A threshold-style decision analysis comparing four preoperative endometrial evaluation strategies before LeFort colpocleisis for detection of occult endometrial cancer.
Should women undergoing LeFort colpocleisis receive preoperative endometrial evaluation, and if so, which diagnostic strategy is most cost-effective?
LeFort colpocleisis obliterates the vaginal canal, making subsequent uterine evaluation difficult. Surgeons worry about missing occult endometrial cancer, but the prevalence is low (0.22%–2.6%). Despite weak evidence for routine screening, 68% of colpocleisis surgeons report performing preoperative endometrial evaluation.
- No routine testing — proceed to surgery without endometrial evaluation
- Selective transvaginal ultrasound — screen higher-risk women; abnormal results trigger office biopsy
- Selective office Pipelle biopsy — direct tissue sampling in higher-risk women; inadequate samples may trigger further workup
- Selective concurrent dilation and curettage — tissue sampling performed during the colpocleisis itself (shares anesthetic); results return postoperatively so they cannot change the surgical plan
- At base-case cancer prevalence (0.56%), no testing is preferred — testing costs far exceed the benefit of early detection
- Only no testing and transvaginal ultrasound survive on the efficiency frontier; Pipelle and concurrent D&C are dominated (higher cost, fewer QALYs)
- Transvaginal ultrasound becomes the preferred strategy when high-risk prevalence exceeds approximately 0.8%, achievable in women with postmenopausal bleeding, obesity, tamoxifen use, or Lynch syndrome
- Results are most sensitive to cancer prevalence and delayed diagnosis cost
colpocleisis_costeff/
├── colpocleisis_selective_testing_model.R # Main model function
├── run_example.R # Quick-start script (runs base case, prints results)
├── generate_figures.R # Generates all 3 publication figures
├── manuscript.txt # Abstract text (423 words) + figure legends
├── output/ # Generated figures and CSV tables (gitignored)
│ ├── figure1_ce_plane.jpeg
│ ├── figure2_tornado.jpeg
│ ├── figure3_threshold.jpeg
│ └── *.csv (strategy and frontier tables)
└── .gitignore
R (tested on 4.4.x) with these packages:
install.packages(c("tibble", "dplyr", "purrr", "rlang", "readr", "ggplot2", "scales", "forcats", "tidyr"))source("colpocleisis_selective_testing_model.R")
strategy_bundle <- run_colpocleisis_selective_testing_model()
strategy_bundle$strategy_table
strategy_bundle$frontier_table
strategy_bundle$summary_sentencestrategy_bundle <- run_colpocleisis_selective_testing_model(save_csv = TRUE, output_dir = "output")Rscript generate_figures.RProduces three JPEG files (300 DPI) in output/.
Rscript run_example.R # model + CSV tables
Rscript generate_figures.R # all 3 figuresAll defaults are literature-based. Key sources:
| Parameter | Default | Source |
|---|---|---|
| Occult cancer prevalence (high-risk) | 0.56% | 2025 prolapse hysterectomy cohort |
| Occult cancer prevalence (low-risk) | 0.22% | 2021 meta-analysis |
| High-risk fraction tested | 30% | Modeled from practice pattern surveys |
| TVUS sensitivity / specificity | 94.1% / 66.8% | Postmenopausal bleeding diagnostic review (4mm threshold) |
| Pipelle sensitivity / specificity | 100% / 98% | Conditional on adequate sample |
| Pipelle inadequate sample rate | 29.1% | 2025 office biopsy cohort |
| D&C sensitivity / specificity | 88% / 98.4% | 2023 systematic review/meta-analysis |
| TVUS cost | $125.23 | 2022 CMS nonfacility estimate |
| Pipelle cost | $172.55 | 2022 prolapse preop evaluation study |
| Concurrent D&C incremental cost | $800 | Marginal OR cost estimate |
| D&C effective detection credit | 50% | Modeled (results return post-procedure) |
| Delayed cancer diagnosis cost | $20,000 | Anchored to 90-day endometrial cancer costs |
| QALY gain per early detection | 0.10 | Conservative estimate |
| Willingness-to-pay threshold | $100,000/QALY | Standard US threshold |
To override any parameter:
run_colpocleisis_selective_testing_model(
high_risk_prevalence = 0.02,
delayed_cancer_cost = 50000
)- Figure 1 — Cost-effectiveness plane with efficiency frontier
- Figure 2 — One-way sensitivity tornado diagram with parameter value labels
- Figure 3 — Threshold analysis showing strategy preference across cancer prevalence (0.1%–5%)
- Efficiency frontier with strong and extended dominance removal
- Sequential ICERs on the frontier
- Net monetary benefit ranking
- Pipelle inadequate-sample follow-up pathway (with recovery detection)
- Concurrent D&C modeled with partial QALY credit (results return after surgery)
- Input validation including cross-parameter consistency checks
This is a threshold-style expected-value model for abstract development and rapid scenario testing. It is not a full Markov model, microsimulation, or lifetime cost-effectiveness analysis. Test accuracy parameters are drawn from postmenopausal bleeding literature, not specifically from asymptomatic colpocleisis candidates.
Manuscript (423 words) is in manuscript.txt with figure legends appended.
The decision-model architecture in this repository — single self-contained R file, explicit parameter defaults, efficiency-frontier analysis with strong and extended dominance removal, one-way sensitivity, and threshold analysis — is patterned after the companion code published with:
Larose TL, Meheus F, Brennan P, Johansson M, Robbins HA. Assessment of biomarker testing for lung cancer screening eligibility. JAMA Netw Open. 2020;3(3):e200409. doi:10.1001/jamanetworkopen.2020.0409.
Companion code: hilaryrobbins/costeff_lung_biom_public
The clinical domain is unrelated (lung cancer screening), but the analytic structure is what we borrowed. Manuscripts based on this repository should cite Larose et al. 2020 in the Methods section when describing the cost-effectiveness framework.
Tyler Muffly