Fallopian Tube Lesions in Women at High Risk for Ovarian Cancer: A Multicenter Study

In the early 2000s, closer examination of the fallopian tube in BRCA1/2 mutation carriers undergoing risk reduction salpingo-oophorectomy (RRSO) informed our understanding of invasive serous carcinomas of the fallopian tube and led to the classification of serous tubal intraepithelial carcinoma (STIC), a noninvasive lesion believed to be a precursor of high-grade serous carcinoma (HGSC; refs. 1, 2). The current theory is that the STIC cells detach from their residence at the fallopian tube surface and disseminate to the nearby ovaries and the peritoneal soft tissue, where they form masses. This theory explains why most HGSCs are not present at an early stage and the recent observation that women can develop peritoneal cancer post RRSO (3). Morphologically, STICs are characterized by a combination of nuclear enlargement, loss of polarity, increased nuclear atypia, aberrant p53 expression, and increased proliferative activity (4, 5). Serous tubal intraepithelial lesions (STIL) and p53 signatures have also been identified in the fallopian tube. STILs are lesions that just fall short of being classified as a STIC due to limited proliferation, whereas p53 signatures have aberrant p53 immunostaining in at least 12 consecutive secretory nuclei in normal-appearing nonproliferative tubal epithelium (6). The role, if any, of the p53 signatures in the carcinogenic process has not been elucidated.

Demonstrating that STICs are a precursor of HGSCs has been extremely challenging due to the infeasibility of lesion biopsies and close monitoring of the tube. Compelling molecular evidence described by our research group and others, however, supports this notion (7–12). STICs have been identified in the fallopian tube of up to 59% of women with HGSC and in primary peritoneal serous carcinomas (13, 14). Etiologically important genes and proteins associated with HGSC such as TP53 mutations and protein aberrant expression of p53, p21, cyclin E1, Rsf-1, laminin γ1 protein, fatty acid synthase, stathmin1, and p16 have been identified in STICs (7–12). Molecular alterations reflective of DNA damage (expression of γH2AX, a marker of double-stranded DNA breaks and pCHK2), oxidative damage (8-OHdG), and short telomere length have also been reported in both STICs and HGSC (8, 15, 16). By examining for predictors of STICs/STILs and invasive carcinoma in the fallopian tube, we believe we can improve the risk stratification and personalize preventive strategies in women at high-risk for ovarian cancer.

The source population for this study is cancer-free women who are at high risk of developing ovarian/fallopian tube cancer because of carriage of a BRCA1/2 mutation or having at least 2 first- and/or second-degree relatives of the same lineage with ovarian cancer. They also had to have undergone a RRSO at ≥30 years of age in one of four academic centers: Johns Hopkins University (JHU; Baltimore, MD), University of Toronto (Toronto, Canada), Memorial Sloan Kettering Cancer Center (MSKCC; New York, NY), or Yale Cancer Center (New Haven, CT). Furthermore, the fallopian tubes and ovaries had to be processed in their entirety using the SEE–FIM Protocol (17). Women with prior histories of cancer were excluded with the exception of nonmelanoma skin cancer or a diagnosis of breast cancer within the previous 10 years. The Institutional Review Board (IRB) at JHU approved the study and each study site obtained local IRB approval.

The study was divided into three parts as outlined in Figure 1. All parts involve women from the same source population described above. In part 1, 400 cancer-free women were identified from two prospective epidemiologic cohorts at JHU (71 cases; ref. 18) and Toronto (195 cases) and a retrospective cohort at MSKCC (134 cases). Serial sections of each tissue block from both tubes were stained for hematoxylin and eosin (H&E), and p53 ± Ki-67 staining was done at their institution on suspicious lesions for confirmation. Slides were reviewed for occult invasive carcinomas and STICs by at least one gynecologic pathologist at each of the three sites and often by their tumor board. Yale was not included in this part of the study due to logistical issues regarding data linking. In part 2, 105 cancer-free women were identified from sequential cases undergoing RRSO at JHU (37 cases), MSKCC (33 cases), and Yale (35 cases) between June 1, 2011 and May 31, 2013. Cases from Toronto were not included because the institutional pathology review committee would not allow flipping of the tissue blocks. A further restriction was added to maximize detection of STICs/STILs and p53 signatures. The tissue blocks of cases could not have previously been used for research. This meant that only 26 cases from part 1 were eligible for part 2. For every case, as shown in Supplementary Figure S1, the top and bottom of each fimbria block were sampled for STICs/STILs and p53 signatures. In part 3, all 37 cases from JHU were included and underwent extensive staining (top and bottom) of both tubes for STICs/STILs and p53 signatures.

