CHOP療程為基礎的化學治療在犬多中心型淋巴瘤的毒性反應以及腫瘤治療效果間的相關性

(1)

國立臺灣大學獸醫專業學院臨床動物醫學研究所碩士論文

Institute of Veterinary Clinical Science School of Veterinary Medicine

National Taiwan University Master Thesis

CHOP 療程為基礎的化學治療在犬多中心型淋巴瘤的毒性反應以及腫瘤治療效果間的相關性

Correlation between Toxicity and Antitumor Efficacy of CHOP-based Chemotherapy

in Canine Multicentric Lymphoma

郭璇 Kuo Hsuan

指導教授：李繼忠博士 Advisor: Jih-Jong Lee, Ph.D.

中華民國 103 年 7 月 July 2014

(2)

Abstract

This study was aimed at investigating the correlation between chemotherapy

toxicity and antitumor efficacy in canine multicentric lymphoma. Medical records of 69

dogs with multicentric lymphoma received CHOP-based chemotherapy at National

Taiwan University Veterinary Hospital were reviewed, and impact of bone marrow

toxicity and GI toxicity on time to tumor progression (TTP), overall survival time

(OST), and short-term tumor-killing effect was evaluated. Neutrophil nadir lower than

5000 /μl improved OST (P =0.045). Presence of GI signs, including anorexia, vomiting,

and diarrhea, or presence of only vomiting improved TTP and OST (P =0.042, 0.007 for

TTP; 0.023, <0.001 for OST). Presence of diarrhea of grade 3 or 4 decreased TTP and

OST (P =0.034, 0.017). More than 10 years old was associated with less low-grade GI

toxicities and less favorable outcome. Occurrence of neutropenia or GI toxicity after a

treatment increased the like hood of effective treatment over ineffective treatment,

implying a positive relationship between toxicity and short-term efficacy. The results of

the study supported the concept of toxicity-adjusted dosing, but prospective trials are

warranted to develop a sophisticated toxicity-adjusted dosing regimen.

Key words: chemotherapy, bone marrow toxicity, GI toxicity, lymphoma, CHOP, dogs

(3)

摘要

本研究目的為探討犬多中心型淋巴瘤的化療毒性以及療效間的相關性。六十九隻罹患多中心型淋巴瘤且在國立台灣大學附設動物醫院接受以 CHOP 療程為基礎的化學治療的犬隻的病歷被回顧，並且分析其中骨髓毒性以及消化道毒性對於疾病進展時間(Time to tumor progression, TTP)，總存活時間(Overall survival time,

OST)，以及短期的腫瘤抑制效果的影響。嗜中性球低點小於5000/μl 可增進 OST (P

=0.045)。出現食慾不振、嘔吐、腹瀉其中一種消化道毒性或者僅出現嘔吐皆可增加 TTP 及 OST (P =0.042, 0.007 對於 TTP; 0.023, <0.001 對於 OST)。出現毒性分級三或四的腹瀉會降低 TTP 及 OST (P =0.034, 0.017)。年齡超過十歲較不容易出現毒性分級低的消化道毒性，同時也有較不理想的治療成果。在單次治療後出現嗜中性球低下或者消化道毒性，會增加該次治療成為有效治療而非無效治療的可能性，

顯示毒性以及短期療效之間存在正向關係。本研究的結果支持依據毒性反應來調整藥物劑量的概念，但是必須靠前瞻性研究才能建立起完善的依據毒性做劑量調整的規範。

關鍵字：化學治療，骨髓毒性，消化道毒性，淋巴瘤，CHOP，犬

(4)

Table of Contents

Abstract --- I

Table of contents --- II

List of Tables --- VI

List of Figures --- VIII

1. Introduction --- 1

2. Literature Review --- 3

2.1 Canine multicentric lymphoma and CHOP-based chemotherapy --- 3

2.2 Correlation between toxicity and antitumor efficacy of chemotherapy --- 4

2.2.1 Related research in human medicine --- 4

2.2.2 Related research in veterinary medicine --- 8

3. Aim --- 11

4. Materials and Methods --- 12

4.1 Patient selection --- 12

4.2 Chemotherapy protocol --- 12

4.3 Diagnosis and staging --- 14

4.4 Response and toxicity assessment --- 15

(5)

4.5 Prognostic factors --- 17

4.6 Long-term analysis --- 18

4.7 Short-term analysis --- 20

4.8 Statistics --- 22

5. Results --- 24

5.1 Patient characteristics --- 24

5.2 Response and toxicity --- 25

6. Discussion --- 36

6.1 Patient characteristics, overall response, and toxicity profile --- 36

6.3.1 Neutropenia and efficacy --- 38

6.3.2 GI toxicity and efficacy --- 41

6.3.3 Timing of toxicity during the protocol and efficacy --- 42

(6)

6.3.4 Toxicity after different drugs and efficacy --- 43

6.3.5 Frequency of toxicity and efficacy --- 44

6.3.6 Age and GI toxicity --- 45

6.3.7 Confounding factors --- 46

6.3.8 Multivariate analysis --- 46

6.9 Toxicity-adjusted dosing --- 48

6.10 Limitations --- 50

7. Conclusion --- 52

8. References --- 54

(7)

List of Tables

Table 1 Human oncology studies establishing relationship between toxicity and

efficacy. --- 65

Table 2 Selected veterinary studies investigating relationship between toxicity and

efficacy. --- 66

Table 3 The 6-month, maintenance-free, modified version of the University of

Wisconsin (UW)-Madison chemotherapy protocol (UW-25) utilized in this study. --- 67

Table 4 Modified neutropenia grading system based on VCOG-CTCAE v1.0. --- 68

Table 5 Five-digit coding system for long-term analysis. --- 69

Table 6 Neutropenia profile presented as the number and percentage of patients

experienced certain type of toxicity of certain grade. --- 70

Table 7 GI toxicity profile presented as the number and percentage of patients

experienced certain type of toxicity of certain grade. --- 71

Table 8 Median TTP and OST and P values for prognostic factor analysis. --- 72

Table 9 P values for TTP and OST of the 154 groupings in long-term analysis. --- 73

Table 10 Median TTP and OST and P values for 15 groupings regardless of factors of

timing, drug, and frequency. --- 75

(8)

Table 11 Intergroup P values for the 4 groupings dividing patients into three groups in

Table 10. --- 77

Table 12 Median TTP and OST and P values for the 3 frequency-adjusted groupings

with P <0.05 and their non-frequency-adjusted counterparts. --- 78

Table 13 Numbers and percentages of patients in different toxicity status of different

age groups, grouping 31110, 41140, 51110, and 51140. --- 79

Table 14 Numbers and percentages of patients in different toxicity status of different

response groups, grouping 11110, 11140, 31110, 41140, 51110, and 51140. --- 80

Table 15 Numbers and percentages of patients in different toxicity status in different

anemia status in grouping 11310. --- 82

Table 16 Results of short-term analysis. --- 83

(9)

