A new computerised tool or nomogram for predicting a patient's risk of cancer recurrence after surgery to remove primary gastrointestinal stromal tumours (GIST) is more accurate than current predictive models. As such, it could help doctors select patients who are likely to benefit from additional treatment such as with the drug imatinib, finds an Article published Online First and in the November issue of the Lancet Oncology.
Knowing how likely a cancer is to recur is important so patients can be counselled about their probable outcome, and also enables patients to be selected for postoperative treatment if effective additional therapy is available. Cancer recurrence in patients with GIST is common, even after seemingly successful surgery. The drug imatinib has been shown to prolong recurrence-free survival (RFS), and has recently been approved for the additional treatment of operable GIST. However, the financial cost and potential toxic effects of imatinib make the ability to calculate the risk of recurrence in individual patients vital.
So Ronald DeMatteo from Memorial Sloan-Kettering Cancer Center, USA, and colleagues developed a nomogram using three established prognostic criteria—tumour size, location (stomach, small intestine, colon/rectum, or other), and mitotic index—and data from 127 patients with primary GIST to assess RFS after surgery. The nomogram works by adding up the risk scores associated with each criterion and predicting the likelihood of RFS at 2 and 5 years.
The nomogram was tested for accuracy in 212 patients with GIST from the Spanish Sarcoma Research Group (GEIS) and 148 patients who had surgery for GIST from The Mayo Clinic. The authors then compared the predictive ability of the nomogram to three commonly used models or staging systems: US National Institutes of Health (NIH)-Fletcher, NIH-Miettinen, and the recently updated Armed Forces Institute of Pathology (AFIP)-Miettinen.
Overall, the nomogram was found to be better at predicting the likelihood of RFS than the NIH and AFIP staging systems.
Findings showed that the predictive accuracy, as measured by the concordance probability of the nomogram, was 0.78 in the original dataset (78% of the time the nomogram accurately predicted the ordering of the outcome between two randomly selected patients), 0.76 in the GEIS, and 0.80 in the Mayo Clinic validation datasets.
Additionally, concordance probabilities of the nomogram were significantly better than both NIH models when tested on patients in the GEIS cohort (0.76 vs 0.70 and 0.66) and in the Mayo cohort (0.8 vs 0.74 and 0.78). The nomogram also had higher but not statistically different concordance probability to that of the AFIP model when tested on patients in both the GEIS (0.76 vs 0.73) and Mayo cohorts (0.80 vs 0.76). Further calculations showed that the nomogram predictions of RFS were better calibrated than predictions made with the AFIP model.
The authors say: "Overall, prognostic nomograms give better prediction of the likelihood of events for individual patients than do staging systems that stratify patients into a few broad groups."
They conclude: "The appeal of the current nomogram is that…the variables of tumour size, location, and mitotic index are routinely reported by many pathologists and, therefore, the nomogram should be broadly applicable…The nomogram might be useful for patient care, interpretation of clinical trial results, and the selection of patients for adjuvant imatinib therapy."
In an accompanying Reflection and Reaction, Heikki Joensuu from Helsinki University Central Hospital in Finland welcomes the new tool and calls it "a step forward in the individualisation of prognostication."
Jeanne D'Agostino, Department of Public Affairs, Memorial Sloan-Kettering Cancer Center, New York, USA. T) +1 212 639 3573 E) dagostij@mskcc.org
Professor Heikki Joensuu, Helsinki University Central Hospital, Helsinki, Finland. T) +358 40 72 10 438 E) heikki.joensuu@hus.fi
For full Article and Reflection and Reaction, see: http://press.thelancet.com/tlogist.pdf
Journal
The Lancet Oncology