Abstract
Objective.
To determine changes in the characteristics of low-grade serous ovarian cancer (LGSOC) and serous borderline ovarian tumor (serous-BOT) in a time-specific manner.
Methods.
We conducted a population-based retrospective study examining the Surveillance, Epidemiology, and End Results Program from 1988 to 2000. Trends, demographics, and outcomes of 775 women with well-differentiated serous ovarian cancer, used as a surrogate for LGSOC, were compared to 3937 women with serous-BOT.
Results.
In the multivariable analysis, women with LGSOC were more likely to be older, have stage II-IV disease, and have undergone hysterectomy at surgery, but less likely to be a Western U.S. resident compared to those with serous-BOT (all, adjusted-P < 0.05). During the study period, the number of LGSOCs decreased by 25.9%, particularly stage I disease (37.6% relative decrease) compared to stage II-IV disease (21.1% relative decrease) (all, P < 0.05). With a median follow-up of 16.9 years, there was a decreasing trend in the 15-year overall survival rates among LGSOC (28.7% relative decrease, P = 0.056) but not in serous-BOT (2.5% relative increase, P = 0.416) as a whole cohort. The magnitude of hazard risk from all-cause death for women with LGSOC compared to those with serous-BOT increased by 68.9% from 1988 to 2000 (P < 0.001). LGSOC remained an independent prognostic factor for decreased overall survival compared to serous-BOT (adjusted-P < 0.05).
Conclusion.
Our study suggests that the decreasing number and survival of LGSOC over time may be due to a diagnosis-shift from LGSOC to serous-BOT. Given the distinct characteristics and outcomes of LGSOC compared to serous-BOT, our study endorses the importance of making the correct diagnosis upfront. Whether this diagnostic-shift supports a hypothesis that serous-BOT is a precursor lesion of LGSOC merits further investigation.
Keywords: Ovarian cancer, Serous borderline ovarian tumor, Low malignant potential, Low grade serous ovarian cancer, Trend
1. Introduction
In 2019, ovarian cancer remains the most deadly gynecologic malignancy in the United States [1]. Ovarian cancer is not a single disease entity, rather it is comprised of various histologic subtypes based upon the cell of origin, of which serous histology is the most common. Serous ovarian cancers are generally classified into either high-grade or low-grade tumors, with distinct clinical and molecular characteristics and outcomes between the two [2-4]. Low-grade serous ovarian cancer (LGSOC) is a relatively new distinction that was proposed in the mid-1980s but not formally defined until 2004 [2,5]. It is considered a rare tumor and has been understudied due to its low incidence.
A recent population-based analysis of the tumor registry in the United States has shown a gradual and steady decrease in the number of LGSOC among serous ovarian cancers starting around the mid-1980s [6]. Serous borderline ovarian tumor (serous-BOT) represent non-invasive type serous ovarian tumors described initially in 1929 and classified in the World Health Organization (WHO) in 1971 [7]. In contrast to the decrease in LGSOC, there has been a concurrent increase in the number of serous-BOT cases in the past few decades [8,9].
Several theories have been proposed to explain this inverse association of the trends among the two diseases (decreasing LGSOC and increasing serous-BOT). One possibility is that there has been a diagnosis-shift from LGSOC to serous-BOT over time. As prior studies demonstrating these populational-level trends were performed in separate continents, it remains unknown if such diagnosis-shift between the two diseases occurred within the same population [6,8,9]. The objective of the current study was to examine changes in the characteristics of LGSOC and serous-BOT in a time-specific manner in the United States.
2. Materials and methods
2.1. Data source
This is a population-based retrospective observational study examining data from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) program [10]. The SEER program is considered the largest population-based tumor registry in the United States, and was launched in 1973. This database has a more than a four-decade history of operation, and currently covers approximately 35% of the U.S. population. The SEER program represents a powerful and suitable tool for the study of rare tumors, as it provides population characteristics and long-term outcomes. The study was exempted by the University of Southern California Institutional Review Board due to the use of de-identified and publicly available data in the program.
2.2. Study eligibility
Women with serous-BOT and those with LGSOC who were diagnosed from 1988 to 2000 were eligible for the analysis. This study period was chosen per prior publications [11,12]. Histology types other than serous-BOT and LGSOC, and patients diagnosed outside of the study period were excluded from the study.
2.3. Data extraction
SEER*Stat 8.3.2. (IMS Inc., Calverton, MD, USA) was used for the data abstraction. The International Classification of Disease for Oncology third edition site/histology validation list and the WHO histological classification were used for histologic coding as described previously [11,13-15]. The datasets for serous-BOT and LGSOC were constructed separately from the SEER program, and then merged to one master data set for the analysis.
2.4. Clinical variables
Information abstracted from the SEER program was patient demographics, tumor characteristics, treatment type, and survival outcome. Patient demographics included age at diagnosis, calendar year at diagnosis (trisected as 1988–1991, 1992–1996, and 1997–2000), race/ethnicity (white, black, Hispanic, Asian, and others), marital status (single, married, and others), and registered area (West, Central, and East). Tumor characteristics included histology type (serous-BOT versus LGSOC), cancer stage (I, II, III, and IV) and tumor size (≤4,4.1–6, 6.1–8, 8.1–10 and ≥10 cm).
Treatment types included adnexectomy (yes versus no), hysterectomy (yes versus no), and pelvic lymphadenectomy (yes versus no) with sampled lymph node number among staged cases. For survival outcome, follow-up time after the diagnosis and vital status (death versus alive) were collected from the database. Among those deceased, cause of death was examined.
