Granulosa Cell Tumors: Novel Predictors of Recurrence in Early-stage Patients

Summary:

Granulosa cell tumors (GCTs) comprise 2% to 5% of ovarian neoplasms, with unpredictable patterns of recurrence. The HER family, GATA4, and SMAD3 genes are reportedly involved in GCT proliferation and apoptosis and may serve as new predictors of recurrence. The aim of the study was to evaluate novel predictors of recurrence in GCT from a large single institution cohort. Patients diagnosed with GCTs (n = 125) between 1975 and 2014 were identified. Clinicopathologic parameters were obtained and immunohistochemical evaluation was performed of calretinin, inhibin, HER2, CD56, SMAD3, and GATA4. Statistical analyses were conducted using Fisher exact test and Kaplan-Meier survival curves and Cox regression analysis. The median follow-up period was 120 months (range, 1–465 mo). Recurrence was noted in 12/125 (9.6%) patients. Kaplan-Meier analysis showed a shorter mean disease-free interval in whites versus blacks (P = 0.001), stage III-IV versus stage I-II (P = 0.0001), patients treated with surgery+chemotherapy versus surgery (P = 0.0001), mitotic rate ≥4 (P = 0.005), severe nuclear pleomorphism (P = 0.013), high HER2 expression (P = 0.001), high CD56 expression (P = 0.001), and high SMAD3 expression (P = 0.001). On Cox regression analysis, SMAD3 and type of treatment received were the only 2 independent prognostic factors for disease-free interval (P = 0.03 and P = 0.007, respectively). On subanalysis for early-stage (stage I) GCTs, the need for adjuvant chemotherapy and high expression of SMAD3 continued to be independent predictors of recurrence (HR = 10.2, P = 0.01 and HR = 8.9, P = 0.001, respectively).

Ovarian granulosa cell tumors (GCTs) represent approximately 2% to 5% of all ovarian cancers. They are the most common type of ovarian sex-cord stromal tumors (1). GCTs are divided histologically into juvenile and adult forms. Adult granulosa cell tumor (AGCTs) account for 95% of GCTs with a median age at diagnosis of 50 to 55 years (2). GCTs have malignant potential but fortunately, the majority of patients present with stage I disease, for which the long-term disease-free survival is approximately 84% to 95% (3,4). Many of these patients are cured by surgery but insidious recurrence is still reported in up to 20% to 30%, even in early-stage disease (5,6). Because complete tumor resection is a significant prognostic factor (5,7) and because of the indolent nature of GCTs, routine staging has been traditionally recommended with the exception when there is a need for fertility-sparing surgery.

Although prognostic indicators have long been studied, predicting the likelihood of recurrence with certainty remains elusive; hence, postoperative management is usually based on stage of disease, completeness of tumor resection/presence of residual tumor, and whether fertility-sparing surgery is performed.

Chemotherapy has been used for extensive residual disease and inoperable metastases/recurrences (8–10). Choice of ideal candidates as well as standard postoperative adjuvant therapy is yet to be determined (11), hence the need for more accurate clinical and pathologic predictors of the clinical behavior of GCT.

Prognostic factors investigated can be divided into clinical as well as pathologic factors. Clinical prognostic factors include age (1,12,13), tumor size (1,12,14), stage of disease (12–16), extent of surgery (3,15), and residual disease/tumor rupture (5,7). Pathologic factors include tumor histology (4,17), mitotic index (MI) (7,17,18), nuclear atypia (19), and expression of inhibin, epidermal growth factor (EGFR), and Ki67 (20,21).

With the recent diagnosis of the missense point mutation c402>G (codon C134W) in the gene coding for Forkhead Box L2 (FOXL2) transcription factor in 90% to 97% of GCTs (22), FOXL2 pathways were investigated with the aim of identifying new prognostic factors. Absence of and increased FOXL2 expression were described as poor prognostic factors (23,24). Aberrant granulosa cell proliferation and apoptosis have been recently implicated in the pathogenesis of GCTs secondary to interactions between C134-mutated FOXL2, SMAD3, and GATA4 (25). SMAD3, a transcription factor in the tumor growth factor-β (TGF-β) signaling pathway, activates NF-_kB which increases cell viability in human GCT cells through a positive feedback loop (26). In murine cell models, SMAD3 interacts with the GATA4, suggesting the involvement of GATA4 in TGF-β signaling pathway as well (27).GATA4 is a zinc finger transcription factor that plays an important role in granulosa cell function and development (27,28), and has recently been associated with advanced stage GCT and increased recurrence risk (29) because of its antiapoptotic effect (30).

