Economic benefit of expanding mammography screening for breast cancer in Colombia: A cost modelling analysis

Abstract

Objectives

Breast cancer (BC) remains a leading cause of mortality, especially when diagnosed at advanced or metastatic stages. Early detection improves survival and reduces treatment costs. This study estimates the economic implications of expanding mammography screening coverage on breast cancer care in Colombia by modelling stage redistribution and associated direct medical treatment costs.

Study design

This is a cost-benefit study to evaluate the impact of breast cancer screening strategies on the cost of care in Colombia.

Methods

A 5-year Markov model estimated costs by cancer phenotype and clinical stage, utilising healthcare resource utilization data from a comprehensive Colombian cancer center. We compared current costs (31% mammography coverage) with projected costs, assuming a 70% screening rate (using the Netherlands' distribution as a reference).

Results

Data from 284 Colombian BC patients were included: Luminal (59%), HER2+ (26%), and Triple-Negative BC (15%). First-year costs varied widely, ranging from early-stage Luminal (USD 10,309) to metastatic HER2+ (USD 60,612). Advanced BC significantly drove healthcare costs. By increasing screening coverage to 70%, the number of early-stage cases detected each year would increase by 56% compared to the current 33%. Increasing screening could shift diagnoses earlier, reducing the average 5-year cost per patient from $68,183 to $58,542. This would translate to a USD 164 million saving for Colombia.

Conclusions

Increasing screening coverage improves patient outcomes by detecting more early-stage cases and significantly reduces healthcare expenditures by decreasing costly advanced-stage diagnoses, yielding substantial savings for the Colombian health system.

1. Introduction

Breast cancer (BC) remains the leading cause of cancer-related deaths among women worldwide and the most frequent cancer in 157 out of 185 countries [1]. In 2022, 1.9 million new BC cases and 495,000 BC-related deaths were reported globally [2]. In Colombia, BC is the leading cause of cancer mortality in women, with an age-standardized incidence rate of 50.7 and 4752 deaths in 2022 [2].

Early diagnosis is crucial to improving survival and reducing treatment-related costs. When detected at a localized stage, 5-year survival reaches 99% [3]. In the United States, mortality decreased by 58% between 1979 and 2019, with 25% of this reduction attributed to screening [4]. Likewise, BC mortality in high-income countries is 17% lower than in low- and middle-income countries (LMICs), mainly explained by stronger healthcare systems and well-organized screening programs [5]. In contrast, 30–75% of LMIC BC cases are diagnosed at advanced stages, increasing mortality and treatment costs [5,6]. In Colombia, 5-year survival remains among the lowest in the region [7]; yet improves among Colombian migrants in the US, suggesting that access to timely mammography is linked to survival outcomes [8].

The World Health Organization (WHO) recommends population-based mammography screening with 70% coverage [6] to achieve a 2.5% annual reduction in BC mortality, saving 2.5 million lives by 2040 [9]. Aligned with this framework, Colombia's Ten-Year Public Health Plan (2022-2031) and the Cancer Shock Plan aim to reach 70% mammography coverage to reduce inequities detect 60% of cases early, confirm diagnosis within 60 days, and ensure continuous treatment for 80% of patients [10,11]. However, only 17% of women diagnosed with BC report a prior mammogram, indicating low adherence to guidelines [12].

Early detection improves survival and generates savings for healthcare systems, with evidence consistently showing that treating early-stage BC is substantially less costly than advanced disease [13,14]. Although prior studies have described the economic burden of breast cancer in Colombia and other Latin American countries, most have reported aggregate cost estimates without incorporating tumour phenotype, stage-specific progression, or screening-induced stage redistribution. Furthermore, no published analysis has modelled the economic implications of achieving the WHO-recommended 70% mammography coverage target within the Colombian health system.

Understanding the costs of early detection is essential to inform investment decisions [6]. This study addresses this gap by combining real-world, phenotype-specific cost data with scenario-based stage redistribution modelling to assess the economic impact of increasing screening coverage.

