Pre-diagnostic trajectories of lymphocytosis predict time to treatment and death in patients with chronic lymphocytic leukemia.

Background: The dynamics of pre-diagnostic lymphocytosis in patients with ensuing chronic lymphocytic leukemia (CLL) need to be explored as a better understanding of disease progression may improve treatment options and even lead to disease avoidance approaches. Our aim was to investigate the development of lymphocytosis prior to diagnosis in a population-based cohort of patients with CLL and to assess the prognostic information in these pre-diagnostic measurements.
Methods: All patients diagnosed with CLL in the Greater Copenhagen area between 2008 and 2016 were included in the study. Pre-diagnostic blood test results were obtained from the Copenhagen Primary Care Laboratory Database encompassing all blood tests requested by Copenhagen general practitioners. Using pre-diagnostic measurements, we developed a model to assess the prognosis following diagnosis. Our model accounts for known prognostic factors and corresponds to lymphocyte dynamics after diagnosis.
Results: We explore trajectories of lymphocytosis, associated with known recurrent mutations. We show that the pre-diagnostic trajectories are an independent predictor of time to treatment. The implementation of pre-diagnostic lymphocytosis slope groups improved the model predictions (compared to CLL-IPI alone) for treatment throughout the period. The model can manage the heterogeneous data that are to be expected from the real-world setting and adds further prognostic information. Conclusions: Our findings further knowledge of the development of CLL and may eventually make prophylactic measures possible.

Introduction

Chronic lymphocytic leukemia (CLL) is characterized by an accumulation of clonal, mature, CD5+ B cells in the peripheral blood, bone marrow, and secondary lymphoid organs[1]. The presence of more than 5000 clonal B cells per µL in peripheral blood is diagnostic for CLL[1, 2]. Prior to the diagnosis, clonal B cells develop from undetectable levels through (undetected) monoclonal B cell lymphocytosis (MBL) to overt lymphocytosis and finally to a state that exceeds the diagnostic criterion for CLL[3]. Consequently, lymphocytosis has often been present years prior to diagnosis, and at diagnosis, the clinicians will occasionally be able to assess this accumulation of the clone in terms of lymphocytosis in blood samples taken years before diagnosis.

Once diagnosed, the clinical course of CLL is very heterogeneous with some patients living for decades without needing treatment, while others have short survival despite multiple lines of intensive treatments[4, 5]. The diagnosis of CLL is often incidental, only 5% of all patients require treatment at the time of diagnosis[5] while the remaining 95% of patients are observed until signs of active disease. Roughly estimated, more than 400 000 individuals with CLL are in observational “watch and wait” for CLL in Europe and the United States[6, 7]. The current and most widely used prognostic model, the international prognostic index (CLL-IPI), is based on five variables—age, stage, TP53 aberration, IGHV mutational status, and beta-2-microglobulin[8]. Unfortunately, CLL-IPI is poor to predict disease progression in patients with asymptomatic early-stage CLL[9].

One important risk factor for almost all cancers is the rate by which it grows; the same is true for CLL. Previous studies have reported that the cancer growth rate carries prognostic information and progressive lymphocytosis with an increase of 50% over a 2-month period or lymphocyte doubling time <6 months are indications for treatment[2, 10, 11]. Cancer growth is, however, difficult to assess because CLL cells are located and proliferate in three different compartments—peripheral blood, lymph nodes, and bone marrow[10]. On one hand, the absolute lymphocyte count (ALC) in peripheral blood is easy to measure, but the proliferation rate for circulating CLL cells is only between 0.1 and 1% per day, not even taking into account the proportion of CLL cells disappearing due to apoptosis and necrosis[10]. Nevertheless, Gruber showed that genomic features correlate to individual patterns of lymphocyte kinetics in peripheral blood after diagnosis of CLL[12]. Thus, the next step toward understanding this heterogonous cancer would be to characterize the clonal development prior to the diagnosis of CLL. Such knowledge can potentially help define a more clinical meaningful transformation point from monoclonal B-lymphocytosis to CLL than the current somehow arbitrary 5000 clonal B cells per µL. Moreover, trajectories of important biomarkers prior to diagnosis of CLL may carry important prognostic information.

