Enrollment and eligibility

Dogs were prospectively enrolled at the University of Missouri, The Ohio State University, and the University of Guelph following informed owner consent, in accordance with IACUC approval (protocol numbers were as follows: University of Missouri, 24140; The Ohio State University, 2020A00000022; and University of Guelph, 4431). Dogs were considered eligible for enrollment if they were scheduled for thyroidectomy with a board-certified surgeon for treatment of naturally occurring thyroid cancer, following preoperative cervical imaging within 1 month of surgery. Acceptable imaging included CT, ultrasound, or MRI. In addition, CBC, serum biochemistry, and thyroid hormone assessment with total T4 (TT4) or free T4 (fT4) were required. Bilateral MRP lymph nodes were resected concurrently with thyroidectomy. All deep cervical lymph nodes identifiable at surgery were resected. The decision to pursue thyroidectomy was made by clients in consultation with attending surgeons/oncologists on the basis of standard of care. Both clinically node-negative dogs and dogs with suspicion of nodal metastasis were enrolled. Dogs were ineligible for enrollment if they had undergone neoadjuvant radiotherapy or chemotherapy. Dogs with clinical evidence of pulmonary metastasis were eligible for inclusion.

Demographic and preoperative clinical data

Signalment, clinical signs and duration, and concurrent illness were recorded at each institution. Total T4 was recorded, with or without TSH and fT4 at the attending clinician’s discretion. Serum biochemistry was reviewed for evidence of hypercalcemia, with subsequent review of ionized calcium assay, if available. Dogs were considered hypercalcemic if a high total calcium was reported without assessment of ionized calcium or with concordant ionized calcium. Dogs with increased total calcium and normal ionized calcium were not considered hypercalcemic. Preoperative cervical staging with either cervical ultrasound or cervical cross-sectional imaging (CT or MRI) was required, in addition to thoracic staging with either thoracic radiographs or CT. For dogs with preoperative thoracic radiographs and thoracic CT, CT was used to determine the presence or absence of pulmonary nodules. All imaging was reviewed by a residency-trained/board-certified radiologist at the participating institution. Cervical imaging was reviewed to determine the dimensions and location of thyroid masses. When available, the clinical team for each institution reviewed cross-sectional imaging to identify visible deep cervical lymph nodes prior to resection.

Surgery

Dogs were anesthetized by or under the supervision of an attending anesthesiologist, with protocols determined by the attending anesthesia team. Following induction of anesthesia, the ventral neck was clipped and aseptically prepared. Dogs were then transferred to a dedicated operating room, and final aseptic preparation was performed. For eutopic thyroid tumors, a ventral cervical midline approach was performed, with subsequent thyroidectomy performed at the attending surgeon’s discretion. Vessels with gross vascular invasion were ligated or sealed and divided distal to the tumor thrombi to allow for en bloc resection. Hyoid ectopic tumors were resected as previously described.²²

Bilateral (MRP) lymphadenectomy was performed as previously described²³ and the neck examined for deep cervical lymph nodes (Figure 1).²¹ Briefly, CrDC lymph nodes were positioned caudal to the MRPs and craniolateral to the thyroid gland. At the surgeon’s discretion, these lymph nodes could be approached directly or identification of CrDC lymph nodes could be facilitated by tracing the cervical lymphatic trunk caudad from the caudal pole of the MRP lymph node. Middle deep cervical (MDC) lymph nodes were identified in the midneck, within the deep cervical potential space ventral to the trachea. Caudal deep cervical (CaDC) lymph nodes were similarly located within the ventral deep cervical potential space, typically at or just cranial to the thoracic inlet. Clearly lateralized lymph nodes (including mandibular, MRP, and CrDC) were considered to lie within the lateral compartment, while non- or minimally lateralized lymph nodes (MDC and CaDC) were considered to lie within the central compartment. Only identified lymph nodes were extirpated, without en bloc excision of deep cervical tissue. Additional lymph nodes could be resected at the attending surgeon’s discretion.