2. Methods

2.1. Study design

A 5-year Markov model was developed to estimate the direct medical cost of BC management in Colombia, stratified by clinical stage and tumour phenotype. The model uses healthcare resource utilization (HCRU) data from the Centro de Tratamiento e Investigación sobre Cáncer (CTIC), a comprehensive cancer center. Current BC stage distribution, based on national reports by the High-Cost Account (Cuenta de Alto Costo–CAC) [15], was compared with a scenario achieving higher screening rates, using a reference country.

The model was stratified independently by phenotype and stage: Luminal BC, including A and B: early-stage (I, IIA), locally advanced (IIB, III), metastatic (IV); HER2+: early-stage (I), locally advanced (II, III), metastatic (IV); TNBC: non-metastatic (I, II, III), metastatic (IV).

2.2. Markov model structure

A Markov modelling approach was selected because breast cancer progression involves transitions between clinically distinct and mutually exclusive health states over time, each associated with different mortality risks and healthcare costs [16]. This framework allows estimation of cumulative economic burden while accounting for time-dependent disease progression and stage-specific transitions.

The model simulated disease progression and costs of breast cancer through four mutually exclusive health states:

• 1.B₀: Alive with no relapse -patients with stages I–III disease who have completed initial treatment and remain in remission.
• 2.B_m: Alive with metastatic disease-patients with active stage IV breast cancer receiving systemic therapy and ongoing management. This state includes patients initially diagnosed with metastatic disease or those progressing from earlier stages.
• 3.B_p: Post-metastatic stage-patients who experience further progression following metastatic disease or require palliative-oriented management.
• 4.B_d: Death.

$$B0\to Bm\to Bp\to Bd$$

B₀ may transition to B_m or remain in that state. B_m could transition to B_p or stay in that state, and all states could transition to B_d. B_p and B_d were considered absorptive states, and reverse transitions were not allowed. Annual cycles were applied over a 5-year horizon with half-cycle correction. Each phenotype-stage subgroup was modelled independently.

2.3. Transition probabilities

State- and phenotype-specific transition probabilities were derived from literature and categorized by phenotype, stage, and year of follow-up (Supplementary Table S1). A directed literature search prioritised Colombian real-world survival data. When local estimates were unavailable, regional Latin American evidence was used. In the absence of national or regional data, transition probabilities were derived from pivotal clinical trials or large international observational studies to ensure robust phenotype-specific progression estimates. Unmodified transition probabilities Kaplan–Meier (KM) survival curves were digitized to extract event-free and overall survival at years 1–5. These unmodified probabilities were converted into annualised progression and mortality rates applying a non-parametric method based on the following formula:

$$P\left(t\right)=\frac{1-S\left(t\right)}{S\left(t-1\right)}$$

Where:

• •P(t) is the probability of transitioning (e.g., to progression) in year t
• •S(t) is the KM survival probability at time t
• •S (t−1) is the survival probability at time t−1

This method assumes a constant hazard within each interval, and is applied in health economic evaluations to derive discrete-time transition probabilities from published KM curves when individual patient data are unavailable [16]. These resulting time-varying probabilities were used to populate the Markov transition matrices by phenotype and stage.

2.4. Healthcare resource utilization (HCRU) and costs

HCRU data were extracted from CTIC's electronic medical record software (TASY®), identifying the frequency of procedures and medications used by BC patients. The HCRU analysis was based on a retrospective cohort of breast cancer patients treated at CTIC between 2022 and 2024. All consecutive patients with a breast cancer diagnosis and complete electronic medical records available during the study period were included. No probabilistic sampling technique was applied; the cohort represents the full cohort institutional census of eligible cases during the study period.

Patient-level HCRU data were directly observed for Years 1 and 2 of follow-up. For Year 2, as some patients had incomplete follow-up at the time of data extraction, utilization rates were calculated using monthly averages among patients who remained under active follow-up to avoid underestimation of resource use.

For Years 3–5, where long-term follow-up data were insufficient in the institutional database, HCRU inputs were extrapolated from observed patterns in earlier years and validated through structured expert review with an oncology panel at CTIC to ensure consistency with expected phenotype- and stage-specific care trajectories. These estimates primarily reflect follow-up care and maintenance therapy rather than intensive treatment phases, which are concentrated in the first two years after diagnosis.