Increased expression of human epidermal growth factor receptor 2 (HER2), a transmembrane tyrosine kinase receptor, is prognostic of aggressive disease in breast, gastric, and epithelial ovarian cancer (31,32) with inconsistency in its expression in GCTs; positive in some studies (33,34) and negative in others (35,36).

CD56 is a membrane-bound cell surface sialoglycoprotein and a member of the immunoglobulin supergene family, which induces cell-to-cell interactions during embryonic development, cell migration, and organogenesis (37).

CD56 is expressed by most neuroectodermal-derived cell lines and different tumors and its expression is associated with poor prognosis in a subset of squamous cell carcinoma and renal cell carcinoma (38). CD56 expression in primordial follicles and theca cells was described to have a role in folliculogenesis (39). CD56 expression in GCTs was suggested to be secondary to acquisition of CD56 immunoreactivity during neoplastic transformation or because GCTs originate from theca cells rather than granulosa cells (40). Expression of CD56 was described in primary and recurrent GCT but not as a predictor of recurrence (41).

We report a review of patients diagnosed with GCTs from a single large institution with the aim of identifying new predictors of recurrence, which could aid in patient counseling and decision making postoperatively and providing more targeted therapy.

MATERIALS AND METHODS

After obtaining Institutional Review Board approval, patients diagnosed with GCTs (n = 125) between 1975 and 2014 were identified. Clinical parameters (age, race, body mass index, treatment, time to recurrence, and vital status) and pathologic variables (histologic type, tumor size, and stage) were obtained from electronic medical records and pathology records. Tumor stage was assigned according to the International Federation of Gynecology and Obstetrics (FIGO) criteria (42). Morphologic features including mitotic rate, nuclear pleomorphism (NP), and histologic patterns were evaluated from review of whole-tissue sections of each tumor. Mitotic rate was assessed by enumerating the number of mitotic figures in 10 consecutive HPFs (× 400, field diameter = 0.55 mm); <4/10 HPF is low and ≥4/10 HPF is high. Immunohistochemical (IHC) evaluation was performed on tissue microarrays for 88 primary GCT cases. Tissue microarrays were constructed from formalin-fixed, paraffin embedded blocks using a duplicate of 1-mm-diameter core obtained from a representative area of each tumor (43). The antibodies used were SMAD3 (Catalog no. 51–1500, 1:100 dilution; Invitrogen, Camarillo, CA), VEGF (sc-152, 1:100 dilution; Santa Cruz Biotechnology, Santa Cruz, CA), Her2 (sc-33684, 1:50–1:500 dilution; Santa Cruz Biotechnology), Inhibin (mcA951ST, 1:25 dilution; Serotec, Raleigh, NC), Calretinin (PA5–16681, dilution 1:100; Life Technologies, Chicago, IL), CD56 (123C307–5603, 1:100 dilution; Invitrogen, Carlsbad, CA), and GATA4 (LS-B7989, 1:100 dilution; Lifespan Biosciences Inc.). Immunostaining was performed using heat-induced epitope retrieval. Slides were blocked with 3% H₂O₂ and then incubated with the antibody for 1 hour at 37°C. Detection of antibody binding was obtained using Optiview (Ventana) after a polymer-based amplification step (Ultraview; Ventana). Slides were then counterstained with hematoxylin and coverslipped. Positive controls were used; for calretinin, mesothelioma tissue section; for inhibin, adrenal gland tissue section; for CD56, brain tissue section; for GATA4, prostate tissue section; for HER2, HER2-positive breast tumor tissue section; and for SMAD3, breast tissue was used. With regards to negative controls, ovarian tissue section without any pathology was used. All diagnoses and IHC reads were confirmed by 2 gynecologic oncology pathologists and in a few cases, where discrepancy was noted, a multihead microscope was used to discuss the areas of discrepancy to reach a consensus on the final read.