The diagnosis of STIC and STIL was based on a previously validated algorithm (4, 5). Slides were initially stained for H&E and laminin γ1 as shown in Figure 2, instead of H&E and p53, to detect STICs and STILs and then cases positive for laminin γ1 and/or suspicious for a STIC/STIL on H&E were further stained for p53 and Ki-67. Laminin γ1 encodes for LAMC1 and has been shown in STICs to be positive when p53 is positive and in STICs that lacked p53 immunoreactivity due to frame shift, nonsense, or splicing junction mutations of TP53-null mutations (19). For p53 signatures, we applied the same criteria used in a prior study (5). Two gynecologic pathologists reviewed each case (R. Kurman, V. Parkash, I.-M. Shih, R. Soslow, and R. Vang) and were blinded to all clinical information. They were not given cases from their own institution. If the review was discordant, additional reviews were conducted by a third pathologist (R. Soslow) and if necessary, a fourth pathologist (R. Kurman) who were blinded to prior reviews and clinical information. Laminin γ1 expression was scored as either positive or negative, p53 was scored as aberrant if diffuse expression (>75% of the cell) was present in at least 12 epithelial cells (with or without intervening ciliated cells) or there was complete absence of staining or nonabnormal pattern, and Ki-67 was categorized as <10% or ≥10% staining. STICs are expected to be laminin γ1 positive, p53 diffuse, or completely negative and Ki-67 ≥ 10%. Quality controls were included in every batch.

IHC for parts 2 and 3 was performed at Johns Hopkins Immunopathology Laboratory to assess the expression levels of p53 (clone Bp53–11, catalog no. 760–2542; Ventana Medical Systems) and Ki-67 (clone 30–9, catalog no. 760–4286; Ventana Medical Systems) in formalin-fixed, paraffin-embedded tissue sections. All sections were immunostained automatically using either Ventana Benchmark Ultra or XT (Ventana Medical Systems, Inc.). Immunoreactivity was detected by iView (catalog no.760-091, Ventana Medical Systems, Inc.). IHC to assess laminin γ1 polyclonal antibody (catalog no. HPA001909, Sigma-Aldrich) was performed manually in the Shih laboratory according to manufacturer's instructions (diluted 1:400) and immunoreactivity was detected by the Dako Liquid DAB+ Substrate Chromogen System (catalog no. K3468, DAKO North America, Inc., Agilent Technologies).

A template and code book was developed on the basis of review of participant questionnaires from JHU and Toronto and provided to all sites. Deidentified data were then sent to JHU where it was reviewed and checked prior to harmonization to create uniform definition across sites. Variables with significant missing information were not included. Exposures examined included race, Ashkenazi Jewish ancestry, BRCA1/2 status, family history of breast and/or ovarian cancer, parity, breast feeding and mean duration, oral contraceptive use and mean duration, hormone therapy use, tubal ligation, hysterectomy, endometriosis, mean age at menarche in years, personal history of breast cancer, fibroids, smoking, talcum powder, and body mass index (BMI). Germline BRCA1/2 mutation status was confirmed from written reports and all surgical procedures were confirmed from pathology reports.

The proportions and frequencies of the three lesions and p53 signatures were calculated overall, based on location (fimbria vs. nonfimbriated tube), BRCA1/2 status, and age. Risk factor associations and their P values were calculated using descriptive statistics. The categories used were based on prior literature (20, 21) and distributions of exposure in women without lesions. Direct age-standardized estimates were calculated to compare characteristics of women with and without lesions (https://www.hsph.harvard.edu/donna-spiegelman/software/table1-for-windows). A number of sensitivity analyses were undertaken with details provided in the results section. Percent agreement and Kappa were calculated for the final diagnosis reported by the first and second pathologist. All tests were considered statistically significant at P < 0.05 and performed using STATA version 14.

Table 1 provides a summary of the tubal lesions identified in 400 high-risk women who underwent RRSO overall and by site. Of these women, 92% had an identified pathogenic mutation in BRCA1/2. The combined prevalence of invasive cancers/STICs/STILs was 6.3% (1.5% for invasive serous carcinomas, 3.5% for STICs, and 1.3% for STILs). The prevalence among BRCA1 mutation carriers was 3% and among BRCA2 mutation carriers was 3.3%. No invasive serous cancers, STICs, or STILs were identified in the 19 high-risk women who tested negative for a pathogenic mutation in BRCA1/2. There was a significant difference in the mean age across the sites (P = 0.02). The mean age for women from JHU was 50 years (SD 8.8), MSKCC was 49 years (SD 9.2), and Toronto was 47 years (SD 7.7).

