniques

    Islam MM, Haque MR, Iqbal H, Hasan MM, Hasan M, Kabir MN. Breast Cancer Prediction: A Comparative Study Using Machine Learning Techniques. SN Computer Science. 2020;1:290. doi: 10.1007/s42979-020-00305-w

    work page doi:10.1007/s42979-020-00305-w 2020

  28. [28]

    Breast Cancer Malignancy Prediction Using Deep Learning Neural Networks

    Prakash SS, Visakha K. Breast Cancer Malignancy Prediction Using Deep Learning Neural Networks. In: IEEE 2020; Coimbatore, India:88–92

    2020

  29. [29]

    International evaluation of an AI system for breast cancer screening.Nature

    McKinney SM, Sieniek M, Godbole V , et al. International evaluation of an AI system for breast cancer screening.Nature. 2020;577(7788):89–94. doi: 10.1038/s41586-019-1799-6

    doi:10.1038/s41586-019-1799-6 2020

  30. [30]

    Latent Markov Models for Longitudinal Data

    Bartolucci F, Farcomeni A, Pennoni F. Latent Markov Models for Longitudinal Data. Chapman and Hall/CRC. 1 ed., 2012

    2012

  31. [31]

    Mammographic density and risk of breast cancer.American Society of Clinical Oncology Educational Book

    Boyd NF. Mammographic density and risk of breast cancer.American Society of Clinical Oncology Educational Book. 2013;33(1):57–62. doi: 10.1200/EdBook_AM.2013.33.e57

    work page doi:10.1200/edbook_am.2013.33.e57 2013

  32. [32]

    Obesity and Breast Cancer Risk: The Oncogenic Implications of Metabolic Dysregulation

    Naaman SC, Shen S, Zeytinoglu M, Iyengar NM. Obesity and Breast Cancer Risk: The Oncogenic Implications of Metabolic Dysregulation. The Journal of Clinical Endocrinology & Metabolism. 2022;107(8):2154–2166. doi: 10.1210/clinem/dgac241

    work page doi:10.1210/clinem/dgac241 2022

  33. [33]

    The impact of body mass index on breast cancer incidence among women at increased risk: an observational study from the International Breast Intervention Studies

    Smith SG, Sestak I, Morris MA, et al. The impact of body mass index on breast cancer incidence among women at increased risk: an observational study from the International Breast Intervention Studies. Breast Cancer Research and Treatment.2021;188(1):215–223. doi: 10.1007/s10549-021- 06141-7

    work page doi:10.1007/s10549-021- 2021

  34. [34]

    Adult weight gain, fat distribution and mammographic density in Spanish pre- and post-menopausal women (DDM-Spain)

    Pollán M, Lope V , Miranda-García J, et al. Adult weight gain, fat distribution and mammographic density in Spanish pre- and post-menopausal women (DDM-Spain). Breast Cancer Research and Treatment.2012;134(2):823–838. doi: 10.1007/s10549-012-2108-3

    work page doi:10.1007/s10549-012-2108-3 2012

  35. [35]

    Association of the Interaction Between Mammographic Breast Density, Body Mass Index, and Menopausal Status With Breast Cancer Risk Among Korean Women.JAMA Network Open

    Tran TXM, Moon SG, Kim S, Park B. Association of the Interaction Between Mammographic Breast Density, Body Mass Index, and Menopausal Status With Breast Cancer Risk Among Korean Women.JAMA Network Open. 2021;4(12):e2139161. doi: 10.1001/jamanetworkopen.2021.39161

    arXiv 2021

  36. [36]

    The effect of weight change on changes in breast density measures over menopause in a breast cancer screening cohort

    Wanders JO, Bakker MF, Veldhuis WB, Peeters PH, Gils vCH. The effect of weight change on changes in breast density measures over menopause in a breast cancer screening cohort. Breast Cancer Research. 2015;17(1):74. doi: 10.1186/s13058-015-0583-2 18

    work page doi:10.1186/s13058-015-0583-2 2015

  37. [37]

    Determinants of Mammographic Density Change.JNCI Cancer Spectrum

    Azam S, Sjölander A, Eriksson M, Gabrielson M, Czene K, Hall P. Determinants of Mammographic Density Change.JNCI Cancer Spectrum. 2019;3(1):pkz004. doi: 10.1093/jncics/pkz004

    work page doi:10.1093/jncics/pkz004 2019

  38. [38]

    Age-related Changes in Mammographic Density and Breast Cancer Risk.American Journal of Epidemiology

    Lokate M, Stellato RK, Veldhuis WB, Peeters PHM, Gils vCH. Age-related Changes in Mammographic Density and Breast Cancer Risk.American Journal of Epidemiology. 2013;178(1):101–109. doi: 10.1093/aje/kws446

    work page doi:10.1093/aje/kws446 2013

  39. [39]

    Mammographic density in women on postmenopausal hormone replacement therapy

    Leung W, Goldberg F, Zee B, Sterns E. Mammographic density in women on postmenopausal hormone replacement therapy. Surgery. 1997;122(4):669–674. doi: 10.1016/S0039-6060(97)90072-6

    work page doi:10.1016/s0039-6060(97)90072-6 1997

  40. [40]

    LMest: An R Package for Estimating Generalized Latent Markov Models.The R Journal

    Pennoni F, Pandolfi S, Bartolucci F. LMest: An R Package for Estimating Generalized Latent Markov Models.The R Journal. 2025;16:74-101. https://doi.org/10.32614/RJ-2024-036doi: 10.32614/RJ-2024-036

    work page doi:10.32614/rj-2024-036doi: 2025

  41. [41]

    Relationships between epidemiological features and tumor characteristics of breast cancer

    Dogan L, Kalaylioglu Z, Karaman N, Ozaslan C, Atalay C, Altinok M. Relationships between epidemiological features and tumor characteristics of breast cancer. Asian Pacific Journal of Cancer Prevention.2011;12(12):3375–3380

    2011

  42. [42]

    personalized

    Amin MB, Greene FL, Edge SB, et al. The Eighth Edition AJCC Cancer Staging Manual: Continuing to build a bridge from a population-based to a more “personalized” approach to cancer staging. CA: A Cancer Journal for Clinicians. 2017;67(2):93–99. doi: 10.3322/caac.21388 19 APPENDIX Group Score statistic df Effective G Bootstrap p-value Overweight 4.910 2 100...

    work page doi:10.3322/caac.21388 2017

  43. [1397]

    doi: 10.7150/ijbs.21635

    doi:10.7150/ijbs.21635