Can Prognostic Scoring Tools Predict Treatment Outcomes?

    loading  Checking for direct PDF access through Ovid

Excerpt

Of the 134,490 individuals in the United States who were expected to be affected by colorectal cancer in 2016, approximately 20% presented with distant metastatic disease.1 For patients who present with resectable stage IV disease, surgical resection of the primary tumor and metastasis is the primary curative treatment modality. However, the majority of patients present with unresectable metastatic disease, and, for these patients, the primary treatment is systemic chemotherapy.
Over the past 2 decades, there have been a number of important treatment advances with the development of new systemic therapies that have greatly improved the prognosis of patients with unresectable disease; median survival with sequential therapies now exceeds 24 months.2 In this setting, there is ongoing interest in the role of elective, noncurative primary tumor resection (PTR) as part of an overall treatment strategy to decrease the risk of primary tumor-related complications and improve survival.
This issue of Diseases of the Colon & Rectum reports a validation study of a prognostic tool developed to “predict survival after PTR” with the expressed goal of guiding therapeutic strategies.3 The study utilized data from patients treated in 1999 through 2007 for derivation and those treated 2008 through 2013 for validation. The scoring model incorporates a number of clinical factors such as tumor grade and site and number of metastases to categorize patients’ prognosis as good (≤3 points), moderate (4–6 points), and poor (7 points). All patients had undergone primary tumor resection and were deemed to have unresectable metastatic disease. Notably, the overall rate of morbidity was low, 17.9% grade 1 to 2 and 3% grade 3 or higher, although patients who died of noncancer causes were excluded from the analysis.
It is not surprising that the major factors associated with survival included site and number of metastases and tumor grade as CEA level. Interestingly, receipt of systemic chemotherapy was not included in the prognostic score, although 92% of patients eventually went on to receive systemic therapy. Finally, over the course of the treatment periods, there was a dramatic improvement in survival for all prognostic score groups, and the median survival overall improved from 11.6 months for the derivation cohort to 32.9 months for the validation cohort and 56.8 months in the good prognosis subgroup.
It is tempting to try to interpret these data to predict which patients are most likely to benefit from PTR. Unfortunately, because the models were developed on patients who all underwent PTR, it is not possible to predict which patients were most likely to benefit from PTR. Moreover, stratification by prognostic cohorts demonstrates that factors unrelated to PTR are the most important determinants of survival, with the most important factor being treatment during the more recent era of drug therapy. The best prognosis was among patients with a score ≤3, which would preclude most patients with more than 1 site of disease or significant comorbidity. In essence, younger patients with low burden and few comorbidities were identified to have the best prognosis. These are also the patients most likely to be able to tolerate sequential systemic regimens, and perhaps most likely to undergo elective noncurative PTR.
As the authors acknowledge, conflicting data exist regarding the survival impact of PTR.4,5 Unfortunately, retrospective comparative survival analysis without randomization or clear explanation of the indications for resection are at great risk for bias and misleading results. In addition, although the prognostic model provides information regarding the likely outcome (in this case, survival) as a function of the patient and tumor characteristics, it is not a predictive factor that can determine the effect of treatment (in this case, the effect of surgery) on survival.
    loading  Loading Related Articles