Prognostic Factors for Survival of Elderly Patients with Acute Myeloid Leukemia after Intensive Chemotherapy: Validation of 3 Popular Prognostic Models

Although epidemiologic surveys indicate that the median age of acute myeloid leukemia (AML) patients is in the range of geriatric age groups, treatment of elderly AML patients still represents an unsolved clinical question mainly due to frailty of elder patients and the disease characteristics. Different prognostic models have been proposed to select suitable elderly AML patients for intensive chemotherapy including 3 recent models; French Cooperative group (ALFA), German Study Alliance Leukemia group and M. D. Anderson Cancer Center (MDACC) prognostic models. The French model suggested patients who had unfavorable karyotypes and at least 2 of the following features: ECOG performance status ≥2, WBC count ≥50x10⁹/L and age ≥75, should not be treated with intensive therapy. The MDACC group identified the following risk factors for adverse prognosis obviating intensive treatment: age ≥80, complex karyotype, ECOG performance status ≥2 and elevated creatinine > 1.3 mg/dl. The German group demonstrated karyotype, CD34 expression, WBC count, age, LDH and NPM1 status as independent prognostic risk factors and developed a scoring system. To the best of our knowledge, these models have not been externally validated.

We retrospectively reviewed 114 non-acute promyelocytic leukemia AML patients older than 60 years treated in two University Hospital Hematology Departments between 1999 and 2014. Median age was 70 years. Eighteen (15.8%) patients received supportive care, 68 (59.6%) intensive and 28 (24.9%) non-intensive chemotherapy regimens. Primary end point of the study was overall survival (OS). OS was calculated from diagnosis to the date of mortality of any reason. The patients who did not die at last follow-up were censored at this time for OS computation. Survival analyses were computed by the Kaplan-Meier method. Comparisons of survival rates were done by the Log-rank test. Statistical Packages for the Social Sciences v17.0 (SPSS Inc., Chicago, IL) software was used for other statistical analyses.

The median OS of all patients was 4.2 months. Median OS was better in patients who received intensive treatment compared to patients who did not receive: 5.3 months (0–11.58 95% CI) vs 2.4 months (0.8–3.9 95% CI), respectively (p= .012). We applied all three models to the patients who received intensive therapy. We could not use German model as NPM1 mutation status and CD34 were frequently missing. According to the French model, median OS was shorter in high risk group (n= 18); 1.5 months (0.1–2.84 95% CI) vs 28.7 months (10.3–47 95% CI) (p= .008). MDACC model also demonstrated inferior prognosis in patients with risk factor(s) against intensive treatment (i.e., high risk patients, n= 28) with a median OS 2.3 months (1.2–3.4 95% CI) vs 28.7 months (7.6–49.7 95% CI) (p= .013).

There were 7 patients who did not receive intensive treatment although they were defined suitable for intensive treatment according to French and/or MDACC models. Median OS was significantly shorter in these cases compared to patients who received intensive treatment: 4.6 months (2.1-7 95% CI) vs 15.9 months (0.9-30.8 95% CI), respectively (p= .02). Intensive treatment did not change median OS in patients who were considered unsuitable for intensive treatment according to French and/or MDACC models (p> .05).

In conclusion, we found that both French and MDACC models were predictive in our clinical practice to identify patients who could benefit from intensive treatment. The German model was difficult to use, as it consisted NPM1 status, which was not generally available.