View Full Size

Schematic diagram of the proximate lymph nodes of the neck for eutopic and hyoid ectopic thyroid tumors. Nodal stations are as follows: I, mandibular; II, medial retropharyngeal; III, cranial deep cervical; IV, middle deep cervical; and V, caudal deep cervical. Asterisks indicate that these nodes were considered to lie within the central compartment. Lateralized lymph nodes were considered to lie within the lateral compartment.

Citation: Journal of the American Veterinary Medical Association 2025; 10.2460/javma.24.03.0223

• Download Figure
• Download figure as PowerPoint slide

Closure for eutopic thyroid tumors was routine, with apposition of the sternohyoideus muscles and subcutis using monofilament, absorbable suture. Cutaneous apposition was achieved with a continuous intradermal pattern using monofilament, absorbable suture, and/or external nonabsorbable skin sutures.

Intraoperative complications were contemporaneously recorded and categorized using the Classification of Intraoperative Complications (or CLASSIC) system.²⁴ Additional hospital visits were not required by the study. Postoperative complications were retrospectively reviewed and classified using the contracted Accordion severity scheme.^24,25 Ventral neck or facial swelling that resolved spontaneously and did not require intervention was considered a sequela. Hypothyroidism following bilateral thyroidectomy or resection of a functional follicular tumor and hypocalcemia following resection of at least 3 parathyroid glands were also considered as sequelae; however, hypothyroidism and hypocalcemia following unilateral thyroidectomy (without hyperthyroidism or hypercalcemia, respectively) were considered complications due to the low index of suspicion of these events.

Postoperative care

Postoperative care was performed at the attending clinician’s discretion, but typically included IV fluid support, an NSAID (typically carprofen [2.2 mg/kg, SC or PO, q 12 h] or meloxicam [0.1 mg/kg, IV or PO, q 24 h]), and opioid analgesia (typically hydromorphone [0.03 to 0.1 mg/kg, IV, q 4 to 6 h], methadone [0.1 mg/kg, IV, q 4 to 6 h], or fentanyl [2 to 3 µg/kg/h, IV, constant rate infusion]) until the dog was considered to be comfortable enough for opioids to be withdrawn.

Histopathology and IHC

All specimens were fixed in neutral-buffered 10% formalin and submitted to the University of Missouri Veterinary Medical Diagnostic Laboratory for review by a single board-certified pathologist (DYK). Primary tumors and lymph nodes were sectioned along the long axis of the specimens and processed for histology, and slides derived from these specimens were stained with H&E. Mitotic count per 2.37 mm² was reported at the regions of most frequent mitoses and surrounding areas (ocular field number, 22 mm). Slides from all primary tumors underwent immunolabeling for thyroid transcription factor-1 (TTF-1; M3575, mouse monoclonal; Dako) and calcitonin (A0576, rabbit polyclonal; Dako).²⁶ Heat-induced antigen retrieval with a citrate buffer (Diva Decloaker; Biocare Medicine) was used. MACH 2 (Biocare Medical) was used for antigen detection of TTF-1, while EnVision+ (Dako) was used for calcitonin detection. The percentage of neoplastic cells with immunolabeling was subjectively scored from 1 to 3 (1, < 33%; 2, 33% to 66%; and 3, > 66%). On the basis of histologic pattern and immunolabeling, carcinomas were considered follicular (with follicular, compact, and follicular-compact subtypes), medullary, or mixed follicular/medullary in origin.^26,27 Briefly, tumors were considered follicular in origin if they had immunolabeling for TTF-1 and/or thyroglobulin, without immunolabeling for calcitonin. Neoplastic populations with immunolabeling for calcitonin were considered of medullary origin. Carcinomas were distinguished from adenomas by use of multiple previously described variables, including tumor size, pleomorphism, necrosis, hemorrhage into follicles, capsular invasion, and vascular invasion.^2,28 One slide for each nonmetastatic lymph node following H&E was assessed with IHC, using the antibody that was associated with the highest percent uptake within neoplastic cells from the primary tumor. Metastatic deposits were classified on the basis of burden of disease, as previously described. Deposits ≥ 2 mm were considered macrometastasis, deposits < 2 mm but ≥ 200 µm were considered micrometastasis, and deposits < 200 µm were considered to be isolated tumor cells.^21,29,30 Nodal metastatic deposits not recognized following examination with H&E but identified following IHC were considered occult.

