Clinicopathological characteristics and prognostic factors for canine multicentric non-indolent T-cell lymphoma: 107 cases

Canine lymphoma, as the most common haematopoietic malignancy, encompasses a group of heterogeneous diseases and even within the T-cell immunophenotype, dif-ferences in clinical presentation and responses to treatment exist. The aim of this retrospective study was to determine outcomes and prognostic factors of 107 dogs with multicentric non-indolent T-cell lymphoma (TCL) receiving lomustine-based (70%) and non-lomustine-based (30%) treatment. The majority were Labradors, Boxers, mixed-breed dogs and Dogue de Bordeaux. Eighty-six percent were substage b, 77% had mediastinal involvement, 15% had suspected bone marrow involvement and 12% had other extra-nodal sites of disease. The overall response rate to induction therapy was 80%; dogs receiving procarbazine in the induction protocol ( P = .042), dogs with neutrophil concentration below 8.7 × 10e 9 /L ( P = .006) and mitotic rate below 10 per 5 high power field ( P = .013), had greater response rates. Median progression-free survival (PFS) for the first remission was 105 days; lack of expression of CD3 on flow cytometry ( P < .0001) and pretreatment with steroid ( P = .012) were significantly associated with shorter PFS. Median overall survival time (OST) was 136 days; co-expression of CD79a ( P = .002), lack of CD3 expression on flow cytometry, presence of anaemia ( P = .007), and monocytopenia ( P = .002) were predictive of shorter OST. Multicentric non-indolent TCL in dogs is an aggressive cancer with new possible prognostic factors.

based on clinical, morphological and immunological features. The World Health Organization (WHO) classification of human Non-Hodgkin Lymphoma has been extrapolated to canine lymphoma. 13 Canine T-cell lymphoma (TCL) is a heterogeneous disease. It represents a wide spectrum of disease entities with varying responses to treatment and prognoses. The two extremes are illustrated by the typically indolent T-zone lymphoma (TZL) 14,15 and the aggressive hepatosplenic lymphoma of gamma-delta T-cell lymphocytes. 16 One study found that the most common types of non-indolent TCLs were peripheral TCL not otherwise specified (PTCLNOS-16%) and T-cell lymphoblastic lymphoma (TLBL-5%). 13 Other forms of non-indolent TCL are extra-nodal, such as gastrointestinal lymphoma with modest response to chemotherapy 17 or cutaneous lymphoma with variable clinical course. 18 Currently, most studies on prognostic factors for canine lymphoma examine a heterogeneous population of dogs with different subtypes of lymphoma. Recently, it has been identified that more defined subpopulations of lymphoma are characterized by different prognostic factors, as in the case of multicentric diffuse large B-cell lymphoma. 19,20 Currently, limited data on prognostic factors specific for TCL as a separate clinical entity is available. Some of the existing studies on canine TCL have included cases of indolent TZL or anatomical presentations other than multicentric, which may be a significant confounding factor for disease-free intervals and survival times as well as for prognostic factors assessments. Brodsky et al, 21 Rebhun et al 22  were more likely to achieve CR than other anatomical forms.
Multiple studies evaluating prognostic significance of haematologic abnormalities in human and veterinary patients with lymphoma have been published. Absolute lymphocyte concentration, 26,27 monocyte concentration (AMC), 28 neutrophil:lymphocyte ratio (NLR), 19,27,29,30 and lymphocyte:monocyte ratio (LMR) 19,31 have been the subjects of investigations. To the authors' knowledge, these have not yet been evaluated in dogs with non-indolent TCL.
The purpose of this study was to describe patient demographics, clinico-pathological abnormalities, and outcome of dogs with multicentric non-indolent TCL. We hypothesized that within the group of non-indolent multicentric TCL treated with chemotherapy, treatment type, haematologic parameters and flow cytometry characteristics would be prognostic and that patient outcomes would be inferior to previously reported studies.

| Study design and case collection
Record of dogs with non-indolent multicentric TCL that were referred to three referral centres in the United Kingdom between January 2009 and June 2018 were retrospectively reviewed. Patients with non-indolent lymphoma were included in the study if they had cytological or histologic diagnosis of lymphoma and T-cell immunophenotype. Dogs were excluded if a definitive diagnosis was not achieved, if immunophenotyping was not available, if cytological or histopathological morphological characteristics were suggestive of a "low grade" or TZL, or if neoplastic cells were lacking expression of CD45. Dogs with non-multicentric forms such as gastrointestinal, cutaneous and primary hepato-splenic lymphomas were excluded.
Clinical records were reviewed and follow-up data were obtained from existing medical records and requested from referring veterinarians by phone calls.