2.5. Study definition
Cutoff values and groupings of variables were based on prior studies [12,15]. Cancer staging was based on the American Joint Committee on Cancer 3rd staging classification schema that was performed between 1988 and 2003 in the SEER database [16]. The extent of pelvic lymphadenectomy was based on the coding for the Regional Nodes that was introduced in 1988 in the SEER database.
In this study, LGSOC was defined as well-differentiated serous ovarian cancer per the prior study [6]. This is based on the rationale that historical well-differentiated serous ovarian cancer correlates well with LGSOC in the 2-tier classification per the MD-Anderson Cancer Center system [3]. Cause-specific survival (CSS) was defined as the time interval between the diagnosis and the date of death due to corresponding malignancy in each cohort (serous-BOT or LGSOC). Overall survival (OS) was defined as the time interval between the diagnosis and the death from any cause. Patients without survival event were censored at the last follow-up visit.
2.6. Study aim
The primary objective of the analysis was to examine temporal trends and differences in characteristics of serous-BOT and LGSOC over time. The secondary aim was to examine the differences in survival outcomes between the two diseases in a time-specific manner.
2.7. Statistical considerations
To identify the independent clinico-pathological factors distinguishing the two diseases (serous-BOT versus LGSOC), a binary logistic regression model was fitted entering patient demographics, tumor characteristics, and treatment types, and the magnitude of statistical significance was expressed with adjusted-odds ratio (OR) and 95% confidence interval (CI). The Hosmer-Lemeshow test was used to assess the goodness of fit in the model, and a P-value of >0.05 was considered a good-fit model.
Changes in temporal trends were assessed with the National Cancer Institute's Joinpoint Regression Program (version 4.4.0.0) [17]. Time point data was examined annually to identify temporal changes, and a linear segmented regression test was used for the analysis. Log-transformation was performed to determine the annual percentage change (APC) and 95%CI in each segment.
Propensity score matching was used to adjust for baseline differences between the two tumor groups [18]. Patient demographics, tumor characteristics, and treatment types were entered into the model, and the propensity score was computed by fitting a multivariable logistic regression model. An automated algorithm was utilized for one-to-one propensity score matching between the two tumor groups. The optimal caliper width for estimating differences was equal to 0.2 of the standard deviation for the logit of the propensity score distribution, resulting in a propensity score difference cutoff of 0.035 in this study [19]. The standardized difference (SD) was assessed to evaluate the effect size in post-matching model between the two groups, and a SD ≤0.10 was considered a good balance [20].
Survival curves were constructed with the Kaplan-Meier method, and differences in curves were assessed with a log-rank test. A Cox proportional hazard regression model was fitted to examine independent association of tumor types (serous-BOT versus LGSOC) and survival, expressed with hazard ratio [HR] and 95%CI [21]. The association between tumor types and survival was adjusted by patient demographics, tumor characteristics, and treatment types. This robust estimator calculation is based on the rationale that this approach is beneficial to correct unspecified unbalanced variables for outcomes of interest in the outcome model, accounting for the possible residual [22].
A series of sensitivity analyses were undertaken to assess the robustness of the analysis. In specific, interaction-term analysis was performed per patient age (<50 versus ≥50) and stage (I versus II-IV). The time-specific change in HR for survival was examined over time comparing LGSOC versus serous-BOT in each calendar year. We hypothesized that a diagnostic-shift of serous-BOT that was classified as LGSOC in the older period to serous-BOT over time would make the difference in HR larger for survival. This is based on the rationale that survival for non-invasive ovarian tumor is distinctively superior to invasive ovarian tumor [12], and exclusion of misclassified serous-BOT from the LGSOC group will make the survival difference larger.
Multicollinearity among the covariates was assessed with the variance inflation factor, defined as a value of ≥2.5 in this study [23]. All statistical analyses were based upon a two-tailed hypothesis, and a P-value of <0.05 was considered statistically significant. Statistical Package for Social Sciences (SPSS, version 24.0, IBM Corp, Armonk, NY, USA) was used for the analysis. The STROBE guidelines were utilized to outline the results of the observational study [24].
3. Results
3.1. Study cohort
Among 6379 cases of BOT in the initial search in the SEER program, there were 3937 cases of serous-BOT identified during the study period. Among 133,481 cases of ovarian cancer in the initial search, there were 10,817 cases of serous ovarian cancer with known tumor differentiation during the study period. Of those, 775 cases were well-differentiated tumor and thus comprised the LGSOC cohort. Altogether, 4712 cases were examined for analysis.
3.2. Cohort-level diagnosis-shift
Temporal trends of the proportion of LGSOC among the study cohort were examined per year (Fig. 1A). During the study period, the number of LGSOC cases decreased by 25.9% from 23.2% in 1988 to 17.2% in 2000 (APC −3.2, 95%CI −5.6 to −0.8, P = 0.015). When stratified by cancer stage (Fig. 1B), the decrease in number of LGSOC was larger in stage I disease compared to stage II-IV disease (37.6% versus 21.1% relative decrease; both, P < 0.05). More specifically, among stage I disease cases, the number of LGSOC was 15.1% in 1988 which decreased to 9.3% in 2000 (APC −4.7, 95%CI −8.4 to −0.8, P = 0.023). Among cases of stage II-IV disease, the number of LGSOC was 46.2% in 1988 which decreased to 36.4% in 2000 (APC −1.9, 95%CI −3.8 to −0.0, P = 0.049).