Statistical analysis

A sample size of 100 client-owned dogs was initially chosen on the basis of an estimated proportion of 45%, margin of error set at 10%, and 95% CI (95 dogs recommended).^21,31 Continuous data were assessed for normality with the Shapiro-Wilk test. Normally distributed continuous data were described with mean and SD. Nonparametric continuous data were described with median and range. Ninety-five percent CIs were provided for percentage metastasis by dog and by tumor. The proportion of tumors with metastasis was compared between follicular carcinoma and medullary and follicular/medullary carcinoma groups with the Fisher exact test. P < .05 was considered statistically significant. Dedicated statistical software was used (Prism, version 10; GraphPad Software).

Population

Seventy eligible dogs were included from 3 institutions (48 at the University of Missouri, 19 at The Ohio State University, and 4 at the University of Guelph), with surgery performed between January 1, 2020, and August 22, 2023. Three other dogs were enrolled but subsequently excluded (mass diagnosed as a thyroid adenoma [n = 1], and failure to resect ipsilateral MRP lymph nodes [2]). Median (range) age and body weight were 10 years (2 years and 1 month to 15 years and 1 month) and 23.7 kg (3.3 to 66.5 kg), respectively. Thirty-two dogs were female (1 intact), and 36 dogs were male (2 intact). The most common breeds were Labrador Retriever (n = 5), Boxer (4), Siberian Husky (3), and Welsh Corgi (3), with 33 dogs of 27 other breeds, in addition to 22 mixed-breed dogs.

Clinical data and preoperative staging

At least 1 clinical sign related to a thyroid mass was identified in 58 (82.9%) dogs, while a thyroid mass was found incidentally during evaluation for unrelated issues in 12 (17.1%) dogs. The most common clinical sign was a palpable mass (55 dogs [78.6%]). Other clinical signs included weight loss (n = 10 dogs), coughing (9), dysphagia (7), lethargy, (5), and dysphonia (3), with 16 instances of other clinical signs in 7 dogs.

Forty-four fine-needle aspirates were obtained from 42 dogs (with 1 bilateral and 1 repeat aspiration). Of these 44 aspirates, 41 (93.2%) were considered diagnostic. Preoperative T4 was recorded in 69 dogs (TT4 in 65, and fT4 in 4). T4 was not assessed for 1 dog that met all other inclusion criteria. Two dogs were receiving levothyroxine to treat hypothyroidism at presentation and were excluded from further assessment of T4. Median (range) TT4 was 2.3 µg/dL (< 0.5 to > 15.0 µg/dL). T4 was increased relative to the laboratory reference range in 16 (19.3%) dogs. Cervical imaging was performed with CT in 62 dogs, MRI in 2, PET-CT in 1, and ultrasound in 5 at a median of 9 days preoperatively. Thyroid masses were left-sided in 29 dogs, right-sided in 29, bilateral in 9, and located at midline (hyoid location) in 3, resulting in a total of 79 thyroid lesions. The mean (± SD) maximum tumor dimension reported contemporaneously was 4.5 ± 2.4 cm. Thoracic staging was performed primarily with radiographs in 17 dogs and CT in 53 dogs. Pulmonary nodules were identified in 5 (7.1%) dogs, with equivocal findings in a further 3 (4.3%).