Expected follow-up care patterns included:

• •Early and locally advanced HER2+: quarterly oncology consultations and annual imaging (one ultrasound and bilateral mammography).
• •Metastatic HER2+: Year 2 management costs were carried forward.
• •Early luminal: quarterly oncology visits, annual ultrasound and bilateral mammography, cholesterol monitoring, bone densitometry (osteodensitometry), and ongoing letrozole therapy.
• •Locally advanced luminal: Year 3 costs matched Year 2; Years 4-5 included quarterly oncology consultations, annual ultrasound, bilateral mammography, cholesterol control, bone densitometry, and ongoing letrozole therapy.
• •Metastatic luminal: Year 2 management costs were carried forward.
• •Non-metastatic TNBC: quarterly oncology consultations, annual ultrasound, and bilateral mammography.
• •Metastatic TNBC: Year 2 management costs were carried forward.

For the post-metastatic stage (relapse after metastasis), the palliative care costs from Gamboa et al. [13] updated to 2024 were used. The death state was valued at half of palliative care costs [13].

Cost data for 2024 were obtained from publicly available sources: Individual Registry of Healthcare Services (Registro Individual de Prestación de Servicios de Salud-RIPS), Capitation Payment Unit Sufficiency database (Estudio de suficiencia de Unidad de pago por capitación-SUF), mandatory traffic accident insurance tariff manual (Seguro Obligatorio de Accidentes de Tránsito-SOAT), and national price medication tracking system (Sistema de Información de Precios de Medicamento-SISMED). RIPS captured service utilization, while SUF, SOAT, and SISMED provided standardized tariffs for procedures, medical services, and medications, respectively.

Each health state was assigned an annual cost varying by cycle and subgroup. The model included direct medical costs: treatment, hospitalization, and management. A 5% discount rate was applied to costs, following the 2014 Colombian Institute for Health Technology Assessment (IETS) guidelines.

The total annual cost was calculated as:

$${Cost}_t=\sum _{i=1}^4\left({StateProportion}_{i,t}X{C}_{i,t}\right)$$

Where:

State Proportion _i,t, represents the proportion of patients in each health state in year t, and C_i,t, is the corresponding cost. Aggregated 5-year costs were estimated per patient.

2.5. Mammography coverage and screening projections

Five-year costs for each phenotype and stage were projected to the current distribution of cases (Table 2) [15], using the 2022 mammography coverage of 31% (CAC) [17]. To simulate the impact of increased coverage on the distribution of disease stages by achieving the national coverage goal [[10], [11]], [[10], [11]] a 70% screening scenario was modelled. The Netherlands was used as the primary reference, due to its high mammography coverage (74%) and well-documented BC stage distribution at diagnosis [18]. The Netherlands stage distribution was adjusted for Colombia's phenotype distribution to estimate the shift to early-stage disease. An additional scenario using Sweden (>80% coverage) was also conducted (Supplementary Table S2).

Table 2.

Cost per patient by type of service (USD) (Year 1).

Phenotype/Service	Early HER2+	LA HER2 +	Metastatic HER2+	Early Luminal	LA Luminal	Metastatic Luminal	No Metastatic TNBC	Metastatic TNBC
Medical Attention	348	774	2136	312	590	1606	448	615
Transfusion Services	-	265	182	8	45	101	42	396
Diagnostic Services	307	1031	1826	1238	861	1472	1053	2556
Hospital Stay	1010	2512	5782	632	1458	5008	1123	7766
Laboratory Diagnostics	193	754	996	475	940	1026	1291	1815
Pharmacological Therapies (Medications)	15,061	32,110	38,376	3772	7260	15,811	28,800	33,118
Other Treatments	3457	5028	10,312	3039	3855	5230	4104	9880
Surgical Procedures	1602	1764	1003	832	1889	1385	1428	2052
Total Cost per Phenotype	21,979	44,239	60,612	10,309	16,898	31,638	38,289	58,198

LA: Locally advanced; TNBC: triple-negative breast cancer.