Table 2 compares selected age-standardized characteristics of women diagnosed with carcinoma (invasive cancer/STICs/STILs; N = 23) to those without carcinoma (N = 377). Additional characteristics are reported in Supplementary Table S2. Older age was the only significant predictor of having a carcinoma (51.9 years vs. 48.1 years; P = 0.036), particularly among women with invasive carcinoma (P = 0.009). Other risk/protective factors for HGSC including oral contraceptive (OC) use, parity, and history of hormone therapy were not found to be significantly different between women with and without invasive carcinoma/STICs/STILs (20). On average, women with a carcinoma reported a lower mean duration of breast feeding (7.4 months) when compared with women with no lesion (14.5 months), but this difference was not statistically significant. These results did not change in sensitivity analyses limited to the two prospective studies (JHU and Toronto), BRCA1/2 mutation carriers, excluding women with tubal ligation or excluding women with breast cancer within the last 10 years. Risk factors associated with other epithelial ovarian cancer subtypes were also not associated with a diagnosis of carcinoma compared with no lesion.

Table 3 provides individual-level data on the prevalence of the fallopian tube lesions and p53 signatures for 105 women extensively sampled in parts 2 and 3. In part 2, the mean number of fimbriae blocks were 1.7 for JHU, 2.3 for MSKCC, and 2.6 for Yale. A review by multiple pathologists increased the number of invasive cancers/STICs/STILs detected by 50%. Additional tissue sampling also more than doubled the number of lesions. In total, 10 carcinomas were detected in the tubes of 6 women. In only half of the cases were the STICs/STILs limited to the fimbria and in 50% of cases, multiple lesions were identified. In 2 of the 3 individuals where a STIC was identified in the top and bottom of the same tissue block, we stained and reviewed every 10th and 11th slide for H&E and laminin γ1 for 90% of the block to determine whether they were independent lesions. On the basis of our evaluation, the two STICs appear to be independent. Unlike STICs/STILs and invasive carcinomas, p53 signatures were identified across the entire tube in 10 women. Six p53 signatures were identified in the fimbria (with 5 of 6 in BRCA1/2 mutation carriers), and 8 p53 signatures in the nonfimbriated tube.

Clinical characteristics of the 105 women are described in Supplementary Table S3. The mean age of women with p53 signatures in the fimbria was significantly older when compared with the women with signatures in the remainder of the tube (54 years vs. 45 years, P = 0.03). Of note, invasive carcinoma/STICs/STILs were only detected in BRCA1/2 mutation carriers, whereas p53 signatures were detected irrespective of mutation status. Figure 3 provides a summary of the prevalence of unique types of lesions per individual for each part of the study. The combined prevalence of invasive cancer/STICs/STILs for parts 1 and 2 was 6.3% and 11.9%, respectively, after extensive sampling and review by multiple pathologists and 27% for unique p53 signatures from part 2.

The percent agreement for the diagnosis of STIC versus non-STIC by the two initial pathologists was 99.3% and the κ = 0.61 based on laminin γ1 expression and H&E. The percent agreement for the diagnosis of p53 signature based on H&E, laminin γ1, p53, and Ki-67 was 96.9% and the κ = 0.19. Every case of p53 signature required at least a third pathology review and sometimes a fourth to reach consensus.

To our knowledge, this is the first multicenter study to extensively examine for risk and/or protective factors associated with STICs/STILs in women at high-risk for ovarian cancer. This is also the first study to conduct sampling of both ends of tumor blocks across the entire tube and have all slides reviewed by multiple pathologists to further characterize the prevalence and location of STICs/STILs and p53 signatures. Comprehensive sampling and screening of the fallopian tube in 400 RRSO cases based on H&E and p53 expression identified invasive serous cancers/STICs/STILs in 6.3% of cases. Older age was associated with having a carcinoma. No other risk/protective factors were significantly associated with invasive carcinoma/STICs/STILs. The observation that mean breast feeding was on average 7 months lower in cases with a carcinoma than no lesion deserves further exploration. There has been only one single institution study that has examined for risk factors associated with STICs (22). Vicus and colleagues observed in 173 BRCA1/2 mutation carriers that older age, tubal ligation, and a BMI > 25 kg/m² were positively associated with STICs and that there was an inverse relationship with increased duration of oral contraceptive use. Some of these women are part of our study. In BRCA1/2 mutation carriers, breast feeding and duration of oral contraceptive use were associated with between 25% and 40% reduction in ovarian cancer incidence. It is plausible that the association between breast feeding, oral contraceptives, and STICs/STILs is not as strong (21) or that known risk/protective factors (21) are associated with progression of STICs to HGSC and not the incidence of STICs. Another possibility is that the development of STICs are related to random mutations occurring by chance that target cancer drivers, including TP53 (23).

