Interobserver agreement of ¹⁸F-Fluorodeoxyglucose Positron-Emission Tomography combined with low-dose Computed Tomography for occult cancer screening in patients with unprovoked venous thromboembolism.

Venous thromboembolism (VTE) can be the earliest sign of cancer. The risk for occult cancer diagnosis in patients with unprovoked VTE is substantially elevated during the first six months of follow-up and declines rapidly thereafter (1). Identifying occult cancers at the time of VTE diagnosis may lead to significant improvement of patients’ care since the cancer may be at a curable stage and early cancer treatment might prevent cancer associated morbidity. This is also an issue of utmost importance for patients, who want to know if an underlying cancer might have triggered the VTE.

The most promising diagnostic modality for extensive occult cancer screening seems to be the ¹⁸F-Fluorodeoxyglucose Positron-Emission Tomography combined with Computed Tomography (FDG-PET/CT). This diagnostic modality is routinely used for the diagnosis, staging and restaging of many cancers (2). It has the advantage of providing non-invasive molecular whole-body imaging. In 2016, our group published in the results of the MVTEP trial which assessed the superiority of an extensive screening strategy comprising of limited screening strategy plus FDG-PET/CT, over a limited screening strategy in unselected patients with unprovoked VTE. Although the MVTEP trial did not show that FDG-PET/CT detected a higher rate of occult cancers at screening, the incidence of missed cancers was significantly lower in patients randomized in the FDG-PET/CT group (3).

Image interpretation of FDG-PET/CT for cancer screening is challenging. Indeed, any cancer type may be diagnosed in patients with VTE. Standardized interpretation criteria have been defined and validated only for a small subset of tumors (e.g. lung nodules, lymphoma) and do not address every possible situation. As such, in the MVTEP study, we used the following definition for a positive FDG-PET/CT: foci of non-physiological FDG uptake, not attributable to the acute VTE (i.e. areas of vascular thrombosis or lung infarcts), with no predetermined cut-off maximum standardized uptake values (SUVmax). FDG-PET/CT was interpreted as negative when physiological FDG uptake or non-physiological FDG uptake attributable to a typical benign disease was observed. FDG-PET/CT readings not falling into any of these categories were classified as equivocal (3). However, how reproducible is this criterion has not yet been determined. To fill this knowledge gap, the aim of this study was to assess the interobserver agreement in the interpretation of FDG-PET/CT for occult cancer diagnosis in patients with unprovoked VTE.

We analysed patients from the FDG-PET/CT group of the MVTEP trial (NCT00964275) (3). For the purpose of this analysis, all FDG-PET/CT scans were centrally reinterpreted a posteriori by two independent board-certified nuclear medicine physicians without knowledge of patients’ clinical information, initial FDG-PET/CT interpretation, and follow-up status. Reader 1 (R1, PYS) was located at the main study site (Brest University Hospital, Brest, France); Reader 2 (R2, RKG) was located in New-York (Memorial Sloan Kettering Cancer Center, NYC, USA). As previously described, FDG-PET/CT were interpreted qualitatively into positive, negative or equivocal categories. Interobserver variability was assessed by calculating kappa (κ) coefficient. We a priori planned 2 analyses: one comparing FDG-PET/CT results with the 3 different interpretations using a linear weighted kappa coefficient; the other grouping positive and equivocal results and assessing agreement for positive or equivocal vs. negative results.

Between March 3, 2009, and August 18, 2012, 200 patients were randomized to the FDG-PET/CT group. Three patients withdrew consent, two patients were found to be ineligible for the trial, and 23 patients did not present for their FDG-PET/CT or refused to undergo the examination (4). Out of the 172 patients analysed, 10 (5.8%) patients were diagnosed with cancer at inclusion (n=9) or during the 24-month follow up period (n=1).

There was agreement on the FDG-PET/CT interpretation in 148 of 172 patients in trinary interpretation (i.e. positive or equivocal or negative). Discrepancy was observed in 24 patients, of them 3 were diagnosed with cancer (table 1). The κ value of agreement among all 3 categories was 0.59 (95%CI 0.44 – 0.74), indicating moderate agreement. Examples of agreements and discrepancies between readers are shown Figure 1. Grouping positive and equivocal results (i.e binary interpretation), there was agreement on the FDG-PET/CT interpretation in 158 of 172 patients, and discrepancy was found in 14 patients. The κ value of agreement grouping positive and equivocal categories was 0.75 (95%CI 0.62 – 0.87), indicating good agreement.

TABLE 1: Readers’ FDG-PET/CT interpretation.

