Introduction

The prevalence of chronic lymphocytic leukaemia (CLL) in adults over the age of 65 has gradually increased in high income countries [1, 2]. CLL disproportionately affects males, and an inferior survival rate in males has been reported in several studies [3,4,5].

Over the last two decades, novel clinical and genetic-based prognostic factors have been identified in patients with CLL [6]. These include age, gender, immunoglobulin heavy chain variable region gene (IGHV) mutation status and cytogenetic abnormalities [7, 8], the aberrant expression of CD38 and ZAP70 [9], TP53 mutation [10], β2-microglobulin [11], and the Eastern Cooperative Oncology Group (ECOG) performance status [6, 7]. The development and implementation of prediction models have allowed for the risk-stratification of patients with CLL based on genetic traits [12].

In patients with CLL, therapy consisting of ibrutinib [13, 14], chlorambucil [15], fludarabine and cyclophosphamide [16, 17] yielded low overall response rates (ORR), with treated patients having an estimated 5-year overall survival (OS) of < 40% [18, 19]. These clinical outcomes in patients with CLL led to a shift towards novel antibody-based therapies in the last decade. These include rituximab, an anti-CD20 monoclonal antibody which when administered in combination with standard chemotherapy, improves the patient response rates and is associated with complete remission (CR) in patients with CLL [20,21,22]. However, despite the benefit of chemoimmunotherapy (CIT) with rituximab, patient outcomes are highly variable [23]. The efficacy of rituximab-based CIT has been demonstrated in cohorts of patients without the associated genetic aberrations such as Del(17p) and TP53 mutations [24].

The advances and refinement of prognostic risk scores has led to improved risk stratification of patients with CLL. The cornerstone of these risk scores, are the revised Rai [25] and Binet [26] staging systems, and novel prognostic indices such as CLL International Prognostic Index (CLL-IPI) [27] which allow for a precise risk stratification. Pertinent challenges in the risk stratification of patients with CLL on CIT include the lack of cumulative evidence on the predictive value of integrated cell and genetic based prognostic models [28]. Moreover, the lack of diverse multi-ethnic cohorts and prevalent risk factors also contribute to the imprecision of these predictive models [29, 30]. Therefore, the current systematic review and meta-analysis sought to identify and evaluate studies reporting on prognostic factors in patients with CLL on CIT or targeted therapy. Moreover, we aimed at providing a comprehensive synthesis and confirmation of prognostic factors associated with poor clinical outcomes in patients with CLL on CIT.

Methods

Eligibility criteria

The eligibility criteria was based on the Population, Index prognostic factor, Comparator prognostic factors, Outcome, Timing and Setting (PICOTS) guidelines [31]. We included randomised controlled trials (RCTs) reporting on prognostic factors in patients with CLL on CIT containing anti-CD20 monoclonal antibodies (rituximab, obinutuzumab, ofatumumab) or targeted therapy such as Bruton’s tyrosine kinase (BTK) inhibitors. We also included studies that aimed at developing or validating predictive models for mortality in CIT-treated patients with CLL. In addition, we included studies reporting on predictive measures at any time point and setting. Reviews, letters, and case-studies were excluded. In this systematic review, predictive models were considered as multivariable models used to predict survival in patients with CLL using selected predictors. We considered index prognostic factors derived from the CLL International Prognostic Index (CLL-IPI) [27], the German CLL Study Group (GCLLSG) [32], and the MD Anderson Cancer Centre (MDACC) nomogram predictive models [33].

Search strategy and selection process

A systematic literature search was performed by two independent reviewers (ZAM and BBN) on the MEDLINE, MasterFILE premier, Health source: Nursing/Academic edition, and clinical trials.gov. We made use of Medical Subject Headings (MeSH) and related synonyms which included, chronic lymphocytic leukaemia, rituximab, ofatumumab, Obinutuzumab, anti-CD20 monoclonal antibody, ibrutinib, venetoclax, acalabrutinib, idelalisib and prognosis. All electronic databases were searched from inception to the 1st of August 2022. A detailed search strategy is presented in Supplementary Table 1. To augment the database search, we screened the bibliographies of relevant reviews and included studies.