H Score

Each marker was scored for % and intensity and H score was established by combining both. Expression of GATA4 was categorized as negative for 0%, low for <80% with +1 nuclear staining, intermediate for <80% with +2/+3 or for >80% with +1, and high for ≥ 80% with +2/+3 staining (25). Expression of inhibin was categorized as negative for 0%, low for ≤50% with +1 cytoplasmic staining, intermediate for ≤50% with +2/+3 or >50% with +1, and high for >50% with +2/+3 staining (44); for SMAD3, negative for 0%, low for any % with +1 cytoplasmic staining, intermediate for <70% with +2/+3, and high for ≥70% with +2/+3 staining (25); for VEGF, negative for <20% with +1 cytoplasmic staining, low for 20% to <60% with +2, intermediate for ≤60% with +1/+3 or >60% with +1/+2, and high for >60% with +3 staining (45). Scoring of membranous antigen for Her2 was negative for 0%, low for <80% with +1 staining, intermediate for <80% with +2/+3 or ≥80% with +1, and high for ≥80% with +2 or +3 staining (19). Scoring for membranous and cytoplasmic antigen for CD56 was negative for 0%, low for <50% with +1 or +2 staining, intermediate for <50% with +3 or ≥50% with +1/+2, and high for ≥50% with +3 staining (46). Expression of calretinin was scored as low for <50% with +1 nuclear and cytoplasmic staining, intermediate for ≤50% with +2/+3 or >50% with +1, and high for >50% with +2/+3 staining (44). Figure 1 illustrates the different IHC stains used. Statistical analyses were conducted using Fisher exact test (2-tailed) and Kaplan-Meier (KM) survival curves. Significance was set at P<0.05.

FIG. 1.

IHC stains of primary GCT showing high mitoitc index and high expression of different the pathologic markers [Calretinin (A), CD56 (B), GATA4 (C), HER2 (D), Inhibin (E), SMAD3 (F)] investigated.

RESULTS

The patient clinical characteristics (125 patients with the diagnosis of primary GCT) with respect to recurrence were analyzed using x² and Fisher exact test to perform univariate analyses and are summarized in Table 1. The median follow-up period was 120 months (range, 1–465 mo). Recurrence was noted in 12/125 (9.6%) patients and was significantly higher in whites and patients who had surgery and chemo-therapy (P = 0.001 and 0.0001, respectively). Early-stage GCT (stages I–II) were significantly less likely to have a recurrence as compared with stages III-IV (P = 0.001). Patients with recurrent GCT had a shorter overall survival (83 vs. 138 mo) and disease-free interval (DFI) (60 vs.135 mo), without a statistical difference noted.

TABLE 1.

Correlation of clinical characteristics of primary GCT to the recurrence status

Parameter	Variable	n (%)		P
		Nonrecurrent (n = 113)*	Recurrent (n = 12)
Age at diagnosis	≤50	64 (59.3)	4 (33.3)	0.47
	> 50	44 (40.7)	8 (66.7)
Race	White	12 (16.7)	8 (66.7)	0.001
	Black	46 (63.9)	3 (25.0)
	Other	14 (19.4)	1 (8.3)
BMI (median)		31.5	27.7	0.41
Stage	Stage I-II	111 (98.2)	8 (66.7)	0.001
	Stage III-IV	2(1.8)	4 (33.3)
Lymph node status	Negative	28 (93.3)	5 (100.0)	1.0
	Positive	2 (6.7)	0 (0.0)
Lymphadenectomy	Yes	30 (27.3)	4 (66.7)	0.66
	No	80 (72.7)	2 (33.3)
Treatment	Surgery only	63 (92.6)	2 (16.7)	0.0001
	Surgery+chemo	3 (4.4)	10 (83.3)
	Chemotherapy	2 (2.9)	0 (0.0)
Type of surgery	USO	27 (40.9)	4 (33.3)	0.54
	BSO	7 (10.6)	1 (8.3)
	TAH/USO	26 (39.4)	7 (58.3)
	TAH/BSO	6 (9.1)	0 (0.0)
Median survival (mo)		138	83	0.94
DFI (mo)		135	60	0.12
Vital status	Alive	77 (68.1)	6 (50.0)	0.21
	Dead	36 (31.9)	6 (50.0)

^*Data is missing in some cases.