Surgery

Fifty-nine dogs underwent unilateral thyroidectomy (29 left, 29 right), 9 underwent bilateral thyroidectomy, and 2 dogs had resection of a hyoid ectopic thyroid tumor. Both hyoid tumor resections involved partial hyoidectomy, including the basihyoid bone. Gross vascular invasion was reported in 24 (31.2%) tumors. A total of 191 lymph nodes were resected, including 6 mandibular, 139 MRP, 31 CrDC, 3 MDC, 7 CaDC, 3 superficial cervical, and 2 other lymph nodes. A median of 2 lymph nodes were resected per dog (range, 2 to 6).

Intraoperative complications were encountered in 5 of 70 (7.1%) dogs. Briefly, complications were as follows: grade I, 1 dog (failure to resect contralateral MRP lymph node); grade II, 3 dogs (hemorrhage); and grade IV, 1 dog (hemorrhage; Supplementary Table S1). Ventral cervical or facial swelling without other associated clinical signs was recorded as a sequela, managed conservatively, in 29 of 69 (42.0%) dogs. Hypocalcemia was recorded as a sequela in 7 dogs, and hypothyroidism was documented in 12 dogs. Postoperative complications were reported in 12 of 69 (17.4%) dogs (grade 1: 2 events in 2 dogs, grade 2: 7 events in 6 dogs, grade 3: 5 events in 4 dogs, and grade 4: 3 events in 3 dogs; Supplementary Table S2).

Histopathology

Nodal metastasis

Of the 70 dogs included, 12 had evidence of nodal metastasis following H&E review, including 21 metastatic lymph nodes (1 to 4 metastatic lymph nodes/dog with nodal metastasis; Figures 2 and 3). Fifteen metastases were classified as macrometastasis, and 6 were classified as micrometastasis. Small metastatic deposits were identifiably subcapsular; however, as the size of the metastatic aggregate increased relative to the lymph node, distinction of a specific location became progressively challenging. One dog was identified to have occult micrometastasis to an MRP lymph node following nodal IHC for TTF-1 (1 mm aggregate) for a final metastatic rate of 13 of 70 (18.6%; 95% CI, 11.2% to 29.2%). Isolated tumor cells were not identified in any lymph node. Medullary carcinoma (5/8 [62.5%]) and mixed follicular/medullary carcinoma (3/7 [42.9%]) were both associated with a significantly higher rate of nodal metastasis than follicular carcinoma (5/61 [8.2%]; P = .001 and P = .03, respectively; Table 2).

View Full Size

Photomicrographs of representative follicular carcinomas, with associated cervical lymph node metastases, including follicular (A through D), compact (E through H), and follicular-compact (I through L) subtypes, with H&E sections of the primary tumor (first column), immunolabeling for thyroid transcription factor-1 (TTF-1; second column), immunolabeling for calcitonin (third column), and H&E sections of an associated metastatic lymph node (fourth column). Scale bars are included in the bottom right corner of each photomicrograph (panels A, E, and H, bar = 200 µm; B, C, F, and G, bar = 100 µm; D, bar = 500 µm; I and L, bar = 1 mm; and J and K, bar = 50 µm). A through C—Follicles could be seen throughout the tumor, with strong nuclear labeling for TTF-1 but no calcitonin immunolabeling. D—A follicular pattern was again seen for this 8 mm nodal metastatic aggregate (black arrowhead) within the left medial retropharyngeal (MRP) lymph node. E through G—Compact cellular lobules divided by fibrovascular stroma were noted on H&E. Strong nuclear TTF-1 immunolabeling was present without calcitonin immunolabeling. H—Multiple metastatic solid cellular lobules (black arrowheads) with a maximum diameter of 600 µm were seen within this right MRP lymph node. A distinct follicular pattern was absent. I through K—Regions with distinct follicles were separated by areas without apparent lumina. Again, TTF-1 was diffusely positive within nuclei, while no calcitonin immunolabeling was seen. L—Interestingly, this multifocal nodal metastasis (black arrowheads) within a right MRP lymph node (largest aggregate measuring 12 mm) had a primarily compact appearance, with minimal evidence of the follicles evident in the primary tumor.