Data extraction

Two reviewers (ZAM and BBN) independently extracted data items from the included studies defined in the critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies for Prognostic factors CHARMS-PF checklist [34]. The extracted study characteristics included, source of data, participant description, sample size, outcomes to be predicted, candidate predictors, type of model.

Risk of bias and quality assessment

The certainty and strength of the evidence was assessed by two independent reviewers (ZAM, SAM) using the Quality In Prognostic Studies (QUIPS) tool [31]. The tool consists of six domains used to appraise studies of prognostic factors (Supplementary Table 2). A third reviewer (BBN) was consulted for arbitration.

Statistical analysis

The Cohen’s kappa was used to assess the inter-rater reliability for the study selection and the study quality and risk of bias assessments [35]. The hazard ratios (HR) or odds ratios (OR) and 95% confidence interval (CI) were pooled to estimate the pooled OS and PFS. The effect estimates of studies were pooled using a random-effects model [36]. The I2 and Chi squared statistical tests were used to assess the levels of statistical heterogeneity [37, 38]. An I2 value of  >50% was considered as substantial [36]. All data analysis was performed using STATA 16.0 (StataCorp LP, TX, USA).

Subgroup and sensitivity analyses

To explore the sources of heterogeneity amongst the included studies, we performed a sensitivity analysis based on the study design and quality.

Confirmation of predictive factors

The reported prognostic factors were confirmed based on the robustness of the overall direction of the effect across all eligible studies. Moreover, adjusted effect estimates that remained statistically significant (p < 0.05) after adjusting for covariates in the multivariate analysis were considered as confirmed.

Results

Included studies

We retrieved a total of 4123 citations through the database search, and after excluding 602 duplicated studies only 3521 studies were eligible for screening. Amongst these, 3320 studies were ineligible and excluded during the abstract screening phase. A total of 201 citations were retrieved and 118 articles with available full-texts were assessed for eligibility. A total of 171 studies were excluded for the following reasons: single arm studies (n = 61), ineligible study design (n = 38), clinical endpoint not reported (n = 26); no suitable comparator group (n = 33); only contained post-trial follow-up data (n = 13). In all, 17 studies [14,15,16,17, 39,40,41,42,43,44,45,46,47,48,49,50,51] met the inclusion criteria and were included in the qualitative and quantitative analysis (Fig. 1). The overall reviewer agreement for study selection, was 89% (kappa = 0.82).

Fig. 1
figure 1

PRISMA flow diagram showing the study selection process

Full size image

Characteristics of included studies

The 17 included studies were published between 2010 and 2021 comprising of a total of 7 349 patients with CLL (Table 1). Most of the included trials were multicentre studies and the study sample size varied from 66 to 817 patients (Median: 389, IQR: 296—532). The age of enrolled participants ranged from 22 – 92 years.

Table 1 Characteristics and outcomes of included CLL studies (n = 17)
Full size table

The geographic distribution of the included studies consisted of Europe, Americas, Asia, Australia (Table 1). The included studies comprised of 64% (n = 4 700) patients who were treatment-naïve, 11% (n = 815) of patients who were previously treated and 22.3% (n = 1 642) who were relapsed/refractory. In addition, 47% (n = 8) of the included studies reported on the Rai staging whereas 41% (n = 7) reported on Binet staging system. One study (6%) reported both Rai and Binet staging systems and another study (6%) did not specify the staging system used.

Prognostic factors in patients with CLL

In the included studies, prognostic factors were analysed before the start of treatment (Table 2). Overall, the studies comprised of 25.5% (n = 1 823) of patients who were 70 years or older, 55.7% (n = 3 984) of patients with an unmutated IGHV status, 17.4% (n = 1 245) with del11q, 6.8% (n = 489) with a del17p, 26.8% of the patients (n = 1 915) had del13q, and 3.9% (n = 264) had TP53 mutation. Notably, 6% (n = 429) patients were reported to have Trisomy 12. In the reported cell-based prognostic factors the included studies reported on ZAP-70 expression in 12.2% (n = 872) of the patients, and CD38 expression was reported in 12% (n = 863) of the included patients, 21.3 (n = 1 526) patients had elevated B2M levels (≥ 3.5 mg/L). In all, 36.7% (n = 2 625) of the included patients with CLL were in the advanced stage of the disease.