Additional sampling of the fallopian tube generated some interesting findings. It doubled the detection of STICs/STILs and p53 signatures. The observation that STICs/STILs occur at equal prevalence in the nonfimbriated part of the tube raises the question of whether all STICs have the same propensity to translocate to the ovary or peritoneum. Similarly, the observation of multiple STICs in 50% of the cases could be indicative of greater risk for STIC migration to occur to the ovary or peritoneum and therefore, the development of HGSC when compared with a woman with a single lesion. The role of p53 signatures in the development of HGSC is less clear. P53 signatures occurred at a greater prevalence than STICs (27% vs. 11.9%) with only 40% occurring in the fimbria. Furthermore, signatures in the fimbria compared with the remaining tube were more prevalent among older age women suggesting that all p53 signatures may not necessarily lead to HGSC. The increase in the number of detected lesions and p53 signatures after review by multiple pathologists and the low agreement across pathologists particularly for p53 signatures suggests that further work is needed for the development of standard reproducible criteria. It is likely that the imbalance caused by few positive lesions and significant discordance between pathologists is only part of the explanation for the low agreement (24).

Our prevalence of STICs/STILs based on comprehensive processing was 4.8%, which is within the range of 2.0% to 6.2% reported by the few studies of more than 200 high-risk women (3, 25–27). Shaw and colleagues’ is the only study to report on STICs/STILs and p53 signatures in 176 high-risk women and the tissue processing was similar to what we did for the initial 400 cases. The prevalence of STICs/STILs was much higher than in our study at 10% whereas the prevalence of p53 signatures was lower at 11% (28). Our estimates of invasive cancer were 1.5%. This estimate is lower than the GOG-0199, a screening study that reported a prevalence of 3.4% after review of 966 RRSO cases but similar to three large (N > 200) observational studies that reported a prevalence ranging between 0.9% and 2.0% (25–27, 29). None of the studies did more extensive sampling. The few pilot studies that have examined the impact of deeper sectioning of tissue blocks on the detection of STICs/STILs from the fimbria in cases with cancer (13, 30, 31) have shown an increase in detection of tubal lesions. None of them have performed extensive sectioning across the entire tube in cancer-free individuals.

Strengths of this study include the multicenter design, standardized tissue processing across sites, review of slides with and without lesions by multiple gynecologic pathologists, and the ability to link pathology tissue to detailed epidemiologic data. In this study, we had adequate power to detect moderate to large differences in exposures between women with and without a lesion. The widespread implementation of standard pathologic criteria across hospitals in conjunction with the prospective collection of epidemiologic data may help elucidate novel risk/protective factors associated with fallopian tube lesions and better understand their natural history.

R. Soslow has received speakers bureau honoraria from Ebix/Oakstone. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Visvanathan, P. Shaw, T.-L. Wang, V. Parkash, R. Soslow, R. Kurman, I.-M. Shih

Development of methodology: K. Visvanathan, P. Shaw, T.-L. Wang, R. Soslow, R. Kurman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Visvanathan, P. Shaw, B.J. May, A. Bahadirli-Talbott, S. Narod, V. Parkash, D.A. Levine, R. Soslow, R. Kurman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Visvanathan, B.J. May, A. Kaushiva, H. Risch, S. Narod, R. Vang, R. Soslow, R. Kurman, I.-M. Shih

Writing, review, and/or revision of the manuscript: K. Visvanathan, P. Shaw, B.J. May, A. Bahadirli-Talbott, A. Kaushiva, H. Risch, S. Narod, V. Parkash, R. Vang, D.A. Levine, R. Soslow, I.-M. Shih

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Visvanathan, B.J. May, A. Bahadirli-Talbott, A. Kaushiva, V. Parkash, D.A. Levine, I.-M. Shih

Study supervision: K. Visvanathan, D.A. Levine

The authors thank Sophia HL George, Ramlogan Sowamber, and Noor Salman for the retrieval and processing of Toronto Cases (University Health Network, Toronto, Canada); and staff and participants of the BOSS cohort at Johns Hopkins.This work was supported by DOD Ovarian Consortium W81XWH-11-2-0230; Breast Cancer Research Foundation (to K. Visvanathan); Richard W. TeLinde Research Fund (to I.-M. Shih); NCI P30CA06973 (to R. Soslow); and NCI P30 CA008748 (to K. Visvanathan, I.-M. Shih, R. Kurman, and T.-L. Wang).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.