Secondary involvement of the uterine cervix by non-gynecologic neoplasms: a detailed clinicopathologic analysis

Abstract

Secondary involvement of the uterine cervix by non-gynecologic neoplasms is rare accounting for <2% of metastases to the gynecologic tract. The aim of this study was to analyze the clinicopathologic features of cervical involvement by non-gynecologic malignancies. A total of 47 cases were identified including 39 (83%) carcinomas, 6 lymphomas (12.8%) and 2 (4.2%) cutaneous malignant melanomas. The most common primary site of origin among carcinomas was the gastrointestinal tract (27, 69.2%), followed by breast and urothelium (5 each, 12.8%), gallbladder and lung (1 each, 2.6%). The gynecologic tract was involved at presentation in 16 patients (34%), including 5 (10.6%) with the cervix being the first site, 7 (14.9%) with synchronous involvement of the cervix and other gynecologic sites, and 4 (8.5%) with involvement of other gynecologic sites prior to cervical presentation. Patients with lymphoma were younger compared to those with carcinoma (43.7 versus >50.5) (p=0.01). Mean time to identification of cervical metastases was <1 year for gallbladder carcinoma, melanomas and gastrointestinal signet ring cell carcinomas (p=0.03). Features that varied with different types of metastatic tumor included lymphovascular space invasion, depth of stromal invasion, growth pattern (glands lacking architectural complexity, cribriforming, solid), presence of goblet cells and signet ring cells, degree of cytologic atypia, and overall findings mimicking a benign/non-invasive process (p≤0.027). Six tumors (12.8%) were initially misdiagnosed as cervical primary. Metastatic non-gynecologic tumors can mimic primary in situ or invasive neoplasms in both ecto- and endocervix. In patients with a known prior malignancy, the clinical history with ancillary studies and a high level of suspicion are crucial to ensure accurate diagnosis.

Metastases to the gynecologic tract most commonly involve ovaries and vagina (approximately 80%) (1). Secondary involvement of the cervix is most frequently due to contiguous extension of endometrial carcinomas or drop metastases from adnexal carcinomas (2, 3). Non-gynecologic metastases to the cervix are exceedingly uncommon, accounting for less than 2% of metastases to the gynecologic tract (1, 4). The most common non-gynecologic primary sites of origin include the gastrointestinal tract and breast. Metastatic urothelial, pancreatobiliary, pulmonary and renal carcinomas as well as cutaneous malignant melanomas have also been described (4–6). A single-institution but tertiary referral center study of 144 patients with metastases to the cervix identified non-gynecologic origin in 14% of cases, of which lymphomas (5.5%) and breast carcinomas (2.8%) were the most common tumor subtypes (7). Distinguishing between metastases and primary cervical carcinomas is important to ensure appropriate treatment and accurate prognostication (8). Careful microscopic evaluation and ancillary studies as well as knowing details regarding the clinical history of the patients are helpful in this differential diagnosis.

Patients with metastatic non-gynecologic tumors to the cervix may present with vaginal bleeding or abnormal cervical cytology. Clinically, the cervix may be unremarkable, exhibit diffuse enlargement, a well-defined mass or a poorly-defined mass (4, 9). Under the microscope, metastatic neoplasms frequently pose diagnostic challenges as they may be mistaken for benign proliferations or more commonly intraepithelial or invasive primary tumors, especially when the past clinical history is unknown to the pathologist. Metastases may replace the preexisting glandular lining epithelium without stromal desmoplastic reaction mimicking a cervical precursor lesion (high-grade squamous intraepithelial lesion [HSIL] or adenocarcinoma in situ [AIS]) or invasive carcinoma. In general, metastatic tumors have been reported to display permeative growth surrounding normal endocervical glands, multifocal involvement, extensive lymphovascular space invasion (LVSI) and, in some metastatic carcinomas, signet ring cells (3, 4, 7, 9, 10). However, no detailed and systematic morphologic characterization of metastatic tumors is provided in any of the published series.

The aim of this study was to analyze the frequency, clinical features and morphologic patterns of involvement of metastatic non-gynecologic neoplasms to the uterine cervix.

MATERIALS AND METHODS

The institutional databases at Memorial Sloan Kettering Cancer Center, Massachusetts General Hospital, Leipzig University Hospital and Mount Sinai Hospital were searched for patients diagnosed with secondary involvement of the uterine cervix by non-gynecologic neoplasms between 2004 and 2018. One case was contributed by The University of Maryland and was included in a prior report (11). Epithelial, melanocytic and hematolymphoid tumors of non-gynecologic origin were included. Pertinent electronic medical records, pathology reports and slides were reviewed in all patients.

The following clinicopathologic features were recorded: patient age, date of first diagnosis of original tumor and first clinical gynecologic manifestation, procedure type, tumor size and presence of LVSI. The following histologic patterns of involvement were assessed: a) architectural and cytologic features including infiltrating growth versus colonization of pre-existing epithelium (pagetoid spread) with replacement of squamous mucosa or endocervical glands; glands lacking architectural complexity versus complex glands with cribriforming or solid growth; goblet cells; signet ring cells; and cytologic atypia (mild versus moderate versus severe); b) stromal features, including superficial (<50%) versus deep (>50%) invasion; desmoplasia versus no desmoplasia; and unifocal versus multifocal involvement. Descriptive statistical analysis was performed in SPSS 26.0.