Weighted 5-year costs under both distributions were multiplied by the number of incident cases of BC reported by Globocan for 2022 [2] to estimate the national cost impact of increasing screening coverage.

Importantly, the screening scenarios did not assume an increase in overall breast cancer incidence. The total number of incident cases was held constant based on Globocan 2022 estimates, and cases were redistributed across stages to reflect earlier detection. Thus, the model estimates the economic implications of stage redistribution rather than additional cancer case detection.

Screening programme costs were estimated for women aged 50–69 years, assuming biennial mammography, ultrasound for 4% of abnormal cases, and biopsy with immunohistochemistry for 3% [[19], [20], [21], [22]].

2.6. Sensitivity analysis

One-way sensitivity analyses assessed the robustness of the model by varying: 1) discount rate (3%-7%), 2) medication costs (±10%) to reflect possible drug price variations.

3. Results

3.1. Cohort characterization

A total of 284 patients with BC were included. Luminal phenotypes was the most frequent (n = 167; 59%), followed by HER2+ (n = 74; 26%) and TNBC (n = 43; 15%) (Table 1). The mean treatment duration ranged from 14.4 to 18.6 months, by phenotype and stage.

Table 1.

Cohort characterization of the study population.

Variable	Frequency	Percentage (%)	Mean treatment duration (Months)
Phenotype and stage (n = 283)
HER2 +
Early HER2+	5	1.76%	14.56
Locally Advanced HER2+	59	20.8%	15.56
Metastatic HER2+	10	3.53%	16.65
Luminal
Early Luminal	91	32.15%	16.46
Locally Advanced Luminal	57	20.14%	15.20
Metastatic Luminal	19	6.71%	18.63
Triple Negative Breast Cancer
No Metastatic TNBC	40	14.3%	14.34
Metastatic TNBC	3	1.06%	15.28

TNBC: triple-negative breast cancer.

3.2. Health care resource utilization

Year 1 HCRU are presented in Fig. 1. Results showed that metastatic HER2+ patients had the highest resource use (923 services per patient), followed by metastatic TNBC (744). Medications and other treatments (chemotherapy, targeted therapy, radiotherapy) accounted for over 65% of total use in advanced stages. Overall, metastatic cases consumed around three times more services than early-stage cases for HER2+ and luminal phenotypes, and 1.8 more for TNBC.

Fig. 1.

Average frequency of services per patient by phenotype and disease stage (Year 1).

Table 2 presents first-year costs per patient by service. Among HER2+ disease, costs ranged from USD 21,979 for early-stage disease to USD 44,239 for locally advanced and USD 60,612 for metastatic disease. In Luminal phenotype, first-year costs were USD 10,309, USD 16,898, and USD 31,638 for early, locally advanced, and metastatic stages, respectively. For TNBC, the estimated cost was USD 38,289 in non-metastatic and USD 58,198 in metastatic disease (Table 2).

3.3. Modelled costs per patient

Fig. 2 presents the 5-year modelled cost per patient, adjusted by phenotype, stage, and transition probabilities. Early luminal tumours had the lowest cost per patient (USD 35,760), followed by early HER2+ (USD 31,961) and non-metastatic TNBC (USD 60,205). Costs increased significantly in advanced diseases, particularly in metastatic HER2+ (USD 156,364) (Fig. 2). For luminal and TNBC phenotypes, metastatic treatment was approximately double that of early-stage disease, whereas for HER2+, management costs were three to five times higher.

Fig. 2.

5-year modelled cost per patient in each phenotype per stage.

As shown in Fig. 3, costs were concentrated in the first two years, reflecting initial intensive treatment. The average per-patient costs during this period were USD 19,493 (early stages), USD 30,931 (locally advanced), and USD 38,205 (metastatic). From year 3, costs declined steeply in early-stage but remained high in advanced disease due to ongoing treatment.

Fig. 3.

Cost of breast cancer per patient by phenotype and stage over 5 years.