Citation: Journal of the American Veterinary Medical Association 2025; 10.2460/javma.24.03.0223

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Photomicrographs of representative medullary and mixed follicular/medullary tumors, with associated cervical lymph node metastases, including medullary (A through D), mixed follicular medullary (E through H), and immunolabeling of mixed follicular medullary nodal metastasis (I and J), with H&E sections of the primary tumor (first column), immunolabeling for TTF-1 (second column), immunolabeling for calcitonin (third column), and H&E sections of an associated metastatic lymph node (fourth column). Scale bars are included in the bottom right corner of each photomicrograph (panels A, D, and H through J, bar = 1 mm; B and C, bar = 50 µm; and E through G, bar = 100 µm). A through C—Note the packets of solid cellular islands surrounded by stroma, with strong nuclear labeling for TTF-1 and cytoplasmic labeling for calcitonin. Vascular invasion was present (black arrow in image A). D—This MRP lymph node (3 cm metastatic aggregate) was largely effaced by neuroendocrine cells in a packeting pattern. E—Much of the tumor demonstrated a tubular appearance; however, patchy areas of packeting could be seen (black arrows). F and G—Strong nuclear immunolabeling was seen diffusely for TTF-1, with patchy cytoplasmic immunolabeling for calcitonin. H—Part of a 1 cm neuroendocrine metastatic aggregate is shown within this caudal deep cervical lymph node. Interestingly, in panels I and J, 2 populations could be found within the metastatic aggregate; one had strong nuclear TTF-1 immunolabeling and absent (background) calcitonin immunolabeling (white arrowheads), and the other had moderate nuclear TTF-1 immunolabeling with strong cytoplasmic calcitonin immunolabeling (gray arrowhead).

Citation: Journal of the American Veterinary Medical Association 2025; 10.2460/javma.24.03.0223

• Download Figure
• Download figure as PowerPoint slide

Table 2

Lymph node metastasis associated with thyroid cancer histologic type/subtype.

Tumor type	No. of tumors	No. with nodal metastasis	Percent metastasis (95% CI)
Follicular*	61	5	8.2% (3.6%–17.8%)
Follicular	24	1	4.2%
Compact	20	2	10.0%
Follicular-compact	17	2	11.8%
Medullary	8	5	62.5% (30.6%–86.3%)
Mixed follicular/medullary	7	3	42.9% (15.8%–75.0%)
Carcinosarcoma	1	0	0%
Total by malignant tumor	77	13	16.9% (10.1%–26.8%)

Data in this table are presented on the basis of individual tumor rather than by patient. Confidence intervals are provided for percent metastasis for overall metastasis and follicular, medullary, and mixed follicular/medullary types.

*Data regarding follicular subtypes are included individually.

Ipsilateral lateral compartment nodal metastasis was seen in 12 dogs (17.1%; 3 MRP alone, 5 CrDC alone, and 4 MRP and CrDC ± central compartment). Contralateral lateral compartment nodal metastasis was not identified in this population. While central compartment metastasis was only identified in 3 dogs (4.3%; 1 alone, and 2 in combination with MRP and CrDC), a high proportion of resected central compartment lymph nodes were metastatic. No metastasis was identified in node centers beyond the MRPs and deep cervical lymph nodes, although the numbers of other lymph nodes resected were low (Table 3).

Table 3

Metastasis distribution from thyroid cancer, assessed by lymph center evaluated. Middle and caudal deep cervical lymph nodes were identified at midline and were not strikingly lateralized.

Lymph center	n	Nodal metastasis	Percent metastasis
Mandibular	6	0	0%
Medial retropharyngeal	139	7	5.0%
Ipsilateral	77	7	9.1%
Contralateral	62	0	0%
Cranial deep cervical	31	9	29.0%
Ipsilateral	23	9	39.1%
Contralateral	8	0	0%
Middle deep cervical	3	2	66.7%
Caudal deep cervical	7	3	42.9%
Superficial cervical (ipsilateral)	3	0	0%
Other cervical	2	0	0%
Total lymph nodes	191	21	11.0%

Among the 3 dogs with nodal metastasis from a mixed follicular/medullary carcinoma, the histologic appearance of the metastatic aggregate included medullary carcinoma in all 4 lesions. One lymph node had metastasis with components of both follicular and medullary origin (Figure 3); however, the remaining nodal metastases exhibited calcitonin immunolabeling, suggesting metastasis specifically of the medullary component of the mixed tumor.