List of Figures

Figure 1 Kaplan-Meier’s curve for TTP of all patients --- 84

Figure 2 Kaplan-Meier’s curve for OST of all patients --- 84

Figure 3 Kaplan-Meier’s curve for TTP, age --- 85

Figure 4 Kaplan-Meier’s curve for OST, age --- 85

Figure 5 Kaplan-Meier’s curve for TTP, WHO clinical stage --- 86

Figure 6 Kaplan-Meier’s curve for OST, WHO clinical stage --- 86

Figure 7 Kaplan-Meier’s curve for TTP, WHO clinical substage --- 87

Figure 8 Kaplan-Meier’s curve for OST, WHO clinical substage --- 87

Figure 9 Kaplan-Meier’s curve for TTP, anemia --- 88

Figure 10 Kaplan-Meier’s curve for OST, anemia --- 88

Figure 11 Kaplan-Meier’s curve for TTP, response --- 89

Figure 12 Kaplan-Meier’s curve for OST, response --- 89

Figure 13 Kaplan-Meier’s curve for TTP, time to finish the first two cycles --- 90

Figure 14 Kaplan-Meier’s curve for OST, time to finish the first two cycles --- 90

Figure 15 Kaplan-Meier’s curve for OST of grouping 11110 --- 91

(10)

Figure 18 Kaplan-Meier’s curve for TTP of grouping 31110 --- 92

(11)

(12)

1 Introduction

Chemotherapeutic agents cause DNA damage or interfere with specific cell

cycles, thus are most effective on rapid-dividing cells, such as tumor cells,

gastrointestinal epithelium cells, and bone marrow cells. As a result, chemotherapy can

induce both favorable antitumor responses and unfavorable adverse events.

Possible positive correlation between toxicity and antitumor effects has long been

noticed in both human and veterinary oncology practice. In human medicine, the

intensity of chemotherapy toxicity and efficacy have been both linked to

pharmacodynamics in numerous studies with various chemotherapeutic agents and

tumor types. There are also several clinical studies showing patients with hematological

toxicity are prone to have better treatment outcomes. However, relevant clinical analysis

in veterinary medicine is little. As more and more evidence suggested that body surface

area-based dosing, the traditional way to dose chemotherapeutic agents, poorly adapts to

inter-patient variation, toxicity-adjusted dosing may be a supplemental method to

improve dosing accuracy, and more investigation is needed before this approach can be

practically performed.

(13)

This study is aimed at retrospectively analyzing the correlation between

chemotherapy toxicity and efficacy in canine multicentric lymphoma, focusing on

gastrointestinal and bone marrow impacts, in both long-term aspect and short-term

aspect, utilizing the medical records of the National Taiwan University Veterinary

Hospital.

(14)

2. Literature Review

2.1 Canine multicentric lymphoma and CHOP-based chemotherapy

Lymphoma is the most common hematopoietic neoplasm of dogs. Eighty-four

percent of dogs with lymphoma developed the multicentric form, which is usually

characterized by the presence of superficial lymphadenopathy. Without treatment, most

dogs with lymphoma will die of their disease in 4 to 6 weeks after diagnosis, although

significant variability exists (Vail et al., 2013).

Over the last 30 years, the standard of care for dogs with multicentric lymphoma

has evolved from single-agent chemotherapy protocols to combination chemotherapy protocols. The duration of protocols also changed from indefinite to 6 months or less.

The standard of care combination protocols are now generally recognized as

“CHOP-based” protocols, consisting of cyclophosphamide, hydroxyl-daunorubicin (doxorubicin), Oncovin (vincristine), and prednisone. There are many variations of this

particular combination of drugs. Variations from this protocol include differences in the

order of drug administration, addition of L-asparaginase or methotrexate to the protocol,

slight differences in drug doses, and increased or decreased protocol duration (Chun,

2009). Currently randomized prospective evidence does not strongly recommend one

(15)

protocol over the other as long as the basic CHOP components are present.

CHOP-based chemotherapy induces remission in approximately 80%~95% of dogs,

with overall median survival times of 10 to 12 months. Approximately 20% to 25% of

treated dogs will be alive after initiation of these protocols (Vail et al., 2013).

Generally, CHOP-based chemotherapy protocols were well tolerated by dogs with

lymphoma. In a clinical trial using a 6-month, maintenance free, CHOP-based protocol,

22 in 53 (41.5%) dogs requiring a treatment delay or dose modification due to bone

marrow or gastrointestinal toxicities, but only 5 (9.4%) dogs needing hospitalization

(Garret et al., 2002).

Many factors have been shown to influence treatment response and survival of

canine lymphoma. The well-established negative prognostic factors included WHO

clinical stage V, WHO clinical substage b, T-cell phenotype, presence of anemia at

diagnosis, hypercalcemia, and prolonged steroid pre-treatment (Vail et al., 1996;

Khanna et al., 1998; Marconato et al., 2011; Jagielski et al., 2002).

2.2 Correlation between toxicity and antitumor efficacy of chemotherapy

2.2.1 Related research in human medicine

(16)

Most studies focusing on both chemotherapy toxicity and antitumor efficacy

investigated the issue of how to dose chemotherapeutic agents precisely. To achieve

maximum tumor-killing effect, the dose should be as high as possible. But concurrent

chemotherapy toxicity sets the limit of dose escalation, since severe toxicity can

compromise life quality or even cause mortalities. In fact, all chemotherapeutic agents

are characterized by a narrow therapeutic window and significant variability in

therapeutic and toxic effects. Current body surface area (BSA)-based dosing regime

fails to adapt to interpatient or intrapatient pharmacodynamic variability, which leads to

unstandardized systemic anticancer drug exposure, despite under the same dose of the

same drug (Hon et al., 1998; Gao et al., 2008).

Hon et al. summarized a substantial amount of studies demonstrating the

relationship between systemic exposure and both efficacy and toxicity (Hon et al.,

1998). Systemic exposure of certain drug was measured by various pharmacokinetic

parameters in different studies, such as systemic clearance, steady-state plasma

concentration (Cpss), and area under the concentration–time curve (AUC). Some of the

studies linked systemic exposure to antitumor efficacy, evaluated by response to

treatment, overall survival time, or disease-free survival time, according to different

(17)

study designs, whereas some of the studies linked systemic exposure to toxicity,

including hematological toxicity, gastrointestinal toxicity, and ototoxicity. The majority

of studies focused on hematological toxicity. All of the studies showed a similar result:

the higher systemic exposure, the higher efficacy or toxicity.

The relationship between systemic exposure and both efficacy and toxicity was

investigated most extensively for 5FU, a widely used chemotherapeutic agent. AUC of

5FU was highly correlated with hematological and gastrointestinal toxicities in patients

with head and neck cancer. An AUC threshold value of 30000 µg/L．h was highly

predictive of toxicity (Thyss et al., 1986). The half-cycle and full-cycle AUC (AUC0 –

3days and AUC0–5days) were also higher in toxic than in nontoxic cycles (Santini et al.,

1989). Patients with an average AUC per cycle for all 3 cycles >29 000 µg/L．h

exhibited longer survival (Milano et al., 1994). Note the optimal threshold AUC for

survival being very close to the maximum tolerated AUC for toxicity.