| Data collection
For each patient, the following data were recorded: signalment, body weight, age at diagnosis, haematology results, calcium measurements, clinical signs at presentation, method of diagnosis (cytology or histopathology) and immunophenotyping.
Haematological abnormalities were defined based on the reference intervals provided by the laboratories concerned. Blood smear reports (if available) were reviewed to assess for the presence of circulating neoplastic cells, and to differentiate between true thrombocytopenia and post-sampling platelet aggregation if automated low platelet count was reported. The abnormalities were recorded prior to any treatment initiation. Median absolute leukocyte concentrations were used to calculate the NLR and LMR ratios. Patients were described as "hypercalcemic" when free calcium levels or when total calcium levels were elevated without concurrent hyperalbuminemia and clinical signs typically associated with hypercalcemia of malignancy, such as polyuria/polydipsia were present.
When available, blood smears and cytological slides of tissue and fluid aspirates were reviewed by a single board-certified clinical pathologist (L. M. P.). For the remaining cases, information was obtained from the original cytopathology reports. Details recorded included: nuclear size, nuclear shape, nucleoli (number, size, prominence), cytoplasmic features, mitotic counts, presence of necrosis or capillaries and prominence of tingible body macrophages. Mitotic rate was defined as number of mitotic figures per five ×40 or ×50 high power fields (hpf). Where sufficient information was available, lymphoma was tentatively classified based on the WHO classification scheme, adapted to canine lymphoma, 13 as well as previous publications on mediastinal TCLs in dogs. 32 Cases were classified as lymphoma of granular lymphocytes (LGL) if they contained magenta cytoplasmic granules, typically located in one focal perinuclear area.
Lymphoma consisting of intermediate size nuclei (1.5-2 RBC in diameter) with finely granular chromatin and inconspicuous nucleoli were tentatively classified as lymphoblastic lymphomas (LBL). The remainder of the cases, having a large nucleus (>2 RBC) and/or prominent nucleoli, were grouped together as TCLs not otherwise specified (TCL-NOS).
The methods of immunophenotyping were recorded for each case. For cases with flow cytometry available, receptor expression patterns were recorded. Aberrant expression pattern on flow cytometry was based by previously published criteria. 33 Peripheral lymph node involvement was defined as enlargement or firm consistency of peripheral lymph nodes on physical examination or cytological/histological confirmation of lymphoma. Dogs presenting with clinical signs associated with systemic illness were classified as substage b. Full staging was performed at the attending clinician's discretion but was not required for inclusion.
Diagnostic imaging modalities and findings were recorded for each case. If internal lymph nodes were enlarged, they were classified as being involved. For liver and spleen, it was recorded whether the involvement was confirmed with cytology or histology.
Where the information was available, presence of neoplastic lymphocytes in bone marrow aspiration or in peripheral blood smears was used to classify cases as stage V according to WHO classification. All dogs included in the study were classified according to WHO staging criteria.
Treatment protocols, clinical response, date of progression, rescue treatments, date and cause of death and necropsy findings (if available) were recorded for each patient. Similar to previous studies, dogs were classified as receiving steroids if there was any history of continuous steroid administration for longer than 10 days prior to commencing chemotherapy. 19 Chemotherapy induction protocols were classified as 1-LOP (lomustine-and vincristine-based, with/without L-asparaginase, procarbazine or cytarabine), 2-COP (vincristine-and cyclophosphamide-based, with/without L-asparaginase or cytarabine), 3-CHOP or CEOP where doxorubicin was replaced by epirubicin, with/without L-asparaginase or cytarabine, 4-lomustine-based without vincristine (lomustine with/without L-asparaginase or procarbazine) 5-prednisolone with/without Lasparaginase only. Subsequently, chemotherapy induction protocols were divided into lomustine-containing protocols vs others.
Response to first-line and rescue chemotherapy treatments was based on the Veterinary Cooperative Oncology Group response evaluation criteria for peripheral nodal lymphoma 34 and diagnostic imaging findings for cases with internal involvement only. Cases were classed as being in CR when lymph nodes (both peripheral or internal) had returned to normal size; partial remission (PR) when lymph nodes remained enlarged but had reduced in size by at least 30% and no new lesions were recognized; progressive disease (PD) was used for occurrence of new lesions or increase in size of enlarged lymph nodes by at least 20%; and stable disease (SD) as a change in size of lymph nodes which was not sufficient to be classified as PD or PR with no occurrence of new lesions. In cases of solely internal involvement where no imaging was available to directly measure the response, a significant improvement in clinical signs was classified as PR. Response had to be sustained for a minimum of 28 days to be classified as a CR or PR. Dogs that died or were euthanized within 1 week of starting a treatment were considered as non-responders.
The objective response rate was defined as the sum of the cases with a CR or a PR.
PFS was defined as the period of time between treatment onset and disease progression or death from any other cause. Dogs were censored for PFS if lost to follow-up before progression occurred, if still alive and in CR at the end of the follow-up period or if progression did not occur before they died from cause other than lymphoma. OST was defined as time between treatment onset and death from any cause. Dogs were censored for OST if lost to follow-up, or still alive at the end of the study period.

