Imaging ovarian cancer – from baseline characteristics to high-risk image factors

Computer tomography (CT) is an established tool in ovarian cancer (OC) diagnosis, clinical decision making regarding surgery and treatment, and follow-up of cancer survivors [1].

The CT provides highly accurate information on ovarian tumor load and abdominal spread (i.e. peritoneal carcinomatosis) [2, 3]. However, the radiological description and presentation is subjective and very much depending on radiologists and local traditions. The established peritoneal carcinomatosis index [4] (PCI) is a structured score which evaluates and quantifies the carcinomatosis in CT (CT-PCI) and/or at surgery (surgical-PCI (S-PCI)).

A recent paper from our research group showed that high CT-PCI was associated with higher clinical stage and impaired OC survival [5], which is in line with previous studies also highlighting the value of CT-PCI in evaluating surgical outcome [6] and patient prognosis [7]. However, from a radiological perspective, a thorough PCI-scoring is a tedious procedure not prioritized in daily clinical work. A previous paper on S-PCI concluded that selected PCI regions (i.e., small intestine and hepatoduodenal ligament) were more predictive of complete resection and survival than the entire PCI-score [8]. With support in previous research, we therefore hypothesize that an abbreviated CT-based image short score might offer the same clinical insights as the full CT-PCI, and at the same time be clinically feasible, and this has to the best of our knowledge not been previously studied. In addition, such a score could include additional high-risk image parameters (e.g., enlarged lymph nodes) and not be limited to carcinomatosis.

At the very beginning of the diagnostic chain, primary prevention is a goal of future OC care. As of today, it is not known whether any of the established OC risk factors in healthy women may lead to high-risk image factors in the later established OC diagnosis. A previous study by Poole et al. (n = 4342 OC cases) investigated differences in the associations with OC risk factors by tumor aggressiveness (rapidly fatal (dead within three years) vs all other tumors) and concluded that rapidly fatal cases were older and tended to have shorter duration of oral contraceptive use [9]. A similar approach but with imaging as endpoint (high risk image being defined as increasing CT-PCI or CT-based short score) has to the best of our knowledge not been studied. Understanding and highlighting such risk factors may add knowledge to how patient factors are potentially linked to the image. Further, deepened knowledge could implicate that the radiologist should pay extra attention to the presence or absence of a particular risk factor at time of the CT reading.

The goal of this study was to analyze a CT-based image short score, with the hypothesis that this would provide similarly strong, or stronger, associations with OC prognosis as the PCI-score. In addition, we wanted to explore if any of the known OC risk factors was linked to image factors included in the score.

The distribution of patient characteristics, tumor characteristics and CT parameters is illustrated in Table 1. As seen in the scatter plot and corresponding correlation coefficient (r = 0.769), there was a clear relation and high correlation between the CT short scores from the two readers (subset analysis n = 20 cases) (Fig. 2).

In this study, we have identified strong relationships between a novel CT short score, advanced clinical stage, and impaired OC-specific survival. The short score, embracing both peritoneal carcinomatosis and CPLN, showed a similar or stronger relation to the OC outcomes than the previously studied CT-PCI in the same cohort [5]. From the radiologist’s perspective, this means that a pragmatic way of evaluating high risk CT parameters can be combined with a clinically trustworthy score.

The interpretation of the CT image is subjective, and the performance of CT in detecting tumor deposits varies between organs and regions [15]. With the attempt to structure CT-evaluation and reporting regarding peritoneal carcinomatosis, the potential implementation of CT-PCI in OC has been extensively studied. Previous work on CT-PCI, partly by our group on a more recent cohort, has shown good agreement between CT-PCI and surgical PCI [16, 17], a link between increased CT-PCI and increased risk of OC residual disease at surgery [6, 16] and impaired progression free survival [18], but a mixed association with overall survival [7, 18]. Rosendahl et al. concluded on surgical PCI, that selected regions rather than the full PCI was more predictive of a favorable prognosis [8]. From a radiological perspective, that would also be more feasible from a workflow perspective. This study shows that our novel CT short score on high-risk image parameters (CT-PCI and CPLN) performs even better than CT-PCI in terms of prognosis, and this has to the best of our knowledge not previously been studied. Another image factor of interest, although not available in this cohort, is CT-ascites, and previous work from our group has shown CT-ascites more than 1000 ml to be associated with high surgical PCI and residual disease [16]. Taken together, the fact that both CT-PCI and short score was strongly associated with advanced OC clinical stages is of high clinical relevance. It gives the possibility to identify patients, before surgery, with high tumor load and therefore high risk of incomplete cytoreductive surgery, that being known as the strongest prognostic factor in OC. Further studies with prospective analyses of CT-PCI, CT short score, CT-ascites, and clinical outcomes are however needed prior to a potential clinical implementation.

