Patterns of clinicopathological features and outcome in epithelial ovarian cancer patients: 35 years of prospectively collected data

Investigate the clinical landscape of ovarian carcinoma (OC) over time.


Introduction
With over 290 000 new diagnoses and 180 000 deaths per year worldwide, ovarian cancer is the most lethal of all gynaecological malignancies. 1 This is attributed, in part, to the high frequency at which these malignancies are diagnosed at an advanced stage, which represents a major clinical challenge. For these advanced stage cases, the 5-year survival rate remains poor at under 30%. 2 It is now recognised that ovarian carcinoma (OC)which represents around 90% of ovarian cancer cases-is a collection of discrete diseases, the five main histotypes of which are high grade serous (HGS), endometrioid, clear cell, mucinous and low grade serous (LGS) OC. 3 These histotypes display distinct clinical characteristics, with differing intrinsic chemosensitivity, typical stage at diagnosis and overall survival outcome. 4 Moreover, these histotypes are now known to arise from distinct gynaecological sites. [5][6][7][8][9] Despite intensive research efforts to find further therapeutic options, the standard-of-care for OC has largely remained static in recent decades, comprising maximal cytoreductive debulking surgery followed by platinum-based chemotherapy, frequently in combination with taxanes. 10 In recent years, the use of anti-angiogenic treatments and PARP inhibitors has been integrated into routine practice, with several trials demonstrating prolonged progression-free survival (PFS), largely in the relapse disease setting. [11][12][13][14][15][16] Recognition of the biologically distinct histotypes within OC has highlighted the need for identifying new histotypespecific therapeutic treatments 17 and has led to rationally designed histotype-specific trials of biological agents. 18,19 It is well established that disease stage at diagnosis, patient age and Eastern Cooperative Oncology Group Performance Status (ECOG PS) 20-23 are associated with differential survival outcomes in OC patients; moreover, optimal surgical cytoreduction has emerged as one of the most important determinants of outcome. [24][25][26] The definition of optimal cytoreduction has evolved alongside our understanding of OC as a disease entity, [27][28][29][30][31] with the goal of surgery evolving from <2 cm maximal dimension of the largest residual disease (RD) lesion to minimal RD (<0.5 cm), and now to the current objective of achieving no visible RD. [29][30][31][32] Here, we investigate the changing clinical landscape of ovarian carcinoma patients from South East Scotland (population 1.4 million) over the last 35 years (1981-2015) using data retrieved from The Edinburgh Ovarian Cancer Database.

Cases
Cases were identified using the Edinburgh Ovarian Cancer Database; patient demographics and survival data, prospectively collected as part of routine clinical care, were retrieved from the database. No independent ethical approval for this study was required, as determined by the South East Scotland Research Ethics Service.
All pathologically confirmed epithelial OC diagnoses of serous, mucinous, endometrioid or clear cell histological type between 1981 and 2015 were included ( Figure S1), including cases recorded as primary fallopian tube or primary peritoneal carcinoma, representing the vast majority of OC cases in the region (for example, cases treated solely within private practice will not have attended at the Edinburgh Cancer Centre). All other histotypes were excluded.
Historically diagnosed grade II serous carcinomas (n = 189) were included with documented grade III serous carcinomas (n = 1010) and HGSOCs (n = 554). Well-differentiated (grade I) serous (n = 107) OCs were included alongside contemporary diagnoses of LGSOC (n = 10). Serous carcinomas with unknown grade or variable differentiation were excluded (n = 96). Of the cases, 51.0% represented either contemporary diagnoses (2010 onwards) or cases where histotype has been confirmed by contemporary pathology review by an expert gynaecological pathologist (CSH). [33][34][35][36][37][38][39] Demographics Patients were classified into 5-year cohorts using date of pathologically confirmed OC diagnosis (1981-1985, 1986-1990, 1991-1995, 1996-2000, 2001-2005, 2006-2010, 2011-2015). Staging information was based on the International Federation of Obstetrics and Gynaecology (FIGO) staging system. Debulking status was classified as <2 and ≥2 cm residual disease (RD). Debulking status could not be resolved beyond <2 cm for all cases due to the retrospective nature of these data and historical classification of <2 cm RD as optimal debulking prior to 2008. ECOG performance status (PS) was categorised discretely from 0 (PS 0) to 4 (PS 4). Due to the low number of cases with PS 4 (n = 6), these cases were excluded from ECOG PS analysis. Five distinct cases were excluded from survival analysis (four from disease-specific survival [DSS] and four from PFS analysis) due to missing outcome data.