Thus, we investigated 10-year trajectories of hemoglobin, C-reactive protein (CRP), ALC, and platelet count prior to diagnosis of CLL based on a matched case-control population, with 15 random controls (CLL-free individuals) per case (CLL patients). We further assessed, compared, and validated the distribution of genetic features in a subpopulation with the observations from Gruber and colleagues on the correlation between post-diagnostic lymphocyte kinetics and genetic aberrations in patients with CLL[12]. Lastly, we developed a model based on the trajectories to stratify individuals at high risk of treatment following the CLL diagnosis.

Methods

Study population

Retrospective CLL cohort

We included all patients diagnosed with CLL in Greater Copenhagen between 2008 and 2016 through the Danish CLL Register (DCLLR). Since 2008, all newly diagnosed CLL patients are registered in the DCLLR[13]. All hematological departments in Denmark are required to report clinical data for patients with CLL and the data completeness in the DCLLR is 98.3%[13]. The DCLLR holds information on birthdays, treating hospitals, prognostic markers, and treatment initiations. From this database, we obtained information regarding age at diagnosis, Binet stage, β-2-microglobulin, fluorescence in situ hybridization data, IGHV mutational status, and date of first-line treatment or death (if applicable). CLL-IPI was calculated using B2M cutoff of 4 and del(17p) as the only TP53 aberration[8].

Matched case–control population

Each CLL patient with at least one measurement within 5 years of diagnosis of hemoglobin, CRP, ALC, or platelet count available were matched on sex and birth year with 15 random controls per measurement. Controls were free from CLL at the index date. Vital information on controls was obtained using the Danish Civil Registration System. The Danish Civil Registration system is for practical purposes complete and no persons are lost to follow-up[14].

Variables

The biochemistry data used in this study come from two different data sources. 1) The Centre of Excellence for Personalized Medicine of Infectious Complications in Immune Deficiencies (PERSIMUNE) and 2) The Copenhagen Primary Care Laboratory (CopLab) database.

PERSIMUNE contains data from all hospitals in Denmark. PERSIMUNE has streamlined and quality-checked the data received from the many different laboratories within Denmark.

The Copenhagen Primary Care Laboratory (CopLab) database. In the Copenhagen area (the Copenhagen Community and the former Copenhagen County), with its approximately 1.2 million inhabitants, there was only one laboratory serving general practitioners and practicing specialists until 2015, the Copenhagen General Practitioners’ Laboratory (CGPL). The CGPL was accredited for International Organization for Standardization (ISO) standards ISO17025 and ISO15189 and has saved all data on the analyses it has performed since July 2000. The Copenhagen Primary Care Laboratory (CopLab) database contains all results (n = 178,000,000) from 1 July 2000 to 31 December 2015 from the CGPL[15].