RESULTS

A total of 47 cases were identified consisting of 39 (83%) carcinomas, 6 (12.8%) lymphomas and 2 (4.2%) cutaneous malignant melanomas. The carcinomas were comprised of 27 (69.2%) gastrointestinal, 5 (12.8%) mammary (3 ductal, 2 classical lobular), 5 (12.8%) urothelial (3 previously reported (11)), 1 (2.6%) gallbladder and 1 (2.6%) lung (acinar type adenocarcinoma) origin. The gastrointestinal carcinomas consisted of 17 (63%) colorectal, 6 (22.2%) appendiceal, 3 (11.1%) gastric and 1 (3.7%) ileal carcinomas, of which 8 (29.6%) had predominant signet ring cell morphology (6 appendiceal, 1 ileal, 1 gastric). Of 5 urothelial carcinomas, there were 2 high-grade invasive tumors, 2 non-invasive high-grade papillary tumors and 1 carcinoma in situ. The lymphomas were comprised of 5 diffuse large B-cell lymphomas (1 categorized as germinal center type) and 1 chronic lymphocytic lymphoma/small lymphocytic lymphoma. Both malignant melanomas were nodular and arose in the lower leg. The clinical characteristics of all 47 patients are summarized in Table 1.

Table 1.

Clinical features of 47 cases with secondary involvement of the uterine cervix by non-gynecologic neoplasms.

	BC (n=5)	CRC (n=17)	GA (n=2)	GC (n=1)	LAC (n=1)	Lymphoma (n=6)	MM (n=2)	SRCC (n=8)	UC (n=5)	Total (n=47)
* Age at diagnosis; mean, median (range), years	55.8, 55 (50–61)	53.5, 55 (30–69)	63, 63 (59–67)	77, 77 (77–77)	66, 66 (66–66)	43.7, 44 (25–60)	67.5, 68 (53–82)	50.5, 51 (25–72)	69.4, 72 (56–86)	55.5, 55 (25–86)
Initial presentation; no. (%)
Gynecologic
Cervix only	0	1 (5.9%)	1 (50%)	0	0	2 (33.3%)	0	1 (12.5%)	0	5 (10.6%)
Cervix synchronous	0	1 (5.9%)	0	1 (100%)	0	3 (50%)	1 (50%)	1 (12.5%)	0	7 (14.9%)
Gynecologic non-cervix	0	1 (5.9%)	0	0	0	0	0	2 (25%)	1 (20%)	4 (8.5%)
Non-gynecologic	5 (100%)	14 (82.3%)	1 (50%)	0	1 (100%)	1 (16.7%)	1 (50%)	4 (50%)	4 (80%)	31 (66%)
*Time to cervical involvement; mean, median (range), years	4.8, 5 (1.3–8)	1.4, 0.8 (0–4)	1, 1 (0–2)	0 (0–0)	6, 6 (6–6)	1, 0 (0–6)	0.5, 0.5 (0–1)	0.8, 0 (0–5)	3, 1.1 (0–9)	1.8, 0.8 (0–9)
Clinically visible lesion; no. (%)	2 (40%)	3 (17.6%)	1 (50%)	1 (100%)	1 (100%)	4 (66.6%)	2 (100%)	2 (25%)	1 (20%)	17 (36.2%)
Status; no. (%)
Alive with disease	2 (40%)	5 (29.4%)	2 (100%)	0	1 (100%)	1 (16.7%)	0	2 (25%)	0	13 (27.6%)
Deceased	3 (60%)	2 (11.8%)	0	1 (100%)	0	0	1 (50%)	2 (25%)	1 (20%)	10 (21.3%)
No evidence of disease	0	3 (17.6%)	0	0	0	1 (16.7%)	0	2 (25%)	0	6 (12.8%)
Unknown	0	7 (41.2%)	0	0	0	4 (66.6%)	1 (50%)	2 (25%)	4 (80%)	18 (38.3%)
Follow-up (29 patients); mean, median (range), years	10.8, 10.5 (1–21)	3.4, 3.5 (0–8)	1.5, 1.5 (0–3)	0	0	10.5, 10.5 (5–16)	1 (1–1)	2.7, 2.5 (0–8)	14, 14 (14–14)	4.9, 3 (0–21)

BC, breast cancer; CRC, colorectal carcinoma; GA, gastric adenocarcinoma; GC, gallbladder carcinoma; LAC, lung adenocarcinoma; MM, malignant melanoma; SRCC, signet ring cell carcinoma; UC, urothelial carcinoma;

^*Statistically significant variables

The mean patient age was 55.5 years (range 25–86) and varied depending on the type of metastatic tumor (p=0.01); patients with lymphoma tended to be younger compared to those with carcinoma (43.7 versus at least 50.5). The primary site of origin was already known at the time of cervical involvement in most patients (66%, 31/47), while the remaining 34% (16/47) presented with involvement of the gynecologic tract at initial diagnosis. Of these 16 patients, the cervix was the first site of presentation in 5 cases including 2 diffuse large B-cell lymphomas, 1 colorectal carcinoma, 1 gastric carcinoma and 1 appendiceal goblet cell-signet ring cell carcinoma; synchronous involvement of the cervix and other gynecologic sites occurred in 7 patients including 3 diffuse large B-cell lymphomas, 1 gallbladder carcinoma, 1 appendiceal goblet cell-signet ring cell carcinoma, 1 colorectal carcinoma and 1 melanoma; non-cervical gynecologic sites were involved prior to cervical presentation in the remaining 4 patients, the ovary being the most frequent site.

