Why This Study Matters
Oligometastatic disease, a state of limited metastatic burden, has been proposed as an opportunity to use metastasis-directed therapy (MDT) to change the natural history of metastatic cancer. SABR-COMET provided foundational evidence for adding MDT to standard-of-care (SOC) systemic therapy, but the authors note a potential criticism: it lacked tumor-type stratification. EXTEND was designed to test whether adding MDT to SOC improves PFS within histology-specific baskets, each independently powered. This report presents the primary aggregated analysis across all randomly assigned patients and the primary findings of the kidney and other baskets; the breast, pancreatic, and two prostate baskets were reported previously, with follow-up updated to a single data cut-off.
Study Design
EXTEND was an investigator-initiated, multicenter, open-label, parallel group, superiority-design phase II randomized trial in oligometastatic solid tumors. Patients had 1–5 metastases amenable to MDT, ≤4 prior lines of systemic therapy for metastatic disease, and 1 of 12 histologies. Randomization was equal (MDT + SOC or SOC) using a dynamic Pocock and Simon method, stratified by prior lines (0–1 vs 2–4) and number of metastases (1–2 vs 3–5), plus basket-specific factors. SOC was systemic therapy or observation at the treating oncologist’s discretion; MDT could include radiotherapy, resection, or ablation.
The primary end point was PFS, defined by RECIST v1.1 radiologic progression on blinded central review, clinical progression, or death (PSA progression also counted in the prostate baskets). Non-prostate baskets were powered to detect improved median PFS from 4 to 8.5 months (40 patients per basket; 80% power, one-sided log-rank, 0.10 type I error); prostate baskets required 87 patients each to detect improvement from 18 to 36 months. PFS significance was defined by one-sided log-rank P < .10, but two-sided P values are presented throughout. The authors describe EXTEND as a signal-finding study.
521 screened; 350 randomly assigned; 334 in the per-protocol population (166 vs 168); 6 baskets; 1:1 randomization; median follow-up 53 months (IQR 38–60).
Number of sites and a dosing schedule: Not specified in source text reviewed. Radiation therapy was the MDT in 98% of metastases (370/379, per the abstract).
Patient Population
Patients were enrolled between September 2018 and August 2023. Twelve patients declined allocation and four no longer met inclusion criteria, leaving 334 in the per-protocol population: breast 40, pancreas 40, prostate iADT 87, prostate cADT 87, kidney 40, and other 40. Median age was 67 years (IQR 61–72) and 74% were male. Most had 1–2 metastases (63%), metachronous presentation (67%), and fewer than two prior systemic therapy lines (82%). Radiation therapy was the most common MDT modality (98%), typically stereotactic body radiation therapy (51%), and SOC included systemic therapy in 97%. Seventy-two percent were enrolled at the Texas Medical Center.
Primary Endpoint Results
After median follow-up of 53 months, PFS was significantly longer with MDT + SOC than SOC (HR 0.54 [95% CI 0.41 to 0.72]; P < .001). Median PFS was 25.40 months (95% CI 19.98 to 33.31) vs 14.00 months (95% CI 12.16 to 16.76). The proportional hazards assumption was not strongly refuted (Schoenfeld residual P = .10), and post hoc restricted mean survival time analyses were consistent. A post hoc modified PFS excluding PSA-only events gave HR 0.63 (95% CI 0.48 to 0.84; P = .002).
“The addition of MDT to SOC improved PFS, meeting the primary end point.”
— Discussion, Sherry et al., J Clin Oncol. MDT = metastasis-directed therapy; SOC = standard of care; PFS = progression-free survival.
Subgroup and Basket Analyses
| Basket | PFS HR (95% CI) | P | Median PFS, MDT + SOC (95% CI), months | Median PFS, SOC (95% CI), months |
|---|---|---|---|---|
| Breast | 0.73 (0.29 to 1.83) | .497 | 23.33 (11.63 to NE) | 24.94 (11.07 to NE) |
| Kidney | 1.03 (0.46 to 2.33) | .931 | 10.66 (6.18 to 24.51) | 11.84 (6.9 to 19.78) |
| Pancreatic | 0.40 (0.19 to 0.84) | .012 | 10.64 (5.88 to NE) | 2.53 (1.74 to 7.36) |
| Prostate cADT | 0.56 (0.28 to 1.09) | .082 | 41.4 (36.34 to NE) | 24.54 (16.76 to 45.27) |
| Prostate iADT | 0.44 (0.25 to 0.77) | .003 | 28.39 (23.59 to 47.9) | 15.77 (14 to 22.6) |
| Other | 0.41 (0.17 to 0.99) | .043 | 17.08 (10.74 to NE) | 8.74 (5.82 to NE) |
| Aggregated | 0.54 (0.41 to 0.72) | <.001 | 25.40 (19.98 to 33.31) | 14.00 (12.16 to 16.76) |
| Aggregated excluding prostate | 0.60 (0.40 to 0.89) | .010 | 14.00 (10.71 to 18.89) | 8.48 (6.08 to 13.17) |
NE = not estimable. The authors state that the aggregate HR point estimate was encompassed within the 95% CI of each basket, suggesting limited evidence for basket-level differential treatment effects. In the kidney basket, control arm outcomes (median PFS 11.8 months) exceeded the powering estimate of 4 months, which the authors say rendered interpretation inconclusive.