Analytical methods of the ALC, platelet count, and hemoglobin are described elsewhere[16]. CRP may correlate with lymphocyte/leukocyte changes due to infection and has been predictive in many cancers[17]. CRP was assessed in serum by commercially available assays from Olympus A600 (Olympus A/S, Denmark) and ADVIA Chemistry System (Bayer®/Siemens®, Denmark) according to the instructions of the manufacturers. Two ADVIA Chemistry assays were used, including the C-reactive protein (CRP). For the Olympus assay, the inter-serial coefficient of variation percentage was 5.1% (at level 7 mg/L), 1.2% (at level 19 mg/L) and 2.5% (at level 31.5 mg/L). For the Advia Chemistry CRP assay, the inter-serial coefficients of variation percentage was 12.7% (at level 12 mg/L), 4.0% (at level 24.7 mg/L) and 2.6% (at level 98.6 mg/L). For the Advia Chemistry CRP_2 assay, the inter-serial coefficients of variation percentage was 1.59% (at level 23.5 mg/L), 1.45% (at level 75.5 mg/L) and 2.6% (at level 86.1 mg/L). The CRP assays were subject to external quality control through participation in the LabQuality External Quality Assessment Scheme (LabQuality, Helsinki, Finland). The assessment scheme included 12 distributions (each distribution comprised 1 sample) or 4 distributions (each distribution comprised 2 samples). The results through the entire CopLab period (from 2001 to 2015) confirmed the reliability of the assays, and the results from CGPL deviated <1.3% from the method mean. Low results were reported as <3 mg/L (until December 1, 2002), <5 mg/L (between December 2, 2002 and May 28, 2008) and <4 (from May 29, 2008). High results were reported as >300 mg/L (until December 1, 2002).

Sequencing and bioinformatic analysis

Next-generation sequencing has been described elsewhere[18]. Twenty-five driver mutations were assigned to signaling pathways according to consensus among two seminal papers[19, 20], and a pathway was considered altered if at least one gene in the pathway was mutated. Mutated genes involved in multiple pathways count as multiple altered pathways.

Statistics and reproducibility

ALC, platelet count, CRP, and hemoglobin

The matched population was used to estimate population-averaged trajectories separately for each biomarker. The marginal models for lymphocytosis, hemoglobin, and platelet counts were fitted using generalized estimating equations. A Gaussian parametric survival model, with a robust variance estimator, was used to account for the left censoring of CRP values below the detection limit. ALC, CRP, and platelet counts were log-transformed. We included time to diagnosis, classification group (control or case), sex, and age at diagnosis as covariates, along with an interaction between time to diagnosis and classification group. Higher-order polynomials of time to diagnosis were included, if significant, to allow for nonlinear trajectories.

Prognostic significance of lymphocyte slope

We fitted a linear mixed model with random intercept and slope to define pre-diagnostic lymphocyte slope groups. We used the ALCs, from three years prior to diagnosis, from the CLL cohort patients. ALC was log-transformed and time to diagnosis was included as a covariate. The three slope groups (low, medium, high) were defined as the tertiles of the estimated random slopes. We then fitted Cox proportional hazard models for the cause-specific hazards of death and treatment. The timescale was time from diagnosis until the first event of death or treatment. The models were adjusted for slope group, sex, and CLL-IPI. We performed model control on all models. We compared the difference in AUC over time between the models including and not including the slope group to assess the predictive performance of the slope group[21]. The study was approved by the Danish National Committee on Heath Research Ethics and informed consent was not required for retrospectively included patients according to Danish legislation.

Table 1

Patient characteristics.

	CLL cohort	Matched population
Level	CLL patients	Case: CLL patients	Control: subjects without CLL
Number of individuals	1123	778	15,124
Age at diagnosis in years (median (IQR))	69 (62–77)	69 (62–77)	69 (62–78)
Sex (number of males (percentage))	668 (59%)	444 (57%)	8726 (58%)
Before diagnosis
Follow-up in years (median (IQR))	2.9 (0.1–8.1)	7.8 (4.5–10.8)	7.9 (4.6–10.8)
Absolute lymphocyte count (number)	4073	3185	48,713
Hemoglobin concentration (number)	—	4700	75,432
Platelet count (number)	—	4687	75,348
CRP (number)	—	4263	76,787
After diagnosis
Biomarker follow-up in years (median (IQR))	2.3 (0.9–4.5)	—	—
Absolute lymphocyte count (number)	16,958	—	—
Event time follow-up in years (median (IQR))	2.6 (1.2–4.7)	—	—

CLL chronic lymphocytic leukemia, IQR interquatile range.

Patient characteristics of the persons who were followed in the CLL cohort and the matched cohort.