Background

Ovarian carcinoma is the fourth most common cancer in women worldwide but it is responsible for the most fatalities due to gynaecological malignancy in the developed world [1]. Known prognostic factors include tumour characteristics (grade, histotype, clinical stage), patient factors (age, fitness, smoking status), treatment factors, and the response to first line treatment [2].

There is evidence that some epithelial ovarian tumours evoke an immune response and that this affects prognosis [3]. Several studies have investigated the effect of various subtypes of immune cell response on prognosis in ovarian cancer [4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20]. These studies have been of modest sample sizes ranging from 9 through to 500 tumours. Immunohistochemistry was used to investigate the immune response, but they have shown conflicting results (outlined in Table 1), though most suggest that CD3(+) and CD8(+) tumour infiltrating lymphocytes are associated with a favourable outcome. Immune infiltration has also been shown to be enriched in some molecular subtypes of ovarian cancer [21]. The anti-tumour immune response is perhaps most easily assessed morphologically using standard histopathology based on haematoxylin and eosin (H&E) staining. Therefore this may represent the most promising avenue for clinical translation. However, no study has yet investigated the effect of total tumour infiltrating lymphocytes (TILs) on prognosis in ovarian cancer.

Table 1 Studies investigating immune cell response and effect on prognosis in epithelial ovarian cancer
Full size table

There have been calls for the use of a standardised method of evaluating TILs in cancer and incorporating it clinically as a prognostic variable [22]. Haematoxylin and eosin (H&E) staining and visual analysis is a recognised cheap and clinically accessible method of evaluating TILs. A standardised method for H&E visual evaluation of TILs has been validated in the context of breast cancer [23].

We investigated the pathological and prognostic significance of TILs in epithelial ovarian cancer. We used a standardised method of TIL evaluation and a large prospective population-based study of epithelial ovarian cancer to achieve this.

Methods

Study population

The SEARCH ovarian cancer study is an ongoing, population-based ovarian cancer case–control study, covering the regions served by the East Anglia and West Midlands cancer registries in the UK. All patients diagnosed in East Anglia with invasive epithelial ovarian cancer under the age of 70 years since 1991, and still alive in 1998 when recruitment started are eligible to participate. Details of this study and its participants have been published previously [24]. To date over 2500 women with epithelial ovarian cancer have been recruited. We have retrieved archival, formalin-fixed, paraffin-embedded (FFPE) tumour tissue from the primary surgery from 707 SEARCH participants with invasive epithelial ovarian cancer. A total of 953 samples have been retrieved including tumour from both primary and metastatic sites.

Survival time data were available through the regional cancer registry which obtains notification of deaths by death certificate flagging through the Office of National Statistics. The lag times for this are a few weeks for cancer deaths and two months to a year for non-cancer deaths. In addition, the cancer registry checks vital status by querying the National Health Service Strategic Tracing Service. Vital status was ascertained at the end of June 2015 and all analyses were censored on 31 December 2014 to allow for delay in reporting of vital status. Ovarian cancer specific mortality was defined as deaths where ovarian cancer was listed as the cause of death on Parts 1a, 1b, or 1c of the death certificate.

Evaluation of tumour infiltrating lymphocytes

Formalin-fixed, paraffin-embedded tumour samples were sectioned and stained using hematoxylin and eosin. The morphology was reviewed by a gynaecological pathologist (MJ-L). Full face stained sections were then scanned and the digital images were used for evaluation of TILs.

Evaluation of TILs was performed by manual visual assessment of percentage area covered by lymphocytes, following a standardised method as described in Salgado et al. [23]. In this method the area within the tumour border is selected, the intra-tumoral and stromal compartment are defined, only areas of mononuclear lymphocytic infiltrate are included, and the percentage area is assessed at low (4× objective) and high magnification (10× objective). The percentage categories used were 1%, 5% and the nearest 10% up to 100%. The intra-tumoral and stromal compartments were scored separately. The intra-tumoral and stromal TILs were classified into three groups for the purpose of analysis – very low (1%), low (5%) and high (≥10%). Representative examples of tumours with very low, low and high TIL are shown in Fig. 1.

Fig. 1
figure 1

Representative examples of tumours with very low, low and high TIL

Full size image

Statistical analysis

The proportion of tumours in the three TIL score categories by other histopathologic characteristics was compared using the Pearson χ2- test. Multi-variable Cox regression was used to assess the association between TILs and disease-specific mortality. Follow-up time was adjusted for delayed study entry by using the date of study entry as the start of time at risk (left truncation) [25]. All models were adjusted for FIGO stage (stage 1–4 treated as an ordinal variable), histotype (seven categories) and age at diagnosis. Analyses were censored at 10-years of follow-up or on death from a cause other than ovarian cancer. Stata 14 (Stata Corp, College Station, Texas, USA) statistical software was used for all analyses.

Results

Data were available for 953 invasive epithelial ovarian tumours (primary and/or metastatic), from 707 women. Of these, primary tumour material was available for 682 cases, omental tumour material for 161 cases and other metastasis material was available for 114 cases (Fig. 2). The clinical characteristics (age at diagnosis and stage) of these cases by histotype are shown in Table 2.