Table 2 Treatment arms and confirmed prognostic factors in studies included in the meta-analysis (n = 17)
Full size table

Risk of bias and quality assessment

We assessed the quality of all included studies using the QUIPS tool for assessing risk of bias in prognostic factor studies [31]. The study-level risk of bias assessment is presented in Supplementary Table 2. Briefly, two studies were scored as high-risk [16, 41], five as moderate risk [39, 40, 47, 48, 50], whilst the rest were deemed to be at low risk of bias [14, 15, 17, 42,43,44,45,46, 49, 51]. Overall, the included studies were scored as low risk for study participation (k = 0.76, minimal agreement), and outcome measurement (k = 0.88, strong agreement), moderate risk for study attrition (k = 0.88, moderate agreement) and confounding measurement (k = 0.65, minimal agreement) and high risk for prognostic factor measurement (k = 0.90, strong agreement) and statistical analysis and reporting (k = 0.76, minimal agreement) (Fig. 2).

Fig. 2
figure 2

Risk of bias assessment of the prognostic factor studies

Full size image

Primary outcomes

Survival outcomes of patients with CLL receiving CIT containing anti-CD20

A total of 5 studies [15,16,17, 42, 47] reported on an improved PFS in patients with CLL, when an anti-CD20 mAbs were concurrently used with standard chemotherapy. CIT in combination with anti-CD20 monoclonal antibodies, was associated with improved PFS (HR = 0.50 Cl [0.35–0.65], p < 0.01). There were high levels of heterogeneity (I2 = 90.78%) in the included studies. Overall, the pooled effect estimate showed no statistically significant difference in OS in patients with CLL treated with CIT and chemotherapy alone (p = 0.22) (Fig. 3).

Fig. 3
figure 3

Meta-analysis of the hazards ratios (HR) for progression-free survival (PFS) for CLL patients treated with Anti-CD20 mAb containing CIT and standard chemotherapy alone or targeted therapy

Full size image

Survival outcomes of patients with CLL on maintenance therapy with anti-CD20

A total of 4 studies [43,44,45, 51] reported on an improved PFS following maintenance therapy with anti-CD20 therapy as compared to patients who did not receive any treatment (observation group). The pooled effect estimate showed improved albeit non-significant PFS (HR = 0.51 [0.42–0.60], p = 0.93). There were no differences in OS between patients receiving maintenance therapy compared to those who were not on treatment. There were no significant differences in the pooled effect estimates (p = 0.96) and there were low levels of statistical heterogeneity amongst included studies, I2 = 0%.

Survival outcomes of patients with CLL receiving targeted therapy

In the meta-analysis, a total of eight studies [14, 39, 40, 46, 48,49,50] reported an improved PFS with novel targeted agents as compared to chemoimmunotherapy. Target therapy containing BTK and BLC2 inhibitors was associated with significantly improved PFS as compared to CIT (HR = 0.25 Cl [0.19–0.30], p = 0.07). OS data was available for seven studies [39,40,41, 46, 48,49,50]. Overall, targeted therapy was associated with improved OS (HR = 0.56 [0.33–0.80], p = 0.05). There were substantial levels of heterogeneity in the included studies (I2 = 51.67%).

Overall, the meta-analysis shows that chemoimmunotherapy and maintenance therapy with anti-CD20 antibodies is superior to chemotherapy, and targeted therapy is superior to CIT in terms of PFS with HR = 0.39 [0.31–0.47], p < 0.01 and OS (HR = 0.66 [0.53–0.78], p < 0.02 (Fig. 4). There were high levels of heterogeneity on studies assessed for PFS (I = 88.16%).

Fig. 4
figure 4

Meta-analysis of the hazards ratios (HR) for overall survival (OS) for CLL patients treated with Anti-CD20 mAb containing CIT and standard chemotherapy alone or targeted therapy

Full size image

Prognostic factors associated with poor patient outcomes in CLL patients

Prognostic markers ranged from host factors, such as age and cytogenetics, whereby 10 (58.8%) studies reported Del(17p) as a prognostic factor for PFS [14, 15, 17, 40,41,42,43,44, 47,48,49,50]. Two studies excluded patients with Del(17p) [45, 46] and in another study, del(17p) and del (11q) did not impact PFS [44]. Whereas 10 studies reported unmutated IGHV as a prognostic factor [17, 39,40,41,42, 45, 46, 48,49,50]. Trisomy 12 was identified as a prognostic factor in three studies [39, 42, 46] and TP aberrations was reported in four studies [40, 41, 48, 49].