Statistical analysis
DSS was evaluated as time from date of diagnosis to disease-specific death. Deaths from other causes were censored. PFS was evaluated as time from date of diagnosis to date of OC progression, 40 where progression was established by radiologically confirmed progressive disease (PD), CA125 PD or clinical deterioration as determined by the treating physician. Statistical analyses were performed using R 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria). Survival analyses were visualised using the Kaplan-Meier method. Survival statistics are presented as median survival with corresponding 95% CI, alongside 5and 10-year survival rates and statistical comparison by restricted mean survival time analysis. Multivariable analyses were performed using Cox proportional hazards regression models, stratified by RD, histotype and age or PS. Differences in frequency were analysed using the Chisquare test. P < 0.05 was considered statistically significant.

Results
In all, 2805 patients met the inclusion criteria ( Figure S1, Table 1). Of these, 51.0% represented contemporary diagnoses (2010 onwards) or had their histotype confirmed by     Table 1). The median age at diagnosis significantly increased across time (57 years in 1981-1985 versus 66 years in 2011-2015, P < 0.0001) ( Figure 3A). When plotted against the mean PS for each 5-year cohort, a strong correlation can be observed against mean patient age (r = 0.86) ( Figure 3B), consistent with the overall correlation between age and PS across the cohort ( Figure S3).

Associations between histological subtype and outcome
HGSOC demonstrated the lowest 5-year DSS (25.0%, 95% CI 22.9-27.2) of the histotypes (Figure 2A) LGSOC showed the highest median survival of 6.76 years in this analysis ( Figure 2E). Corresponding early stage (stage I and II) DSS analysis mirrored the results of the overall DSS analysis (Figure 2A).

Associations between other clinicopathological features and outcome
Low RD volume following surgical debulking, lower PS and earlier stage were all associated with significantly prolonged DSS ( Figure 2B-D). Patients with <2 cm RD demonstrated significantly higher median DSS (7.33 years, 95% CI 6.46-8.80, P < 0.0001) than did those with ≥2 cm of RD (1.46 years, 95% CI 1.32-1.55). Each increase in performance status (reduction in ECOG PS score) saw a significantly increased median survival ( Figure 2D, Table S2). PS 3 was associated with a median DSS of 0.67 years (95% CI 0.43-1.01), whereas PS 0 was associated with a median DSS of 5.52 years (95% CI 4.81-6.80) (P < 0.0001, PS 3 versus PS 0). Similarly, stages I, II and III showed higher DSS compared with stage IV (P < 0.0001 for all) ( Figure 2B, Table S2).
Multivariable analysis of disease stage at diagnosis, histotype, time period of diagnosis, RD volume, ECOG PS and age at diagnosis reflected the univariable analyses (Tables S4  and S5). Notably, these data highlight an independent association of both age and PS with DSS, despite the observed correlation between these two factors ( Figure S3).    (Table S8). Increase in median PFS across the same period was not significant in stage III HGSOC (0.95 years versus 1.25 years, P = 0.0601) but was statistically significant in stage IV HGSOC (0.69 years versus 1.14 years, P = 0.0003) (Table S9). These data mirror the results from the pan-histotype DSS and PFS analysis for stage across the cohort (Tables S6 and S7).

Main findings
We have demonstrated and quantified the improvement in the DSS of women with epithelial OC across time at the Edinburgh Cancer Centre. A similar improvement in PFS was not seen. Differences in survival based on histotype, RD volume following debulking, ECOG PS and stage were consistent with previous research. An increase in advanced stage HGSOC incidence and survival was seen. A strong correlation was found between increasing age at diagnosis and ECOG PS across time, indicating a shift in the clinical demographic towards an older patient population with more frequent co-morbidities.

Strengths and limitations
Strengths of the study include the large number of cases and the high granularity of the prospectively collected clinical and treatment data; few similarly extensive longitudinal analyses of real-world OC data have been reported to date. Data were collected as part of routine care, almost exclusively by a single individual, optimising consistency.
We recognise several limitations of this study. First, criteria for defining progression have changed over time 40 and were heterogeneous across the periods defined in our study. Our samples are therefore subject to varying definitions of progression over time-including CA125 and radiological evidence as well as more subjective clinical assessment. Second, contemporary pathology review was not carried out for all cases; lack of review for all LGSOC and high grade endometrioid cases, which have historically been poorly differentiated from HGSOCs, is a particular weakness. Moreover, the mucinous OC group had a lower rate of pathology review or contemporary diagnosis, likely a reflection of the increasing rarity of true primary mucinous OC by modern pathological criteria. However, across the whole of our OC cohort, over half of cases were confirmed by pathology review in previous studies or represented contemporary diagnoses, in contrast to previous investigations performing no such review, 41-43 representing a major strength of this study over previous work. Differences in practice between treating physicians and the impact of ascertainment bias also represent potential limitations.