Fig. 1.
Diagnosis-shift from sBOT to LGSOC between 1988 and 2000. Proportion of LGSOC among the study population (LGSOC and sBOT) per year of diagnosis between 1988 and 2000 is shown for (A) all stage and (B) stage I and II diseases. Line indicates modeled value, and circles represent observed value with 95% confidence interval.
3.3. Patient-level characteristics
Patient demographics are shown in Table 1. On multivariable analysis, women with LGSOC were more likely to be older (adjusted-OR per age 1.03, 95%CI 1.03–1.04), have stage II-IV disease (adjusted-ORs for stages II, III, and IV, 3.50, 6.11, and 9.22, respectively), and have undergone hysterectomy at the time of surgery (adjusted-OR 1.68, 95%CI 1.32–2.13) but less likely to be Western U.S. residents (adjusted-OR 0.76, 95%CI 0.60–0.95) compared to those with serous-BOT (all, adjusted-P < 0.05). A recent year of diagnosis was independently associated with a lower likelihood of a LGSOC diagnosis (adjusted-OR in middle and last third time period, 0.30 and 0.31, respectively; both, P < 0.001).
Table 1.
Clinico-pathological characteristics between serous-BOT and LGSOC (pre-matching).
| Characteristic | Serous-BOT | LGSOC | P-value | Adjusted-OR (95%CI) | P-value† |
|---|---|---|---|---|---|
| Number | n = 3937 | n = 775 | |||
| Age (cont) | 48.0 (±16.5) | 55.9 (±16.2) | <0.001 | 1.03 (1.03–1.04) | <0.001 |
| Year | <0.001 | <0.001* | |||
| 1988–1991 | 347 (8.8%) | 178 (23.0%) | 1 | ||
| 1992–1996 | 1889 (48.0%) | 305 (39.4%) | 0.30 (0.23–0.39) | <0.001 | |
| 1997–2000 | 1701 (43.2%) | 292 (37.7%) | 0.31 (0.24–0.41) | <0.001 | |
| Race/ethnicity | <0.001 | 0.032* | |||
| White | 2848 (72.3%) | 605 (78.1%) | 1 | ||
| Black | 318 (8.1%) | 54 (7.0%) | 0.96 (0.68–1.36) | 0.835 | |
| Hispanic | 515 (13.1%) | 80 (10.3%) | 1.29 (0.97–1.74) | 0.085 | |
| Asian | 171 (4.3%) | 33 (4.3%) | 1.22 (0.79–1.89) | 0.370 | |
| Others | 85 (2.2%) | 3 (0.4%) | 0.20 (0.06–0.69) | 0.011 | |
| Registry area | <0.001 | <0.001* | |||
| West | 2514 (63.9%) | 435 (56.1%) | 0.76 (0.60–0.95) | 0.017 | |
| Central | 854 (21.7%) | 179 (23.1%) | 1 | ||
| East | 569 (14.5%) | 161 (20.8%) | 1.29 (0.98–1.70) | 0.068 | |
| Marital status | <0.001 | 0.055* | |||
| Single | 880 (22.4%) | 115 (14.8%) | 1 | ||
| Married | 2145 (54.5%) | 437 (56.4%) | 1.24 (0.96–1.60) | 0.098 | |
| Others | 738 (18.7%) | 208 (26.8%) | 1.16 (0.85–1.57) | 0.352 | |
| Unknown | 174(4.4%) | 15 (1.9%) | 0.62 (0.33–1.14) | 0.122 | |
| Hysterectomy | <0.001 | <0.001* | |||
| No | 1380 (35.1%) | 141 (18.2%) | 1 | ||
| Yes | 2249 (57.1%) | 435 (56.1%) | 1.68 (1.32–2.13) | <0.001 | |
| Unknown | 308 (7.8%) | 199 (25.7%) | 3.01 (2.20–4.11) | <0.001 | |
| Adnexectomy | <0.001 | 0.002* | |||
| Yes | 3865 (98.2%) | 739 (95.4%) | 1 | ||
| No | 52 (1.3%) | 32 (4.1%) | 2.75 (1.50–5.03) | 0.001 | |
| NOS | 20 (0.5%) | 4 (0.5%) | 0.46 (0.13–1.67) | 0.235 | |
| Lymphadenectomy | 0.001 | 0.343 | |||
| No | 2579 (65.5%) | 451 (58.2%) | 1 | ||
| Yes | 1331 (33.8%) | 318 (41.0%) | 1.15 (0.95–1.39) | 0.144 | |
| Unknown | 27 (0.7%) | 6 (0.8%) | 1.01 (0.33–3.09) | 0.991 | |
| Tumor size (cm) | <0.001 | <0.001* | |||
| ≤4.0 | 493 (12.5%) | 85 (11.0%) | 1 | ||
| 4.1–6.0 | 171 (4.3%) | 54 (7.0%) | 1.43 (0.92–2.22) | 0.111 | |
| 6.1–8.0 | 198 (5.0%) | 49 (6.3%) | 0.95 (0.61–1.50) | 0.836 | |
| 8.1–10 | 149 (3.8%) | 43 (5.5%) | 1.18 (0.74–1.88) | 0.478 | |
| >10 | 410 (10.4%) | 104 (13.4%) | 1.01 (0.70–1.44) | 0.964 | |
| Unknown | 2516 (63.9%) | 440 (56.8%) | 0.72 (0.54–0.95) | 0.021 | |
| Cancer stage | <0.001 | <0.001* | |||
| I | 3012 (76.5%) | 277 (35.7%) | 1 | ||
| II | 287 (7.3%) | 84 (10.8%) | 3.50 (2.61–4.69) | <0.001 | |
| III | 468 (11.9%) | 278 (35.9%) | 6.11 (4.87–7.66) | <0.001 | |
| IV | 100 (2.5%) | 123 (15.9%) | 9.22 (6.61–12.86) | <0.001 | |
| Unknown | 70 (1.8%) | 13 (1.7%) | 2.27 (1.13–4.53) | 0.021 |
A binary logistic regression model for multivariable analysis. Significant P-values are emboldened.