| Statistical analysis
Frequency and proportion were used to summarize categorical variables; median (minimum and maximum) was used for numerical data. Fisher's exact test was used to compare sex/neutering status, stage and substage of the dogs between institutions. Kruskal-Wallis tests were used to compare body weight and age distributions between institutions. Kaplan-Meier curves were used to depict survival curves and estimated median survival time reported.
Analytes with skewed results were divided into thirds (using tertile cut-off points) prior to further analysis. Univariable and multivariable Cox regression were employed to evaluate predictors of PFS and OST. Results were presented as hazard ratio (HR) and 95% confidence intervals (CI). Two sets of univariable and multivariable binary logistic regression were used to assess predictors of responders (CR + PR) vs non-responders (SD + PD), and CR vs PR of disease. Results were presented as odds ratio (OR) and 95% CI. Results of univariable analysis are placed in a supplementary material (Table S1). Variables that had P values <.1 in the univariable analysis were included in the multivariable analysis.
Backward elimination method was used to obtain final multivariable models. Because of missing data, several models were developed with or without mitotic rate, CD3 or CD79a expression in the analysis. Significance level was set as 5%. Analyses were carried out in R 3.5.1-R Core Team (2018).

| Staging
Results of staging are summarized in Table 1. According to WHO staging system, 58 dogs (54%) were categorized as at least stage III, 18 (17%) as stage IV and 31 (29%) as stage V.

| Treatment and response rates
Two (2%) dogs were euthanized without any treatments and these dogs were excluded from the PFS and OST analysis. Seven dogs (7%) received prednisolone prior to commencing chemotherapy treatment.
Only nuclear shape (P = .017) and L/M ratio above 2.3 (P = .01) remained statistically significant in the multivariable analysis ( Table 2).
Univariable analysis of the factors significantly associated with shorter PFS is included in the supplementary file (Table S1). Since not all of the lymphomas were investigated with flow cytometry, expression of CD3 was recorded in 56/107 dogs (52%) only. Risk factors associated with shorter PFS in the multivariable models of either including or excluding CD3 expression are presented in Table 3.  Factors associated with shorter OST on univariable analysis are presented in the supplementary file (Table S1). Multivariable analyses of excluding either CD3 or CD79a or both factors were presented in Table 4.

T A B L E 2 Multivariable analysis of induction response
OR (95% CI) P value

Model 1 responders vs non-responders (n = 58)
Lack of procarbazine in the induction protocol 8

| DISCUSSION
In this study, we describe the clinical and pathological findings from a homogenous population of dogs with non-indolent multicentric TCL.
We found that the median PFS and OS of the whole group were lower than previously reported for dogs with TCL, [21][22][23][24][25] which suggests that prognosis for this lymphoma subtype is worse than described for TCL in general. However, a selection bias in the present study owing to case recruitment solely from specialist referral centres cannot be rule out, and could explain why our study population contained a large proportion of patients in clinical stage IV and V, with a majority in substage b (86%).
Stage and substage have been previously reported as prognostic in canine lymphoma 35,36,57 but we were unable to confirm this in our study. Results from previous studies may have been confounded by a variability of heterogeneous lymphoma subtypes included but it is also possible that the number of dogs with certain factors was too low in our study to show a statistical significance.
Not surprisingly, one of the breeds most commonly represented in our study was the Boxer. Several studies showed increased prevalence of TCL in this breed. 21,22,37,38 In a recent study by Morgan et al, 25 the PFS was shorter in the Boxer breed, this association was not seen in our study, possibly because of low numbers of Boxers included. More interestingly, Dogue de Bordeaux's predisposition to TCL has been previously reported among canine population in Poland, 38 and more recently in Australian population, 39  showed that presence of a mediastinal mass negatively impacted PFS.
Mediastinal involvement did not influence any of the outcomes in our study, similarly to others. 22 The response to induction chemotherapy treatment was high (80% for all chemotherapy protocols and 86% for lomustine-based protocols). This rate was similar in some studies [21][22][23]25 and lower in others. 24 Multivariable analysis showed that inclusion of procarbazine in the induction protocol was associated with higher likelihood of response to chemotherapy. Increased expression of one of the drug transporters of the ATP-Binding Cassette superfamily, BCRP (ABCG2), was shown in canine TCL. 42 As a main mechanism of lymphoma resistance to anthracyclines, this might explain low response rates to doxorubicin in dogs with TCL. 43 Lomustine and procarbazine, as alkylating agents, are not typical substrates for the ABC transporters, which might explain their potential benefit. Although there was a benefit in PFS and OST for dogs receiving lomustinebased induction chemotherapy in our study, this difference was not statistically significant comparing to other types of treatment.
In the multivariable analysis, lack of expression of CD3 was associated with a shorter PFS. Deravi et al 44  This study has limitations, such as its retrospective nature. Inclusion criteria aimed to identify a homogenous population of multicentric TCL in dogs, however in one third of the study population, T-cell immunophenotype was determined only by demonstration of a clonal population of T cells by PARR. PARR is not a preferred tool for immunophenotyping as cross-lineage rearrangement has been reported. 55 Flow cytometry was shown superior over PARR for immunophenotyping, 56

ACKNOWLEDGEMENT
The author (K. P.) thanks Professor Christopher R. Lamb for advice and support throughout the residency and in writing this manuscript.