Six FDG-PET/CT were interpreted as positive by both readers, a cancer was found in 5 patients (colon, lung/pleura, oropharynx, ovary/uterus, testicle). Twelve FDG-PET/CT were interpreted as equivocal by both readers, and a cancer was found in 1 patient (prostate cancer). One patient with a negative FDG-PET/CT was diagnosed with colon cancer 18 months after inclusion. Seven FDG-PET/CTs were interpreted as equivocal by R1 and as positive by R2, one cancer was found (prostate cancer). Three FDG-PET/CT were interpreted as positive by R1 and as equivocal by R2, one cancer was found (pancreas cancer). One FDG-PET/CT was interpreted as positive by R1 and as negative by R2, one cancer was found (prostate cancer).

R2	Positive	Equivocal	Negative	Total
R1
Positive	6*****	3*	1*	10
Equivocal	7*	12*	4	23
Negative	3	6	130*	139
Total	16	21	135	172

^*indicates cancer diagnosis. R1: reader 1; R2: reader 2.

Figure 1: Examples of agreements and discrepancies between readers in four patients.

A. Axial CT (1), FDG-PET (2), fused FDG-PET/CT (3) through the oropharynx: R1 and R2 = positive FDG-PET/CT. Patient was diagnosed with an uvula cancer; B. Axial slices through the abdomen: R1 and R2 = positive FDG-PET/CT. Patient was diagnosed with colon cancer; C. Axial slices through the abdomen: R1 = equivocal FDG-PET/CT; R2 = positive FDG-PET/CT. Patient was diagnosed with gallstone without complication; D. Axial slices through the pelvis: R1 = positive FDG-PET/CT; R2 = equivocal FDG-PET/CT. Patient was diagnosed with diverticulosis. (R1: reader 1; R2: reader 2).

PET/CT interpretation in the setting of occult cancer screening is challenging. This is particularly true for patients with unprovoked VTE in whom any type of tumor can be diagnosed. Moreover, the acute VTE might be responsible for non-malignant uptakes, confounding further the interpretation (5). Our results showed that agreement between nuclear medicine physicians when using the 3 classes interpretation (i.e positive, negative or equivocal) was only moderate κ value =0.59, 95%CI 0.44 – 0.74). However, there were only few extreme discrepancies. Most discrepancies were cases in which both physicians have identified the same region of interest but classified it differently. Reassuringly, agreement was better when grouping equivocal and positive results κ value =0.75, 95%CI 0.62 – 0.87). If the threshold between positive and negative FDG-PET/CT may be obvious in case of FDG-avid tumors, differentiation between positive and equivocal FDG-PET/CT is challenging. Some nuclear medicine physicians might consider any uptake as positive or at least equivocal findings in order not to miss a possible cancer, whereas other physicians may restrict positivity to clearly positive FDG-PET/CT (e.g. intense focal uptake associated with morphologic abnormalities on CT). On the one hand, restricting positivity only to patients with a positive FDG-PET/CT for possible malignancy results in higher specificity and positive predictive value (PPV) at the expense of a lower sensitivity (4), but exposes patients and clinicians to the risk of misdiagnosing some low FDG-avid cancer. On the other hand, considering positivity in all patients deemed as having a positive or equivocal FDG-PET/CT results in higher rate of false positive findings. However, despite the lack of detailed interpretation criteria, the overall PPV of a positive or equivocal FDG-PET/CT appears to be high enough, 27% in a previous analysis from our study (4), to deserve thorough investigation. From a clinical management perspective, the objective is to stratify patients into a group with a very low risk of cancer who may be safely reassured on one hand, and a group with a significant risk of cancer who deserve further investigations on the other hand. Therefore, the good reproducibility of FDG-PET/CT interpretation when using a binary interpretation – negative vs. equivocal/positive – supports the reliability of the test in this indication.

The use of simple, easy, standardized organ specific objective interpretation criteria would be ideal if available, and would ensure optimal interobserver reproducibility. Quantitative metabolic parameters such as SUVmax have been shown to be able to differentiate malignant versus benign tissue (6). However, a number of physiologic FDG uptakes variants are commonly encountered, including normal physiologic uptake in the head and neck, myocardium, breast, liver, spleen, gastrointestinal tract, genital system, urinary collecting system, bone marrow, or muscles (7). Moreover, benign nonphysiologic lesions with increased FDG uptake are also frequently seen and can be misinterpreted as malignancies (8). Given that patients with unprovoked VTE may develop any cancer type (including non FDG avid tumor), location or stage (i.e early or localized), defining dedicated SUVmax cut-off in this particular setting might not be possible. Another way to improve specificity of PET/CT reporting might be to better characterize lesions on the CT portion of the PET/CT study, especially for lesions with moderate or high FDG uptake (9). Reproducibility could be further improved by establishing a lexicon of diagnostic certainty (10).

In summary, the interobserver agreement of FDG-PET/CT for occult cancer screening in patients with unprovoked VTE was good in binary interpretation supporting the reliability of the test in this indication.