P-value for multivariable model.
P-value for interaction. Hosmer-Lemeshow test, P = 0.187 indicates a good-fit-model. Abbreviations: serous-BOT, serous borderline ovarian tumor; LGSOC, low-grade serous ovarian cancer; OR, odds ratio; CI, confidence interval; and NOS, not otherwise specified.
When examined by chronologic sequence (Supplemental Tables S1-S2), patient age was unchanged in the two tumor groups during the study period (both, P > 0.05). For surgical procedures, lymphadenectomy rates increased in both groups, and hysterectomy rates decreased in the serous-BOT group (all, P < 0.05). In the LGSOC group, the number of stage I disease slightly declined from 39.3% to 34.2% over time (13.0% relative decrease, P < 0.001). In contrast, serous-BOT tumors became less likely to be stage IV disease from 4.6% to 1.6% over time (65.2% relative decrease, P = 0.002).
3.4. Cohort-level survival
The median follow-up was 16.9 (IQR 14.0–20.3) years, and there were 520 women who died of disease and 1491 women who died from any cause. Temporal trends of the 15-year OS rates are shown in Supplemental Fig. S1. There was a decreasing trend in 15-year OS rates among those with LGSOC from 53.3% in 1988 to 38.0% in 2000 (28.7% relative decrease) although it did not reach statistical significance (P = 0.056). The 15-year OS rate remained unchanged in the serous-BOT cohort: 76.7% in 1988 and 78.7% in 2000 (2.5% relative increase, P = 0.416).
3.5. Patient-level survival
Results of propensity score matching are shown in Table 2. All the baseline characteristics are well balanced after the matching (all, SD ≤0.10). In the univariable model, the 15-year CSS rates were 59.4% for LGSOC and 94.4% for serous-BOT (P < 0.001; Fig. 2A), and the 15-year OS rates were 43.7% for LGSOC and 77.9% for serous-BOT (P < 0.001; Fig. 2B). On multivariable model controlling for patient demographics, tumor characteristics, and treatment type (Table 3), LGSOC remained an independent prognostic factor for decreased CSS (adjusted-HR 3.15, 95%CI 2.45–4.04) and OS (adjusted-HR 1.74, 95%CI 1.48–2.04) (both, P < 0.05). This association was also seen in young and old women as well as those with stage II-IV disease (Table 3).
Table 2.
Clinico-pathological characteristics between serous-BOT and LGSOC (post-matching).
| Characteristic | Serous-BOT | LGSOC | SD (pre) | SD (post) |
|---|---|---|---|---|
| Number | n = 669 | n = 669 | ||
| Age (cont) | 54.3 (±16.8) | 54.4 (±16.0) | 0.480 | 0.006 |
| Year | 0.300 | 0.012 | ||
| 1988–1991 | 127 (19.0%) | 131 (19.6%) | ||
| 1992–1996 | 288 (43.0%) | 274 (41.0%) | ||
| 1997–2000 | 254 (38.0%) | 264 (39.5%) | ||
| Race/ethnicity | 0.141 | <0.001 | ||
| White | 512 (76.5%) | 512 (76.5%) | ||
| Black | 49 (7.3%) | 48 (7.2%) | ||
| Hispanic | 75 (11.2%) | 74 (11.1%) | ||
| Asian | 27 (4.0%) | 32 (4.8%) | ||
| Others | 6 (0.9%) | 3 (0.4%) | ||
| Registry area | 0.188 | 0.022 | ||
| West | 399 (59.6%) | 383 (57.2%) | ||
| Central | 132 (19.7%) | 152 (22.7%) | ||
| East | 138 (20.6%) | 134 (20.0%) | ||
| Marital status | 0.141 | 0.077 | ||
| Single | 126 (18.8%) | 104 (15.5%) | ||
| Married | 368 (55.0%) | 377 (56.4%) | ||
| Others | 162 (24.2%) | 174 (26.0%) | ||
| Unknown | 13 (1.9%) | 14 (2.1%) | ||
| Hysterectomy | 0.573 | 0.050 | ||
| No | 124 (18.5%) | 124 (18.5%) | ||
| Yes | 389 (58.1%) | 410 (61.3%) | ||
| Unknown | 156 (23.3%) | 135 (20.2%) | ||
| Adnexectomy | 0.146 | <0.001 | ||
| Yes | 649 (97.0%) | 648 (96.9%) | ||
| No | 16 (2.4%) | 18 (2.7%) | ||
| NOS | 4 (0.6%) | 3 (0.4%) | ||
| Lymphadenectomy | 0.150 | 0.035 | ||
| No | 399 (59.6%) | 385 (57.5%) | ||
| Yes | 263 (39.3%) | 279 (41.7%) | ||
| Unknown | 7 (1.0%) | 5 (0.7%) | ||
| Tumor size (cm) | 0.073 | 0.042 | ||
| ≤4.0 | 66 (9.9%) | 77 (11.5%) | ||
| 4.1–6.0 | 47 (7.0%) | 42 (6.3%) | ||
| 6.1–8.0 | 30 (4.5%) | 41 (6.1%) | ||
| 8.1–10 | 38 (5.7%) | 34 (5.1%) | ||
| >10 | 98 (14.6%) | 88 (13.2%) | ||
| Unknown | 390 (58.3%) | 387 (57.8%) | ||
| Cancer stage | 0.942 | 0.026 | ||
| I | 277 (41.4%) | 277 (41.4%) | ||
| II | 78 (11.7%) | 81 (12.1%) | ||
| III | 219 (32.7%) | 227 (33.9%) | ||
| IV | 76 (11.4%) | 71 (10.6%) | ||
| Unknown | 19 (2.8%) | 13 (1.9%) |
SD ≤0.10 indicates good-balance between the two groups. Abbreviations: SD (pre), standardized difference in the pre-matching model; SD (post), standardized difference in the post-matching model; NOS, not otherwise specified; serous-BOT, serous borderline ovarian tumor; and LGSOC, low-grade serous ovarian cancer.
