In this study with patients with neuroendocrine tumors grade I-II treated with several classes of tumor targeted treatments, we did not find any statistically differences in TTP between those who achieved OR and those who achieved SD. There was, however, a trend towards shorter TTP for those who achieved OR. We have not been able to find any other study that systematically compare the TTP in response groups (OR or SD) in tumor targeted therapies. We found no differences in patient- or tumor characteristics that separated the two response groups. Age, Ki 67%, site of primary tumor location and stage were comparable within the two response groups. This may indicate that there could be other biological factors than Ki67% and the known behavior of the different primaries that could influence both the response and the duration of the response to tumor targeted treatment modalities.
The strength of this study was that all patients were well characterized, treated in a single center and none of the patients were lost to follow up. The study has, however, several limitations. The numbers of patients in each group were few and the power to detect differences between the response groups low. Some patients were included in more than one group, i.e. 83% in the everolimus group had previously been treated with stz/5-FU or PRRT, and 38% had received both treatments. This might lead to a selection bias reproducing the same pattern with those with SD tending to have longer TTP for the different treatment modalities studied. Still, this did not alter the main observation that the patient and tumor characteristics recorded could not explain why those who achieved OR did not obtain a longer TTF, but rather a tendency towards a shorter TTP. Differences in tumor grade is theoretically the most plausible explanation for any differences in TTP between the groups. We found no such difference between the groups with regards to the proliferation marker Ki67% although this could be due to the rather limited number of patients. We cannot know for sure whether the recorded Ki-67% estimates are representative for each of the patients. We know that there is significant intratumor heterogeneity [
10] and that there are differences in Ki-67% between primaries and metastases [
11]. We only have one to three Ki-67% estimates from each patient, and with disseminated disease, this estimate might not be representative for their disease. Some of the Ki-67% assessments were performed by less experienced pathologists and not all samples were reexamined by our institution’s pathologists specialized in neuroendocrine neoplasms. We do not believe, however, that occasional suboptimal evaluation of the proliferation index would systematically bias the assessment, but tend to both over- and underestimate the Ki67%, probably at the same extent.
The method used to assess radiological response in this study is both a strength and a limitation. The most widely used radiological response criteria for radiological response evaluation in treatment studies on neuroendocrine tumors are the Response Evaluation Criteria In Solid Tumors (RECIST 1.0 or 1.1) [
12] the Southwest Oncology Group standard response criteria (SWOG) [
13] and the WHO criteria [
14]. These criteria were introduced to evaluate the effect of chemotherapy on tumor burden and are based on measuring the diameter of predefined target lesions as well as detection of any new lesions. In the RECIST-criteria, the most widely used assessment method, the diameters of the target lesions are added, and an increase from the start of treatment, or after initial therapy-induced tumor shrinkage, of 20% or more is defined as progressive disease. A reduction of 30% or more is defined as an objective response. Any change between 20% increase and 30% reduction is classified as stable disease. If new lesions emerge, or if preexisting non-target lesions grow, even if there is no change in the target lesions, the patient has progressive disease. RECIST is far from optimal for evaluating treatment response in slow-growing malignancies such as neuroendocrine tumors [
15]. We have previously shown that assessing treatment response with RECIST gives an unrealistic positive impression of the treatment effect compared to assessing the treatment response with our “conventional method” [
9]. The treatment response in the SD-group based on these criteria varies from a 19% increase to 29% decrease in added target lesion diameter. The heterogeneity in this group restricts our possibility to detect clinically interesting features as demonstrated in our study where the survival curves for those treated with PRRT overlaps when the RECIST criteria are used and diverges when evaluated with the “conventional method” (where any unequivocal change was regarded significant). The “conventional method” is, however, far from optimal. It lacks standardization and it is based on one or two radiologist’s overall impression of the tumor status. It is therefore not suitable in treatment trials or for reproducing results made by other investigators. Our results indicate, however, that radiological response evaluation systems that are more sensitive to response changes in neuroendocrine tumors are highly needed. With the high resolution of today’s radiological procedures one could argue that the thresholds used for classifying the different overall response groups in RECIST could be redefined. For example, 5% increase in the sum of diameters of target lesions instead of 20% could define progressive disease. Decreased tumor density as an effect of treatment secondary to tumor necrosis is not taken into account in the above mentioned response evaluation systems. Sometimes reduction in tumor viability, recognized as changes in contrast uptake, is the only initial sign of treatment effect. To meet these challenges in response evaluation irRECIST has been introduced [
16]. The Choi criteria combine density and size with a lower size threshold than RECIST, and has been proposed for use in response evaluation for neuroendocrine tumors [
17]. It has been shown to be more accurate compared to RECIST in a trial of sunitinib for gastroenteropancreatic neuroendocrine tumors [
18]. Other response evaluation systems incorporating density and size as mRECIST, Chun and MASS [
19‐
21] have so far not been validated for evaluation of treatment effect in neuroendocrine tumors. In the future we will probably also see that artificial intelligence with its ability to detect and interpret minor changes in size and density in the CT, MRI and PET examinations will be used in routine evaluations [
22].