Interpretation
The 5-year DSS rate observed in this study for the 2011-2015 period was 46% (95% CI [41][42][43][44][45][46][47][48][49][50][51]; this is consistent with data reported by Siegel et al. 2 A significant improvement was seen from 1981 to 1985 where the 5-year survival rate was 31%. The median DSS improved significantly from 1.73 to 4.23 years. This improvement represents the culmination of changes in management over time, including the movement toward centralised care in centres with specialist expertise, more robust histopathological classification, improvements in disease monitoring such as imaging technology, and the introduction of additional therapeutic options. Most notably, platinum-taxane combination chemotherapy was introduced as standard of care within the study time period, and there has been a paradigm shift toward extensive cytoreductive surgery to maximise the chances of complete first-line macroscopic resection of disease, 25,44 aided by neoadjuvant chemotherapy in some patients. Despite the significantly prolonged DSS observed over time, observed improvement in PFS time failed to meet statistical significance (Table S3, Figure S2). This suggests that while treatment has improved for recurrent disease, there has been little improvement in preventing or significantly prolonging relapse. This is consistent with the largely static standard of care for first-line OC treatment in the study period. Recent studies of first-line olaparib treatment for BRCA1 or BRCA2 mutant HGSOC patients 13 and hormone maintenance for LGSOC patients 45 indicate that the coming years may see an improvement in OC PFS with the routine use of these agents. Notably, however, these regimens will be limited to subsets of patients.
A change in proportions of different histotypes was observed over the last 35 years with significant increases in HGSOC cases and decreases in mucinous and endometrioid cases. It is now recognised that many previously diagnosed high grade endometrioid carcinomas in fact represent variants of HGSOC 46 ; this may explain the relative depletion of endometrioid diagnoses over time. Moreover, historical misclassification of metastatic malignancies of the gastrointestinal tract as primary mucinous OC may explain the decline in mucinous cases over time. 47,48 It is therefore likely that the change in proportions of histotypes observed in this study is, at least in part, a result of a refinement in classification of tumour types.
A significant increase in the proportion of HGSOC patients presenting with stage IV disease was also observed, alongside a corresponding decrease in stage I patients. This indicates that despite efforts to increase awareness of OC symptoms, these efforts have thus far failed to increase the proportion of early stage diagnoses. However, median DSS for these cases has increased significantly overall, and for HGSOC patients specifically (Tables S4 and S9), indicating post-relapse management has improved. It is also feasible that the observed increased incidence and survival in advanced stage cases is a consequence of the Will Rogers phenomenon, 49 whereby advances in diagnostic techniques (such as more sensitive imaging) leads to up-staging of cases that would otherwise have been earlier stage. Certainly, increased ability of contemporary imaging to detect features such as epicardial nodes could account for a significant amount of stage shift over the time cohorts analysed. The improved outcome observed in advanced stage cases within our study is consistent with recent SEER analysis demonstrating improved outcome in this patient group. 50 The proportion of cases with <2 cm of RD remained within the 50-60% range for 1981-2010, showing a large increase to 88% in the 2011-2015 year range. It is likely that the emphasis on optimal debulking surgery for OC patients in recent years, driven by the recognition that complete macroscopic cytoreduction is associated with markedly favourable outcome, 31 has led to this increase. Moreover, this may account for decreases in median DSS and PFS seen in the <2 cm RD cohort at later diagnosis periods, as modern efforts to achieve complete macroscopic tumour resection-including radical debulking surgery and introduction of neoadjuvant chemotherapy-has enriched this cohort for poor prognosis cases over time. Difference in survival between histotypes observed in this study was generally consistent with results of previous studies 42,46,51 ; the LGS and endometrioid histotypes displayed better survival compared with HGSOC. Peres et al. 51 found that mucinous OC displayed favourable survival at early stage, but dismal prognosis when diagnosed at advanced stage. As the majority of mucinous cases were stage I (196/282 = 70%), the overall trend for favourable survival seen in this study, across time and within each 5year cohort, is consistent with data previously reported. 51 Our data show that early-stage mucinous cases show a favourable outcome, whereas advanced-stage cases perform poorly. Clear cell OC cases demonstrated poor survival at early and advanced stages, consistent with previous reports of intrinsic chemoresistance in clear cell and mucinous OC, 4,52,53 highlighting the need for targeted therapies aimed at the underlying biology of these malignancies.
Previous studies have uncovered and emphasised the importance of FIGO stage 54,55 and extent of RD following debulking 54,56-58 as prognostic factors in OC. This study confirms the importance of these two factors in OC survival, as well as ECOG PS. Although there have been recent reports that ECOG PS is of limited importance, 20 we observed a clear delineation in survival based on ECOG PS. Moreover, an adjusted multivariable model indicated an association with survival independent of other clinicopathological factors.
We observed a significant increase in median patient age across time (57 years in 1981-1985 versus 66 years in 2011-2015, P < 0.0001), reflective of the UK's ageing population. A similar increase was seen on comparing the mean PS of cases across time. We show a correlation between increased age and PS (r = 0.86) across time. Multivariable analysis indicated the independent adverse associations of both of these factors on survival (Tables S4 and  S5). This is indicative of the shift towards an older and frailer clinical demographic, representing a clinically challenging population characterised by co-morbidities, chemotherapy delays and poorer survival outcome. 59 Collectively, these data shed new light on the shifting clinical landscape of OC management, demonstrating survival improvement across time as management of OC patients has evolved. They also highlight the current areas of greatest unmet clinical need, where new therapeutic options are urgently required to improve outcome.

