Prognostic value of lymphocyte-monocyte ratio at diagnosis in Hodgkin lymphoma: a meta-analysis

Hodgkin lymphoma (HL) is a type of lymphoma of B-cell origin. About 15% of lymphomas are HL [1]. Two major subtypes of HL are classical HL and nodular lymphocyte predominant Hodgkin’s lymphoma (NLPHL). Classical HL is further subclassified into four histological subtypes: nodular-sclerosis classical HL, lymphocyte-rich classical HL, mixed-cellularity classical HL, and lymphocyte-depletion classical HL [2]. HL is one of the most common malignancies in young adults aged 20–40 years [3]. Only about 15–35% of HL patients are older than age 60 years [3‐6]. This variation in incidence by age at diagnosis is represented in Western countries by the bimodal age-incidence curve showing two peaks, first at around age 20 years, and second at around age 65 years [7]. While the bimodal curve is a defining epidemiological feature of HL, its shape varies significantly by race, socioeconomic status, geography, time, sex, and histological subtype [8‐10]. For example, for calendar years 2010 to 2014 the age standardised incidence rates for Asian and Pacific islanders and non-Hispanic whites in the US were 1.2 and 2.9 per 100,000, respectively [9].

Standard treatment depends on stage and other clinical information. It usually consists of two to 8 cycles of chemotherapy followed by radiotherapy in selected patients [3]. The treatment algorithm is largely determined by clinical parameters such as age, stage and size of disease bulk. More than 95% of early stage HL, and up to 80–90% of intermediate or advanced stage patients can achieve permanent remission and can be considered cured with modern therapy [11]. However, at least 10–20% of patients in all stages may be under- or over-treated [11]. It is important at the time of diagnosis to determine the prognosis accurately. This allows clinicians to refine and tailor the treatment strategy, to avoid undertreatment (such as inadequate cycles of chemotherapy or exposure to ineffective cytotoxic agents) for patients at higher risk of disease relapse or increased resistance to chemotherapy and radiotherapy, and prevent overtreatment for those with a high chance of having their lymphoma cured, who may be suitable for less toxic therapies [11‐13]. Patients who have their diseases relapsed after standard first line therapy need salvage treatment with high dose chemotherapy followed by autologous stem cell transplantation, and only half are successful treated [14]. On the other hand, HL long-term survivors have a two to four times increased risk of a second malignancy and cardiovascular disease compared with healthy members of the general population. This is important especially when most HL patients are young adults, and these long-term toxicities are associated with the anti-cancer treatment [15, 16]. One study evaluating the outcomes of decreased treatment intensity in a subgroup of early stage HL patients showed that the treatment effectiveness was not compromised when therapy was de-escalated, and more than 50% of all deaths during long-term follow-up were possibly related to the delivered treatment [17].

The goal of using prognostic markers to predict outcomes is to achieve a personalised approach: allowing us to provide more intensive or novel therapies (or avoid exposure to ineffective treatments) to patients with more aggressive disease, and to de-escalate therapy to patients with a high probability of achieving long-term remission, to spare the treatment toxicity.

The international prognostic score (IPS) is a standard stratification system for advanced classical HL [18]. Prognostic factors affecting clinical outcomes of NLPHL are similar to those included in the IPS [19‐21]. For early stage HL, IPS is a less appropriate risk stratification system and other prognostic scoring systems can be used, such as those from the German Hodgkin Study Group (GHSG) [17], European Organisation for Research and Treatment of Cancer (EORTC) [22], and National Cancer Institute of Canada (NCIC) [23]. These prognostic systems are based mainly on clinical parameters, such as Ann Arbor staging, and tumour sizes [17, 18, 22, 23]. They do not consider the host immune status and tumour microenvironment, which can be variable among patients with similar clinical characteristics.