3.4. Impact of increased screening coverage

Based on national data (CAC 2022) [15], locally advanced and metastatic cases accounted for more than 67% of the disease stage distribution (Table 3). Luminal locally advanced disease represents the largest subgroup (46.4%), followed by non-metastatic TNBC (17.8%) and locally advanced HER2+ (13.3%).

Table 3.

5-Year breast cancer cost by phenotype and stage per patient (USD).

Phenotype/Stage	5-year cost (modelled)	Current distribution (%)	5-year current distribution-weighted cost	Distribution of 70% screening (%)	5-year 70% screening distribution-weighted cost
Early HER2+	31,961	3.3	1055	8.4	2684
Locally advanced HER2+	96,228	13.3	12,798	8.6	8275
Metastatic HER2+	156,364	1.4	2189	1.0	1564
Early luminal	35,760	11.5	4112	29.2	10,441
Locally advanced luminal	67,913	46.4	31,512	29.9	20,304
Metastatic Luminal	80,278	4.8	3853	3.6	2890
Non-metastatic TNCB	60,205	17.8	10,716	18.2	10,956
Metastatic TNBC	129,844	1.5	1948	1.1	1428
All phenotypes	-	100	68,183	100	58,542

TNBC: triple-negative breast cancer.

Increasing screening coverage to 70%, generates a substantial shift in the stage at diagnosis, notably decreasing advanced and metastatic cases across phenotypes. Early-stage cases would increase from 33% to 56%. Early-stage luminal cases would increase to 29.2% (+17.7%), while locally advanced and metastatic luminal cases are expected to decrease to 29.9% (−16.7%) and 3.6% (−1.2%), respectively. Similarly, HER2+ early-stage cases would increase to 8.4% (+5.1%), accompanied by a decrease in locally advanced cases from to 8.6% (−4.7%) and metastatic HER2+ cases to 1.0% (−0.4%). Among patients with TNBC group, non-metastatic cases would have a slight increase from to 18.2% (+0.4 %), while metastatic TNBC cases are expected to decrease to 1.1% (−0.4%).

Increasing breast cancer screening coverage to 70% and achiving a higher redistribution of cases toward earlier stages results in economic implications. With the current distribution, the average 5-year treatment cost per patient was USD 68,183. With 70% screening, costs drop to $58,542 (∼$10,000 less per patient). Extrapolated to all 17,018 new BC cases in Colombia (2022) this represents potential savings of USD 164 million over five years (from USD 996 million to USD 832 million).

Using Sweden's >80% coverage as reference yielded similar results (Supplementary Table S2). The proportion of early cases would increase to 55.1%, with a weighted cost per patient of USD 58,153, leading to USD 170 million savings in the 5-year treatment of incident cases in Colombia (2022).

3.5. Screening program costs

Increasing screening coverage from 31% to 70% would generate additional costs of USD 28,633,578, for mammograms, ultrasounds, and biopsies in women aged 50-69 years. After accounting for these expenses, net savings remained USD 135 million.

3.6. Sensitivity analyses

Results for sensitivity analyses are presented in Supplementary Table S3. Applying a 3% discount rate resulted in greater overall savings, exceeding USD 400 million, driven by lower per-patient costs in locally advanced stages, which represent the majority of the population. Adjusting drug prices by ±10% for possible negotiated price changes, proportionally changed total costs; however, in both scenarios, increasing screening coverage resulted in cost savings (USD 151-177 million).

4. Discussion

Our study modelled the economic impact of increasing mammography screening in Colombia from 31% to the national target of 70%, estimating potential savings of USD 164 million over five years. Higher coverage would increase early-stage breast cancer (BC) diagnoses from 33% to 56%, shifting treatment away from higher-cost advanced-stage care. Consistent with global and regional evidence, late-stage breast cancer incurs significantly higher costs, especially for HER2+ and TNBC. In our model, metastatic HER2+ costs were three to five times higher than early-stage cases, reflecting longer and more intensive systemic therapy. Costs were concentrated in the first two years, then declined, especially for early-stage disease.

Global comparisons reinforce our results. In the Netherlands, population-based screening reduced in late-stage diagnoses by 20–30% decreasing mortality. And was cost-effective over a 10-year horizon through avoided costs [23], [24]. Similar findings from the United States and Europe, report that early-stage BC is not only associated with better prognosis but also significantly lower treatment expenditures [14].