Bold values are statistically significant.

BMI indicates body mass index; DFI, disease-free interval; GCT, granulosa cell tumors.

Histomorphologic Features and Tumor Characteristics Associated With Recurrent GCT

AGCT comprised the majority of patients (113/125, 90%) and 12 (10%) patients were juvenile type (JGCT) (Table 2). Tumors with a mitotic rate ≥4/10 HPF and severe NP were found to have a significantly higher likelihood of recurrence (P = 0.021 and 0.016, respectively). JGCT were more likely to recur (2/12, 17%) as compared with AGCT (10/113, 9%). Tumor subtype, size, and capsule rupture did not show a significant effect on recurrence.

TABLE 2.

Correlation of histomorphologic features and tumor characteristics of primary GCT with recurrence

Parameter	Variable	n (%)		P
		Nonrecurrent (n = 113)*	Recurrent (n = 12)
Tumor type	Adult	103 (91.2)	10 (83.3)	0.32
	Juvenile	10 (8.8)	2 (16.7)
Tumor subtype	Solid	39 (61.9)	5 (50.0)	0.55
	Cystic	9 (14.3)	1 (10.0)
	Solid cystic	15 (23.8)	4 (40.0)
Tumor size	<10 cm	43 (47.8)	6 (60.0)	0.52
	≥10 cm	47 (52.2)	4 (40.0)
Capsule rupture	Yes	13 (28.3)	2 (50.0)	0.57
	No	33 (71.7)	2 (50.0)
Histopathologic type	Diffuse	45 (61.6)	7 (70.0)	0.98
	Nondiffuse types	28 (38.4)	3 (30.0)
Mitotic rate	<4/10 HPF	56 (70.0)	4 (33.3)	0.021
	≥4/10 HPF	24 (30.0)	8 (66.7)
Nuclear pleomorphism	Mild	42 (63.6)	2 (20.0)	0.016
	Moderate	19 (28.8)	5 (50.0)
	Severe	5 (7.6)	3 (30.0)

^*Data is missing in some cases.

Bold values are statistically significant.

GCT indicates granulosa cell tumors.

Univariate analysis for IHC assessment of protein expression is illustrated in Table 3. High levels of Inhibin, HER2, CD56, and SMAD3 expression were significantly associated with increased recurrence (P = 0.02, 0.014, 0.001, and 0.0001, respectively). On multivariate logistic regression analysis of clinicopathologic variables, treatment type continued to have a significant impact on recurrence, with the addition of chemotherapy to surgery predicting increased likelihood of recurrence by 59-fold (P = 0.001; 95% CI, 5.39–646.34). Mitotic count ≥4/10 HPF and high levels of expression of CD56 and SMAD3 were also found to be independent predictors of recurrence (OR = 13.3, P = 0.038;OR = 9.8, P = 0.04; and OR = 14.2, P = 0.001, respectively) (Table 4). HER2 expression was omitted from the multivariate analysis because of not having nonrecurrent patients with HER2 expression.

TABLE 3.

Correlation of IHC markers with recurrent and nonrecurrent GCT

Marker	H Score	n (%)		P
		Nonrecurrent (n = 71)*	Recurrent (n = 12)
Calretinin	Negative	15 (21.1)	1 (10.0)	0.71
	Low: ≤50%	25 (35.2)	4 (40.0)
	High: >50%	31 (43.7)	5 (50.0)
Inhibin	Negative	41 (59.4)	3 (30.0)	0.02
	Low: ≤50%	24 (34.8)	4 (40.0)
	High: >50%	4 (5.8)	3 (30.0)
Her2	Negative	71 (100.0)	8 (80.0)	0.014
	High≥80%	0 (0.0)	2 (20.0)
CD56	Negative	28 (38.4)	0 (0.0)	0.001
	Low: ≤80%	17 (23.3)	0 (0.0)
	High: >80%	28 (30.4)	10 (100.0)
VEGF	Negative	2 (2.8)	0 (0.0)	0.74
	Low: ≤80%	2 (2.8)	0 (0.0)
	High: >80%	67 (94.4)	10 (100.0)
SMAD3	Negative	51 (72.9)	0 (0.0)	0.0001
	Low: ≤80%	11 (15.7)	2 (20.0)
	High: >80%	8 (11.4)	8 (80.0)
GATA4	Negative	53 (74.6)	6 (60.0)	0.15
	Low: ≤80%	15 (21.1)	2 (20.0)
	High: >80%	3 (4.2)	2 (20.0)