Discussion

Lymph node metastasis was uncommon in dogs with follicular carcinomas but was seen frequently in dogs with medullary carcinoma and mixed follicular/medullary carcinoma. As such, we partially accepted our first hypothesis. Nodal IHC identified 1 additional lymph node metastasis, leading us to accept our second hypothesis, although the low frequency of occult metastasis did not provide strong support for routine nodal IHC in the clinical setting.

Our study identified a striking difference in metastatic rate between follicular and medullary carcinoma. Differences in clinical behavior between follicular carcinoma and medullary carcinoma have previously been poorly defined in the veterinary literature, perhaps due to inconsistent use of IHC. Medullary carcinoma has been suggested to have a less aggressive disease course or similar prognosis when compared to follicular carcinoma.^5,6 However, Enache et al⁹ more recently reported a 4-fold higher risk of disease-specific death in dogs with nonfollicular carcinoma compared to follicular carcinoma. While these were likely to be primarily medullary carcinomas, inclusion criteria for this subset were not clearly described and IHC to confirm diagnosis was not reported.²⁴ This may have resulted in inclusion of other tumors with aggressive behavior, such as anaplastic carcinoma or tumors with a mesenchymal component.³² The high rate of nodal metastasis among dogs with medullary carcinoma or mixed follicular/medullary carcinoma in our study supports routine use of IHC to define the cell of origin, especially as mixed follicular/medullary carcinoma has previously only recently been reported in a single dog.²⁷ While the clinical impact of nodal metastasis in canine thyroid carcinoma is currently unclear, failure to resect metastatic lymph nodes could lead to regional failure, with a very similar presentation to local recurrence if proximate lymph nodes are involved. Despite a high rate of nodal metastasis, medullary carcinoma in people may often progress slowly and it is possible that this somewhat indolent behavior, in combination with nodal dissection, may be sufficient to control disease for many dogs.^16,33 Interestingly, despite the high rate of nodal metastasis, none of the dogs with medullary carcinoma had pulmonary nodules at preoperative staging. While the funding for this study could not extend to covering costs of routine restaging, we intend to report outcomes in these dogs longitudinally to explore the impact of nodal metastasis and tumor type on outcome.

Our data support that deep cervical lymph nodes are an important potential site of metastasis. Given the high proportion of metastasis among identified ipsilateral CrDC and central compartment lymph nodes, surgeons should thoroughly review diagnostic imaging for these lymph nodes prior to surgery and assess the deep cervical nodal stations during thyroidectomy, with extirpation of identified deep cervical lymph nodes. Although surgeons have advocated against such a “berry-picking” approach in people due to the potential to overlook sites of disease,^33–35 extirpation of identifiable deep cervical lymph nodes may be a reasonable approach in dogs until more specific guidelines are available to direct more, or less, aggressive approaches. Management of MRP lymph nodes is likely to be more controversial, given the relatively infrequent metastasis identified in this study. This contrasts strikingly with the findings of our previous retrospective series regarding dogs managed with elective MRP ± deep cervical lymphadenectomy.²¹ In that study,²¹ 9 metastatic MRP lymph nodes were identified among the 22 dogs included, in addition to 4 metastatic CrDC lymph nodes. A combined dataset of 92 dogs from both studies would include 23 dogs with metastasis (25.0%). In this combined population, contralateral MRP lymphadenectomy (excluding hyoid carcinoma) appears to be low yield (1.3% combined metastatic rate; 1/79 dogs with contralateral MRP resection). Ipsilateral MRP metastasis was seen in 17.4% (16/92) of dogs in the combined dataset. Given the proximity of the ipsilateral MRP and the minimal additional time required for resection at the time of thyroidectomy, we recommend routine ipsilateral MRP resection until sufficiently granular data are available to guide more specific recommendations. Preoperative identification of tumor type, nodal size or characteristics based on imaging, and primary tumor features, such as maximum length, volume, or vascular invasion, may all play a role in risk stratification for nodal metastasis. By restricting elective lymphadenectomy to the ipsilateral side, impacts on lymphatic flow could be diminished to limit potential patient morbidity. Limiting morbidity must, however, be balanced with the goal of locoregional control. Due to the challenges of repeat cervical surgery in people, including increased difficulty in identification of recurrent laryngeal nerves and parathyroid glands, Moley et al¹⁶ advocated for “adequate dissection” of lymph nodes at the first procedure during management of medullary carcinoma. A subsequent manuscript will detail review of CT findings to establish the diagnostic accuracy of contrast-enhanced CT for detection of nodal metastasis in this population. If CT is shown to be a sensitive predictor of metastasis, this may allow a more targeted approach in certain populations of dogs.

