Salivary Gland Cancer Model for Survival ± PORT (Advanced Disease)

Learn the SEER-based Cox model for advanced salivary gland malignancy: OS and CSS with surgery and radiotherapy (PORT), staging scope, and how to interpret adjusted hazard and registry radiotherapy data.

Advanced salivary gland cancer — survival model ± radiotherapy

SEER-based Cox model (Ma et al., 2025) for AJCC 6th stage III/IV disease. Enter demographics, SEER historic stage, AJCC T/N/M, histology, surgery, and radiotherapy (includes postoperative RT when recorded).

Outcome *

Overall survival (OS)Cancer-specific survival (CSS)

Age group *

Sex *

Race *

Tumor category (OS model) *

Histologic type *

SEER historic stage A *

AJCC 6th T *

AJCC 6th N *

AJCC 6th M *

Surgery *

Not performed / referencePerformed

Radiotherapy (PORT / adjuvant RT in SEER) *

None (reference)Yes

Results

Favorable vs baseline profile

Relative hazard vs Cox baseline (all reference categories): 0.486 (LP = -0.722)

Adjusted effect of radiotherapy

In the multivariate model, RT yes vs none: HR ≈ 0.762 (95% CI 0.605–0.96) for OS. Interpret with indication, margins, and guideline context.

How to read relative hazard

Values < 1 suggest lower hazard than the Table 3 reference profile; values > 1 suggest higher hazard. This is not an absolute survival percent; see the Formula tab for calibration limits.

Model source and scope

This tool applies the multivariable Cox proportional hazards model (hazard ratios as Exp(β)) reported in Table 3 by Ma C, Nong X, Discover Oncology 2025 (doi:10.1007/s12672-025-02072-7). The analysis used SEER 8.4.3 patients with AJCC 6th TNM stage III or IV malignant salivary gland tumors. Training (2010–2015) and test (2004–2009) cohorts were used to build and validate nomograms for overall survival (OS) and cancer-specific survival (CSS); validation C-indices were approximately 0.71–0.78.

What you are computing

Linear predictor (LP): sum of natural logs of the adjusted hazard ratios for the selected categories vs the reference row in each variable (Cox baseline pattern).
Relative hazard vs baseline: exp(LP), interpreted as the model-adjusted hazard relative to the reference combination (each factor at its reference level in Table 3).
Radiotherapy (PORT / adjuvant RT): SEER records radiotherapy as a binary yes/no; in surgically treated patients this commonly reflects postoperative radiotherapy when given, but the database does not separate timing or dose. Adjusted HR for RT “yes” vs “none” is 0.76 for OS and 0.74 for CSS in the published model.

Important limitations

Designed for locoregionally advanced (stage III/IV) disease in the SEER era studied — not for early-stage tumors or non–salivary histologies outside the table.
Outputs are not individualized absolute survival percentages unless paired with baseline survival from the original nomogram calibration; this calculator emphasizes relative hazard and the modeled RT association.
Staging is AJCC 6th as in the paper; do not mix with 7th/8th edition labels without conversion.
Does not replace multidisciplinary judgment, pathology review, or guideline-based indications for PORT.

Disclaimer

Educational decision support only. Verify critical values against the primary publication and current practice guidelines.

Understanding prognostic modeling in advanced salivary gland cancer

Malignant tumors arising from the major salivary glands are heterogeneous in histology, grade, and patterns of spread. For patients with locoregionally advanced disease, treatment typically combines oncologic resection when feasible with selective use of postoperative radiotherapy (PORT) and, in some settings, systemic therapy. Because both tumor biology and treatment exposure influence long-term outcomes, clinicians often seek tools that integrate multiple prognostic factors beyond TNM stage alone.

What this calculator implements

This tool is aligned with a population-based multivariable Cox proportional hazards model published for patients with AJCC 6th edition TNM stage III or IV salivary gland malignancy identified in the Surveillance, Epidemiology, and End Results (SEER) program. The underlying work constructed nomograms to summarize how demographic factors, SEER “Historic stage A”, AJCC 6th T, N, and M categories, histologic subtype, surgery, and radiotherapy associate with overall survival (OS) and cancer-specific survival (CSS) after variable selection and regression modeling.

In the calculator interface, you can switch between OS and CSS. The cancer-specific model additionally includes chemotherapy (yes versus no or unknown) where that factor was retained in the published multivariable table. The overall survival model includes a tumor category contrast (epithelial versus other) where that factor was reported for OS in the primary tabulated model.