Apart from systemic exposure, some clinical studies directly investigated the

relationship between hematological toxicity and efficacy. Gao et al summarized several

studies associating neutropenia during chemotherapy with increased survival in patients

with several different tumors (Gao et al., 2008). For instance, three studies of

(18)

node-positive early breast cancer treated with either cyclophosphamide, methotrexate

and fluorouracil or cyclophosphamide, doxorubicin, and ftorafur demonstrated that low

leukocyte nadir after chemotherapy can lead to increased distant disease-free survival or

overall survival (Poikonen et al., 1999; Saarto et al., 1997; Colleoni et al., 1998).

Studies directly linked non-hematological toxicity with efficacy was sparse. One

study of advanced hepatocellular carcinoma treated with sorafenib showed that

development of grade 3 to 4 diarrhea was associated with increased overall survival

(Koschny et al., 2013). However, sorafenib is a target therapy drug, not a traditional

anticancer drug, thus not fully corresponded to the scope of the current study. Three

studies demonstrated that the occurrence of hand-foot syndrome (HFS), a particular

presentation of skin toxicity, was associated with better outcome in colorectal cancer

patients treated with capecitabine, with or without other chemotherapeutic agents

(Stintzing et al., 2011; Hofheinz et al., 2012; Twelves et al., 2012). Interestingly, in one

of the three studies, gastrointestinal toxicity and diarrhea were significantly more

common in patients with HFS but not often co-incident with hematological toxicities

(Hofheinz et al., 2012).

(19)

Human oncology studies establishing relationship between toxicity and efficacy

are summarized in Table 1.

Some studies investigating correlation between toxicity and efficacy explained the

connection by interpatient pharmacodynamic variations (Stintzing et al., 2011; Rankin

et al., 1992; Mayers et al., 2001; Cameron et al., 2003; Di Maio et al., 2005): The

response of cancer cells to chemotherapy depends on a sufficient amount of active drug

reaching the target. These factors also apply to healthy cells. The availability of active

drug at tumor cells or healthy cells is affected by pharmacokinetic factors (ie, the

metabolism, distribution, and catabolism) of drugs, which produce a similar effect in

tumor cells and healthy cells. In addition, some studies also proposed a view of

interpatient genetic variations (Di Maio et al., 2005): The sensitivity of tumor cells and

healthy cells is affected, by genetic predisposition, which can similarly affect both cell

types on the same patient, but is also modified by tumor-specific acquired resistance.

2.2.2 Related research in veterinary medicine

(20)

To the author’s knowledge, only two studies in veterinary medicine have

established a positive correlation between chemotherapy toxicity and antitumor efficacy.

The first study retrospectively investigated the impact of chemotherapeutic dose

intensity and hematologic toxicity on first remission duration in dogs with lymphoma

treated with a chemoradiotherapy protocol (Vaughan et al., 2007). The study result

showed that development of grade III or IV neutropenia during chemotherapy was

associated with prolonged first remission duration. The second study implemented a

dose-intense CHOP-based chemotherapy protocol for canine lymphoma, and found that

dogs required dose reductions and treatment delays had significantly longer time to

tumor progression and lymphoma-specific survival times (Sorenmo et al., 2010).

Some other veterinary clinical studies of various cancers included the occurrence

of toxicity in prognostic value analysis, but none have found a relationship between

toxicity and outcome. Four studies of canine lymphoma treated with CHOP-based

chemotherapy showed that toxicity had no influence on either disease-free interval or

overall survival (Simon et al., 2006; Keller et al., 1993; Zemann et al., 1998; Garrett et

al., 2002). The definition of occurrence of toxicity differed from one study to another.

One study regarded toxicity of any grade, according to the Veterinary Co-Operative

(21)

Oncology Group’s common terminology criteria for adverse events v1.0 (VCOG-

CTCAE v1.0), as occurrence of toxicity, and recorded number of neutropenic episodes

and degree and number of gastrointestinal toxicosis episodes for evaluation (Simon et

al., 2006). Two studies regarded toxicities that caused treatment change as occurrence

of toxicity (Keller et al., 1993; Garrett et al., 2002), whereas the other study defined

occurrence of toxicity as neutrophil less than 1000 /µl or hospitalization for GI adverse

events (Zemann et al., 1998). As we can see, the criteria for toxicity were relatively

loose in the first study, and stricter in the three latter studies. These differences could

affect the result of analysis. There are also one study of canine appendicular

osteosarcoma and two studies of canine urinary bladder transitional carcinoma finding

that hematological and gastrointestinal toxicity not correlated with outcome (Bacon et

al., 2008; Chun et al., 1997; Marconato et al., 2011). Except one transitional carcinoma

study defined neutropenia as neutrophil less than 2000 /µl, the other two studies utilized

loose toxicity criteria by including toxicities of all grade, according to VCOG- CTCAE

v1.0, into analysis.

Selected veterinary studies investigating relationship between toxicity and efficacy

are summarized in Table 2.

(22)

3. Aim

The purpose of this study was to investigate the correlation between chemotherapy

toxicity and antitumor efficacy in canine multicentric lymphoma, in either long-term or

short-term aspect.

(23)

4. Materials and Methods

4.1 Patient selection

Medical records of dogs with multicentric lymphoma between January 2000 and

December 2014 at The National Taiwan University Veterinary Hospital were reviewed

retrospectively. Dogs that were cytologically or histologically diagnosed with

multicentric high-grade lymphoma and received CHOP-based chemotherapy for the

treatment of lymphoma without any chemotherapy prior to CHOP were included in the

present study. Dogs that failed to finish at least the first two cycles of CHOP due to

reasons other than lymphoma-related death (ie, tumor progression or severe

chemotherapy toxicity) were excluded from the study.

4.2 Chemotherapy protocol

This study utilized a 6-month, maintenance-free, modified version of the

University of Wisconsin (UW)-Madison chemotherapy protocol (UW-25), which is

provided in Table 3 (Garrett et al., 2002). Dose reductions of 20% to 30% and

(24)

3000 /µl or severe gastrointestinal toxicity, depending on the clinician’s preferences and

the patient’s whole body status.

Because of the retrospective nature of the study, protocol adjustments were

frequently seen for the included patients. The standard UW-25 protocol administered

one treatment of vincristine, followed by cyclophosphamide, and again vincristine, and

then doxorubicin; so one full cycle of UW-25 can be abbreviated as VCVA (ie, V for

vincristine, C for cyclophosphamide, and A for doxorubicin). But in this study, dogs

were sometimes treated by a “VCVCA” protocol, or even “VCVCVCA”. The other

frequent protocol adjustments were prolonged treatment intervals and dose reductions

without actual occurrence of hematological or gastrointestinal toxicity. In addition,

there were many premature treatment cessations, by which many patients did not

receive four cycles of chemotherapy. All these protocol adjustments were attributed to

practical requirements, such as client compliance, economic considerations, patient age,

and clinician’s preferences.

