A Critical Analysis of Prognostic Factors in Patients with HTLV-1 Adult T-Cell Leukemia/Lymphoma: A Multicenter Clinicopathologic Experience and New Prognostic Score.

Purpose: Adult T-Cell Leukemia/Lymphoma (ATLL) is a rare aggressive Human T-cell Lymphotropic Virus Type-I (HTLV-I) associated peripheral T-cell neoplasm with 4 recognized clinicopathologic subtypes: acute, lymphomatous, chronic, and smoldering. Since the initial description of these variants, several studies have sought to identify additional prognostic factors. We assessed prognostic models already in use for aggressive non-Hodgkin lymphomas to develop a novel risk stratification scheme.

Methods: Data regarding patients with ATLL were collected from 3 medical centers between 8/92 and 5/07. Descriptive statistics were used to assess categorical and continuous variables. Overall survival (OS) was defined as time from diagnosis to death. Survival curves for OS were estimated using the Kaplan-Meier method. Univariate associations between individual clinical factors and OS were evaluated using the log-rank test for categorical variables and the Cox model for continuous variables. Maximum logrank analysis was used to select the optimal cut-off for calcium. In order to develop a simple risk model and allow for interactions of factors independently associated with OS, we used recursive partitioning analysis.

Results: 89 patients with ATLL were identified; 37 males (41.6%) and 52 females (58.4%) and median age 50 years (range 22 to 82). The acute subtype of ATLL predominated (68.5%), followed by lymphomatous (20.2%), chronic (6.8%) and smoldering (4.5%). Median OS for all sub-types was 24 weeks (range 0.9 to 315). According to the International Prognostic Index (IPI), 8 patients (9.1%) were classified as low risk, 11 patients (12.5 %) as low intermediate risk, 13 patients (14.8 %) as high intermediate risk, and 56 patients (63.6 %) as high risk, 1 patient could not be evaluated due to missing data. Median OS by IPI risk group was 271, 65, 31 and 16 weeks, respectively (p<0.01). The Prognostic Index for PTCL-U (PIT) could be determined in 68 patients; 10 patients (14.7 %) had a score of 0–1 (group 1), 19 patients (27.9 %) had a score of 2 (group 2), 31 patients (45.6 %) had a score of 3 (group 3), and 8 patients (11.8 %) had a score of 4 (group 4). Median OS by PIT risk group was 61.1, 28, 24, and 11.3 weeks respectively (p<0.01). A new risk model was developed using the variables of the IPI and PIT. In addition, calcium level at diagnosis was also included as it had independent prognostic value. Recursive partitioning of OS based on these variables gave a tree with 5 nodes, which fell into three risk categories: low risk patients with Stage I–II disease and a performance status <2; the medium risk group composed of two sets of patients:

• those with Stage III–IV disease with an ECOG performance status < 2 or

• those with an ECOG performance status ≥ 2 with calcium ≤ 11 mg/dL and age ≤ 60;

and the high risk group (also comprising 2 sets of patients):

• those with a performance status ≥ 2 with calcium ≤ 11 mg/dL and age > 60 or

• those with a performance status ≥ 2 and calcium > 11 mg/dL. There were 10 patients (11.2%) in the low risk (median survival= 156.6 weeks), 31 (34.8%) in the intermediate risk (median survival = 45.4 weeks), and 48 (53.9%) in the high risk (median survival= 13 weeks) categories (p<0.01).

Conclusion: This retrospective series confirms a poor outcome for North American patients with HTLV-1 related ATLL. Although the IPI and PIT identified subsets of patients, these models had liabilities. We propose a new prognostic model based on recursive partitioning analysis that successfully identifies three prognostic categories based on performance status, stage, age and calcium level at diagnosis in a more robust and distinct fashion.

Table 1. Comparison of Prognostic Scores and Kaplan Meier Survival Estimates (%) of patients with ATLL