Prognostic Value of Blood Eosinophilia in Primary Cutaneous T-Cell Lymphomas

Objective  To investigate the prognostic value of initial characteristics including blood eosinophilia in patients with primary cutaneous T-cell lymphoma.

Design  A retrospective inception cohort, patients included from date of diagnosis (1982-1998).

Setting  Two dermatology departments of a university hospital.

Patients  A total of 104 patients with cutaneous T-cell lymphoma, including patients with mycosis fungoides (n = 69), Sézary syndrome (n = 13), and nonepidermotropic cutaneous lymphoma (n = 22). The following variables were recorded: age, sex, diagnosis according to the European Organization for Research and Treatment of Cancer (EORTC) classification, type of skin involvement at the time of diagnosis, initial eosinophil absolute count, lactate dehydrogenase value, date of disease progression, and cause and date of death or date of last contact.

Main Outcome Measures  Time from diagnosis to disease progression and to disease-specific death.

Results  The median follow-up was 43 months (range, 7-197 months). Estimated rates of disease progression and disease-specific death for 3 years were 19.5% (95% confidence interval [CI],11.3%-27.6%) and 9.9% (95% CI, 2.8%-13.6%), respectively. Univariable analysis of initial variables possibly influencing disease progression revealed significant prognostic value for diagnosis according to EORTC classification (hazard ratio [HR], 2.77; 95% CI, 1.04-7.41; P = .04), type of skin involvement (HR, 2.70; 95% CI, 1.00-7.25; P = .04), raised blood eosinophil absolute count (HR, 7.33; 95% CI, 2.84-18.91; P<.001), and raised serum level of lactate dehydrogenase (HR, 3.72; 95% CI, 1.58-8.78; P = .001).Concerning disease-specific death, significant prognostic indicators were diagnosis according to the EORTC classification (HR, 6.62; 95% CI, 1.68-26.12; P = .007) and a raised blood eosinophil absolute count (HR, 10.57; 95% CI, 2.28-49.0; P<.001). In multivariable analysis, only blood eosinophilia was associated with disease progression and disease-specific death.

Conclusion  These results strongly suggest that blood eosinophilia at baseline is a prognostic factor in patients with primary cutaneous T-cell lymphoma.

Primary cutaneous T-cell lymphomas (CTCLs) are characterized by a proliferation of T lymphocytes in the skin, most often of the CD4⁺ helper subtype. They comprise various entities, distinguishable on the basis of clinical, histologic, and immunophenotypic criteria, with variable prognoses, as defined by the European Organization for Research and Treatment of Cancer (EORTC) classification for primary cutaneous lymphomas¹: mycosis fungoides (MF), the most common group of CTCLs, and CD30⁺ large T-cell lymphomas (CD30⁺) have an indolent clinical course; Sézary syndrome (SS) and CD30⁻large T-cell lymphomas (CD30⁻) have an aggressive clinical course; and pleomorphic small- to medium-sized CTCLs (PSMs) have a prognosis not yet well defined. Several studies, mostly focusing on the epidermotropic lymphomas of the MF and SS groups, have attempted to identify clinical, biological, histopathologic, or immunophenotypic characteristics that can predict outcome. So far, the main prognostic factors identified in this group are the type and extent of skin involvement, extracutaneous spread of the disease, initial response to treatment, histologic transformation, high serum level of lactate dehydrogenase (LDH), and the detection of a cutaneous or peripheral blood T-cell clone by polymerase chain reaction.^2-14

As with other T-cell malignancies and Hodgkin disease, CTCLs are frequently associated with eosinophilia.^3,13-18 In this context, eosinophilia has been related to the predominant secretion of T helper cell type 2 (T_H2) eosinophilopoietic or eosinophilotactic cytokines (interleukin [IL] 3, IL-5, and sargramostim) by neoplastic cells.^19-26 Since T_H1 cells are likely to play a key role in the initiation and the persistence of an antitumoral response, while a predominant T_H2 differentiation has been associated with a relative defect of the antitumoral and anti-infectious response,²⁷ the hypothesis that eosinophilia might be an indicator of poor prognosis was raised. Accordingly, several studies investigating eosinophilia in Hodgkin disease found a worse relapse-free survival rate in patients with eosinophilia than in those without,^28,29 despite 1 report of better overall survival;²⁹ in a more recent study, tissue eosinophilia has been shown to be the strongest prognostic factor for poor relapse-free survival and overall survival in nodular sclerosing Hodgkin disease.³⁰ In patients with CTCL, the relevance of blood eosinophilia has been suggested in 1 study,¹³ but so far, results from this early report have not prompted large-scale studies investigating this latter variable. In the present study, we retrospectively investigated the prognostic value of blood eosinophilia in a cohort of 104 patients with CTCL.