Fig. 2.
Survival curves based on tumor types (post-matching). Log-rank test for P-values. (A) Cause-specific survival and (B) overall survival based on tumor types. Abbreviations: sBOT, serous borderline ovarian tumor; and LGSOC, low-grade serous ovarian cancer.
Table 3.
Hazard ratio for survival outcomes (propensity score matching).
| Outcome | Characteristic | Whole cohort | Age < 50 | Age ≥ 50 | Stage I | Stage II-IV |
|---|---|---|---|---|---|---|
| CSS | Tumor types | |||||
| Serous-BOT | 1 | 1 | 1 | 1 | 1 | |
| LGSOC | 3.15 (2.45–4.04) | 4.66 (3.01–7.21) | 2.69 (1.97–3.68) | 3.27 (1.61–6.62) | 3.23 (2.45–4.26) | |
| OS | Tumor types | |||||
| Serous-BOT | 1 | 1 | 1 | 1 | 1 | |
| LGSOC | 1.74(1.48–2.04) | 2.80 (1.99–3.954) | 1.45 (1.22–1.74) | 1.10 (0.85–1.426) | 2.09 (1.70–2.58) |
A Cox proportional hazard regression model for hazard ratio estimates with 95% confidence interval. Abbreviations: serous-BOT, serous borderline ovarian tumor; LGSOC, low-grade serous ovarian cancer; CSS, cause-specific survival; and OS, overall survival.
Differences in survival between LGSOC and serous-BOT were examined every year (Fig. 3 and Supplemental Fig. S2): HR for LGSOC versus serous-BOT was assessed as it represents the survival difference between the two diseases, and the larger value indicates the larger survival difference. Between 1988 and 2000, the magnitude of HR for OS for women with LGSOC compared to those with serous-BOT increased by 68.9% as the whole cohort level (P < 0.001); similarly, HR for CSS has increased significantly during the same time (P < 0.001). On propensity score matching models of the two disease groups in each study period (1988–1991,1992–1996, and 1997–2000), similar trends were observed and the survival difference between the two diagnoses became significantly enlarged over time (Supplemental Table S3).
Fig. 3.
Trends in hazard ratio for LGSOC relative to serous-BOT between 1988 and 2000. Hazard risk for LGSOC relative to serous-BOT per year from 1988 to 2000 is shown. Relative increase in hazard ratio for CSS and OS was 42.7% and 68.9%, respectively (P < 0.001), indicating the survival difference between the two groups enlarged during the study period. Results of trend analysis are shown in Supplemental Fig. S1A-B. Circles represent observed hazard ration with 95% confidence interval. Abbreviations: serous-BOT, serous borderline ovarian tumor; LGSOC, low-grade serous ovarian cancer; CSS, cause-specific survival; and OS, overall survival.
4. Discussion
Various hypotheses can support the diagnosis-shift observed in our study. First, a possible compelling plausible explanation is that earlier detection of precursor lesions of invasive cancer may result in more cases of serous-BOT and fewer cases of LGSOC. Various studies have demonstrated increasing population-based incidences of BOT over the past few decades [8,9], particularly, of the serous type [8]. A plausible hypothesis is that serous-BOT serves as a precursor of LGSOC, and increased use of transvaginal ultrasonography during the study period may have contributed to the early detection of serous-BOT prior to its progression to LGSOC [25-28]. This emerging precursor hypothesis is based on the notion that serous-BOT can recur as LGSOC and that there is genomic evidence to support the continuum of serous-BOT and LGSOC [29-32]. Taken together, more study is necessary to identify if serous-BOT can progress to LGSOC.
Another possibility for the diagnosis-shift from LGSOC to serous-BOT may be cross-classification of the two diagnoses. As described earlier, BOTs were firstly described in 1929 but had not been categorized as a separate entity of ovarian tumor by the WHO system until 1971 [5,7]. Thus, it is likely that some serous-BOT may have been previously classified as well-differentiated serous ovarian cancer [6]. Recognition of LGSOC in the mid-1980s may have enhanced the discrimination of serous-BOT from other serous tumors, resulting in acceleration of diagnosis-shift towards serous-BOT in the United States.