Of 12 patients with cervical involvement at initial presentation (7 with synchronous involvement of other gynecologic sites, 5 with cervical only involvement), vaginal bleeding was the most frequent presenting symptom (33%, 4/12), occurring in 2 patients with diffuse large B-cell lymphoma, and 1 each with gastric carcinoma and appendiceal goblet cell-signet ring cell carcinoma. The remaining 67% (8/12) had clinically and/or radiologically detected lesions suspicious for cervical cancer, including 3 diffuse large B-cell lymphomas, 2 colorectal carcinomas, 1 appendiceal goblet cell-signet ring cell carcinoma, 1 gallbladder carcinoma and 1 melanoma. The mean time to cervical involvement was 1.8 years (range 0–9) and varied depending on the type of metastasis (p=0.03), typically being <1 year for gastrointestinal signet ring cell carcinoma, melanoma and gallbladder carcinoma. In addition, of the 29 patients with available follow-up data, 10 (34.5%) were deceased at the time of this study, 13 (44.8%) were alive with disease, and 6 (20.7%) had no evidence of disease (median 3 years, range 0–21).

Morphologic growth patterns were assessed on cervical biopsy samples (n=18) including polypectomy in 1 melanoma, cone biopsies (n=4), hysterectomy or pelvic exenteration (n=24), and autopsy (n=1). Tumor size could be measured in 41 cases, with a mean and median size of 17.3 mm and 12 mm (range 3–80), respectively. LVSI was present in 28 of 41 (68.3%) tumors consisting of melanomas and carcinomas. Detailed morphologic features of all 47 cases are described in Table 2.

Table 2.

Morphologic features of 47 cases with secondary involvement of the uterine cervix by non-gynecologic neoplasms.