An aggressive cancer generates an aggressive CT image with high-risk image factors. Specific base line characteristics potentially leading up to such an image, or other adverse outcome [9], are worth highlighting in order to increase the understanding of the underlying biology illustrated in the image. Even if not modifiable, these factors may aid in creating radiological risk profiles, where images may be interpreted differently or with higher suspicion based on a woman’s risk profile. There are several established OC risk factors with family history being the most important [19], and with oral contraceptives having a protective role [20]. Regarding parity patterns, nulliparity [21] and infertility [22] have been shown to increase the risk of OC, but data on BMI and OC risk varies with menopausal status. However, there are no previous studies on OC risk factors and image findings. With a rather explorative approach, we detected three base line factors related to parity and menopausal status, possibly connected to CT-PCI and short score, but no association held true in adjusted analyses. There may be several reasons for this, but one limitation is the long time span (i.e., between study inclusion/risk factor collection and later OC diagnosis) wherein several factors may interact, and another reason may be the limited sample size. Larger studies are needed in order to identify differences in risk factor associations between the different image patterns (aggressive vs less aggressive), which in turn could improve our understanding of ovarian carcinogenesis.

Some methodological issues require consideration. The established CT-PCI score has been shown with high intra- and inter-observer reliability [3] and the CT short score can be seen as an abbreviated variant. In this present study, only one radiologist (and at one time point) interpreted the images, which is a shortcoming. However, a subset of 20 cases with abdominal CT images was re-read and scored (CT short score) by a second radiologist, showing high correlation between readers which we believe is a promising finding. However, to fully test the important clinical aspect of reliability between radiologists, two readers interpreting all cases independently is of the essence. At our institution, a prospective study analyzing CT short score and surgical short score is ongoing (ten women with ovarian cancer enrolled up until now), and a second radiologist reading all cases is planned for. We hope that such study will allow for further in-depth analyses regarding CT short score feasibility.

This is a retrospective study including cases over a long time span, and information on potential neoadjuvant therapy, surgical method, and surgical outcome was not part of the gathered clinical data. Adequate surgical cytoreduction is the most important independent factor affecting survival in epithelial OC and the lack of this parameter is especially noticeable. The survival analyses in this study were however adjusted for histological subtype/grade and clinical stage, both factors that strongly contribute to selecting the method of surgery and oncological treatment and thus may be seen as proxy variables for surgical method and outcome.

Our study has several clinical implications ranging from the general perspective to the very practical everyday work. An image-based short score would be an effective (for the radiologist) and comprehensive (for the gynecologist and oncologist) way of communicating important and aggressive image information in the setting of for instance a multidisciplinary conference. As of today, the structured surgical report from the European Society of Gynecological Oncology (ESGO) already includes surgical PCI, so the addition of CT-PCI or short score would be both logical and beneficial. As a future and visionary goal, the image short score could be even sharper if risk factor information could be used interpreting the image in dubious findings, and perhaps the presence or absence of a particular risk factor could help the radiologist deeming a particular finding as pathological or not. Taken together, imaging is central in OC care, and future research is needed.

In conclusion, we have identified strong relationships between an image short score and advanced clinical stages and impaired OC survival. A pragmatic approach (based on CT) to evaluate high risk image findings in OC could reduce the radiologist’s workload and at the same time provide structured reports to surgeons and oncologists involved in OC care.