Strengths of the current study include that this is a population-based study with large sample size. Restricting the study population to the same time period and area allowed us to assess and interpret the diagnosis-shift between LGSOC and serous-BOT more reliably. Propensity score matching enriched the statistical rigor. Moreover, the relatively long follow-up period of 15 years strengthened the survival analysis.
There are several limitations in the study. First, possible unmeasured bias inherent to this type of study likely exists in the current analysis. For example, factors such as information regarding the interpreting pathologist, including experience and gynecologic subspecialty training, the number of sectioned slides, and the criteria used for making the histopathological diagnosis were not available in the SEER database but would likely affect the diagnosis. Another salient confounder missed in the analysis includes whether a patient received chemotherapy and hormonal therapy, and if so, which regimen was utilized. Similarly, this database does not have information on comorbidity such as obesity. This may have confounded the analysis as obesity is endemically increased in the United States and is also associated with an increased risk of serous-BOT [33,34].
Second, central pathology review of archived specimens was not feasible for this study, and the accuracy of histology diagnosis is unknown. In this study, well-differentiated serous ovarian cancer was used as a surrogate for LGSOC. While it may be possible that some of the well-different serous ovarian tumors were indeed high-grade serous ovarian cancer, the FIGO grading system has been shown to correlate well with the 2-tier system of low- and high-grade tumors (95% agreement rate) [3]. Moreover, reproducibility and inter-observer agreement among pathologists can be a concern for diagnosis BOT: a prior large-scale study demonstrated that nearly 5% of BOT were upgraded to invasive cancer following re-review of archived slides by expert gynecologic pathologists [35]. Thus, it may be possible that certain fractions of our study population may have been misclassified between the two groups.
Third, this study examined cases from 1988 to 2000, and the clinical relevance and utility from such old data may be limited. Surgical approaches, perioperative care, chemotherapeutic regimens, and hormonal therapy have advanced over the period since the study, and it is likely that the prognosis of women with these diseases in the current era is superior that from 1988 to 2000. Fourth, information regarding disease recurrence is not available in the SEER database, making complete outcome analysis not assessable. Fifth, this study was conducted in the U.S. population only, and generalizability in other geographic and racial/ethnicity populations remains unknown.
Lastly, a possible methodological flaw needs to be recognized in our analytical approach. In our study, we used the enlarging survival difference between the two diseases as a surrogate marker for diagnosis-shift from LGSOC to serous-BOT. Thus, one may argue that this could be simply due to stage-shift in each disease group with tumors of the LGSOC group becoming less likely to be stage I disease whereas tumors of the serous-BOT group becoming less likely to be stage IV disease. However, robustness of the analysis was re-demonstrated and survival difference between the two diagnoses remained enlarging over time even after clinico-pathological demographics are matched (Supplemental Table S3).
The results of our study clearly demonstrate that oncologic outcomes of LGSOC and serous-BOT are distinctively different, with both early and advanced disease. Thus, making a proper diagnosis for LGSOC versus serous-BOT is paramount as it impacts surgical and postoperative management. More specifically, comprehensive lymphadenectomy is indicated for LGSOC whereas the decision for lymphadenectomy is made on an individual basis in the setting of serous-BOT [36]. Postoperative chemotherapy is recommended for stage II-IV LGSOC whereas chemotherapy is generally withheld for serous-BOT, even with advanced disease [36].
In order to enhance the accuracy of histopathological diagnosis, review by expert gynecological pathologists is necessary given that LGSOC is a rare tumor. One such approach may be a central pathology review. For example, live review of histopathology slides was used in a recent trial of ovarian clear cell carcinoma, another rare gynecologic tumor, to ensure the proper diagnosis [37]. Another example is a phase II/III trial for recurrent LGSOC (GOG-0281) where prospective, central digital pathology review was employed to determine the trial eligibility [38]. Centralizing care for rare tumors including LGSOC would likely prove similarly beneficial [39].
Whether or not more sectioning of ovarian tumors would improve the diagnostic accuracy of LGSOC over serous-BOT is of interest. In mucinous ovarian tumors, the current recommendation is to section the ovarian tumor at least every 1 cm as there is a well described continuum of BOT to invasive cancer existing in the same tumor [40]. Similarly, as described earlier, serous-BOT may be a precursor lesion of LGSOC. Therefore, a similar concept to that utilized for mucinous tumors may be applicable to serous ovarian tumors.
In summary, there was a decreasing number of LGSOC with a concurrent increasing number of serous-BOT observed in 1988–2000 in the United States. Moreover, differences in survival between LGSOC and serous-BOT have gotten larger. These results possibly support that there was a diagnosis-shift from LGSOC to serous-BOT during the study period, and endorse the importance of making the correct diagnosis.
Supplementary Material
HIGHLIGHTS.
The incidence of low-grade serous ovarian cancer (LGSOC) decreased from 1988 to 2000.
There was a concurrent increase in the incidence of serous borderline ovarian tumor (BOT).
Survival difference increased between the two diagnoses during the study period.
These results possibly support a diagnosis-shift from LGSOC to serous-BOT.
Acknowledgments
Funding source
Ensign Endowment for Gynecologic Cancer Research (K.M.)
Footnotes
Disclosure statement
Scientific consulting, Kiyatec, Merck, share holder, Biopath, research funding, M-Trap (A.K.S.); honorarium, Chugai, textbook editorial, Springer, investigator meeting attendance expense, VBL therapeutics (K.M.); consultant, Quantgene (L.D.R.); advisory board, Tesaro, GSK (M.K.); Stock and other ownership interests, Celgene, Johnson & Johnson, Biogin, consulting or advisory role, Clovis Oncology, research funding, Novartis, royalties from Elsevier as book editor, royalties from UpToDate for authorship (D.M.G.); none for others.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.org/10.1016/j.ygyno.2019.08.030.