This analysis also reinforces evidence from the Latin American (LATAM) region. Studies from Colombia, Mexico, and Brazil have highlighted that late-stage BC care absorbs a disproportionate share of oncology budgets and that lack of early detection contributes to both higher costs and worse survival [13,14,25]. However, most LATAM analysis lacked phenotype-level costing or linkage to specific detection strategies and real-world treatment patterns [13,25]. By integrating phenotype-level costing with stage redistribution modelling based on real-world resource utilization, our study provides a more granular assessment of the economic implications of earlier detection in the Colombian context.

Similarly, in Latin America, mammography coverage remains heterogeneous and generally below 50%, with a substantial proportion of cases diagnosed at advanced stages. Regional evidence from Colombia, Mexico, Brazil, and Peru highlights persistent structural barriers to early detection, reinforcing the relevance of modelling screening expansion within this context. However, empirical evidence on stage redistribution following large-scale screening expansion is limited in the region. Therefore, evidence from other settings was used as an external reference to inform the expected direction of stage shift.

Unlike top-down costing methods or international transfer estimates, our model leverages patient-level resource use and stage-specific treatment durations, strengthening the external validity. Particularly for HER2+ disease, where targeted therapy costs are substantial, our phenotype stratification aligns with Brazilian findings [26]. Furthermore, our model contrasts with uniform-treatment approaches in other global models, underscoring substantial cost differences by phenotype and stage. Metastatic HER2+ remains more costly, consistent with real-world studies from the US, Canada, and Europe [27,28].

The leveraged Dutch stage distribution data (74% coverage) to model screening expansion aligns with national health policy targets, Ten-Year Public Health Plan 2022-2031 and the Cancer Shock plan [10,11] Although cross-country extrapolation introduces uncertainty, adjusting Dutch data to Colombia’’s phenotype distribution helped mitigate bias. Comparable UK models projecting biennial screening (50–70 years) estimated annual savings of £140 million through earlier-stage detection [29]. While direct comparisons are limited by healthcare financing and pricing differences, the directionality and magnitude of impact reinforce the policy relevance of our findings.

Beyond economic considerations, expanding screening coverage has important implications for equity in cancer care. Structural barriers within Colombia's health system, including healthcare fragmentation, geographic disparities, socioeconomic inequalities, and differences across insurance regimes, limit timely BC care. Analyses from the High-cost account (CAC) report that women in rural areas and subsidised regimes are more often diagnosed at advanced stages and face higher mortality [15]. Organized screening can help reduce these disparities by improving early detection and care continuity, making it both a cost-saving strategy and equity-promoting tool. Reaching 70% coverage, however, will require operational investments in mobile mammography, referral systems, and human resources. Experiences from other LMIC from the region, such as Mexico and Panama, demonstrate that infrastructure and service integration are crucial for ensuring timely care after screening [6]. Sensitivity analyses confirm the robustness of our findings. Outputs remained stable under variations in discount rates and variations in systemic therapy costs. These results suggest that the estimated cost savings from earlier detection are not overly sensitive to modest variations in unit costs or discounting assumptions.

Importantly, our findings align with WHO's Global Breast Cancer Initiative, which highlights early detection as a critical strategy to reducing mortality [30]. WHO modeling suggests that scaling early diagnosis could reduce breast cancer mortality by approximately 2.5% annually, particularly in LMICs where late-stage presentation predominates [27].

Together, these findings affirm that earlier detection is both clinically and economically advantageous. Implementing population-wide screening can reduce advanced-stage diagnoses, lower healthcare costs, and potentially improve equity in cancer care. As only direct medical costs were included, additional societal benefits, including productivity gains and reduced financial hardship, likely amplify the value proposition. Moreover, our methodology is transferable to other upper-middle-income countries facing similar challenges in screening coverage. By combining national data with international benchmarks, this model supports dynamic, context-specific planning that remains relevant across diverse settings in LATAM and other regions.