^*Data is missing in some cases.

Bold values are statistically significant.

GCT indicates granulosa cell tumors; IHC, immunohistochemistry.

TABLE 4.

Multivariate analysis using forced stepwise logistic regression and cox regression analyses

Variable	OR	95% CI	P	HR	95% CI	P
Treatment	59.04	5.39–646.34	0.001	7.02	1.69–29.17	0.007
SMAD3	14.23	2.87–70.65	0.001	5.02	1.16–21.64	0.03
Mitotic rate	13.33	1.15–154.41	0.038	0.86	0.95–7.78	0.89
CD56	9.88	1.06–91.44	0.04	5.54	0.26–118.66	0.27

Bold values are statistically significant.

CI indicates confidence interval; HR, hazard ratio; OR, odds ratio.

Using KM analysis, shorter mean DFI was noted in whites (142.7 vs. 212.9 and 199.52 mo in blacks and others, respectively, P = 0.001), stage III-IV versus stage I-II (77.2 vs. 432.4 mo, P = 0.0001), and in patients treated with surgery+chemotherapy versus surgery (73.5 vs. 227.2 mo) (P = 0.0001). It is noted that the chemotherapy-only group had longer DFI, but these were 2 patients only and the results were censored.

Also, shorter mean DFI was noted in patients with mitotic rate ≥4 (162.3 vs. 432.9 mo, P = 0.005), severe NP (104.2 vs. 181.2 vs. 440.5 mo, P = 0.013), high HER2 expression (89.0 vs. 401.6 mo, P = 0.001), high CD56 expression (156.8 vs. 453.9 mo, P = 0.001), and high SMAD3 expression (220.6 vs. 441.5 mo, P = 0.001) (Fig. 2). On Cox regression analysis, SMAD3 and the need for adjuvant chemo-therapy were the only 2 independent prognostic factors for DFI (P = 0.03 and P = 0.007, respectively) (Table 4).

FIG. 2.

A–H, Kaplan-Meier plots of disease-free interval to first recurrence according to race, stage, treatment, mitotic rate, nuclear polymorphism, and expression of HER2, CD56, SMAD3.

To evaluate the utility of the prognostic factors in early-stage (stage 1) GCTs, a subanalysis was performed (Table 5 and 6 and Fig. 3). Treatment (need for adjuvant chemotherapy), and high expression of CD56 and SMAD3 were found to be significant predictors of recurrence on univariate analysis (Table 5), but only treatment was found to be an independent predictor of recurrence on multivariate analysis (P = 0.01) (Table 7).

TABLE 5.

Univariate subanalysis (early stage: Stage I)

Parameter	Variable	n (%)		P
		Nonrecurrent (n = 79)*	Recurrent (n = 6)
Treatment	Surgery only	62 (95.3)	1 (16.7)	0.0001
	Surgery+chemo	3 (4.6)	5 (83.3)
Mitotic rate	<4/HPF	55 (69.6)	3 (50.0)	0.377
	≥4/HPF	24 (30.4)	3 (50.0)
CD56	Neg+low: ≤80%	44 (61.1)	0 (0.0)	0.012
	High: >80%	28 (38.9)	5 (100.0)
SMAD3	Neg+low: ≤80%	62 (89.9)	1 (20.0)	0.001
	High: >80%	7 (10.1)	4 (80.0)
CD56 and SMAD3	Low-low	22 (78.6)	1 (20.0)	0.007
	Low-high	2 (7.1)	0 (0.0)
	High-high	4 (14.3)	4 (80.0)

^*Data is missing in some cases.