Despite inclusion of follicular, medullary, and mixed follicular/medullary carcinomas, TTF-1 was consistently the antibody associated with the most extensive labeling of neoplastic cells. Given the low rate of occult nodal metastasis in this population, there are currently insufficient data to justify routine immunolabeling of resected lymph nodes in dogs with thyroid carcinoma. It is possible that additional sections would have resulted in identification of additional occult metastases; however, these sections impose a workload burden on pathologists and, thereby, a cost burden. We elected not to include additional sections in this study due to both considerations, as well as due to concerns regarding the feasibility of widespread implementation of fine sectioning in routine clinical assessment. The higher incidence of metastasis in dogs with medullary or mixed follicular/medullary tumors may suggest a greater benefit of nodal IHC in this subset, although data identifying an impact on prognosis or subsequent management would be necessary before routine IHC could be advocated even in this group.

Due to slow enrollment, we did not enroll 100 dogs as planned during the study design. Eight institutions had initially committed to participate; however, the study opened in early 2020 and, likely due to challenges of the COVID-19 pandemic, ultimately only 3 institutions contributed. While any sample population can only provide an estimate of metastatic rate, the larger the population, the greater the confidence that any identified rate is reflective of the true rate. The proportion identified also affects power and the associated margin of error. Post hoc sample size calculation using our final nodal metastatic rate of 18.6%, with the same margin of error (10%) and CI (95%), recommended a sample size of 59 dogs. Identification of deep cervical lymph nodes anecdotally varied within our study. Although not directly assessed, we feel there is likely a learning curve with identification of smaller deep cervical lymph nodes both on imaging and during surgery. In addition, we did not take an exhaustive approach to cervical lymphadenectomy. While we aimed to be inclusive with MRP lymph nodes and deep cervical lymph nodes as the proximate nodes for a eutopic thyroid tumor, we chose not to routinely take mandibular and superficial cervical lymph nodes. It is possible that other sites of metastasis may have been missed as a result; however, we did not feel that the likely increase in morbidity involved in such an extensive node dissection was justified given the limited evidence of metastasis to these sites. Clinicians were free to resect additional lymph nodes as considered indicated.

Cervical lymph node metastasis was uncommon in dogs with follicular carcinoma but affected most dogs with medullary carcinoma and nearly half of dogs with mixed follicular/medullary carcinoma. Immunohistochemistry only revealed 1 additional metastatic lymph node in this population, suggesting its routine use is low yield in this context. Deep cervical lymph nodes are an important site of lymph node metastasis in canine thyroid carcinoma and should be extirpated when identified in these dogs. Ipsilateral MRP lymphadenectomy is currently justified, although routine contralateral MRP is not recommended. Further data to stratify risk of metastasis including assessment of preoperative imaging and the impact of nodal metastasis on prognosis are needed to guide patient-specific recommendations.