	BC (n=5)	CRC (n=17)	GA (n=2)	GC (n=1)	LAC (n=1)	Lymphoma (n=6)	MM (n=2)	SRCC (n=8)	UC (n=5)	Total (n=47)
Type of specimen
Autopsy	0	0	0	0	0	0	0	1 (12.5%)	0	1 (2.1%)
Biopsy	3 (60%)	4 (23.5%)	2 (100%)	1 (100%)	0	4 (66.6%)	1 (50%)	1 (12.5%)	2 (40%)	18 (38.3%)
Cone biopsy	1 (20%)	1 (5.9%)	0	0	0	1 (16.7%)	0	0	1 (20%)	4 (8.5%)
Hysterectomy	1 (20%)	12 (70.6%)	0	0	1 (100%)	1 (16.7%)	1 (50%)	6 (75%)	2 (40%)	24 (51.1%)
Tumor size; mean, median (range), mm (41 patients)	16.2, 20 (5–30)	13.4, 10.5 (3–50)	40, 40 (40–40)	12, 12 (12–12)	4, 4 (4–4)	35.4, 20 (10–80)	15, 15 (10–20)	15, 16 (4–24)	15.5, 15.5 (8–23)	17.3, 12 (3–80)
* Lymphovascular invasion
Yes	3 (60%)	12 (70.6%)	2 (100%)	1 (100%)	1 (100%)	0	2 (100%)	5 (62.5%)	2 (40%)	28 (59.6%)
No	2 (40%)	5 (29.4%)	0	0	0	0	0	3 (37.5%)	3 (60%)	13 (27.6%)
Not applicable	0	0	0	0	0	6 (100%)	0	0	0	6 (12.8%)
Architectural and cytologic features
Growth pattern
Infiltrating	2 (40%)	11 (64.7%)	1 (50%)	0	1 (100%)	5 (83.3%)	1 (50%)	8 (100%)	1 (20%)	30 (63.8%)
Infiltrating and pagetoid, total	3 (60%)	5 (29.4%)	1 (50%)	1 (100%)	0	0	1 (50%)	0	2 (40%)	13 (27.7%)
Replacing endocervical glands	1 (20%)	1 (5.9%)	0	0	0	0	1 (50%)	0	0	3 (6.4%)
Replacing squamous epithelium	1 (20%)	1 (5.9%)	0	0	0	0	0	0	0	2 (4.3%)
Replacing both	1 (20%)	3 (17.6%)	1 (50%)	1 (100%)	0	0	0	0	2 (40%)	8 (17%)
Pagetoid, total	0	1 (5.9%)	0	0	0	0	0	0	2 (40%)	3 (6.4%)
Replacing endocervical glands	0	1 (5.9%)	0	0	0	0	0	0	0	1 (2.1%)
Replacing squamous epithelium	0	0	0	0	0	0	0	0	2 (40%)	2 (4.3%)
Perivascular	0	0	0	0	0	1 (16.7%)	0	0	0	1 (2.1%)
* Type of glands/growth
Non-complex glands	1 (20%)	5 (29.4%)	0	1 (100%)	0	0	0	2 (25%)	0	9 (19.1%)
Cribriforming	0	4 (23.5%)	1 (50%)	0	0	0	1 (50%)	0	0	6 (12.8%)
Solid growth	1 (20%)	2 (11.8%)	0	0	0	5 (83.3%)	2 (100%)	2 (25%)	1 (20%)	13 (27.6%)
None/other	3 (60%)	9 (52.9%)	1 (50%)	0	1 (100%)	1 (16.7%)	0	4 (50%)	4 (80%)	23 (48.9%)
Cell type
*Goblet cells	0	4 (23.5%)	0	0	0	0	0	1 (12.5%)	0	5 (10.6%)
*Signet ring cells	1 (20%)	3 (17.6%)	0	0	0	0	0	8 (100%)	0	12 (25.5%)
Not applicable	0	0	0	0	0	6 (100%)	2 (100%)	0	0	8 (17%)
* Cytologic atypia
Mild	2 (40%)	0	0	0	1 (100%)	4 (66.6%)	0	0	0	7 (14.9%)
Moderate	3 (60%)	11 (64.7%)	2 (100%)	0	0	1 (16.7%)	1 (50%)	3 (37.5%)	3 (60%)	24 (51.1%)
Severe	0	6 (35.3%)	0	1 (100%)	0	1 (16.7%)	1 (50%)	5 (62.5%)	2 (40%)	16 (34%)
Stromal features
* Depth of stromal invasion
Deep (>50%)	4 (80%)	13 (76.5%)	2 (100%)	0	1 (100%)	4 (66.7%)	1 (50%)	7 (87.5%)	1 (20%)	33 (70.2%)
Superficial (<50%)	0	1 (5.9%)	0	0	0	2 (33.3%)	1 (50%)	1 (12.5%)	2 (40%)	7 (14.9%)
None	0	0	0	0	0	0	0	0	2 (40%)	2 (4.3%)
Not applicable	1 (20%)	3 (17.6%)	0	1 (100%)	0	0	0	0	0	5 (10.6%)
Stromal desmoplasia
Present	1 (20%)	12 (70.6%)	2 (100%)	1 (100%)	0	0	1 (50%)	5 (62.5%)	2 (40%)	24 (51.1%)
Absent	4 (80%)	3 (17.6%)	0	0	1 (100%)	6 (100%)	1 (50%)	3 (37.5%)	2 (40%)	20 (42.5%)
Not applicable	0	2 (11.8%)	0	0	0	0	0	0	1 (20%)	3 (6.4%)
Multifocal growth	0	0	0	0	0	0	1 (50%)	0	0	1 (2.1%)
* Mimics other process
Benign/non-invasive process	0	1 (5.9%)	0	0	1 (100%)	3 (50%)	0	0	2 (40%)	7 (14.9%)
Primary malignant neoplasm	3 (60%)	11 (64.7%)	2 (100%)	1 (100%)	0	2 (33.3%)	0	6 (75%)	3 (60%)	28 (59.6%)
* Ancillary studies	4 (80%)	5 (29.4%)	1 (50%)	1 (100%)	1 (100%)	6 (100%)	1 (50%)	1 (12.5%)	3 (60%)	23 (48.9%)
Misdiagnosed	1 (20%)	1 (5.9%)	0	0	0	2 (33.3%)	0	0	2 (40%)	6 (12.8%)

AIS, adenocarcinoma in situ; BC, breast cancer; CRC, colorectal carcinoma; GA, gastric adenocarcinoma; GC, gallbladder carcinoma; HSIL, high-grade squamous intraepithelial lesion; LAC, lung adenocarcinoma; MM, malignant melanoma; SRCC, signet ring cell carcinoma; UC, urothelial carcinoma;

^*Statistically significant variables

Among the architectural patterns noted, 3 (6.4%) tumors demonstrated exclusive pagetoid growth, including 2 urothelial carcinomas with discohesive cells mimicking HSIL and 1 colorectal carcinoma partially involving endocervical glands and mimicking AIS (Fig. 1A, 2A). A purely infiltrating pattern was identified in 30 (63.8%) cases (Figs. 1B, 3A–C, 4A–C, 5A–D), while infiltrating and pagetoid growths coexisted in 13 tumors (27.6%) (Figs. 1C–F, 6A–D). The chronic lymphocytic leukemia/small lymphocytic lymphoma (1/47, 2.1%) had a distinctive predominantly perivascular pattern of spread (Fig. 5E–F). The morphology and growth patterns of the neoplastic glands were associated with type of metastatic tumor (p=0.014). Neoplastic glands lacking architectural complexity were identified in 9 (19.1%), including 5/17 colorectal carcinomas, 2/8 gastrointestinal signet ring cell carcinomas, 1/1 gallbladder carcinoma and 1/5 mammary carcinoma. Complex cribriform glands/growth were identified in 6 tumors, including 4/17 colorectal carcinomas and 1/2 gastric carcinoma; pseudo-cribriforming was noted in 1/2 melanoma (Figs. 4A–C). Solid growth was present in 13 (27.6%) tumors including 5/6 lymphomas, both melanomas, 2/17 colorectal carcinomas, 2/8 gastrointestinal signet ring cell carcinomas, and 1/5 each urothelial and mammary carcinoma. Goblet cells and signet ring cells were present in 10.6% (5/47) and 25.6% (12/47) of tumors, respectively, and associated with type of metastatic tumor (p<0.0001). In addition to all gastrointestinal signet ring cell carcinomas, focal signet ring cells were also seen in 3/17 colorectal carcinomas and 1/5 mammary carcinoma. Cytologic atypia was mild in 7 (14.9%), moderate in 24 (51.1%) and severe in 16 tumors (34%), and also associated with type of metastatic tumor (p=0.005). Severe cytologic atypia was seen in 6/17 colorectal carcinomas, 5/8 gastrointestinal signet ring cell carcinomas, 2/5 urothelial carcinomas, 1/1 gallbladder carcinoma, 1/2 melanoma, and 1/6 lymphoma. At least focal intraluminal “dirty necrosis” was identified in 13/17 colorectal carcinomas.