Bold values are statistically significant.

TABLE 6.

Multivariate analysis using logistic regression (forced stepwise) method and cox regression analyses (forced stepwise) regarding disease-free survival (subanalysis: early stage: stage I)

Variable	Odds Ratio	95% CI	P	Hazard Ratio	95% CI	P
Treatment	17.25	1.73–172.02	0.01	10.26	1.69–62.40	0.01
High CD56+high SMAD3	14.87	0.93–237.42	0.05	7.7	0.8–37.9	0.78

Bold values are statistically significant.

CI indicates confidence interval.

FIG. 3.

A–E, Kaplan-Meier plots of disease-free interval to first recurrence according. treatment, mitotic rate, and expression of CD56 and SMAD3: subanalysis (stage I).

TABLE 7.

Clinical characteristics of the 12 cases of GCT recurrence

Case No.	Age	Race	Surgery	Chemotherapy	Stage	LN Status	DOR (mo)	Vital Status	Recurrence Site
1	25	CA	USO (Lt), staging	BEP	II	0/5	24	Alive	Ascitis, rt ovarian pelvic-abdominal mass. Large necrotic left inguinal LN or possibly a direct hernia
2	68	CA	TAHBSO, omentectomy	Depolupron	IIA		90	Alive	Pelvic side wall, ischeorectal fossa
3	60	CA	USO, staging	BEP	IIIA	0/10, 0/1	31	Alive	Pelvis, diaphragm/liver
4	58	CA	TAHBSO	None	IIB	0/15, 0/5	85	Dead	Pelvic, pararectal
5	50	O	BSO	BEP	IC	0/2, 0/8	38	Alive	Pelvic, lung, liver
6	63	AA	TAHBSO	None	IA		30	Dead	Abdomen, pelvis, peritoneum
7	22	AA	USO	BEP	IIIC		26	Dead	Ovarian
8	36	AA	TAHBSO	BEP, 2nd cytored, cytox/Avast	IIIA	0	49	Alive	Pelvis, omentum
9	63	CA	TAHBSO	Vinbl/BP	NA		60	Dead	Ascitis, intermediate iliac LN, pelvis, abdominal, lung, liver
10	57	CA	TAHBSO	Carbo/taxol	IIIC		64	Alive	Diaphr, peritoneum, pelvis, abdominal wall, omentum
11	51	CA	TAHBSO	CT-BEP-Cytox/Av-2nd cytored-Gem-Doxil-phase I	IC		42	Alive	Liver, bowel
12	80	CA	TAHBSO	Yes (NA)	IIIC		240	Dead	Pelvis, peritoneum, omentum

Bold values are statistically significant.

Avast. indicates avastin; BEP, bleomycin, etoposide, platinum; BP, bleomycin, platinum; Carbo, carboplatin; CT, carboplatin, taxol; Cytored, cytoreduction; cytox, cytoxan; DOR, date of recurrence; Gem, gemcitabine; NA, not available; O, other; TAHBSO, total abdominal hysterectomy with bilateral salpingo-oophorectomy; Vinbl, vinblastine.

In KM analysis, a significantly shorter DFI was noted with the surgery+chemotherapy treatment group and with high expression of CD56 and SMAD3 (P = 0.0001, P = 0.008, and P = 0.0001, respectively) (Fig. 2). We further investigated the value of high simultaneous expression of CD56 and SMAD3 in stage I GCT. Although significant on univariate and KM analysis (P = 0.007 and P = 0.04), simultaneous high expression of CD56 and SMAD3 showed a tendency toward significance on multivariate analysis (P = 0.05). Given that treatment is not an inherent characteristic in patients, repeat Cox regression analysis in stage I (excluding treatment) showed that high expression of SMAD3 was significantly associated with increased risk of recurrence (HR = 8.9; 95% CI, 2.34–34.5; P = 0.001). Tables 7 and 8 summarize the clinicopathologic characteristics of the 12 patients who developed a recurrence.

TABLE 8.