The included dogs were treated by 5 different veterinarians, respectively, among

which there was a senior veterinarian being the mentor of all the other 4 veterinarians,

leading to a similar treatment approach for all dogs.

(25)

For dogs that went out of remission or became resistant to CHOP protocol, various

rescue drugs were offered, including L-asparginase, lomustine, dacarbazine,

actinomycin D, and DMAC protocol (ie, dexamethasone, melphalan, actinomycin D,

and cytosine arabinoside).

4.3 Diagnosis and staging

Of the 69 dogs included in the study, the diagnosis of lymphoma was made by

biopsy of lymph node in 4 dogs, by aspiration cytology of lymph node in 65 dogs.

All dogs were clinically staged at diagnosis by means of a modification of the

World Health Organization (WHO) 5-stage criteria for canine lymphoma (Owen, 1980).

Dogs were assigned to stage V if neutropenia or circulating lymphoblasts were detected

in peripheral blood. Dogs were assigned to stage IV if no evidence of stage V disease

and hepatomegaly or splenomegaly noted in radiography or ultrasonography, or

heterogeneous texture of liver or spleen noted in ultrasonography. Because of lack of a

standard staging manner of all dogs, some dogs with more advanced lymphoma (stage

V or IV) could be classified as less advanced stage.

(26)

4.4 Response and toxicity assessment

Overall response to treatment was classified as complete remission (CR), partial

remission (PR), and no response (NR). The status at which a patient had the least tumor

burden during treatment was taken for evaluation. Complete remission was defined as

disappearance of all target lesions, and any pathological lymph nodes must have

reduction in short axis to <10 mm. Partial remission was defined as at least a 30%

decrease in the sum of diameters of target lesions. Responses other than complete

remission and partial remission was defined as no response.

Besides overall response to treatment, this study documented detailed toxicity and

response during the treatment course. For each treatment, the following data was

recorded: drug, time from diagnosis, gastrointestinal toxicity, hematological toxicity,

response to that treatment, and whether dose reduction or not.

Gastrointestinal toxicities recorded included anorexia, vomiting, and diarrhea, and

graded based on the Veterinary Co-operative Oncology Group common terminology

criteria for adverse events v1.0 (VCOG-CTCAE v1.0). Because anorexia of grade 1 was

difficult to be identified from medical records, none of it was documented. For

(27)

treatments that medical records were too obscure to tell the occurrence of GI toxicity, or

actual cause of GI adverse event could not be determined, the corresponding GI toxicity

to that treatment would be assigned as unavailable, and excluded from analysis.

Hematological toxicities recorded included only neutropenia, and graded according

to VCOG-CTCAE v1.0, except that the criteria of grade 1 neutropenia was adjusted to

meet the needs of the study and further sub-graded to grade 1.0 and grade 1.1, as

provided in Table 4. The presence of neutropenia after a treatment would be assigned as

unavailable and excluded from analysis if blood work was not performed in two weeks

after vincristine, cyclophosphamide, and chlorambucil, or four weeks after doxorubicin,

mitoxantrone, lomustine, dacarbazine, and actinomycin D.

Response to a single treatment was classified as complete remission (CR), partial

remission (PR), stable disease (SD), progressive disease (PD), and uncertain response.

Definitions of complete remission and partial remission were similar to that for overall

response evaluation mentioned above. Progressive disease was defined as at least a 20%

increase in the sum of diameters of target lesions, or occurrence of new target lesions,

such as lymphoblasts in peripheral blood. Responses with neither sufficient shrinkage to

(28)

Responses not clearly recorded in the medical records were assigned as uncertain

response.

4.5 Prognostic factors

Two endpoints for response were evaluated for prognostic significance. Time to

tumor progression (TTP) was defined as the period of time (in days) from diagnosis to

progressive disease or relapse. Overall survival time (OST) was defined as the period of

time (in days) from diagnosis to death. Patients who lost follow-up were censored on

the last day of contact. Patients still in remission or alive at the end of the study were

censored on the last day of data collection (2014/5/30) for TTP or OST analysis,

respectively.

The clinical factors evaluated for potential prognostic significance included: sex

and neuter status, age (grouped as ≤5 y/o, 6~10 y/o, and ≥11 y/o), body weight (<15 kg

or not), immunophenotype of neoplastic cells if available, WHO clinical stage, WHO

clinical substage, presence of hypercalcemia at diagnosis, presence of anemia at

diagnosis, pre-treatment with steroid, response to treatment (complete remission, partial

(29)

remission, or no response), and time to finish the first two CHOP cycles (<80 days or

not).

4.6 Long-term analysis

Long-term analysis investigated the relationship between toxicity and overall

outcome, using time to tumor progression (TTP) and overall survival time (OST) as

endpoints.

Chemotherapy toxicity could happen with different presentation forms (ie,

neutropenia, anorexia, vomiting, and diarrhea), after different drugs, at different time

during the protocol, with various severities (ie, grade 1 to 4), and at various frequencies.

By shifting these variables, the criteria of occurrence of toxicity also changed. For

instance, defining toxicity of any grade after any drugs at any time during the protocol

as occurrence of toxicity is a loose criterion, whereas defining toxicity of higher than

grade 3 happened more than 3 times during the first two cycles as occurrence of toxicity

is a very strict criterion. Setting appropriate criteria might be crucial to establish a

relationship between toxicity and efficacy.

(30)

To thoroughly examine the influences of these variables on the appropriateness of

toxicity criteria, this study developed a 5-digit coding system, as provided in Table 5.

The first digit represents different forms of toxicity: 1 stands for neutropenia; 2 stands

for anorexia; 3 stands for vomiting; 4 stands for diarrhea; 5 stands for combined GI

signs, consisting of anorexia, vomiting, and diarrhea. The second digit represents

different timing in the protocol: 1 stands for full course; 2 stands for the first two cycles.

The third digit represents different drugs: 1 stands for all drugs, 2 stands for vincristine;

3 stands for cyclophosphamide; 5 stands for doxorubicin. The fourth digit represents

various toxicity grades: 1 stands for taking toxicity of all grade as occurrence of toxicity;

3 stands for taking only toxicity of high grade as occurrence of toxicity; 4 stands for

dividing patients into no toxicity, low-grade toxicity and high-grade toxicity, with

slightly different threshold between neutropenia and the other GI signs, as shown in

Table 5. Toxicity grade for a patient was assigned as the highest grade along all

treatments under consideration. The fifth digit represents various frequencies: 0 stands

for toxicity occurrence at least once; 3 stands for more than twice; 5 stands for more

than four times.