Figure 1.

Urothelial carcinoma. Pagetoid involvement of cervical squamous epithelium with discohesive tumor cells, resembling high-grade squamous intraepithelial lesion (A), and invasive urothelial carcinoma with infiltrating growth pattern involving the vagina 2 years later (B); Pagetoid (asterisks) (C, E) and infiltrating (C-F) growth patterns of urothelial carcinoma. Hematoxylin-eosin stain; magnification x400 (A, D, F), x100 (B, C), x200 (E).

Figure 2.

Gastrointestinal carcinomas. Colorectal carcinoma with an adenocarcinoma in situ/AIS-like pattern with partial involvement of an endocervical gland (A), associated with invasive pattern and intraluminal “dirty necrosis” in the endometrium (B), ovary (C) and colon (D); Appendiceal signet ring cell adenocarcinoma with scattered signet ring cells in cervical os biopsy (E), positive for cytokeratin 20 (F). Hematoxylin-eosin stain (A-E), immunohistochemistry (F); magnification x200 (A, E, F), x100 (C, D), x40 (B).

Figure 3.

Breast carcinomas. Ductal carcinoma with infiltrating nests (A) and positive immunohistochemical staining for mammaglobin (B); Lobular carcinoma infiltrating cervical stroma in single files (C). Hematoxylin-eosin stain (A, C), immunohistochemistry (B); magnification x200 (A, C), x400 (B).

Figure 4.

Malignant melanoma. Nests of markedly atypical epithelioid cells infiltrating cervical stroma with adjacent normal endocervical glands (A, B). Melanin pigment and lymphovascular invasion is present (C). Hematoxylin-eosin stain (A-C); magnification x40 (A), x200 (B, C).

Figure 5.

Lymphomas. Diffuse large B-cell lymphoma (A, B), positive for CD20 (C) and Bcl-6 (D); Chronic lymphocytic leukemia/small lymphocytic lymphoma with perivascular lymphoid infiltrate composed of small, monomorphic B-cells (E, F). Hematoxylin-eosin stain (A, B, E, F), immunohistochemistry (C, D); magnification x40 (A, E), x400 (B, F), x200 (C, D).

Figure 6.

Gastrointestinal carcinomas. Fragments of colorectal carcinoma in cervical biopsy (A, B), positive for cytokeratin 20 (C) and CDX2 (D). Hematoxylin-eosin stain (A, B), immunohistochemistry (C, D); magnification x100 (A), x200 (B-D).

Depth of stromal invasion was associated with type of metastatic tumor (p=0.027), with deep invasion seen in 33 tumors (70.2%), including colorectal (13/17), gastrointestinal signet ring cell (7/8), mammary (4/5), gastric (2/2), pulmonary (1/1) and urothelial (1/5) carcinomas, melanoma (1/2) and lymphomas (4/6). Stromal desmoplasia was identified in 24 tumors (51.1%), including 12/17 colorectal carcinomas, 5/8 gastrointestinal signet ring cell carcinomas, 2/2 gastric carcinomas, 2/5 urothelial carcinomas, 1/1 gallbladder carcinoma, 1/5 mammary carcinoma and 1/2 melanoma. Multifocal growth was identified in only 1 melanoma (2.1%).

Overall, the morphologic features mimicked a primary cervical neoplasm in 28 (59.6%) cases including 26 carcinomas mimicking usual-type or gastric-type endocervical adenocarcinoma or poorly differentiated squamous cell carcinoma and 2 lymphomas mimicking sarcoma, and a benign/non-invasive process in 7 tumors (14.9%) including 3 lymphomas (mimicking a reactive lymphoid infiltrate), 1 lung adenocarcinoma (non-complex glands lined by cuboidal epithelium with nuclear hyperchromasia and abortive cilia admixed with cells with basophilic or clear cytoplasm resembling secretory cells and mimicking tuboendometrioid metaplasia), 2 urothelial carcinomas (pagetoid spread mimicking HSIL) and 1 colorectal carcinoma (pagetoid spread mimicking endocervical AIS) (p=0.002).

