New blood test shows promise in triaging patients with vague symptoms for faster cancer workups

Diagnosing cancer remains tough when patients show up with vague, non-specific complaints like fatigue, unexplained weight loss, malaise or low-grade fever, symptoms that could stem from anything from infections to autoimmune issues. A study from last month in Nature Communications offered a different angle, demonstrating how analyzing a broad panel of plasma proteins could help flag those truly at risk for cancer, potentially speeding up referrals and cutting down on unnecessary tests.

Fredrika Wannberg, from the Danderyd Hospital in Stockholm, Sweden, and colleagues, drew on real-world data from patients referred to fast-track diagnostic pathways in Sweden. They profiled 1,463 plasma proteins using proximity extension assay technology, a high-throughput method that measures proteins in tiny blood volumes. Samples came from 456 patients in a discovery cohort (median age 71, about 55% female) before any cancer workup began, plus an independent replication group of 238 similar patients.

In the discovery set, 160 patients turned out to have cancer, mostly hematologic malignancies (28%), pancreatic or biliary adenocarcinomas (11%) and lung adenocarcinomas (8%), while 296 did not. The team identified 29 proteins linked to new cancer diagnoses, with 22 holding up in both cohorts. Key players in their model included anterior gradient 2 (AGR2), cytokeratin 19 (KRT-19), carcinoembryonic antigen-related cell adhesion molecule 5 (CEACAM5), ribonucleotide reductase subunit M2 (RRM2), poly ADP-ribose polymerase 1 (PARP1) and PC4 and SFRS1 interacting protein 1 (PSIP1), proteins often tied to cancer biology in tissue studies.

Wannberg and team also built a classification model on a training subset, then tested it. It achieved an area under the curve (AUC) of 0.80 in distinguishing cancer from non-cancer cases in the discovery cohort. Performance held strong at 0.82 in the replication set (35 cancer cases out of 238 patients). The model also separated cancer from specific non-malignant groups. AUCs ranged from 0.67 for inflammatory conditions to 0.86 for cases with no clear diagnosis in discovery and 0.82 to 0.84 across subgroups in replication.

A component analysis showed clear separation between cancer and non-cancer samples, supporting the idea that these protein patterns capture meaningful biological differences.

Nonspecific symptoms flood primary care and urgent referral pathways, often leading to broad, resource-heavy investigations, imaging, endoscopies and biopsies that often turn up nothing sinister. This proteomics-based approach, the authors argue, could serve as a triage tool. A simple blood draw to calculate cancer probability and prioritize high-risk patients for sensitive diagnostics like PET-CT.

In a system where delays hurt outcomes, especially for aggressive cancers like pancreatic or hematologic, the potential to streamline referrals could warrant more investigation. The model performed well even against tricky mimics like autoimmune or infectious diseases, suggesting it might reduce false positives and ease pressure on specialists. Avoiding low-yield tests in low-risk patients while accelerating workups for those flagged high-risk might trim diagnostic costs and improve resource allocation.

Nonspecific symptoms disproportionately affect older adults and those with chronic conditions, groups already underserved in cancer screening. A blood-based triage could catch cases outside traditional programs (e.g., no routine screening for pancreatic cancer) and address disparities if rolled out broadly.

The study has caveats, however. Cancer prevalence was high in these fast-track cohorts, about 35% in discovery, reflecting referral bias, so performance might dip in lower-prevalence primary care settings. Many cancers were advanced or metastatic, limiting insights on early-stage detection. The panel relied on a research-grade assay covering 1,463 proteins; translating to clinical use would require simplification to fewer markers, validation in diverse populations and head-to-head comparisons with emerging multi-cancer detection tests.

If future studies confirm utility in real-world primary care, it could mean earlier catches, better outcomes and more efficient use of healthcare dollars.