Each 5-digit code corresponds to a particular toxicity criterion, and patients were

(31)

grouped according to these criteria. For example, code 11110 divides patients into two

groups: Patients who experienced at least once neutropenia of any grade at any time

after any drug, and patients who did not. Another example, code 31540 divides patients

into three groups: Patients who experienced at least once vomiting of grade 1 or 2 at any

time after doxorubicin, patients who experienced at least once vomiting of grade 3 or 4

at any time after doxorubicin, and patients who did not experience vomiting at any time

after doxorubicin. Total 360 (5 x 2 x 4 x 3 x 3 =360) groupings were made. After

excluding groupings dividing patients into two groups with less than 5 patients in one

group or dividing patients into three groups with less than 5 patients in two groups, 154

groupings remained. Kaplan-Meier curves of TTP and OST were plotted for the 154

groupings to detect outcome differences.

4.7 Short-term analysis

Short-term analysis investigated the relationship between toxicity and efficacy

caused by the same single treatment, using the toxicity and response records for each

treatment.

(32)

Neutropenia of more than grade 1.1 was considered as occurrence of toxicity.

Anorexia, vomiting and diarrhea were evaluated together as combined GI signs, and

toxicity grade of a treatment was assigned as the highest grade within the three. GI

toxicity of any grade was regarded as occurrence of toxicity.

Treatments were categorized as effective treatment, ineffective treatment, and

uncertain treatment based on tumor response. Effective treatments cause partial

remission or the very first complete remission in a row of complete remissions.

Ineffective treatments lead to progressive disease. Treatments with uncertain response,

stable disease, and subsequent complete remission after the initial complete remission

were defined as uncertain treatments.

Whether occurrence of toxicity had impact on the like hood of effective treatment,

ineffective treatment, and uncertain treatment was examined. Treatments were analyzed

either all together, or separately according to drugs, including vincristine,

cyclophosphamide, and doxorubicin.

(33)

4.8 Statistics

Differences in outcome (TTP and OST) according to potential prognostic factors

(sex and neuter status, age, body weight, immunophenotype, WHO clinical stage and

substage, presence of hypercalcemia at diagnosis, presence of anemia at diagnosis,

pre-treatment with steroid, response to treatment, and time to finish the first two CHOP

cycles) were assessed by the Kaplan-Meier log-rank test. Factors with a P value <0.05

were regarded as significant prognostic factors.

For long-term analysis, differences in outcome (TTP and OST) according to the

154 groupings were assessed by the Kaplan-Meier log-rank test. For groupings with P

value <0.05, chia-square test was performed to examine the correlation between

grouping and significant prognostic factors. Multivariate Cox regression analysis was

used to evaluate selected groupings with a P value <0.05 and significant prognostic

factors for their independent association with TTP and OST.

For short-term analysis, chia-square test was performed to investigate the

association between occurrence of toxicity and response. Odds ratio of the odds of

effective treatment to ineffective treatment was calculated.

(34)

Statistical significance was defined as P < 0.05. All analyses were performed using

SPSS statistical software Version 18.

(35)

5. Results

5.1 Patient characteristics

Total 69 dogs were included in this study. Medical records of 175 dogs with

lymphoma were reviewed, and 106 dogs was excluded due to not receiving CHOP

protocol, dropping out of CHOP protocol before the end of the first two cycles, or

incomplete medical record.

≤5 y/o, 6~10 y/o, and ≥11 y/o

Thirty-six dogs were male (24, 34.8%, neutered and 12, 17.4%, intact) and 33 dogs

were female (10, 14.5%, neutered and 23, 33.3%, intact). There were 49 purebred dogs:

Golden Retriever (n=15) was the most common breed represented. Other breeds were

Beagle (n=5), Chihuahua (n=4), Maltese (n=3), Schnauzer (n=3), Shi Tzu (n=3), Bull

terrier (n=2), English Cocker Spaniel (n=2), Corgi (n=2), Labrador Retriever (n=2),

Pomeranian (n=2), Yorkshire Terrier (n=2), Dachshund (n=1), Pug (n=1), Rottweiler

(n=1), and Bichon Frise (n=1). Nineteen dogs were mixed breed dogs. Breed was

unknown for 1 dog. The mean age was 7.5 years (range, 2–14 years), with 21 (30.4%)

dogs ≤5 years old, 34 (49.3%) dogs within 5 to 10 years old, and 14 (20.3%) dogs ≥11

(36)

years old. The mean body weight was 18.8 kg (range, 1.9–57.1 kg), with 33 (47.8%)

dogs <15 kg and 36 (52.2%) dogs >15 kg. The immunophenotype of the lymphoma

cells were B-cell type in 22 (32%) dogs, T-cell type in 2 (3%) dogs, and not determined

in 45 (65%) dogs. By WHO clinical staging standards, 33 (48%) dogs were in stage III,

23 (33%) dogs were in stage IV, and 13 (19%) dogs were in stage V. Forty-four (64%)

dogs were in substage a and the 25 (36%) remaining dogs were in substage b. Two (3%)

dogs had hypercalcemia at diagnosis. Twenty-five (36%) dogs were anemic at diagnosis.

Eight (12%) dogs received steroids prior to CHOP protocol.

5.2 Response and toxicity

Forty-six dogs achieved CR, and 19 dogs achieved PR. No response to treatment

was observed in 4 dogs. Total response rate (CR + PR) was 94%. Median time to tumor

progression (TTP) was 185 days (range, 16-831 days). Median overall survival time

(OST) was 282 days (range, 32-841 days). Kaplan-Meier curves of TTP and OST are

shown in Figure 1~2.

Forty-eighty (69.6%) dogs were dead due to lymphoma. Five (7.2%) dogs were

euthanized due to lymphoma progression. Two (2.9%) dogs were dead after seizure and

(37)

lymphoma involvement was highly suspected. Nine (13%) dogs lost follow-up, with

median and mean follow-up time 518 and 795 days, respectively. Three (4.3%) dogs

were still alive at the end of the study, with follow-up time 1318, 1745, and 2358 days,

respectively.

Twenty-four dogs (35%) finished the first two cycles less than 80 days, whereas 40

(58%) dogs more than 80 days. Five (7%) dogs were dead due to lymphoma or severe

chemotherapy toxicity before finishing the first two cycles.

Toxicity profile was presented as the number of patients experienced certain type

of toxicity of certain grade, as listed in Table 6~7. The protocol was generally well

tolerated. Anorexia, vomiting, and diarrhea of grade 4, which necessitated

hospitalization, occurred in only 0 (0%), 1 (1%), and 2 (3%) dogs, respectively.

Neutropenia of more than grade 1.1, which necessitated treatment delay, occurred in 47

(50%) dogs, whereas grade 2 and 3 neutropenia occurred in only 10 (15%) dogs. 2 (3%)

dogs were dead due to chemotherapy toxicity.