Ancillary studies were performed in a total of 23 tumors (48.9%), including 6/6 lymphomas, 5/17 colorectal carcinomas, 4/5 breast carcinomas, 3/5 urothelial carcinomas, 1/2 gastric carcinoma, 1/1 lung adenocarcinoma, 1/1 gallbladder carcinoma, 1/8 gastrointestinal signet ring cell carcinoma and 1/2 melanoma (p=0.03). The stains included epithelial markers (cytokeratins 7, 8/18 and 5/6, pan-cytokeratin, epithelial membrane antigen, CAM5.2, carcinoembryonic antigen), markers of gynecologic (PAX8, WT1), gastrointestinal (CK20, CDX2, SATB2), pulmonary (TTF-1, napsin-A), mammary (GATA3, mammaglobin, estrogen and progesterone receptors, HER2), melanocytic (S100, SOX10, HMB45, melan-A) and hematolymphoid (CD45, CD34, PAX5, CD68, CD20, CD3, CD23, CD30, ALK1, c-myc, Mum1, Bcl-6) lineage, markers of squamous (p63, p40) and neuroendocrine (synaptophysin, chromogranin, CD56) differentiation, and others (CD10, p16, vimentin, Ki-67, p53).

Six consultation cases (12.8%) were initially misdiagnosed by the original pathologists as cervical primaries. One invasive lobular carcinoma was diagnosed as a squamous cell carcinoma, 1 non-invasive high-grade urothelial carcinoma as HSIL (showing diffuse, block-like p16 staining typical of HSIL), 1 urothelial carcinoma with mixed infiltrating and pagetoid growth as endocervical AIS and vulvar Paget disease, 1 colorectal carcinoma as endocervical AIS, and 2 diffuse large B-cell lymphomas as “sarcoma not otherwise specified” and “endometrial stromal sarcoma, low-grade with focal high-grade features.” Immunohistochemical studies were subsequently performed in these cases leading to the correct diagnosis.

DISCUSSION

Our study describes the clinical and morphologic features of 47 non-gynecologic neoplasms metastatic to the uterine cervix, including 39 carcinomas, 6 lymphomas and 2 malignant melanomas, representing the largest single study from 4 institutions. Of note, 34% (16/47) of patients had involvement of the gynecologic tract at initial diagnosis, including 5 presenting with only cervical involvement. In addition, 12.8% of tumors were initially misdiagnosed as cervical primaries, emphasizing the need for awareness of metastases mimicking primary cervical carcinomas. The low incidence of metastases involving the cervix has been attributed to its relatively small size combined with abundant fibromuscular tissue, scarce vasculature and centrifugally draining lymphatic network (12). The potential mechanisms for metastasis are hematogenous, retrograde lymphatic and transperitoneal (transtubal or direct extension from the cul-de-sac) routes depending on the primary site (1, 3).

The most common site of origin for carcinomas in our cohort was the gastrointestinal system (27/39, 69.2%), both the upper and lower tracts. The majority (13/17) of tumors showed deep stromal invasion with infiltrating or pagetoid and infiltrating growth patterns and “dirty necrosis,” and 1 case was misdiagnosed as endocervical AIS. Colorectal carcinoma usually metastasizes to the liver, lung and peritoneum and less frequently to the gynecologic tract, with ovarian involvement in up to 7% of all cases. Uterine involvement accounts for less than 10% of all non-gynecologic metastases to the gynecologic tract, including 3.4% with cervical metastases (1). Cervical involvement may result from either direct extension or hematogenous spread (6, 7, 13–17). Metastatic colorectal carcinoma may mimic intestinal-type invasive or in situ endocervical adenocarcinoma morphologically (neoplastic glands with pseudostratified columnar cells) and immunophenotypically (expression of CK20 and CDX2). In addition to the clinical history, useful distinguishing features include the presence of a mass invading the cervical wall from outside usually with mucosal sparing, a transmural mass extending deeply into the cervical wall, and the finding of intraluminal necrotic debris (“dirty necrosis”) when present (6, 7, 14, 17). A SATB2 immunohistochemical stain may also be helpful as most colorectal carcinomas are positive, although 5% of endocervical adenocarcinomas have also been shown to express SATB2 (18).

Signet ring cell carcinoma is a variant of poorly differentiated mucin-producing adenocarcinoma most frequently arising in the stomach and commonly metastasizing to the peritoneum and ovaries (Krukenberg tumor). Thus, not surprisingly, most reported metastatic signet ring cell carcinomas to the cervix have originated from the stomach and displayed extensive LVSI (7, 19). Only 1 appendiceal signet ring cell carcinoma (7) and 2 appendiceal goblet cell carcinoids metastatic to the cervix have been reported (20), both described to have an infiltrating growth, 1 with associated LVSI. The signet ring cell carcinomas in our study were more frequently gastric but also appendiceal or ileal in origin, with an infiltrating growth pattern in all tumors and LVSI in most (5/8). Given that signet ring cell morphology is not site-specific, clinical history and immunohistochemical studies are essential to distinguish between gastrointestinal signet ring cell carcinomas (CK20, CDX2, SATB2) and those originating in the lung (CK7, TTF-1, napsin-A), breast (CK7, GATA3, mammaglobin, hormone receptors) or urinary bladder (CK7, CK20, p63) (21, 22).