Fig. 2
figure 2

Number of patients by available tumour sample

Full size image
Table 2 Clinical characteristics of ovarian cancer cases included in this study
Full size table

There was a moderately strong correlation between the extent of stromal and intra-tumoral TIL levels in the primary tumour (n = 679, Spearman’s rank correlation = 0.60, P < 0.001) with a similar correlation in secondary tumours (n = 224, Spearman’s rank correlation = 0.62, P < 0.001). Secondary tumour samples were either omental metastases or metastases at other sites. We used the data for the omental metastases in preference to data for other metastases if present; otherwise the data from the other metastases was used. We also evaluated the correlation between TIL levels in primary and metastatic tumours using data from the 196 women for whom both a primary and a secondary tumour sample was available. There was a statistically significant but weak correlation between stromal TIL levels in primary and secondary tumour samples (Spearman’s rank correlation = 0.29, P < 0.001) and intra-tumoral TIL levels in primary and secondary tumour samples (Spearman’s rank correlation = 0.19, P = 0.0094) (Table 3 ).

Table 3 Association between tumour infiltrating lymphocyte levels in primary and secondary tumours for stromal and intratumoral infiltrating lymphocytes
Full size table

The extent of stromal TIL levels differed between histotypes (Table 4, Pearson chi2 (12d.f.) 54.1, P < 0.0001) with higher levels of stromal infiltration in the high-grade serous and endometriod cases. However, there was little difference in the degree of intra-tumoral infiltrating lymphocytes by histotype (Pearson chi2 (12d.f.) = 15.0, P = 0.24).

Table 4 Intra-tumoral and stromal tumour infiltrating lymphocyte levels by histotype
Full size table

We assessed the impact of tumour infiltrating lymphocytes on ovarian cancer specific mortality after diagnosis using multi-variable Cox proportional hazards regression. The extent of tumour infiltrating lymphocytes in three categories (<5%, 5–9%, ≥10%) was treated as an ordinal variable in the model. An increase in stromal TIL levels was associated with a slightly better prognosis but the difference was not statistically significant (HR 0.87 95% CI 0.74–1.02 p = 0.091). A slightly stronger and nominally significant association was observed for intra-tumoral TIL levels (hazard ratio = 0.74, P = 0.047). When the extent of both stromal and intra-tumoral TIL levels were included in a multi-variable model the effect of the intra-tumoral TIL levels was similar (HR 0.76 95% CI 0.58–1.00 p = 0.047), but the association of the stromal lymphocytes was attenuated (HR 0.97 95% CI 0.79–1.19 p = 0.78). There was little evidence for heterogeneity of effect of intra-tumoral TIL levels by histotype, although the effect was strongest in the high-grade serous cases (HR 0.70 95% CI 0.50–0.97 p = 0.034).

Discussion

This study is the largest collection of epithelial ovarian tumour samples evaluated for TILs, with almost double the number of tumour samples as the largest published study to date [12]. This is also the first time this standardised H&E visual assessment method of TIL evaluation has been used in ovarian tumours.

We have shown that TILs as assessed by stromal lymphocyte levels correlate with clinical variables such as tumour histological subtype. However, intratumoral TILs did not. We have also shown that increased levels of both intratumoral and stromal TILs are associated with a better prognosis; however, this is only statistically significant for intratumoral TILs.

These results add weight to the argument that the immune system plays an important role in ovarian cancer. They also support the idea that different subtypes of ovarian cancer are variably immunogenic and that the immune response can help define different subtypes.

The difference between stromal and intratumoral TIL effects, despite their correlated levels in both primary and secondary tumours, is interesting, and warrants further investigation. The greater prognostic effect of intratumoral TILs may explained by the biological rationale that T-cell activation necessitates physical contact with target cells in order to engage the T cell receptor [26]. In contrast, the lesser prognostic effect of stromal TILs might be explained in two ways; either they are in-transit to tumour cells hence will eventually become intratumoral TILs, or that they have been ‘excluded’ from the tumour microenvironment by factors secreted by tumour cells and/or other stromal cells, as a means of evading immune attack [26]. Further work to elucidate the distinct roles of intratumoral and stromal TILs is needed.

The results suggest that a clinically useful immune prognostic indicator in epithelial ovarian cancer could be developed using this technique, but larger scale studies are needed to replicate these results. Thus, further studies evaluating the potential of TILs as predictors of treatment response are warranted.

Conclusion

This study is the largest collection of epithelial ovarian tumour samples evaluated for TILs. We have shown that stromal and intratumoral TIL levels are correlated and that their levels correlate with clinical variables such as tumour histological subtype. We have also shown that increased levels of both intratumoral and stromal TILs are associated with a better prognosis; however, this is only statistically significant for intratumoral TILs. This study suggests that a clinically useful immune prognostic indicator in epithelial ovarian cancer could be developed using this technique.