A total of 154 cases of CTCL were recorded in the hospital pathology register between January1982 and December 1998. For the present study, only patients with sufficient clinical information and follow-up were selected. For each patient, the diagnosis was based on the combination of clinical, histologic, and immunophenotypic criteria of the EORTC classification for primary cutaneous lymphomas. The study group included patients with MF, SS, and nonepidermotropic lymphoma (NEL): CD30⁺, CD30⁻, and PSM. Diagnostic criteria for SS, on which there is at present no consensus, were as follows: erythroderma, generalized lymphadenopathy, the presence of Sézary cells in skin and lymph nodes, and the presence of at least 5% circulating large Sézary cells.

The files of the included patients were retrospectively analyzed. The following variables were recorded: age at diagnosis, sex, diagnosis according to the EORTC classification, type of skin involvement at the time of diagnosis, date of disease progression, cause and date of death or date of last contact, whether there was a raised blood eosinophil absolute count (>700 × 10⁹/L, upper limit of the 95% confidence interval [CI] of normal) and a raised serum level of LDH (>200 U/L, 10% or more above the upper limit of the normal range) at the time of diagnosis. Disease progression was defined as lymph node, visceral, or blood involvement in patients with previously more limited disease and/or histologic transformation.

For survival analysis, we considered only disease-specific death, defined as death related to CTCL or its specific treatment. Unrelated deaths (3 patients) were censored. All censored criteria were calculated from the date of diagnosis to the occurrence of an event (disease progression or disease-specific death) or last contact. Distributions over time were estimated by the Kaplan-Meier product limit method.³¹ The log-rank statistic³² was used to test the prognostic value of patient characteristics at the time of diagnosis for the occurrence of the event, and hazard ratios (HRs) with 95% CIs were obtained by use of univariable Cox models.³³ Multivariable Cox models were used to determine independent prognostic variables.

Because of a limited number of events, no more than 2 variables were tested jointly in multivariable Cox models. Two-sided tests were computed, and P values of .05 or lower were considered significant. We used SAS software, version 8.2 (SAS Institute, Cary, NC), for our analysis.

Of the 154 cases initially recorded, 50 were excluded for insufficient information (n = 34), diagnosis of lymphomatoid papulosis (n = 6), or a doubtful diagnosis (n = 10); thus, 104 patients were finally included: 66 men and 38 women with a median age of 56.5 years (range, 21-92 years). According to the EORTC classification, the diagnoses were 69 MF, 13 SS, and 22 NEL (7 CD30⁺, 4 CD30⁻, and 11 PSM). The main characteristics of patients and the type of skin involvement at the time of diagnosis are listed in Table 1. None of the patients with MF or NEL had evidence of extracutaneous disease at the time of diagnosis. Of the 104 patients, 24 (23%) had an increased serum level of LDH, and 28 (27%) presented with blood eosinophilia at the time of diagnosis (Table 2).

The initial treatment modality most commonly used for epidermotropic CTCL was topical mechlorethamine hydrochloride. Most patients with solitary or localized indolent NEL received radiotherapy, while patients with more generalized or aggressive NEL or epidermotropic CTCL with visceral involvement mostly received multiagent chemotherapy as first-line treatment.

The median follow-up was 43 months (range, 7-197 months). Twenty-one patients were lost to follow-up, their median follow-up being 13 months (range, 6-64 months).

The initial treatment resulted in complete clinical remission in 42 patients; 12 (28%) of them had subsequent relapse, with a median delay of 21 months (range, 4-84 months); 35 patients had partial or no remission. In 27 patients, information on the effect of initial therapy was insufficient. Estimated rates of disease progression and disease-specific death are listed in Table 3.

Univariable analysis of initial variables possibly influencing disease progression revealed that diagnosis according to EORTC classification (SS vs MF), type of skin involvement (erythroderma vs patches), raised blood eosinophil absolute count, and a raised serum level of LDH at the time of diagnosis were significant prognostic factors (Table 4). The prognostic influence of eosinophilia is presented in the Figure 1. In multivariable analysis, only eosinophilia at the time of diagnosis was associated with disease progression.