Non-signet ring cell gastric adenocarcinoma may also mimic gastric-type endocervical adenocarcinoma, a human papillomavirus independent cervical tumor characterized by columnar cells with eosinophilic or foamy cytoplasm containing gastric-type mucin and basally located nuclei with inapparent apical mitoses (23). There are 2 reports of poorly differentiated gastric adenocarcinoma metastatic to the cervix (24, 25), described to infiltrate around benign endocervical glands (25). Both gastric carcinomas in our study were characterized by an infiltrating growth (1 admixed with pagetoid spread). Gastric carcinoma has overlapping morphology and immunophenotype (positive CK7, CK20, HIK1083) with gastric-type endocervical adenocarcinoma. Furthermore, p16 is usually not useful in this context given that most (although not all) gastric-type endocervical adenocarcinomas lack p16 expression. PAX8 may be of value, when positive, to confirm a Müllerian origin; however, up to a third of endocervical gastric-type adenocarcinomas are PAX8 negative (26). Given the overlapping morphology and immunophenotype, it is crucial to inquire about a prior history and/or ongoing clinical suspicion for gastric cancer before the diagnosis of gastric-type endocervical adenocarcinoma is made.

There was 1 patient with metastatic gallbladder carcinoma in our study, with a known history of advanced gallbladder carcinoma, deep stromal invasion, infiltrating and pagetoid growth patterns replacing both squamous and endocervical glandular epithelium, non-complex glands with desmoplastic stromal reaction and LVSI. This tumor is usually diagnosed at advanced stage and characterized by frequent visceral metastases (27–29). Only 8 metastatic gallbladder carcinomas to the cervix have been reported to date (9, 30–35), most patients with a history of gallbladder cancer at the time of cervical presentation. One patient was described to have non-mass forming symmetrical enlargement of the uterus (33), 2 presented with cervical erosion/ulceration (30, 35), 1 with cervical mass (9), while 2 had no gross abnormality (34); 3 patients showed diffusely infiltrating adenocarcinoma around benign endocervical glands (9, 30, 33), 2 with desmoplastic stromal reaction (9, 33), and 1 with extensive LVSI (34); no additional description is available in the remaining patients. As demonstrated in our study, deep cervical stromal invasion with non-complex glands, infiltrating and pagetoid growth patterns in association with LVSI may be suggestive of a gallbladder primary.

Our cohort included 5 patients with mammary carcinoma (3 ductal, 2 classical lobular). Two ductal carcinomas showed an infiltrating growth, while infiltrating and pagetoid growth patterns were seen in 2 lobular and 1 ductal carcinomas, and 1 lobular carcinoma was initially misdiagnosed as squamous cell carcinoma. Although the ovary is the most common gynecologic site for metastasis in breast cancer (36), breast cancer is the most common non-gynecologic neoplasm to metastasize to the uterine corpus, accounting for 43% of cases in a study of 63 patients (37). Approximately 40 breast cancers metastatic to the cervix have been reported to date, most with known history and presenting with either widespread disease or isolated cervical involvement (6, 7, 24, 38–44). Prognosis is generally poor with a mean survival of 28 months (range 2–12) (16, 38, 39, 41, 43, 46), and close gynecologic surveillance has therefore been suggested in breast cancer patients (41, 46). Visceral metastases are particularly characteristic for lobular carcinoma where cells tend to be discohesive due to loss of intercellular adhesion molecule e-cadherin, invading in single files and usually sparing the cervical epithelium (7, 38, 40, 41, 45). Cervical adenocarcinoma and squamous cell carcinoma with acantholytic changes may undergo dedifferentiation resulting in morphologic mimicry of classical lobular carcinoma (47, 48). The finding of discohesive cells infiltrating in single files and/or small nests within the cervical stroma, especially with sparing of cervical epithelium, should raise the possibility of lobular carcinoma and trigger the appropriate immunohistochemical studies (GATA3, mammaglobin, ER, PR, HER2). Ductal carcinoma may be more difficult to recognize in the absence of a prior history, typically showing infiltrating nests or glands and mimicking a poorly differentiated usual-type endocervical adenocarcinoma (42, 43).

We report 5 urothelial carcinomas, 3 published previously (11). Variable growth patterns were seen, including 2 tumors with pagetoid spread of dyscohesive cells mimicking HSIL, 2 with concurrent pagetoid and infiltrating growth patterns, and 1 with an infiltrative growth. Two of 5 tumors were misdiagnosed. Patients with urothelial carcinoma typically present with multiple recurrences affecting the urothelial tract as well as with distant metastases (49). Involvement of the gynecologic tract may occur in a synchronous or metachronous manner, most frequently involving the vagina, uterus and vulva (50–57). Cervical involvement is very rare with fewer than 10 reported cases and typically follows vaginal metastasis (4, 6, 11, 53). Distinguishing urothelial carcinoma from primary cervical squamous lesions can be particularly difficult due to overlapping architectural, cytologic and immunohistochemical (diffuse CK7, variable CK20 and p16 expression) features (11). Both tumor types may show papillary, in situ, and invasive growth. Abundant eosinophilic cytoplasm (lower-grade tumors) and enlarged, hyperchromatic, irregular nuclei (high-grade tumors) also represent shared cytologic features (11). Although most patients have a known history of urothelial carcinoma, the gynecologic tract may be the first manifestation in some (11). Furthermore, the presence or absence of invasion in the original tumor does not necessarily determine the pattern of growth in the gynecologic tract (11, 50–54). In the absence of a prior history, pagetoid growth along the cervical epithelium will undoubtedly mimic HSIL (11). However, discohesive nature of neoplastic cells in the areas of pagetoid growth may be a clue to urothelial origin at least in some cases.