Pathologic characteristics of the 12 cases of GCT recurrence

Case No.	Type	Solid/Cyst	Size	Rupture	ME	MR	NP	CD56	SMAD3
1	JGCT	C	16	No	Diffuse	3	Moderate	High	High
2	AGCT	S	3.5		Diffuse	9	Severe	High	Low
3	AGCT	S/C	20	Yes	Diffuse	5	Severe	High	High
4	AGCT	S	16		Micronodular	5	Moderate	High	High
5	AGCT	NA	7	NA	NA	2	NA	NA	NA
6	AGCT	S	0.4		Trabecular	4	Severe	High	High
7	JGCT	S/C	34	No	Diffuse	12	Mild	High	Low
8	AGCT		2.5	Yes	NA	2	NA	NA	NA
9	AGCT	S/C			Diffuse	1	Mild	High	Low
10	AGCT	S/C			Diffuse	9	Moderate	Intermediate	High
11	AGCT	S	8		Diffuse	9	Moderate	High	High
12	AGCT	S	1.5		Trabecular	4	Mild	High	High

Bold values are statistically significant.

AGCT indicates adult granulosa cell tumor; C, cystic; JGCT, juvenile granulosa cell tumor; ME, microscopic exam; MR, mitotic rate; NP, nuclear pleomorphism, S, solid; S/C, solid cystic.

DISCUSSION

GCTs are low-grade malignant neoplasms, in which intraoperative and postoperative management decisions are challenging often times. This stems from the rarity of GCTs and the need for prolonged follow-up periods, which limit the availability of high-quality evidence. Recurrences after therapy have typically occurred within 5 to 10 years (47,48), and up to as long as 37 years (49), which makes the establishment of reliable prognostic indicators or predictors of recurrence a challenge. This may account for the discrepancy noted in the overall 10-year versus the 25-year survival in AGCTs in general and in stage I; 60% to 90% versus 40% to 60% and 85% to 95% versus 69% to 90%, respectively (47,48,50,51). As the majority (60%–90%) of GCTs present with stage I disease (52), the development of reliable predictors of recurrence may help in identifying early-stage GCTs with a higher likelihood for recurrence that may have gone without adjuvant therapy or inadequately followed after surgery. We sought to investigate some of the previously studied clinicopathologic prognostic factors and to explore novel markers of recurrence.

Of the clinical factors investigated, patient’s age and tumor size/rupture remain to be of controversial prognostic value. Although some studies support worse prognosis with older age (1,12), tumor size >10 to >15 cm (53,54), and tumor rupture (16), others as in our study do not (16,18,51,55).

In contrast, stage of disease has been consistently shown to be the most important independent prognostic factor with long-term follow-up studies showing the 10-year survival to be 84% to 95%, 50% to 65%, and 17% to 33% for stage I, II, and III-IV, respectively (3,13).

Although in our study, a significant difference was only noted on univariate and KM analyses for stage but not on multivariate analysis, this may have been explained by the relatively small sample size and the majority of patients being in early stage. In addition, the significant association of chemotherapy administration (reflective of disease stage) and recurrence may support the anticipated effect of stage on prognosis/recurrence.

However, in a subanalysis of stage I, patients receiving adjuvant chemotherapy (stage IC, residual disease) were significantly more likely to have a recurrence and to have a shorter DFI (OR = 17.25; 95% CI, 1.73–172.02; P = 0.01 and 104.3 vs. 58.9 mo, P = 0.0001).

In some studies, postoperative chemotherapy showed a clinical benefit in advanced stage and recurrent disease (56–59), but others failed to show any improvement in progression-free survival (PFS) and overall survival with the addition of chemo-therapy (6,53,60–62). In agreement with our study, Meisel et al (63) reported that patients who did not receive adjuvant chemotherapy had a significantly longer median PFS and a significantly higher 3-year PFS rate than those that did.