(38)

Age, WHO clinical stage, presence of anemia at diagnosis, response to treatment,

and time to finish the first two CHOP cycles were identified as significant for TTP. Age,

WHO clinical stage, WHO clinical substage, presence of anemia at diagnosis, response

to treatment, and time to finish the first two CHOP cycles were identified as significant

for OST. More than 10 years old was related to worse outcome. WHO clinical stage III,

WHO clinical substage a, absence of anemia at diagnosis, and finishing the first two

cycles >80 days were associated with better outcome. Patients in the three categories of

response to treatment exhibited different outcome, with CR being the best and NR being

the worst. Median TTP and OST and P values according to above factors are provided

in Table 8. Kaplan-Meier curves for TTP and OST of above factors are demonstrated in

Figure 3~14.

Immunophenotype was available for only 24 (35%) dogs. Presence of

hypercalcemia at diagnosis, and pre-treatment with steroid were only noted in 2 (3%)

and 8 (12%) dogs. These three factors were excluded from analysis due to low case

numbers.

(39)

In the 154 groupings, statistical significance was found in 9 and 15 groupings for

TTP and OST, respectively. All P values for TTP and OST of the 154 groupings are

provided in Table 9. To best illustrate these results, some groupings were selected and

compared in order to elucidate the influences of the five variables (form of toxicity,

timing during the protocol, drug, toxicity grade, frequency) on the correlation between

toxicity and efficacy.

Form of toxicity and toxicity grade

Table 10 included P values for 15 groupings which set timing during the protocol

as full course, drugs as all drugs, and frequency as at least once (ie, first, second and

fifth digit of coding system fixed to 1). In other words, these 15 groupings focused on

only form of toxicity and toxicity grade, neglecting the other three variables. The

continued part of Table 10 is composed of groupings dividing patients into three groups,

and the exact P values between each group are provided in Table 11.

Based on results of grouping 11110, patients who experienced neutropenia of more

than grade 1.0 (ie, neutrophil <5000 µl) had longer OST (P =0.045) than patients who

(40)

did not experience any episode of neutropenia, whereas there was no difference in TTP

(P =0.094). If shifting definition of occurrence of toxicity to neutropenia of more than

grade 2 (ie, neutrophil <1500 µl), as in grouping 11130, differences could be only

observed in the Kaplan-Meier curve for OST (Figure 16), but no statistical significance

was detected in either TTP (P =0.151) or OST (P =0.063). If dividing patients into no

toxicity, low-grade toxicity (grade 1.0 and 1.1) and high-grade toxicity (grade 2 and 3),

as in grouping 11140, statistical difference was found in OST (P =0.044), with actually

only low-grade toxicity group superior to no toxicity group (P =0.039) and high-grade

toxicity group superior to no toxicity group (P =0.005), but no significant differences in

low-toxicity group and high-toxicity group (P =0.055). The Kaplan-Meier curves for

OST for the three neutropenia groupings were demonstrated in Figure 15~17.

Groupings with first digit assigned as 2 examined anorexia and efficacy. The three

anorexia groupings (21110, 21130, and 21140) in Table 10 were all statistical

insignificant. In fact, no statistical significance was found in any anorexia groupings in

the study.

Groupings with first digit assigned as 3, 4, and 5 focused on vomiting, diarrhea and

combined GI signs. In groupings 31110, 41110, and 51110, toxicity of any grade was

(41)

considered as occurrence of toxicity, and statistical significance was found in vomiting

and combined GI signs, but not diarrhea, for TTP (P =0.042 and 0.007) and OST (P

=0.023 and <0.001). If shifting definition of occurrence of toxicity to toxicity of more

than grade 3, as in groupings 31130, 41130, and 51130, differences were only detected

in diarrhea groupings for both TTP (P =0.034) and OST (P =0.017), and in contrary to

the study’s hypothesis of toxicity improving efficacy, patients who experienced diarrhea

of more than grade 3 exhibited poorer outcome. The Kaplan-Meier curves of the above

groupings with P value <0.05 were shown in Figure 18~23. After dividing patients into

no toxicity, low-grade toxicity (grade 1 and 2) and high-grade toxicity (grade 3 and 4),

as in grouping 31140, 41140, and 51140, statistical significance was found in diarrhea

and combined GI signs for TTP (P =0.015 and 0.017) and OST (P =0.004 and 0.001),

and in vomiting for OST (P =0.04). Looking into the Kaplan-Meier curves, as shown in

Figure 24~29 for these groupings, one can tell that low-toxicity groups performed

better than no toxicity groups, but high-toxicity groups diminished the favorable

prognostic value of toxicity, leading to similar or worse outcome comparing to no

toxicity or low-grade toxicity groups. Detailed intergroup P values for those groupings

are provided in Table 11. Patients with high-grade vomiting or combined GI signs had

(42)

similar outcome to patients with low-grade vomiting or combined GI signs, although

poorer outcome in high-grade toxicity group could be subjectively observed in the

Kaplan-Meier’s curves; Patients with high-grade diarrhea had poorer outcome than both

patients with low-grade diarrhea and no diarrhea.

Timing during the protocol

No groupings focusing on only toxicities appeared in the first two cycles of the

protocol (ie, the second digit assigned as 2) showed statistical significance.

Drugs

Among groupings that focusing on particular drug (ie, the third digit assigned as 2,

3, or 5), statistical significance was found for TTP and OST in grouping 11310 (P

=0.042 and 0.019), 41230 (P =0.022 and 0.015), and 41240 (P =0.031 and 0.014), and

for only OST in grouping 51210 (P=0.03). The results of grouping 11310 suggested that

patients who experienced neutropenia of more than grade 1.0 after cyclophosphamide

had better outcome. The results of grouping 41230 and 41240, which defined

occurrence of toxicity as diarrhea of more than grade 3 after vincristine or divided

patients into 3 groups according to diarrhea grading after vincristine, were similar to the

(43)

results of diarrhea groupings in the previous section: diarrhea of high grade was

associated with poorer outcome. The results of grouping 51210 suggested that patients

who experienced anorexia, diarrhea, or vomiting of more than grade 1 after vincristine

had longer OST, but not TTP. The Kaplan-Meier curves for these four groupings with P

value <0.05 were provided in Figure 30~36.

Frequency

Among groupings that examined if frequency of toxicity is a determinant (ie, the

fifth digit assigned as 3 or 5), statistical significance was found for OST in grouping

31113 (P =0.015) and 31143 (P =0.014), and for TTP and OST in grouping 41143 (P

=0.045 and 0.019). The results of these three groupings were similar to their

non-frequency-adjusting counterparts (ie, 31110 for 31113, 31140 for 31143, and 41140

for 41143): In grouping 31113, patients who experienced vomiting of any grade more

than three times exhibited better outcome, as in grouping 31130; In groupings 31143

and 41143, patients who experienced vomiting or diarrhea of grade 1 or 2 more than

twice had better outcome than patients who did not experience vomiting or diarrhea

more than twice, but patients who experienced vomiting or diarrhea of grade 3 or 4

(44)

low-grade toxicity group, as in grouping 31140 and 41140. Table 12 listed P values of

the three above-mentioned groupings and their no frequency-adjusting counterparts.