The reported prognostic factors associated with early disease progression included elevated B2M levels (levels of ≥ 3.5 mg/L) [17, 43], thymidine kinase (concentration of 10 µ/L), white cell count (10 × 109 per L) and ECOG PS of 2 [17] and advanced disease stage III/IV [17]. After adjusting for covariates, Del(17p), unmutated IGVH status and elevated B2M (Table 4).

Discussion

We conducted a systematic review and meta-analysis of prognostic factors associated with poor survival in patients with chronic lymphocytic leukemia on CIT and novel targeted agents. The available data on the use of ICIs and targeted therapy in the management of CLL is limited to predominantly European and American populations (Table 1). The current study also highlights the lack of multi-ethnic RCTs with diverse population with CLL. The included studies reported on various candidate predictors of survival in patients with CLL on CIT and targeted therapy (Table 3).

Table 3 Characteristics of studies reporting on PFS/OS in patients with CLL on rituximab-containing regimens (n = 17)
Full size table

Amongst the reported prognostic factors only one protein factor (β2-microglobulin) retained predictive value in patients with CLL on anti-CD20-containing CIT, after multivariable analysis. Only two other prognostic factors met our criteria for confirmed prognostic factors and these included, cytogenetic factors (deletion 17p, IGHV status). Notably, in our meta-analysis we pooled studies that reported on adjusted estimates and the levels of statistical heterogeneity were high (I2 > 70%) for the confirmed cytogenetic factors and for β2-microglobulin (Table 4). Interestingly, the value of β2-microglobulin as an independent prognostic marker has not been extensively assessed in patients with CLL on CIT and targeted therapy, although in a previous study its predictive value for treatment-free survival was retained after adjusting for factors such as CD38 expression and IGHV mutation status [52].

Table 4 Overview of confirmed prognostic factor included in the meta-analysis
Full size table

The cut-off levels of B2M associated with poor prognosis remain unclear and in untreated CLL patients a value of 2 mg/L [54] while in our analysis B2M levels ≥ 3.5 mg/L [17, 43] were associated with disease progression in treated patients with CLL. Notably in the current analysis, we report on the retained predictive value of B2M in CLL patients on rituximab-containing CIT and maintenance therapy with rituximab. Future studies comprised of diverse patient populations are needed especially in minority ethnic groups to allow for validation of this prognostic marker in the era of CIT and novel targeted therapy. In the era of CIT, and chemotherapy-free CLL management, future studies evaluating the correlations between B2M levels and expression of CD20 and other immune checkpoints in patients with CLL, may assist in the stratification of patients who are most responsive to immunotherapy.

To the best of our knowledge this systematic review and meta-analysis provides the first analysis of prognostic factors in anti-CD20-containing CIT and targeted therapy. The current study has several limitations, firstly these findings are mainly derived from American and European populations. This limits the extrapolation of these findings into other low-to-middle income countries. Lastly, due to the low number of studies reporting on these prognostic factors in patients with CLL on CIT and targeted therapy, we could not explore the sources of heterogeneity in a subgroup analysis based on the potential differences in disease stage and duration of follow-up.

Conclusion

A plethora of novel prognostic factors have been described in untreated patients with CLL. However, in the era of CIT there is a lack of adequate studies exploring the predictive value of the conventional and novel prognostic factors in a multi-ethnic cohort of patients with CLL. In this systematic review and meta-analysis of prognostic factors, classical cytogenetic factors such as deletion 17p retained predictive value in patients with CLL on CIT. Lastly, the white cell count and conventional prognostic markers such as B2M and LDH levels were also regarded as confirmed prognostic factors in patients with CLL on rituximab containing CIT. These factors should be included in future prognostic factors in the era of CIT and chemotherapy-free era of CLL patient management.