A gene expression profiling study has shown that a raised number of tumour-associated macrophages (TAMs) in a pre-treatment lesional tissue sample is associated with lower survival in patients with HL [24]. These macrophages are derived from peripheral blood monocytes [25] and have been positively associated with the percentage of peripheral blood monocytes [26]. They can secrete trophic factors (that affect tumour microenvironment), which have a role in the process where tumour cells interact with stromal and immune cells, including macrophages, B-cells and T-cells. The interaction in turn leads to neovascularisation and tumour growth [27‐30]. Lymphocytes also have a role in immunosurveillance: the absolute lymphocyte count (ALC) has been shown to be a surrogate of host immune status and is an independent prognostic factor in HL [31].

Many of the advanced techniques for prognostication, including gene expression profiling [24] and immunohistochemical analysis [32, 33] have been studied. These are, however, costly and difficult to perform and interpret. A prognostic factor that is easily determined and widely available is needed.

Peripheral blood lymphocyte-to-monocyte ratio (LMR) at diagnosis may reflect the interaction between host immunity, represented by lymphocytes, and the tumour microenvironment, represented by monocytes. The peripheral blood count and cell count ratio can be determined readily and inexpensively by a standard automated complete blood count machine. Recent studies have indicated that peripheral blood LMR at diagnosis can predict long-term outcomes in haematological malignancies, including follicular lymphoma [34], diffuse large B-cell lymphoma [35], and NK/T cell lymphoma [36].

In HL, the consistency and magnitude of the prognostic value of LMR is controversial. Some studies have demonstrated a correlation between baseline LMR and survival outcomes, while one did not [37]. Different studies have, however, used different LMR cut-off values. One meta-analysis reported the prognostic value of LMR in various cancer types, and found that low LMR at diagnosis was associated with poorer cancer-specific survival and PFS in HL [38]. Since its publication in 2016, more studies on the prognostic role of LMR have been published, and three out of the seven analysed papers on HL were abstracts only. An updated analysis is needed to appraise and summarise the evidence.

This study aims to quantify the relationship between LMR at diagnosis and survival outcomes in adult HL, and to explore the impact of study characteristics on the prognostic value of LMR, by using meta-analytic techniques.

In this study, we used meta-analysis to investigate the relationship between LMR at diagnosis and survival outcomes in adult HL. Specific outcomes considered were overall survival (OS), progression-free survival (PFS), event-free survival (EFS), lymphoma-specific survival (LSS), and time-to-progression (TTP). Definitions of the different survival endpoints are summarised in Table 1 [39].

Analysis and reporting were performed according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) and Meta-analyses of Observational Studies guidelines [40, 41]. Two reviewers (SFL and TYN) independently performed the literature search, assessed study eligibility, extracted the relevant data, and performed risk of bias assessment, following the strategy set out below. Any disagreement between the reviewers was resolved through discussion and consensus.

LMR may reflect the interplay between the host immunity and tumour microenvironment. The strong association identified through the meta-analyses between baseline LMR and survival outcomes indicates that LMR could be a good prognostic biomarker.

Low LMR at diagnosis may be related to more aggressive disease nature or to poorer tolerance to anti-cancer treatment. Low peripheral blood ALC and/or high AMC can cause low LMR. Lymphopenia, defined in IPS as < 0.6 × 10⁹/L or < 8% of total white cell, is a known prognostic factor for HL [18, 31].

This is compatible with our results because low ALC can lead to low LMR if the AMC is unchanged, since LMR is a ratio. Peripheral blood ALC is thought to be a surrogate of the host’s immunity status. On the other hand, the role of monocytes is different. One study showed that the peripheral blood AMC is positively associated with TAM content in lesional tissue samples, [26]. This is consistent with the evidence that TAMs originate from blood monocytes. TAM is important in cancer cell survival and proliferation and its quantity has been shown to be associated with poorer outcomes in a number of cancers [24]. Monocyte chemoattractant protein (MCP-1) belongs to the cytokine superfamily, and its binding with receptors triggers downstream signals that regulate macrophage adhesion, and recruit TAMs. MCP-1 has been associated with tumour growth and disease progression in several kinds of cancer [59]. Furthermore, a subpopulation of immune cells called monocytic myeloid derived suppressor cells (MDSCs) has been identified in marrow and peripheral blood. Their accumulation has been associated with cancer related inflammation and immunosuppression. MDSCs are also correlated with TAMs, and an increased amount of MDSCs confers a poorer prognosis in many cancer types including HL [60‐62]. The available evidence thus demonstrates the prognostic role of non-malignant immune cells, and the possible utility of these in determining accurate prognoses.