Concerning disease-specific death, the diagnosis according to the EORTC classification and a raised blood eosinophil absolute count at the time of diagnosis were statistically significant prognostic factors in univariable analysis (Table 5). In multivariable analysis, only eosinophilia at the time of diagnosis was associated with disease-specific death.

We further restricted the prognostic analysis to the epidermotropic MF and SS groups and found the same prognostic variables as in the whole group. Univariable analysis of initial variables possibly influencing disease progression revealed the following significant prognostic factors: diagnosis according to the EORTC classification (SS vs MF) (HR, 2.80; 95% CI, 1.05-7.48; P = .03); type of skin involvement (erythroderma vs patches) (HR, 2.91; 95% CI, 1.05-8.03; P = .04); raised blood eosinophil absolute count (HR, 7.31; 95% CI, 2.60-20.55; P<.001); and a raised serum level of LDH at the time of diagnosis (HR, 3.65; 95% CI, 1.44-9.27; P = .004). In multivariable analysis, only eosinophilia at the time of diagnosis was associated with disease progression.

In univariable analysis, significant predictive factors for disease-specific death were diagnosis according to the EORTC classification (HR, 7.30; 95% CI, 1.74-30.59; P = .001) and a raised blood eosinophil absolute count at the time of diagnosis (HR, 7.77; 95% CI, 1.65-36.6; P = .002). In multivariable analysis, only eosinophilia at the time of diagnosis was associated with disease-specific death.

In the present study, the prognostic value of the main initial characteristics of 104 patients with primary CTCL was investigated. Our results confirmed that diagnosis according to the EORTC classification, type of skin involvement, and increased LDH values are important prognostic factors in univariable analysis in patients with CTCL. In multivariable analysis, only blood eosinophilia at diagnosis was an independent prognostic factor for disease progression and disease-specific death in these patients. Direct comparison with previous prognostic studies is not feasible because these studies did not include eosinophilia in their analyses.^{2,3,6,7,10,13,14}

Our study population was similar to most studies in age and male-female ratio.^5,10,34 The estimated disease-specific death rate was also similar to that reported in the more recent studies.^6,10,34,35

In our patients, primary CTCL was associated with blood eosinophilia in up to 20% of cases, comparable with previous reports.^3,13,14 As in other types of lymphomas, it is presumed that in this context eosinophilia is due to the secretion of cytokines of the T_H2 pattern by neoplasic cells.^19-26 Indeed, it has been shown that patients with advanced CTCL exhibit a depressed cell-mediated immunity and that lymphocytes from these patients are deficient in IL-2 and interferon γ production while showing an increased production of IL-4, IL-5, and IL-10.^36-42 On the other hand, the T_H1 signal observed in the early stages of the disease is interpreted to indicate either normal trafficking lymphocytes or a tumor-infiltrating host response with a predominance of T_H1 cytokines.^37,39,41

In view of these notions, one might hypothesize that eosinophilia in patients with CTCL would correlate with the increased tumor burden and lead to excessive IL-5 production. However, since only eosinophilia at baseline was investigated in the present retrospective analysis, this latter hypothesis must be tested in further prospective, large-scale studies; changes in blood eosinophilia with clinical evolution or after treatment as well as tissue eosinophilia remain to be investigated. Such research might help to identify subgroups of patients who might benefit from immunotherapeutic approaches such as interleukin 12, interleukin 2, and interferon gamma, which are likely to correct the abnormal cytokine production observed in CTCL.^25,43-47 However, our results strongly suggest that blood eosinophilia at baseline should now be considered a prognostic factor of poor outcome in patients with CTCL.

Correspondence: Emmanuelle Tancrède-Bohin, MD, Policlinique de Dermatologie, Hôpital Saint-Louis, Assistance Publique, Hôpitaux de Paris 1, avenue Claude Vellefaux, 75475 Paris CEDEX 10, France (emmanuelle.tancrede-bohin@wanadoo.fr).

Accepted for publication September 24, 2003.

This work was presented in part at the annual meeting of the French Society of Dermatology, Journées Dermatologiques de Paris;December 6-9, 2000; Paris, France.