Figure 37~40 showed the Kaplan-Meier curves with P value <0.05.

For all grouping with P value <0.05, excluding drug-specific groupings (ie, the

third digit assigned as 2, 3, or 5) and frequency-adjusting groupings (ie, the fifth digit

assigned as 3 or 5), chia-square test was performed to examine the correlation between

grouping and significant prognostic factors. Age, response to treatment and anemia

were found to be associated with some of these groupings: Age was correlated to

grouping 31110, 31140, 41140, 51110, and 51140 (P =0.015, 0.016, 0.016, 0.028, and

0.009); Response to treatment was correlated to grouping 11110, 11130, 31110, 41110,

51110, and 51140 (P =0.004, <0.001, 0.006, 0.001, < 0.001, and 0.001); Anemia was

correlated only to grouping 11310 (P =0.022). P values and percentages of the above

comparisons, are listed in Table 13~15. As demonstrated in Table 13~14, more than 10

years old was associated with lower percentage of toxicity in grouping 31110, and

lower percentage of low-grade toxicity in grouping 31140, 41140, and 51140, whereas

CR was related to higher percentage of toxicity in grouping 11110, 31110, and 51110,

higher percentage of low-grade toxicity in grouping 41140 and 51140, and higher

(45)

percentage of both low-grade and high-grade toxicity in grouping 11140. These results

indicated that the connections found was plausible: More than 10 years old, a negative

prognostic factor, was with less patients in the favorable toxicity group; CR, a positive

prognostic factor, was with more patients in the favorable toxicity group.

In multivariate analysis, all groupings with P value <0.05, excluding drug-specific

groupings and frequency-adjusting groupings, and all significant prognostic factors

were included for Cox regression. For both TTP and OST, WHO clinical stage, time to

finish the first two cycles of the protocol, and grouping 11140 (ie, neutropenia, dividing

into no toxicity, low-grade toxicity and high-grade toxicity groups) remained statistical

significance (P <0.001, =0.011, and 0.008 for TTP; P <0.001, = 0.002, and 0.001 for

OST).

5.5 Short-term analysis

In short-term analysis for neutropenia, statistical differences were detected in

comparisons including treatment with all drugs (P =0.008) and with only vincristine (P

=0.013). Odds ratio of the odds of effective treatment to ineffective treatment were 0.33

(46)

and 0.23, respectively, demonstrating that with occurrence of neutropenia, effective

treatment was more likely to happen than ineffective treatment.

In short-term analysis for combined GI signs, statistical differences were detected

in comparisons including treatment with all drugs, with vincristine, or with

cyclophosphamide (P <0.001, <0.001, and =0.003). Odds ratio of the odds of effective

treatment to ineffective treatment were 0.53, 0.56, and 0.42, respectively, demonstrating

that with occurrence of GI toxicity, effective treatment was more likely to happen than

ineffective treatment.

Table 16 listed all the P values and odds ratios in the short-term analysis.

(47)

6. Discussion

6.1 Patient characteristics, overall response, and toxicity profile

The patient characteristics of this study resembled clinical experiences at the

author’s hospital, with stage III and substage a the most common, anemia sometimes

observed, and hypercalcemia very rare. Besides the low incidence of hypercalcemia, the

patient characteristics were also comparable to many lymphoma studies (Zemann et al.,

1998; Hosoya et al., 2007; Garrett et al., 2002; Moore et al., 2001; Chun et al., 2000;

Simon et al., 2006).

The response rate of this study was 94%, coincident with 80-95% by other

CHOP-based protocols, where as median OST was 282 days, similar to or slightly less

than the survival time of 10-12 months by other CHOP-based protocols (Vail et al.,

2013).

The toxicity profile was comparable to the other study using the same 6-month,

maintenance free, CHOP-based protocol as this study (Garret et al., 2002). In this study,

only 2 (3%) dogs developed GI sign of grade 4 and were hospitalized, and treatment

delay consequent to neutropenia occurred in 47 (50%) dogs. In the study by Garret et al.,

(48)

5 (9.4%) dogs needed hospitalization and 53 (41.5%) dogs required treatment delays. In

addition, the toxicity profile also resembled clinical experiences at the author’s hospital.

The previously proved prognostic factors, such as WHO clinical stage, WHO

clinical substage, anemia, and response to treatment were also established as of

prognostic significance in this study. The other two prognostic factors in this study, age

and time to finish the first two cycles of the protocol, were somehow inconsistent

findings comparing to previous studies.

The result for WHO clinical stage was slightly different from the well-confirmed

connection of stage V disease and poor outcome. Instead, better outcome was associated

with stage III, superior to stage IV and V. This deviation could due to lack of

standardized staging tests for each patient, thus falsely assigning patients to lower

stages. Particularly, because no bone marrow aspiration was performed in any patients

in this study, many stage V diseases could be underestimated as stage IV.

It is a common clinical observation at the author’s hospital that patients with very

(49)

young age (ie, < 2 y/o) were prone to have poor outcome. But only few studies have

proposed that older patients were more likely to have poorer outcome (Zemann et al.,

1998; Hosoya et al., 2007; Myers et al., 1997). The majority of lymphoma studies

showed that age was not a significant prognostic factor (Valerius et al., 1997; Hahn et

al., 1994; Keller et al., 1993; Kiupel et al., 1999; Price et al.,1991; MacEwen et al.,

1987; Greenlee et al., 1990; Garrett et al., 2002; Simon et al., 2006; Moore et al., 2001).

Finishing the first two cycles of the protocol >80 days was associated with better

outcome. This finding was in contrast to some theories that supported high-intensity

chemotherapy (Sorenmo et al., 2010). Finishing the first two cycles >80 days could be

caused by toxicities and consequent treatment delays, but the chia-square tests

performed in long-term analysis did not find any correlation between toxicity and time

to finish the first two cycles. More aggressive treatment could also be implemented for

more advanced disease, but hard to be verified from the current data.

6.3.1 Neutropenia and efficacy

(50)

Based on the findings of this study, setting a threshold of neutrophil <5000 /µl

could differentiate patients from having longer OST to shorter OST (P =0.043), whereas

a threshold of neutrophil <1500 /µl had weaker power of differentiation (P =0.063).

Setting an appropriate neutropenia threshold could be crucial to finding a connection

between neutropenia and efficacy. In the other two veterinary studies establishing a

relationship between toxicity and efficacy, thresholds of neutropenia were set as <1000

/µl (Vaughan et al., 2007) and <1500 /µl (Sorenmo et al., 2010), respectively. It would

be informative to know how the results of the two previous studies would change by

shifting threshold to 5000 /µl.

The result that neutrophil <1500 /µl was a weaker threshold than neutrophil <5000

/µl might suggest that high-grade neutropenia is not more favorable than low-grade.