In the single lung adenocarcinoma in our series, the diagnosis had been established prior to the gynecologic presentation and LVSI, deep stromal invasion and infiltrating growth with well-formed glands mimicking tuboendometrioid metaplasia was seen. Adenocarcinoma is the most prevalent histotype among non-small cell lung carcinomas, typically metastasizing to the liver, bones, brain and adrenal glands. Prognosis depends on tumor stage, grade, and presence of epidermal growth factor receptor (EGFR) mutation (58). Metastases to the gynecologic tract are rare, accounting for 4.8% (3/63) of non-gynecologic metastases to the uterine corpus (37). Only 3 cervical metastases have been reported (59–61), including 1 diagnosed on routine cytology (61), 1 presenting with an exophytic mass, infiltrating growth and deep cervical wall invasion (60), and 1 with imaging findings suggestive of cervical carcinoma (59). As demonstrated in our study, deep cervical stromal invasion with well-formed glands in association with LVSI may be suggestive of pulmonary origin. Clinical history and immunophenotype (positive TTF-1, napsin-A) can further be helpful in establishing the correct diagnosis.

There were 2 metastatic cutaneous malignant melanomas to the cervix in our series, both with LVSI, one with an infiltrating growth, and the other with infiltrating and pagetoid growth patterns. Melanoma can arise in the gynecologic tract as a primary tumor, and usually affects the vulva, less frequently the cervix and vagina (62), with approximately 80 primary cervical melanomas (63, 64) and only 9 metastatic cutaneous melanomas to the cervix published to date (1, 7, 16, 65–67). Most patients are asymptomatic but some may present with vaginal bleeding due to ulceration or a mass (68). Distinguishing between primary and metastatic melanoma can be difficult. The absence of junctional activity and melanin pigment in the squamous epithelium and a prior history or concurrent melanoma elsewhere favor metastatic origin (68, 69). Melanoma can also mimic poorly differentiated primary or metastatic carcinomas due to an infiltrating nested or solid growth of variably atypical epithelioid cells (16, 66). In general, the presence of melanin pigment and LVSI combined with the expression of melanocytic markers (S100, HMB-45, melan-A, SOX10, MITF) and negative epithelial markers should allow accurate diagnosis of melanoma (either primary or metastatic).

Five of the 6 lymphomas in our series were diffuse large B-cell lymphomas, of which the most recent tumor was categorized as a germinal center type (associated with better overall survival) (70). All tumors had an infiltrating solid growth and required extensive workup. In addition, 4 showed deep stromal invasion, 3 mimicked a reactive lymphoid infiltrate, and 2 were misdiagnosed as sarcoma, most likely due to associated sclerotic stroma. Lymphomas involving the uterine cervix can be part of a systemic process or originate in the cervix. Diffuse large B-cell lymphoma is the most common type of primary cervical lymphoma (71). A systematic literature review from 2014 identified 143 primary uterine diffuse large B-cell lymphomas, including 108 (75.5%) with cervical involvement (7, 72–78). The mean patient age was 49.6 years and vaginal bleeding was the most common clinical symptom (64%), while 13% were asymptomatic (72). Diffuse large B-cell lymphomas may mimic primary cervical or metastatic poorly differentiated carcinomas and sarcomas and usually show a diffuse growth of markedly atypical cells, deep stromal invasion and sometimes sclerotic stroma (76). The absence of an associated intraepithelial lesion in the cervix, diffuse/solid and/or infiltrative growth patterns in a poorly differentiated neoplasm lacking morphologic evidence of specific lineage differentiation and/or a prior history of lymphoma should raise suspicion and trigger additional work-up for possible lymphoid origin in cervical specimens.

One patient in our study had chronic lymphocytic leukemia/small lymphocytic lymphoma which was an incidental finding and showed a prominent perivascular growth pattern. This is the most common adult leukemia, characterized by the accumulation of small B lymphocytes in the blood, bone marrow and lymphoid tissues, frequently diagnosed incidentally. An autopsy study reported uterine involvement in 13% of 109 patients (79). Cervical involvement has been described in a few case reports and small series (7, 76, 80–85), predominantly in patients with advanced pelvic disease (80, 85). This low-grade lymphoma is not usually confused with primary or metastatic epithelial neoplasms and may raise a differential diagnosis with a reactive lymphoid infiltrate in the uterine cervix. If combined with a prior history, recent lymphocytosis and/or older age, the possibility of cervical involvement by chronic lymphocytic leukemia/small lymphocytic lymphoma should be entertained. In the absence of such history, the finding of a monotonous cell population with proliferation centers (81, 84), deep location and perivascular distribution within the cervix are helpful features that argue against a reactive process and should trigger immunohistochemical studies with lymphoid markers (CD3, CD20, CD5, CD23) by the pathologist.

In conclusion, involvement of the uterine cervix by non-gynecologic tumors can mimic primary invasive or intraepithelial neoplasms in both the ecto- and endocervix and cervical stroma, including carcinomas and sarcomas. A high level of suspicion is often necessary to work up the tumor beyond initial morphologic impressions. Obtaining pertinent clinical history combined with careful histologic examination and ancillary studies for targeted biomarkers are crucial to ensure an accurate diagnosis.