Individual life expectancy estimation using validated prognostic scores for patients with cancer of unknown primary

To implement 3 published prognostic scores in an independent set of patients with cancer of unknown primary (CUP), and compare their performance on individual life expectancy prediction.

The survival of 430 consecutive patients with CUP was measured after they had allocated to their prognostic group (good prognosis vs. poor prognosis) according to each prognostic score. Using a 2 x 2 contingency table, we measured the sensitivity, specificity, positive predictive value (PPV) and accuracy of each score in predicting individual outcome (survival <90 days or >180 days).

The median overall survival was 189 days (1-4,801 days). Survival was <90 days in 143/421 cases and >180 days in 208/413 cases. The three PPVs were within the same range for prediction of survival <90 days (from 43 to 49%) as well as for prediction of survival >180 days (from 70 to 80%), and underestimate individual life expectancy of 40-50% of the patients. None of the 3 scores appeared significantly better.

The main finding of this retrospective analysis is that the published prognostic scores cannot be used for rational decision making.