References
- [1].Siegel RL, Miller KD, Jemal A, Cancer statistics, 2019, CA Cancer J. Clin 69 (2019) 7–34. [DOI] [PubMed] [Google Scholar]
- [2].Malpica A, Deavers MT, Lu K, Bodurka DC, Atkinson EN, Gershenson DM, Silva EG, Grading ovarian serous carcinoma using a two-tier system, Am. J. Surg. Pathol 28 (2004) 496–504. [DOI] [PubMed] [Google Scholar]
- [3].Bodurka DC, Deavers MT, Tian C, Sun CC, Malpica A, Coleman RL, Lu KH, Sood AK, Birrer MJ, Ozols R, Baergen R, Emerson RE, Steinhoff M, Behmaram B, Rasty G, Gershenson DM, Reclassification of serous ovarian carcinoma by a 2-tier system: a Gynecologic Oncology Group Study, Cancer 118 (2011) 3087–3094. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Gershenson DM, Low-grade serous carcinoma of the ovary or peritoneum, Ann. Oncol 27 (2016) i45–i49. [DOI] [PubMed] [Google Scholar]
- [5].Bell DA, Low-grade serous tumors of ovary, Int. J. Gynecol. Pathol 33 (2014) 348–356. [DOI] [PubMed] [Google Scholar]
- [6].Matsuo K, Machida H, Grubbs BH, Sood AK, Gershenson DM, Trends of low-grade serous ovarian carcinoma in the United States, J. Gynecol. Oncol 29 (2018) e15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Matsuo K, Sood AK, Gershenson DM, Management of early-stage ovarian cancer, in: Bristow RE, Karlan BY (Eds.), Surgery for Ovarian Cancer: Principles and Practice. Chapter 3, 3rd edit, vol. 3, Informa Healthcare, New York: 2015, pp. 67–104. [Google Scholar]
- [8].Hannibal CG, Huusom LD, Kjaerbye-Thygesen A, Tabor A, Kjaer SK, Trends in incidence of borderline ovarian tumors in Denmark 1978–2006, Acta Obstet. Gynecol. Scand 90 (2011) 305–312. [DOI] [PubMed] [Google Scholar]
- [9].Skirnisdottir I, Garmo H, Wilander E, Holmberg L, Borderline ovarian tumors in Sweden 1960–2005: trends in incidence and age at diagnosis compared to ovarian cancer, Int. J. Cancer 123 (2008) 1897–1901. [DOI] [PubMed] [Google Scholar]
- [10].National Cancer Institute Surveillance, Epidemiology, and End Results Program, https://seer.cancer.gov/, Accessed date: 15 June 2019.
- [11].Sherman ME, Mink PJ, Curtis R, Cote TR, Brooks S, Hartge P, Devesa S, Survival among women with borderline ovarian tumors and ovarian carcinoma: a population-based analysis, Cancer 100 (2004) 1045–1052. [DOI] [PubMed] [Google Scholar]
- [12].Matsuo K, Machida H, Mandelbaum RS, Grubbs BH, Roman LD, Sood AK, Gershenson DM, Mucinous borderline ovarian tumor versus invasive well-differentiated mucinous ovarian cancer: difference in characteristics and outcomes, Gynecol. Oncol 153 (2019) 230–237. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].ICD-O-3 Coding Materials. https://seer.cancer.gov/icd-o-3/ (accessed 6/15/2019).
- [14].Lesieur B, Kane A, Duvillard P, Gouy S, Pautier P, Lhomme C, Morice P, Uzan C, Prognostic value of lymph node involvement in ovarian serous borderline tumors, Am. J. Obstet. Gynecol 204 (2011) 438 (e1-7). [DOI] [PubMed] [Google Scholar]
- [15].Matsuo K, Machida H, Takiuchi T, Grubbs BH, Roman LD, Sood AK, Gershenson DM, Role of hysterectomy and lymphadenectomy in the management of early-stage borderline ovarian tumors, Gynecol. Oncol 144 (2017) 496–502. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].American Joint Committee on Cancer 3rd Staging Classification, <http://seer.cancer.gov/seerstat/variables/seer/ajcc-stage/3rd.html> (accessed 6/15/2019).
- [17].Joinpoint trend analysis software. National Cancer Institute. <https://surveillance.cancer.gov/joinpoint/> (accessed on 6/15/2019).