Some limitations warrant mention. First, transition probabilities were derived from published Kaplan-Meier curves digitized by phenotype and disease stage, a validated but an approximate method which may introduce uncertainty. A hierarchical evidence approach was applied, prioritising Colombian real-world data and regional Latin American estimates when available; however, such data remain limited, and several inputs were derived from international studies. Differences in healthcare access and treatment patterns across settings may influence absolute survival estimates; however, the model's primary conclusions are driven by relative cost differences across stages rather than precise survival levels.

Second, HCRU data were derived from a single high-complexity oncology centre. The cohort reflects consecutive patients treated at this institution rather than a nationally representative sample. Patients treated in tertiary referral settings may require more intensive treatment and follow-up, which could lead to a slight overestimation of absolute treatment costs, particularly for advanced-stage disease. However, unit costs were obtained from national public tariff sources rather than institution-specific cost data, and model results are driven primarily by the relative cost differences between early and advanced stages. Therefore, the projected economic impact is expected to remain broadly consistent across healthcare settings.

Addionally, long-term utilization beyond Year 2 was not directly observed due to limited follow-up duration in the institutional database; for Years 3–5, utilization patterns were extrapolated from earlier observed data and validated through an expert panel clinical review. While this approach introduces uncertainty in later-year cost estimates, most breast cancer–related expenditures occur during the initial years following diagnosis, when active treatment is concentrated and resource use was directly observed in this study. Later years are primarily characterised by follow-up care and maintenance therapy, which generally involve lower and more stable resource utilization. Therefore, although extrapolation may affect absolute long-term cost estimates, it is unlikely to materially change the overall cost differences between early and advanced disease reported in the model.

Another limitation is that quality-adjusted life years and non-medical costs were not modelled; therefore results reflect only direct medical costs and do not constitute a full cost-utility or formal budget impact evaluation. While treatment costs were carefully estimated from national tariffs and real-world data, we did not include capital costs (e.g., infrastructure investment for mammography expansion). Nevertheless, when adjusting for the additional costs of mammograms, ultrasounds, and biopsies to reach the 70% goal, the results continued to show a savings of more than $130 million, related to fewer advanced-stage and metastatic cases. Conversely, broader societal benefits such as productivity gains, out-of-pocket costs, and improvements in quality of life were not captured, potentially underestimating the total value of earlier detection. Health status has been shown to be closely associated with labour market participation and productivity outcomes, highlighting the potential relevance of these broader societal effects in economic evaluations [31]. Future research could extend the analysis by incorporating productivity impacts and quality-of-life outcomes to provide a more comprehensive assessment of the value of earlier cancer detection.

The model adopted a 5-year time horizon, reflecting evidence that breast cancer costs are concentrated in the initial years following diagnosis, particularly during active treatment. However, survival beyond five years is increasingly common with modern therapies. While advanced-stage disease typically entails sustained systemic therapy and monitoring, extending the horizon may alter cumulative cost estimates. Importantly, because long-term management costs remain higher for advanced disease, extending the time horizon would likely increase cumulative costs in this group rather than reduce the stage-related cost gap. Lifetime modelling could provide additional insight into long-term economic implications.

Additionally, the 70% screening scenario was informed by high-coverage international benchmarks. The Netherlands was selected as the primary comparator due to its 74% screening coverage and well-documented stage distribution at diagnosis. Alternative high-coverage scenarios (e.g., Sweden) yielded similar directional findings, however, extrapolation across health systems introduces structural uncertainty.

OECD data show that mammography coverage in Sweden approached 95% in 2019, while most Latin American countries remained below 50% [32]. In the absence of regional high coverage comparators, with documented stage redistribution, European screening systems were used as policy aligned benchmarks. Weighting stage distribution using Colombia's current phenotype epidemiology reduced overestimation risk. However, future local data on stage shifts under expanded screening would further refine estimates.