Why postoperative radiotherapy appears as “radiotherapy” in registry data

Administrative and registry fields rarely document radiotherapy with the same granularity as a radiation oncology chart. In SEER, radiotherapy is commonly represented as a binary indicator (for example, received versus not recorded as received). For patients who undergo definitive surgery, radiotherapy recorded in this way frequently corresponds to adjuvant or postoperative treatment, but the database does not consistently distinguish PORT timing, technique, dose, fractionation, or target volumes. For that reason, the radiotherapy variable should be interpreted as a coarse treatment covariate reflecting whether radiation was part of the recorded care pathway, not as a verified prescription of PORT for a specific indication.

How the statistical model is used in the app

Multivariable Cox models estimate adjusted hazard ratios comparing categories to reference levels within each factor (for example, a T category relative to T1, or radiotherapy yes relative to none), holding other modeled variables constant at their selected values. In proportional hazards form, each hazard ratio can be expressed as exp(β) for a coefficient β. A linear predictor is formed by summing the natural logarithms of the hazard ratios for the chosen categories, which is equivalent to summing the β coefficients for those indicator contrasts.

The application reports a relative hazard versus the Cox baseline profile (the combination of reference categories in the published table). Values below one suggest a lower modeled hazard than that baseline profile; values above one suggest a higher modeled hazard. This is useful for comparing profiles and for understanding the direction and approximate magnitude of adjusted associations, particularly when discussing how radiotherapy and surgery covary with other strong prognostic drivers such as age, T stage, and metastatic classification.

Interpreting the radiotherapy association in context

In the published multivariable results, radiotherapy is often described as associated with improved survival in adjusted analyses, but the measured association is not a randomized treatment effect. Patients who receive radiotherapy differ systematically from those who do not, including indications related to margin status, perineural spread, nodal burden, histology, performance status, and institutional practice. The model therefore reflects adjusted observational associations within a defined SEER cohort and era, and it should not be read as proof that radiotherapy benefits every individual patient regardless of pathology and multidisciplinary assessment.

Despite those limitations, the radiotherapy term is clinically meaningful as a decision-support anchor when paired with guideline-based indications for PORT in salivary gland cancers, particularly for high-risk pathologic features where adjuvant radiation is commonly recommended. The calculator highlights the model’s adjusted radiotherapy hazard ratio and confidence interval as reported, so users can juxtapose the statistical association with contemporary indications and patient-specific tradeoffs such as xerostomia, dental risks, and wound healing.

Scope and staging assumptions

The model was developed for advanced-stage salivary gland malignancy using AJCC 6th T, N, and M fields as encoded for analyses in the source study. Mixing staging editions (for example, applying AJCC 8th labels directly without careful crosswalk) can misclassify extent of disease and distort predicted risk. Likewise, SEER historic stage groupings summarize extent at a population level and may not map one-to-one to every contemporary clinical staging workflow.

Histology categories in registry analyses aggregate clinically distinct entities under pragmatic labels (for example, carcinoma not otherwise specified or combined groups). Users should confirm that the selected histology option best represents the case at hand, recognizing that granular pathology review may refine risk beyond registry categories.

What this tool does not provide on its own

Nomograms in the primary publication also convey absolute survival probabilities at fixed time horizons through calibration to baseline survival. This calculator emphasizes the composed Cox linear predictor, relative hazard versus the reference profile, and the adjusted radiotherapy association as tabulated, because reproducing fully individualized absolute percentages requires the baseline survival function and calibration details tied to the original nomogram scales.

The output is intended for education and shared decision-making, not as a substitute for pathology review, radiation oncology consultation, or guideline-directed staging workup. Patients with rare variants, incomplete staging, or multimodal therapy not captured in binary fields may not be well represented by registry-derived effect estimates.

Practical use in multidisciplinary discussion

In tumor board or clinic conversations, a transparent prognostic scaffold can help align expectations when weighing PORT. A higher modeled hazard driven predominantly by T stage, distant metastasis, or aggressive histology may shift discussions toward intensified systemic therapy eligibility, clinical trial screening, or supportive care planning, whereas a profile dominated by resectable anatomy with favorable histology may frame surveillance and rehabilitation priorities. In all cases, the modeled estimates should be integrated with performance status, comorbidity, patient goals, and pathology details that registries only partially capture.