- [18].Hershman DL, Wright JD, Comparative effectiveness research in oncology methodology: observational data, J. Clin. Oncol 30 (2012) 4215–4222. [DOI] [PubMed] [Google Scholar]
- [19].Austin PC, Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies, Pharm. Stat 10 (2010) 150–161. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Yao XI, Wang X, Speicher PJ, Hwang ES, Cheng P, Harpole DH, Berry MF, Schrag D, Pang HH, Reporting and guidelines in propensity score analysis: a systematic review of Cancer and Cancer surgical studies, J. Natl. Cancer Inst 109 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Cox DR, Regression models and life-tables, J R Stat Soc Series B Stat Methodol 34 (1972) 187–220. [Google Scholar]
- [22].Austin PC, The performance of different propensity-score methods for estimating differences in proportions (risk differences or absolute risk reductions) in observational studies, Stat. Med 29 (2010) 2137–2148. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [23].Mansfield ER, Helms BP, Detecting multicollinearity, Am. Stat 36 (1982) 158–160. [Google Scholar]
- [24].von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP. Strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ 2007;335:806–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [25].Romero I, Sun CC, Wong KK, Bast RC Jr., Gershenson DM. Low-grade serous carcinoma: new concepts and emerging therapies. Gynecol. Oncol l2013;130: 660–6. [DOI] [PubMed] [Google Scholar]
- [26].Timor-Tritsch IE, Is office use of vaginal ultrasonography feasible? Am. J. Obstet. Gynecol 162 (1990) 983–985. [DOI] [PubMed] [Google Scholar]
- [27].Higgins RV, van Nagell JR Jr., Donaldson ES, Gallion HH, Pavlik EJ, Endicott B, Woods CH, Transvaginal sonography as a screening method for ovarian cancer, Gynecol. Oncol 34 (1989) 402–406. [DOI] [PubMed] [Google Scholar]
- [28].van Nagell JR Jr.., Higgins RV, Donaldson ES, Gallion HH, Powell DE, Pavlik EJ, Woods CH, Thompson EA, Transvaginal sonography as a screening method for ovarian cancer. A report of the first 1000 cases screened, Cancer 65 (1990) 573–577. [DOI] [PubMed] [Google Scholar]
- [29].Longacre TA, McKenney JK, Tazelaar HD, Kempson RL, Hendrickson MR, Ovarian serous tumors of low malignant potential (borderline tumors): outcome-based study of 276 patients with long-term (> or =5-year) follow-up, Am. J. Surg. Pathol 29 (2005) 707–723. [DOI] [PubMed] [Google Scholar]
- [30].Wong KK, Gershenson D, The continuum of serous tumors of low malignant potential and low-grade serous carcinomas of the ovary, Dis. Markers 23 (2007) 377–387. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [31].Bonome T, Lee JY, Park DC, Radonovich M, Pise-Masison C, Brady J, Gardner GJ, Hao K, Wong WH, Barrett JC, Lu KH, Sood AK, Gershenson DM, Mok SC, Birrer MJ, Expression profiling of serous low malignant potential, low-grade, and high-grade tumors of the ovary, Cancer Res. 65 (2005) 10602–10612. [DOI] [PubMed] [Google Scholar]
- [32].Crispens MA, Bodurka D, Deavers M, Lu K, Silva EG, Gershenson DM, Response and survival in patients with progressive or recurrent serous ovarian tumors of low malignant potential, Obstet. Gynecol 99 (2002) 3–10. [DOI] [PubMed] [Google Scholar]
- [33].Harris R, Whittemore AS, Itnyre J, Characteristics relating to ovarian cancer risk: collaborative analysis of 12 US case-control studies. III. Epithelial tumors of low malignant potential in white women. Collaborative Ovarian Cancer Group, Am. J. Epidemiol 136 (1992) 1204–1211. [DOI] [PubMed] [Google Scholar]
- [34].Heath, United States, https://www.cdc.gov/nchs/data/hus/hus16.pdf#053> 2016. , linked from https://www.cdc.gov/nchs/data/hus/hus16.pdf> (accessed 7/10/2019).
- [35].du Bois A, Ewald-Riegler N, de Gregorio N, Reuss A, Mahner S, Fotopoulou C, Kommoss F, Schmalfeldt B, Hilpert F, Fehm T, Burges A, Meier W, Hillemanns P, Hanker L, Hasenburg A, Strauss HG, Hellriegel M, Wimberger P, Keyver-Paik MD, Baumann K, Canzler U, Wollschlaeger K, Forner D, Pfisterer J, Schroder W, Munstedt K, Richter B, Kommoss S, Hauptmann S, Borderline tumours of the ovary: a cohort study of the Arbeitsgmeinschaft Gynakologische Onkologie (AGO) study group, Eur. J. Cancer 49 (2013) 1905–1914. [DOI] [PubMed] [Google Scholar]
- [36].Ovarian cancer including fallopian tube cancer and primary peritoneal cancer, NCCN clinical practice guidelines in oncology (NCCN guidelines) https://nccn.org, Accessed date: 9 July 2019. [Google Scholar]
- [37].Sugiyama T, Okamoto A, Enomoto T, Hamano T, Aotani E, Terao Y, Suzuki N, Mikami M, Yaegashi N, Kato K, Yoshikawa H, Yokoyama Y, Tanabe H, Nishino K, Nomura H, Kim JW, Kim BG, Pignata S, Alexandre J, Green J, Isonishi S, Terauchi F, Fujiwara K, Aoki D, Randomized phase III trial of irinotecan plus cisplatin compared with paclitaxel plus carboplatin as first-line chemotherapy for ovarian clear cell carcinoma: JGOG3017/GCIG trial, J. Clin. Oncol 34 (2016) 2881–2887. [DOI] [PubMed] [Google Scholar]
- [38].Trametinib in treating patients with recurrent or progressive low-grade ovarian Cancer or peritoneal cavity Cancer. <https://clinicaltrials.gov/ct2/show/NCT02101788> ((accessed 07/14/2019)). [Google Scholar]
- [39].Aletti GD, Cliby WA, Time for centralizing patients with ovarian cancer: what are we waiting for? Gynecol. Oncol 142 (2016) 209–210. [DOI] [PubMed] [Google Scholar]
- [40].Morice P, Gouy S, Leary A, Mucinous Ovarian Carcinoma, N. Engl. J. Med 380 (2019) 1256–1266. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.