Background: Most prognostic and predictive models for survival of persons with chronic lymphocytic leukaemia (CLL) are based on clinical and laboratory co-variates and have AUCs of 0.65 to 0.74 indicating considerable inaccuracy. No current prognostic model uses data on immune state of the bone marrow micro-environment to predict survival.

Methods: We downloaded relevant data from International Cancer Genome Consortium (CLL-ES; N = 485) and Gene Expression Omnibus (GSE22762; N = 195) with survival data. We used a LASSO Cox regression model to build a survival prediction model based on the expression of immune-related genes in the bone marrow micro-environment. Immune cells corresponding to these immune genes were deduced Using CIBERSORT and compared in cohorts with different survival probabilities.

Results: An immune signature based on 9 immune-related genes were constructed in the training cohort (CLL-ES) and tested in a validation cohort (GSE22762). The AUCs of 1, 3 and 5 -year survivals were 0.83 [95% Confidence Interval, 0.63, 0.97], 0.79 [0.67, 0.89] and 0.82 [0.73, 0.90] in training and 0.66 [0.55, 0.79], 0.60 [0.50, 0.69] and 0.66 [0.58, 0.76]in the validation cohort. Subjects in the high‐risk cohort identified by immune signature had significantly worse 5-year survival compared with those in low‐risk cohort group (training cohort: 91% [87, 95%] vs. 99% [98, 100%], P <0.001; validation cohort: 61% [50, 71%] vs 81% [80, 82%], P = 0.003). Subjects with a low-risk immune signature had a higher proportion of CD4-postive T-cells, activated NK-cells compared with those in the high-risk cohort (p < 0.05).

Conclusion: The immune score model of the bone marrow micro-environment we developed accurately predicts survival of persons with CLL. These data suggest a role for immune cells in the bone marrow micro-environment on survival. The immune score we describe can be combined with other prediction models to improve accuracy.


ICGC, GEO, chronic lymphocytic leukemia, immunity, prognostic model


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.

Sign in via your Institution