This review has several strengths. The median follow-up periods in the included studies of patients with a well-defined disease entity were likely adequate (range 3.4–8.9 years), since most of the relapses of HL typically occur within the first 3 years [63]. LMR as a prognostic factor is biologically sound, and is determined by a simple, widely available blood test. In the subgroup analyses stratified by sample size and cut-off value, the prognostic value of LMR remained significant.

We do, however, need to be cautious in interpreting our results as all included studies were retrospective observational studies and it is possible that not all sources of heterogeneity have been adjusted for. Each included study had adjusted for some different covariables, and was conducted in different countries (one from Asia and seven from the Western world, including Israel). The differing ethnic groups and patient characteristics could be one reason for the variable effect sizes found. The optimal LMR cut-off values determined also varied across studies, possibly related to sample variability and differences in baseline patient characteristics. For example, a decrease in lymphocyte counts has been observed more frequently in the elderly [64]. The different treatment regimens (types and doses of chemotherapy agent) may also interact with other factors, such as age and performance status, resulting in variable survival outcomes.

An important limitation of this meta-analysis is that the number of included studies is modest, only eight. This limited the kinds of subgroup analyses we could perform because the number of available studies in some subgroups was low. For instance, subgroup analysis by study region could not be conducted because only one Asian study was available. This could have been useful because although ethnicity was not detailed in the studies, study regions could be surrogates of ethnic groups. It would have been useful to know whether ethnic (regional) differences modify the results, and if the findings can be applied to different ethnic groups.

Importantly, we do find evidence of publication bias in reporting of OS. Publication bias arises when the publication of findings is influenced by the nature and direction of the study results [65]. Although tests for publication bias are optimal only if the number of studies included in a meta-analysis is at least 10 [42], the interpretation of the effect estimate needs caution, as studies seem to be missing in the bottom left-hand side of the funnel plot (Fig. 6). Such studies would probably be trials with large standard errors and small sample sizes, with an overall survival for high LMR lower than that for low LMR. However, it is only an assumption that these studies were ever undertaken [66]. Asymmetry in a funnel plot can also result from poor methodological design in the identified studies, because poor methodological design typically results in the estimated effect size being spuriously magnified.

Potential problems with prognostic studies are manifold [67]. The most relevant ones for our study are: a) primary outcomes might not always be well defined, and b) there is no methodological and reporting standard, so that important information may be omitted and the importance of conclusions may be artificially inflated.

When assessing risk of bias and study quality we did find that five studies had a moderate risk of bias. Although each of the included studies had different risks of bias, their respective estimated HRs for OS and PFS were, however, similar at around 2–3. In the sensitivity analysis where we sequentially omitted each study, the results were largely stable, and no significant difference between studies of low risk and moderate risk was found.

Prior to conducting the pooled analysis, we assumed that effect sizes of the association between LMR and survival time may differ according to the methodological quality of the studies, so we decided to use a random-effects model for analysis. Furthermore, measures were taken to minimise the impact of other sources of reporting biases: we screened non-English papers and grey literature, including abstracts of studies presented at conferences and published in conference proceedings, and academic experts on lymphoma were consulted to identify further data. However, given the strong association between LMR and OS and PFS, respectively, and the observation that there was no significant change in the magnitude and direction of the pooled HRs in subgroup analyses and sensitivity analyses, the findings suggest that baseline LMR could be a significant prognostic marker in HL.

Low LMR at diagnosis is associated with poor OS and PFS. LMR therefore has the potential to be a practical biomarker that predicts survival outcomes in adult HL. A more accurate prediction in prognosis could allow clinicians to better tailor the treatment regimens according to individual risks. The test is inexpensive and easy to perform in outpatient or inpatient settings. Future prospective studies are needed to confirm its exact role in the disease management of HL, and the possibility of combining it with other established prognostic factors.