A New PE and an Unresected Stomach Tumor: Which DOAC Trial Actually Applies to Him
Two major cancer-VTE trials of two different DOACs reached opposite-looking conclusions about GI bleeding risk — and the drug he actually gets should depend on which trial's population he really resembles, not on habit.
K.A., a 67-year-old man whose grandchildren have started a group chat just to send him a photo every morning so he has something to look forward to during chemotherapy, was two weeks into neoadjuvant treatment for a locally advanced gastric adenocarcinoma — still sitting in his stomach, unresected, awaiting restaging — when he came to the emergency department short of breath and was found to have a segmental pulmonary embolism on CT. Cancer-associated venous thromboembolism carries its own dedicated treatment literature separate from ordinary provoked VTE, largely because the recurrence risk on standard anticoagulation is meaningfully higher in cancer patients and the bleeding risk profile differs by both cancer type and anticoagulant class in ways that general VTE guidelines don't capture.
The specific tension in front of the team is which DOAC trial's own findings actually generalize to him. The CARAVAGGIO trial found apixaban non-inferior to dalteparin for cancer-associated VTE with no significant increase in major bleeding overall, including in its gastrointestinal-cancer subgroup — a genuinely reassuring result for exactly his tumor type. The Hokusai VTE Cancer trial, testing edoxaban, found the opposite signal specifically in luminal GI cancers: a real increase in major bleeding driven by gastrointestinal bleeds in that subgroup, and the SELECT-D trial testing rivaroxaban showed a similar GI-bleeding signal. His tumor is still in place, unresected, ulcerated on endoscopy at diagnosis — precisely the anatomic situation the Hokusai and SELECT-D bleeding signal is thought to reflect, a luminal tumor bleeding directly into the GI tract once anticoagulated, rather than a generic cancer-patient bleeding risk that would apply equally across DOACs.
His performance status remains good — he is still able to walk to his own appointments and manage his own medications — which matters directly to today's decision, since a patient too debilitated to reliably self-administer a twice-daily oral regimen or attend injection appointments would need the anticoagulant choice weighted by feasibility as much as by bleeding-risk data; for him, both options remain genuinely practical.
His wife has been managing his medication schedule since his diagnosis and was present for today's discussion, asking pointed questions about injection-site reactions and pharmacy logistics that the team treated as seriously as the clinical evidence itself, since a regimen that fails at home for a practical reason helps him no more than one chosen on incomplete data.
In the ED, weighing which trial's population he actually matches
CARAVAGGIO found apixaban non-inferior to dalteparin with no significant increase in major bleeding overall, and its gastrointestinal-cancer subgroup specifically was reassuring. That's the most relevant randomized data we have for exactly this decision, and an oral drug spares him daily injections through what could be a long treatment course.
I'd want us to look past 'GI cancer' as one category and at his tumor specifically — unresected, ulcerated on endoscopy. That's precisely the anatomic picture that produced real, not theoretical, GI bleeding in Hokusai VTE Cancer and SELECT-D. One favorable trial doesn't erase two unfavorable ones in the exact population he belongs to.
Being reassuring in a subgroup analysis isn't the same as being proven safe for his specific tumor type — subgroup results are hypothesis-generating, not the last word.
I think the honest position is that neither trial result should be treated as fully decisive here — CARAVAGGIO's GI-cancer subgroup is real but wasn't independently powered as a dedicated bleeding trial, and Hokusai's signal, while concerning, came from a different DOAC's pharmacology. Given that genuine uncertainty and his tumor's specific anatomy, I'd default to dalteparin while the tumor remains unresected and ulcerated, and revisit a DOAC — apixaban specifically — once restaging shows either resection or meaningful tumor response.
Agreed: dalteparin started today for the segmental PE, with an explicit plan to revisit apixaban as an oral alternative once restaging after this chemotherapy cycle shows either surgical resection or clear tumor response reducing the luminal bleeding concern.