Finally, national savings projections were based on GLOBOCAN 2022 incidence estimates, which incorporate Colombian population-based cancer registry data and international modelling methods. Variations in national incidence would proportionally affect the magnitude, but not the direction, of projected savings. The screening scenarios held total breast cancer incidence constant, redistributing existing cases across stages to isolate the economic effect of stage shift. In real-world implementation, expanded screening may lead to transient increases in observed incidence due to earlier detection and some degree of overdiagnosis. This dynamic was not explicitly modelled and may influence absolute cost projections. However, given Colombia's current low screening coverage and high proportion of advanced-stage presentation, stage redistribution is expected to represent the dominant short-to medium-term effect. Uncertainty exploration was limited to deterministic sensitivity analyses, and formal probabilistic sensitivity analysis was not conducted due to the scenario-based nature of key structural inputs.

4.1. Conclusions

Expanding mammography screening coverage to 70% in Colombia could markedly increase early-stage breast cancer detection and shift care away from more resource-intensive advanced-stage treatment. These findings reinforce the national case for scaling up screening programs and offers a framework for other middle-income countries facing similar challenges in early cancer detection.

As health systems worldwide transition toward value-based care, investments in early detection represent a high-impact strategy to reduce disease burden, improve outcomes, and optimize resource use. This analysis highlights the importance of incorporating real-world, locally sourced data into health economic evidence to inform national policies that achieve equitable, efficient, and sustainable oncology care and resource allocation.

Ethics statement

This study used anonymized secondary data sources and did not involve human subjects; therefore, ethics committee approval was not required.

Authors’ contributions

DFS, JS, LCP, MDP, and WAM conceived and designed the study. Data collection, as well as data curation and the development of analysis tools, were conducted by all authors. AMR, DFS, JS, LCP, and MDP conducted the formal analyses and statistical analyses. The original draft of the manuscript was prepared by AMR, DFS, JS, LCP, MDP, and WAM. All authors contributed to the review and editing of the manuscript and provided critical feedback. Resources, including study materials and patient data, were provided by all authors. Project administration and technical or logistic support were conducted by AMR, DFS, JS, LCP, MDP, and WAM. WAM provided supervision of the project.

Data sharing statement

All data relevant to the study and all data sources are included in the article or uploaded as supplementary information.

Key objective

What is the economic impact of expanding mammography screening coverage in Colombia, and how does it vary by breast cancer phenotype and stage at diagnosis?

Knowledge generated

Using real-world data from a Colombian cancer center and an economic model, increasing screening coverage from 31% to 70% shifted 23% of cases toward early stages. This expansion generated estimated savings of USD 164 million over five years, as late-stage treatment costs were three to five times higher than early-stage care.

Funding

This study was conducted as part of a collaborative research agreement between CTIC and Roche Colombia. No funding was received.

Declaration of competing interest

AFC reports grants or contracts, payments, honoraria, support for attending meetings, and advisory roles with Merck Sharp & Dohme, Boehringer Ingelheim, Roche, Bristol-Myers Squibb, Foundation Medicine, Roche Diagnostics, Thermo Fisher, Broad Institute, Amgen, Flatiron Health, Teva Pharma, Rochem Biocare, Bayer, INQBox, EISAI, Merck Serono, Jannsen Pharmaceutical, Pfizer, Novartis, Celldex Therapeutics, Eli Lilly, Guardant Health, and Illumina JSZ reports employment at Roche in a co-creation project without direct payments. MAB reports honoraria, advisory roles, and travel support from MSD, GSK, Adium, Novartis, Pfizer, Roche, Knight, AstraZeneca, Bristol-Myers Squibb, Ipsen, and participation in ACHO melanoma training at Johns Hopkins. SXF reports honoraria, travel support, and advisory participation with Novartis, MSD, Eli Lilly, AstraZeneca, Roche, and Pfizer. SQ reports speaker honoraria from MSD and travel support from Eli Lilly. WAM reports research support, consulting, and honoraria from AstraZeneca, Productos Roche, MSD, Pfizer, Novartis, and Eli Lilly. AR, DS, LCP, JS, and MDP are employees of Roche Colombia. AMO, JEC, NSP, and SCB declare no conflicts of interest. None of the authors received any compensation for the authorship of this manuscript.

Appendix A. Supplementary data

The following are the Supplementary data to this article: