One‐year conditional survival of dogs and cats with invasive mammary carcinomas: A concept inspired from human breast cancer

Abstract Numerous studies have described the prognostic factors of canine and feline mammary carcinomas (MCs), that is, variables that predict patient survival after diagnosis. But how does survival estimation evolve in patients that escaped early death from their cancer? In human oncology, conditional survival (CS), the probability of surviving X further years when cancer patients have already survived Y years, is used to analyse cancer outcomes in a long‐term perspective. In this cohort of 344 dogs and 342 cats with surgically removed stage I to III invasive MCs, with a minimal follow‐up of 2 years, we calculated the 1‐year CS, that is, the probability for patients that have survived 1 year, to survive or to die from cancer during the subsequent year. The 1‐year conditional specific survival probabilities were 59% and 48% at diagnosis of invasive MC respectively in dogs and cats, and 80% and 52% in 1‐year surviving dogs and cats respectively, suggesting that 1‐year surviving dogs were relatively protected from cancer‐related death, whereas feline MCs remained life‐threatening cancers for longer periods of time. Among the most significant parameters associated with CS in surviving dogs and cats were the nodal stage and lymphovascular invasion, as well as patient age, cancer stage and margin status in surviving dogs. By comparison, tumour size and the histological grade did not significantly alter CS probabilities in surviving dogs and cats. Conditional survival may be considered a very interesting tool for veterinary practitioners to estimate the likely outcome of cancer survivors.

the histological stage, 21,22 surgical margin status, 13,14,23 and immunophenotype 24,25 are robust factors significantly associated with survival probabilities or disease-free interval after diagnosis of invasive MCs. In human oncology and particularly in breast cancer, all of these factors also provide prognostic information, [26][27][28][29][30][31][32] given to patients at time of diagnosis using median survival times or 5-and 10-year survival probabilities.
In human oncology however, survival probabilities are better than in veterinary medicine because of earlier diagnosis and more efficient therapies, and follow-up is more rigorous. 33 In this context, medical oncologists use the tool of conditional survival (CS) that corresponds to the probability of surviving X further months, given that a patient has already survived Y months after cancer diagnosis. 34 For example, in patients with pancreatic cancer, at diagnosis, the probability of surviving 5 years is 7%, but for patients who survive to the first year, this probability increases to 27%, and reaches 63% if they survive 3 years. 33 CS can be divided into absolute CS and relative CS. Absolute CS is calculated from a single cohort of patients, whereas relative CS compares survival probabilities between a cohort and an age-matched healthy reference population.
To our knowledge, very few studies have described CS in veterinary medicine: Bonnett et al described the probability for dogs of living at least 5, 7, 8 or 10 years according to their breed, 35 as well as the probability for an 8-year-old dog of a given breed to be alive by age 10, data that can be considered as "conditional survival" although these dogs were not affected by a given disease entity. Kass

| Patients
This retrospective study included 344 canine and 342 feline invasive MCs that were diagnosed between 2004 and 2010 and have been previously described. 8,15,24 The owners' written consent and approval from the local animal welfare committee of Oniris were obtained prior to inclusion. Inclusion criteria comprised female dogs and cats, diagnosed with an invasive MC, surgically removed, with a minimal followup of 2 years. Patients were excluded if records were incomplete, if distant metastases or any other malignant tumours were present at diagnosis (based on clinical examination, and medical imaging when available), or if adjuvant treatments were administered before or after surgery. Age, breed, reproductive and medical history, and outcomes were obtained through written questionnaires or telephone interviews with referring veterinarians and owners.

| Follow-up and CS definition
The outcome data were overall survival (OS, time from mastectomy to death from any cause), and cancer-specific survival (SS, time from mastectomy to death attributable to the MC). Because the follow-up duration was 2 years in this study, we restricted our analyses to dogs and cats that had survived for 1 year, in order to calculate their probability of being alive, dead, or dying from cancer in the subsequent year. Thus, in this study, the 1-year CS is defined as the probability of surviving one further year depending on the number of months (0-12) already survived after diagnosis. More specifically, we separately calculated the 1-year conditional overall survival (COS) and the 1-year conditional cancer-specific survival (CSS). Either natural or by euthanasia, death was considered "cancer-related" in this study if it followed locoregional recurrence, distant metastasis (proven by medical imaging or necropsy, or highly suspected on clinical examination), or cancer cachexia, in the absence of any identified intercurrent disease.

| Histopathology
Formalin-fixed paraffin-embedded tumour samples were cut into 3 μm-thick sections and stained with haematoxylin-eosin-saffron (HES). Compared with the more widely used HE stain, the use of saffron in HES stains collagen in orange, but otherwise does not modify the histologic interpretation. In case of multiple/multicentric invasive MCs, the carcinoma with the largest diameter upon histological section was selected for analysis; if tumour size was identical between two MCs in a given patient, the MC with the highest histological grade was considered. Histological subtypes were described according to the adapted World Health Organization classification system. 16,23 In cases that demonstrated more than one histological architecture, the less differentiated was chosen. Histological stages were defined as previously described 21,22 according to the pathological tumour size (pT with pT1 ≤ 20 mm and pT2 > 20 mm), pathological nodal stage (pN with pN1, presence of nodal metastasis; pN0, absence of nodal metastasis, and pNX, lymph node not removed), and lymphovascular invasion (LVI, with LVI+ indicating presence, and LVI− indicating absence of lymphovascular invasion). Other recorded data included the Elston and Ellis histological grade, local invasion of dermis or muscle, margin status, tumour-associated inflammation, central necrosis, ulceration, and squamous differentiation, as previously described. 8,15 2.4 | Immunohistochemistry Immunohistochemistry (IHC) was performed using a Benchmark XT automated instrument (Ventana Medical Systems, Roche Diagnostics) as previously described, 8,21,22,24,37 and further detailed in Table S1.
The antibodies used were p63 (used as myoepithelial marker), pancytokeratin (used as marker of metastatic epithelial cells in lymph nodes), LMO2 (lymphendothelial marker to better assess tumour significantly older (P = .003), were more likely to harbour a multicentric MC (P = .0197), with a larger tumour size (P = .0002), with nodal metastasis (P < .0001) and/or lymphovascular invasion (P < .0001), and thus a more advanced histological stage at diagnosis (P < .0001). Their MC was more likely to be inflammatory or anaplastic (P = .0025), of higher histological grade (P < .0001), with more severe tumour-associated inflammation (P < .0001), more common infiltrated surgical margins (P < .0001) and muscle infiltration (P = .0126), and a higher Ki-67 proliferation index (P = .0039).

| One-year conditional overall survival of dogs with MCs
In dogs, the probability of being alive 1 year after mastectomy was 49 ± 3% at diagnosis ( Figure 1A), then the 1-year conditional overall survival (COS) increased during the first 6 months after diagnosis, reaching 63 ± 3% in 6 month-survivors, with a mean COS gain of 2.4% per month. From the 6th to the 12th month of survival, the COS probability levelled off (comprised between 59 ± 4% and 64 ± 4%).
We then analysed which parameters influenced conditional over-   Table 1 and Figure S1H).

| One-year conditional specific survival of dogs with MCs
The 1-year conditional CSS increased from time of diagnosis (59 ± 3%) to 6 months after (75 ± 3%) with a mean CSS gain of 2.75% per month ( Figure 1B). After the sixth month, the CSS tended to stabilize (between 74 ± 3% and 80 ± 3%).
At diagnosis, the same eight parameters associated with overall survival were significantly associated with CSS ( Table 2) Table 2 and Figure S2H).

| Characteristics of female cats with MCs
Three hundred and forty-two queens fulfilled the inclusion criteria (Table S3) One hundred and seventy-five (51%) cats died in the first year post-diagnosis. At initial presentation, compared with queens that survived at least 1 year, cats that died within the first year of follow-up had a MC of larger pathologic tumour size (P < .001), more likely to be pN+ (P < .0001) and/or LVI+ (P < .0001), and thus of more advanced histological stage (P < .0001), and also more likely to be grade III (P = .0168), with dermal infiltration (P < .0001), cutaneous ulceration (P = .0025), and/or positive margins (P = .0309) (Table S3).

| One-year conditional overall survival of cats with MCs
The 1-year conditional overall survival probabilities were stable from diagnosis (48 ± 3%) to 12 months of survival post-diagnosis (52 ± 4%) ( Table 3 and Figure S3B,C,E).

F I G U R E 1
One-year conditional survival of dogs according to the number of months survived after diagnosis of an invasive mammary carcinoma. A, 1-year conditional overall survival (COS). The probability of being alive 1 year later slightly increased from diagnosis (49 ± 3%) to 12 months post-diagnosis (59 ± 4%). B, 1-year conditional cancer-specific survival (CSS). The probability of dying from cancer during the following year was relatively high at diagnosis (41 ± 3%), but decreased with time, and was only 20 ± 3% in 12-month survivors At diagnosis, the same seven parameters associated with overall survival were significantly associated with the risk of dying from cancer during the subsequent year (Table 4) (Table 4 and Figure S4B,C,E).  Age at diagnosis (years) ≤ 11.7 years, OR = in animals. The 5-year relative CS of breast cancer patients only marginally increases with years of survival, with a gain comprised between +3% and +5% in 5-year survivors. 33,[39][40][41][42] In that respect, breast cancer markedly differs from lung, pancreatic or gastric cancer, characterized by relative CS gains comprised between +57% and +82% in 5 yearsurvivors. 33,40 In the cats of the present study, the absence of COS . The probability of being alive 1 year later was stable from the time of diagnosis (t0, 48 ± 3%) to t0 + 12 months (52 ± 4%). B, 1-year conditional cancer-specific survival (CSS). Twelve-month surviving cats had almost the same cancerspecific survival probability (64 ± 4%) as at diagnosis (56 ± 3%) which were also strongly associated with CS in the present study, including in 12-month survivors. As LVI and pN reflect the metastatic spread of cancer, one can hypothesize that even long-term surviving cats and dogs remain at high risk of cancer-related death because of distant metastases if their MC was LVI+ and/or pN+.

| DISCUSSION
In dogs and cats, the histological stage of invasive MCs, derived from pT, pN and LVI, 21,22 was significantly associated with SS at diagnosis (we compared stage I-II MCs: any pT, pN0-pNX and LVI−, with stage III MCs: any pT, pN+ and/or LVI+). In both species, stage I to II MCs showed a relatively stable CSS probability over time (+5% gain in CSS in dogs and cats, from diagnosis to 12 months after). By comparison, stage III MCs were associated with a +27% gain in CSS in dogs from diagnosis to 12 months after, whereas in cats stage III MCs showed also a relatively stable CSS probability over time (+5% gain in CSS from diagnosis to 12 months after); in both species, cancer stage still impacted CSS in 12-month survivors. These observations in dogs correlate with those reported in human breast cancer, in which lower stages do not significantly affect the relative CS, whereas women with higher-stage breast cancers have a significant improvement of their relative CS if they have survived 5 years. [42][43][44][45]55,56 In other words, the more advanced stage at diagnosis, the more survival time means "good news" for the patient, that is, decreased probability of dying from cancer afterwards.
In our cohort as in human breast cancer, 45 the histological grade was significantly associated with specific survival at diagnosis, but not in surviving patients. Grade I to II MCs showed a relatively stable CSS over time (+8% in dogs and +3% in cats within 12 months post-diagnosis), whereas grade III MCs were associated with a +28% gain in CSS in dogs and +15% in cats during the first year of follow-up. In women, the CSS improvement over time observed for high grade T A B L E 3 Clinical-pathologic parameters that influenced conditional overall survival of cats according to duration of survival (in months) after diagnosis breast cancers could be related to more aggressive therapy and better response to treatment of grade III compared with grade I breast cancers. 45 In the present cohort of animals that had surgery as a single therapy, this increase in CSS with survival time could be because of an earlier death of patients with higher grade MCs.
In dogs (but not cats) with MCs, the Ki-67 proliferation index was significantly associated with CSS probabilities during the first 10 months post-diagnosis (difference in CSS comprised between −15% and −25% for highly proliferative compared with slowly proliferative MCs). Afterwards, Ki-67 was not significantly associated with CSS because there had been a +25% gain in CSS in 12 months for highly proliferative MCs. In women with breast cancer, the Ki-67 proliferation index (at 14% cut-off) is not significantly associated with relative CS of 5-year surviving patients. 56 With respect to immunophenotype, PR positivity of feline MCs was a protecting factor during the first 4 months of follow-up, and at months 9 and 10, with higher CSS probabilities associated with PR+ MCs than ER+ PR− and ER− PR− MCs. This is in agreement with the favourable prognostic value of PR in feline MCs. 37 In dogs, luminal (ER+ and/or PR+) MCs showed a late increase in 1-year CSS, from 76 ± 5% at month 5 to 92 ± 4% at month 12 (+16% gain in CSS in 7 months): the probability for a dog of dying from a luminal MC during the subsequent year became negligible if the dog had survived 1 year. Human breast cancers behave differently, maybe because patients with luminal tumours benefit from 5 to 10 years of hormone therapy: ER-positive breast cancers are associated with a cancer-specific mortality peak at 4 years followed by a plateau, whereas ER-negative breast cancers are associated with a mortality peak at 2 years post-diagnosis. 57 T A B L E 4 Clinical-pathologic parameters that influenced conditional specific survival of cats according to duration of survival (in months) after diagnosis As retrospective in nature, this study has limitations. Firstly, the minimal follow-up period was 2 years, a duration not sufficient enough to identify at which time the MC has no more influence on life expectancy. In human oncology, when the 5-year relative CSS probability exceeds 95%, one considers that the given cancer does not cause excess mortality anymore; this is achieved in patients who have survived 10 years after diagnosis of a gastric, colorectal, cervical or thyroid cancer, whereas this is not achieved for breast cancer, 33 probably because of the late mortality peak observed 8 to 10 years after breast cancer diagnosis. [57][58][59] Secondly, exclusion criteria of this study included the presence of distant metastases at diagnosis, but this parameter could play a role in the CS of dogs and cats, as it does in breast cancer, 43,55 and it would be interesting to analyse this parameter in a larger cohort. Thirdly, the lack of an age-matched reference

| CONCLUSION
CS appears as an interesting tool to refine the prognosis of dogs and cats with invasive MCs that have survived a few months after mastectomy. Twelve-month surviving dogs had a 2-fold decreased risk of dying from cancer in the subsequent year compared with their risk at diagnosis, whereas cats remained at the same high risk of dying from cancer even if they had already survived 1 year. Some prognosticators, such as lymphovascular invasion and nodal stage, durably affected the probability for canine and feline patients of living one further year, even in 12-month survivors. By comparison, tumour size and histological grade were well associated with early cancer-related death, but not significantly in surviving patients. This study thus highlights the differences between prognosis establishment at diagnosis, and 1 year later for survivors. Finally, we thank the referring veterinarians and the owners of the animals included in this study, who gave us the clinical and followup data.

CONFLICT OF INTEREST
The authors declare no conflict of interest. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author.

ETHICS STATEMENT
The study design was reviewed and approved by the CERVO (Comité