Conclusion
OC patient survival in South East Scotland has improved markedly over the last 35 years. Histology, stage, extent of RD and ECOG PS are strongly associated with survival outcome. Advanced stage disease has seen an increase in incidence and survival, both within HGSOC cases specifically and across all histotypes. Despite this, PFS has not seen a corresponding increase. Recent trials of first-line agents for specific subgroups of OC 13,45 indicate that PFS improvement may be seen over the coming years in these groups. However, to see a large PFS increase in the overall OC population there is an urgent need for further improvements in first-line management. Advanced stage clear cell and mucinous OCs represent those patients with greatest unmet need. Moreover, the changing clinical demographic towards an older population with more comorbidities highlights a growing patient group that represents a more complex clinical challenge.
Future work should aim to investigate the impact of recently introduced therapeutic options, such as anti-angiogenic therapies and PARP inhibitors, on outcome in OC. In particular, whether the use of these agents in the firstline setting leads to an improvement in the currently stagnant PFS of OC patients should be investigated.

Disclosure of interests
MM: honoraria from Tesaro, BristolMyersSquibb and Roche. FN: Non-personal interests in AstraZeneca and Tesaro. CG: research funding from AstraZeneca, Aprea, Nucana, Tesaro and Novartis; honoraria/consultancy fees from Roche, AstraZeneca, Tesaro, Nucana, MSD, Clovis, Foundation One, Sierra Oncology and Cor2Ed; named on issued/pending patents relating to predicting treatment response in ovarian cancer beyond the scope of this work. AI, RLH, KH, TR, MC, CB and CSH declare no conflicts of interest. Completed disclosure of interests forms are available to view online as supporting information.  data obtained as part of routine care do not require NHS ethical review. As such, no independent ethical approval for this study was required.

Funding
RLH is supported by an MRC-funded research fellowship. This study was supported by charitable donation from the Nicola Murray Foundation.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Figure S1. Flow diagram for case inclusion. Figure S2. Scatterplot of median DSS and PFS increase over time. Figure S3. Boxplot of median age at diagnosis for discrete levels of ECOG performance status. Table S1. Cases by histotype confirmed by contemporary pathology review. Table S2. Median disease-specific survival (DSS) times by year of diagnosis, histotype, RD following debulk, ECOG performance value and stage. Table S3. Median progression-free survival (PFS) times by year of diagnosis, histotype, RD following debulk, ECOG performance value and stage Table S4. Hazard ratios for multivariable model including stage, year of diagnosis, age at diagnosis and stratified for histotype, RD following debulk, ECOG PS. Table S5. Hazard ratios for multivariable model including stage, year of diagnosis, ECOG PS and stratified for histotype, RD following debulk, age at diagnosis. Table S6. Median disease-specific survival (DSS) times by histotype, stage, RD following debulk, ECOG performance value across time. Table S7. Median progression-free survival (PFS) times by histotype, stage, RD following debulk, ECOG performance value across time. Table S8. Median disease-specific survival (DSS) times by stage of HGSOC disease across time.