This was similar to the conclusion of a human lung cancer study (Di Maio et al., 2005)

that the presence, but not severity, of chemotherapy-induced neutropenia were

prognostic for increased survival. However, in the Kaplan-Meier curve for OST of

patients dividing into no toxicity, low-grade toxicity, and high-grade toxicity (Figure

17), the trend of high-grade toxicity performing the best and no toxicity performing the

worst could be observed, while statistical significance was only detected between no

(51)

toxicity group and low-grade toxicity group (P= 0.039), and no toxicity group and

high-grade toxicity group (P=0.005), but not low-grade toxicity group and high-grade

toxicity group (P =0.055), with a P-value slightly exceeded 0.05. These findings

provided weak evidence of severe toxicity ensuring even better outcome, and strictly

speaking left the question unanswered. Nevertheless, since no neutropenia-related

illness or mortality occurred in this study, and theoretically severity of neutropenia

parallels systemic exposure and thus tumor-killing effect, high-grade toxicity leading to

even better outcome could be a plausible result.

Another study of human breast cancer (Cameron et al., 2003) showed that only

moderate neutropenia (grade 1~3, neutrophil ranged from lower normal limit to 500 /µl)

was associated with increased survival, while grade 4 (neutropenia <500 /µl)

neutropenia produced similar outcome as no neutropenia. It should be noted that grade 4

neutropenia never occurred in this study, and the range of moderate neutropenia of the

human breast cancer study was identical to low-grade plus high-grade neutropenia in

this study, so that the results of the two studies were actually comparable.

(52)

6.3.2 GI toxicity and efficacy

Vomiting and combined GI signs were found to be strong determinants for better

outcome, while diarrhea was a weak determinant and anorexia was not a determinant.

To the author’s knowledge, this is the first report in the veterinary literature establishing

the correlation between GI toxicity and antitumor efficacy. The inconsistence between

anorexia and other GI signs could be explained by the fact that the actual causes of

anorexia are difficult to decide, so that episodes of anorexia not secondary to

chemotherapy may be easily misinterpreted as chemotherapy toxicities. The intense

negative impact of high-grade diarrhea on outcome, as illustrated in the following, made

diarrhea a weaker determinant than vomiting and combined GI signs.

The strongest statistical significance (ie., P<0.001) of all groupings was found in

two combined GI signs groupings, grouping 51110 and 511140. The result could

suggest that evaluating all GI signs together was a better method to assess toxicity. As

the same injuries to GI tract could present as various clinical signs, assessing different

GI signs separately could fail to demonstrate true significance.

(53)

In contrast to neutropenia, high-grade GI toxicity was proved not to produce even

better outcome. High-grade vomiting and combined GI signs led to shorter TTP or OST

than low-grade toxicity, and high-grade diarrhea led to shorter TTP and OST outcome

than both no diarrhea and low-grade diarrhea. These findings illustrated the intrinsic

difference between bone marrow toxicity and GI toxicity: GI toxicity is related to more

morbidity and mortality, and overall body condition would be compromised following

severe GI clinical signs, thus prevent favorable outcome.

6.3.3 Timing of toxicity during the protocol and efficacy

Tumor cells, but not normal cells, are recognized as capable of developing

resistance to chemotherapeutic agents gradually along treatments (Gupte et al., 2013).

According to this theory, a drug at certain dose could cause sufficient systemic exposure

and thus toxicity and antitumor efficacy after the initial treatments, but as resistance

developed, only toxicity remained without concurrent efficacy. However, this scenario

did not take place in this study, because timing of toxicity was not proved to be

significant for outcome. An identical result was also found in a human colorectal cancer

(54)

study (Hofheinz et al., 2012), showing that patients developing skin toxicity during the

first two cycles of treatment had no better outcome than patients with late skin toxicity.

In fact, several patients in this study exhibited frequent toxicity after initial treatments,

but gradually no toxicity in the remaining protocol. This change could be the

consequences of disease stabilization, but it was uncertain whether resistance to

chemotherapy of normal cells did exist and played a role.

6.3.4 Toxicity after different drugs and efficacy

In current literature, the principles of toxicity enhancing efficacy are mostly

believed to be associated with either interpatient pharmacodynamics variations or

common sensitivity of neoplastic and normal cells to various drugs. Consequently, it

was not anticipated that toxicity by a particular drug would have more significance than

by other drugs. In this study, only 4 drug-specified groupings had P values <0.05.

Three of the 4 groupings were focused on vincristine, and all had non-drug-specified

counterparts with P value <0.05. Considering that vincristine was the most commonly

(55)

administered drug, the findings in these 3 drug-specified groupings could be just

reflecting the significance found in the 3 non-drug-specified counterparts.

6.3.5 Frequency of toxicity and efficacy

By examining the influence of frequency of toxicity on the relationship between

toxicity and efficacy, this study tried to answer two questions: Firstly, whether

frequency was a determinant for better outcome, in other words, whether only single

episode of certain toxicity was sufficient to improve survival; Secondly, if single

episode of certain toxicity ensures favorable outcome, whether more episodes ensure

even longer survival.

Since the no groupings changed from insignificant to significant after adding

frequency conditions, a conclusion could be derived that single episode of toxicity could

be sufficient. In the three frequency-adjusting groupings with P values <0.05, no

survival benefit was observed over their non-frequency-adjusting counterparts.

However, because setting higher frequency leads to fewer patients in the toxicity groups,

sampling bias could be amplified, and this limitation leaves the seconds question not

(56)

elucidated. Theoretically, if occurrence of toxicity equals effective tumor killing, then

frequent toxicity equals multiple effective tumor killings. It is straightforward that

multiple effective tumor killings are beneficial to patients with gross disease, but for

patients in remission, whether multiple effective tumor killings are advantageous or

redundant was hard to determine.

6.3.6 Age and GI toxicity

In the chia-square tests performed in long-term analysis in the study, older patients

were observed to have less low-grade GI toxicity. Baum et al demonstrated that the

number of proliferating cells in canine intestine epithelium decreased during aging, but

only weak correlation was found (Baum et al., 2007). It was speculated that due to

lower proliferation rate of GI mucosa cells of older patients, they were more resistant to

chemotherapy. However, only less low-grade GI toxicity, but not less GI toxicity of all

grades was found in older patients, so that the above explanation is still far from

satisfactory.

(57)

6.3.7 Confounding factors

More than 10 years old was associated less low-grade GI toxicity and poorer

outcome. If more than 10 years old was a solid negative prognostic factor for canine

lymphoma, then the result of low-grade GI toxicity producing better outcome could be

just caused by confounding, with age being the confounding factor. Under this

circumstance, the established relationship between GI toxicity and efficacy would be

less credible. However, since more than 10 years old is not a solid negative prognostic

factor for canine lymphoma, GI toxicity could also be the confounding factor, creating a

spurious connection between age and outcome. The actual causal relationship could not

be determined in this study.

CR was associated with more bone marrow and GI toxicity and better outcome.

Since the impact of response to treatment on survival is rather straightforward, and in

some studies response to treatment were also chosen as endpoints, it was not a concern

if response to treatment was a confounding factor of toxicity and outcome.

6.3.8 Multivariate analysis