The effcacy of pathologic factors (MI, NP) as predictors of recurrence remains controversial partly because of the subjectivity (interobserver variation), use of different cutoffs, and not controlling for stage. Some studies reported a worse prognosis with increased NP (12,47) and MI (different cutoffs >3 to >10) (12,14,16,17,64), whereas others reported no prognostic significance (1,34,51,55,65). In agreement with the findings of Fox and colleagues and Lassus and colleagues, the subanalysis performed in our study showed no predictive value for MI in early-stage (stage I) GCTs. However, the mitotic rate was found to be an independent predictor of recurrence on logistic regression analysis of the whole data with a greater likelihood for recurrence with a mitotic rate ≥4. This may be explained by possible skewness of the data by the higher stage GCTs, and thus highlights the importance of performing subanalysis on early-stage GCTs.

Although high serum levels of and high expression of VEGF were noted in GCT (45), 1 study failed to show a prognostic significance in predicting recurrence/PFS as shown in our study (66). Also, even though GATA4 was recently found to be an independent predictor of recurrence (19), our data failed to show any prognostic utility for GATA4 as well as HER2 (33,35). However, expression of HER2 may be of importance in predicting resistance to treatment with bevacizumab.

CD56 has been recently found to be expressed in primary and recurrent GCTs (39), with a high degree of sensitivity for sex-cord stromal tumors (46). In our study, high expression of CD56 independently predicted a higher likelihood of recurrence and a significantly longer DFI.

The expression of SMAD3 in GCTs has not been studied before. High expression of SMAD3 independently predicted a higher likelihood of recurrence and a significantly longer DFI. In addition, on a subanalysis for early-stage (stage I) GCTs, high expression of CD56 or SMAD3 showed a significant prediction of recurrence on univariate analysis only; however, simultaneous high expression of CD56 and SMAD3 predicted a higher likelihood of recurrence and significantly shorter DFI on multivariate analysis. More importantly, the true utility of SMAD3 in predicting recurrence in early-stage GCTs was illustrated upon exclusion of treatment (which is not an inherent characteristic of patients) from the analysis. High expression of SMAD3 was the only independent predictor of recurrence. This may prove to be a useful tool in counseling patients postoperatively about recurrence risk, closer surveillance, and need for chemotherapy and may prove to be a future therapeutic target.

Ideally, a predictive model for recurrence of early-stage (stage I) GCTs could be developed based on combining clinical and pathologic factors. An example of such model was suggested by Nosov et al. (20), where it was suggested that patients younger than 48 years whose tumor has a higher expression of inhibin and Ki67 are at a higher risk for recurrence and hence would benefit from adjuvant chemotherapy. Another model predicting recurrence-free survival in a retrospective cohort of 127 patients from multiple institutions created a nomogram based on clinical stage, body mass index, tumor diameter, and MI (67).

Appreciable differences are noted among different studies as well as our study with regards to the degree of expression of different pathologic prognostic factors, which are potential therapeutic targets (such as VEGF, HER2). Also potential novel pathologic prognostic factors/therapeutic targets (SMAD3 and other TGF-β pathway targets) are being investigated. In an era of precision medicine, individualizing patient care postoperatively and making a choice of adjuvant chemotherapy based on which of these IHC pathologic factors is highly expressed, patients (determined to be at higher risk for recurrence) can be treated with bevacizumab (45,68), trastuzumab/imatinib (69), peptide aptemers, and ligand traps/antisense oligonucleotides (70) if they have GCTs that have a high expression of VEGF, HER2, SMAD3, and TGF-β receptor, respectively.

Our study is not without limitations; this is a retrospective review with inherent bias but because of the rare nature of GCTs and tendency for late recurrence, this makes accruing patients in a prospective randomized study very diffcult, even from multiple institutions. Also, the percentage of recurrence was relatively low (12/125, 9.6%), compared with higher reported rates of recurrence by other authors.

However, this review is one of the larger series published on GCTs from a single institution and spans over 39 years. Because many studies investigating prognostic factors in AGCTs do not include a pathologic review of the cases, one of the major strengths of our study is the stringent pathology review by the same gynecologic oncology pathologist to confirm the diagnosis. Also performing a sub-analysis to predict recurrence and DFI in early-stage patients was rarely done in studies evaluating GCTs because of small sample size.

To the best of our knowledge, this is the first report of the utility of CD56 and SMAD3 IHC expression in primary GCTs as independent predictors of recurrence. This will need to be validated on a larger scale, possibly through multisite collaboration, to confirm the utility of these prognostic factors.