KEY POINTS
- This retrospective study used the US National Cancer Database to evaluate 63,261 patients with nonmetastatic pancreatic adenocarcinoma managed without definitive surgery between 2004 and 2023. Among these, 45,838 received chemotherapy alone and 17,423 received chemotherapy plus radiotherapy. Survival analyses excluded patients diagnosed in 2023 because outcomes were unavailable, and multivariable modeling accounted for demographic, socioeconomic, treatment-era, and recorded comorbidity variables.
- Chemotherapy plus RT was associated with longer overall survival than chemotherapy alone. Median OS was 14.92 months versus 10.38 months, respectively, an absolute difference of 4.54 months. Multivariable Cox regression yielded HR 0.71 (95% CI 0.69–0.72; p<0.001), and propensity score matching produced the same estimated HR. The matched analysis included 13,874 chemoradiotherapy recipients and 26,823 chemotherapy-only patients, but did not eliminate unmeasured differences in disease extent, treatment eligibility, or systemic therapy.
- The apparent survival association changed substantially over follow-up. In a time-dependent Cox model accounting for RT initiation, the adjusted mortality HR was 0.66 at six months, 0.85 at 12 months, 0.98 at 18 months, and 1.09 at 24 months. The interaction between treatment and follow-up time was significant (p<0.0001), indicating nonproportional hazards. Therefore, the overall HR of 0.71 does not describe a constant treatment-associated benefit throughout the disease course.
- Higher-dose five-fraction SBRT was associated with longer survival than lower-dose five-fraction treatment. Registry-defined prescriptions of 50–80 Gy in five fractions were associated with lower mortality than 30–49 Gy in five fractions (HR 0.86, 95% CI 0.76–0.97; p=0.01). Conventional fractionation and lower-dose 15-fraction regimens showed less favorable associations in cross-regimen comparisons. However, the registry did not establish comparable tumor anatomy, OAR exposure, target coverage, or treatment indications across dose categories.
- Higher biologically effective dose was associated with improved survival, particularly in the contemporary treatment era. Compared with BED10 53.1–89 Gy, prescriptions corresponding to BED10 90–99 Gy and 100–151 Gy had mortality HRs of 0.78 (p=0.006) and 0.77 (p<0.001), respectively. The associations persisted among patients diagnosed in 2014–2023, but not among those diagnosed in 2004–2013. The authors also identified an exploratory BED10 threshold around 75 Gy, which should not be interpreted as a validated prescription criterion.
- The magnitude of the RT-associated survival difference varied by treatment era and patient characteristics. For patients diagnosed in 2014–2023, chemotherapy plus RT was associated with HR 0.65, compared with HR 0.77 among patients diagnosed in 2004–2013. Older patients and those with lower recorded comorbidity burden showed favorable high-BED associations in selected analyses. These findings cannot establish that treatment intensification itself accounts for survival differences, given the evolution of systemic therapy, imaging, and selection criteria over two decades.
- The database lacked several variables essential to causal interpretation and treatment safety. Detailed pancreatic tumor anatomy, vessel involvement, functional status, chemotherapy regimen and sequencing, actual dose distributions, and radiation-related toxicity were not available for comprehensive adjustment. Survival was measured from diagnosis, while RT could begin later, creating potential time-related selection effects despite the supplementary extended Cox analysis. The exceptionally broad 50–80 Gy in five fractions category also should not be converted into a routine prescription recommendation without patient-level planning and toxicity data.
CLINICAL TAKEAWAY
Adding radiotherapy to chemotherapy was associated with longer survival in a large US nonsurgical pancreatic cancer cohort, with additional associations favoring higher biologically effective doses. However, the survival relationship varied over time, and registry limitations prevent determining the independent causal effect of RT or the optimal dose. These findings support prospective evaluation of carefully selected dose-escalation strategies rather than routine adoption of higher prescriptions based on database associations alone.