Post hoc analyses in the aggregate population suggested potential heterogeneity by prior lines of therapy (greater relative benefit with 0–1 vs 2–4 prior lines; interaction P = .006) and liver involvement (potentially greater benefit with liver metastases; interaction P = .02). The prior-lines interaction within baskets was seen only in kidney (interaction P = .002). These are post hoc and descriptive.
Circulating Tumor DNA (Exploratory)
| Analysis | HR (95% CI) | P |
|---|---|---|
| Enrollment ctDNA(+) vs PFS (adjusted) | 2.51 (1.73 to 3.66) | <.0001 |
| Enrollment ctDNA(+) vs DSS | 2.69 (1.43 to 5.06) | .002 |
| Enrollment ctDNA(+) vs OS | 2.05 (1.20 to 3.67) | .009 |
| MDT + SOC benefit on PFS, ctDNA(+) | 0.61 (0.38 to 0.98) | .04 |
| MDT + SOC benefit on PFS, ctDNA(−) | 0.41 (0.22 to 0.76) | .005 |
| ctDNA clearance vs landmarked OS | 0.25 (0.08 to 0.64) | .003 |
Of 237 patients profiled at enrollment, ctDNA was detected in 115 (49%). The treatment-by-ctDNA interaction was not significant (P = .6). ctDNA clearance was numerically higher with MDT + SOC (15/46, 33%) than SOC (10/44, 23%; odds ratio 1.65 [95% CI 0.65 to 4.20]; P = .3).
Safety Profile
No significant difference was observed in grade 3 or higher adverse events: n = 10 (6%) with MDT + SOC vs n = 4 (2%) with SOC (P = .11; risk difference 3.6% [95% CI −0.7% to 8%]). Quality of life was not worsened by MDT + SOC. Local failures in the MDT arm numbered 21 (13% per patient; 6% per lesion).
Overall survival and the kidney basket. OS (HR 0.96; 95% CI 0.65 to 1.41; P = .815) and disease-specific survival (HR 0.82; 95% CI 0.53 to 1.25; P = .351) were not significantly improved in the aggregate. In the kidney basket, OS and DSS were shorter after MDT + SOC (OS HR 5.38 [95% CI 1.04 to 27.8]; DSS HR 5.34 [95% CI 1.07 to 26.97]); there were 10 deaths in the MDT + SOC arm and four in the SOC arm, with 10 and 3 attributed to disease. At first progression after SOC, 51% of patients (66/129) received crossover MDT.
Interpretation and Broader Context
The authors conclude that EXTEND supports adding MDT to SOC for oligometastatic disease and identified histology-specific efficacy signals for phase III testing; a phase III EXPAND trial in pancreatic cancer (NCT06593431) has been initiated. They also caution:
“for certain disease sites, they caution against the routine use of MDT in clinical practice outside of a clinical trial.”
— Discussion, Sherry et al. The preceding sentence refers to the trial findings; MDT = metastasis-directed therapy.
Immune profiling showed induction of CD8+ T cells with proliferation and activation markers after MDT + SOC but not SOC, most pronounced in baskets with PFS superiority, which the authors frame as hypothesis-generating.
What the Study Could Not Establish
Limitations stated by the authors include the signal-finding design, crossover and heterogeneous post-progression therapy that may have obscured secondary end points, differing surveillance protocols between baskets, poor patient-reported outcome response rates, and exploratory, non-prespecified translational assays without multiplicity control. The trial was open-label, PFS significance was defined by a one-sided P < .10 threshold, the heterogeneity analyses by prior lines and liver involvement were post hoc, and the kidney basket was inconclusive because control-arm PFS far exceeded the powering assumption. Overall survival was not significantly improved.