Long-term outcomes and recurrence pattern of 18F-FDG PET-CT complete metabolic response in the first-line treatment of metastatic colorectal cancer: a lesion-based and patient-based analysis

Colorectal cancer is the third commonest cancer worldwide with nearly 1.4 million new cases diagnosed annually. [1] Nearly 20% of patients are diagnosed late and curative surgery is not possible due to extensive metastatic disease [2]. For this group of patients, palliative chemotherapy with or without biological agents are recommended. [3, 4] As the aims of treatment in these patients are to prolong survival and improve quality of life, it is prudent to identify responders from non-responders to avoid futile treatments and excessive toxicities.

Currently, contrast enhanced computed tomography (CT) is the imaging modality of choice and the Response Evaluation Criteria In Solid Tumors (RECIST) is commonly applied in assessing treatment response. [5] However, as RECIST is based on anatomical changes in size of the tumor, its correlation with morphological alterations and patient outcomes, particularly in the era of novel combination chemotherapy and biological agents are unclear. [6]

¹⁸F-2-fluoro-2-deoxy-D-glucose (¹⁸F-FDG) positron emission tomography-computed tomography (PET-CT) is an imaging modality that measures and quantifies metabolic avidity in cancer cells, thus serving as a proxy for the underlying cellular activity and viability. [7] It is widely used in the diagnosis, prognostication and treatment response in many cancers although its use in colorectal cancer remains controversial. [8] Previous studies have shown its sensitivity in detecting treatment response and those with metabolic response (mR) might have improved clinical outcome. [9‐11]

Complete metabolic response (CMR) is a distinct clinical entity and it is increasingly encountered with the advent of systemic therapy in patients with metastatic colorectal cancer (mCRC). However, little is known whether achieving CMR, with or without RECIST response, confers survival advantage or merely represents a transient quiescence in FDG-uptake with limited clinical impact. To study the clinical significance of CMR in mCRC, we performed a retrospective analysis using both lesion-based and patient-based approaches on consecutive patients in a prospectively-maintained database.

Metabolic response to systemic therapy is a well-recognized prognostic marker for improved survival but the clinical impact of achieving CMR in patients with mCRC remains to be elucidated [11]. To our knowledge, this represented the largest cohort of CMR reported in the literature with uniform imaging technique and mature follow-up data (median follow-up time of 47 months) that highlighted the long-term outcomes and recurrence patterns. The PFS and OS of 15.6 and 44.6 months, respectively, compared favorably to contemporary literature in the biomarker-selected population. [13‐15]

Despite significant improvement in survival, there remains a discrepancy between achieving CMR on PET-CT and complete eradication of disease. Normalization of ¹⁸F-FDG uptake in metastases after chemotherapy has been described as CMR but the long term outcomes and recurrence pattern of these lesions remain unclear. [16‐19] Our results have shown that sustained CMR is possible in 40% of CMR patients and those who achieved sustained CMR have significantly prolonged survival. The outcomes did not appear to be the result of post-CMR or maintenance therapy as the median number of cycles of systemic therapy were similar for patients with sustained CMR and PMD (Table 1).

In accordance with our study, a previous study has also shown that pre-treatment SUVmax of the main metastatic lesion was associated with OS [20]. Furthermore, we have shown that high SUVmax (> 4.4) of individual lesion was predictive of subsequent PMD while sustained CMR was more frequent in individual lesions with RECIST CR on a lesion-based analysis. Thus, initial SUVmax and the corresponding RECIST response of individual lesion may guide decisions for radical local therapy e.g. resection or stereotactic body radiotherapy, on a lesion-by-lesion basis after achieving CMR with systemic therapy. Further research is required to decipher the relationship between tumor metabolism and sensitivity to systemic treatment, taking into account of other established PET parameters not analyzed in this study such as SUVmean, metabolic tumor volume and total glycolytic volume. [21]

In our cohort, there was considerable discordance between PET-based and CT-based response evaluation with 45% of patients considered PR/SD only on CT. This was consistent with previous studies by Monteil et al. and Skougaard et al. whereby better overall metabolic responses on PET-CT were seen than best overall response on CT. [22, 23] We have analyzed the outcomes of these two groups and shown no statistical difference but the small number of patients did not allow a definitive conclusion. Nevertheless, this suggested that PET-CT was complementary, if not superior, to CT-based analysis for identifying a subgroup of patients that may benefit substantially from systemic treatment alone. A caveat to this finding was that, upon lesion-based analysis, metastasis which only achieved PR on CT are statistically more likely to have PMD. PMD on previously indexed locations implied residual tumor although they did not affect OS in our cohort as 31/41 (75.6%) of lesions identified on PET-CT at the time of subsequent PMD per patient-based analysis were new lesions. It was beyond the scope of this retrospective study to address whether early intervention to those RECIST PR lesions in the presence of CMR would effectively prevent subsequent systemic progression. Furthermore, our results supported the current understanding of mCRC as a systemic disease and aggressive local therapy in the presence of initial widespread disease have to be highly-selective despite CMR, especially in the presence of high initial SUVmax and RECIST non-CR of individual lesion. On the other hand, durable disease control was more likely in those lesions with low baseline SUVmax on PET-CT and RECIST CR after systemic therapy alone, and they might be spared of more aggressive therapy.

Imaging biomarkers aside, we have also demonstrated the prognostic value of serum CEA levels. Pre-operative CEA has previously been shown to be an independent prognostic factor for outcomes [24, 25] and in our cohort, patients with normal serum CEA levels at diagnosis have significantly longer PFS, OS and were more likely to achieve sustained CMR compared to those with a raised serum CEA. Furthermore, normalization of CEA was seen in the majority of patients when CMR was achieved and raised again in those who had subsequent PMD. This added credence to its use for comprehensive treatment response assessment and detection of recurrence in those patients who have already achieved CMR and might only require a less intensive schedule for PET/CT scan. Radiological imaging could be better scheduled when a rising CEA trend is established thus reducing ionizing radiation to the patients and costs to the healthcare system.

There were known limitations intrinsic to a retrospective study, however, due to unpredictability and relatively infrequent occurrence of CMR in mCRC prospective recruitment or randomization of post-CMR treatment of these patients is not feasible.

To minimize selection bias, only patients treated with fluoropyrimidine-based systemic therapy in the first-line setting were included, although some factors that have shown prognostic values such as serum CA 19–9 or LDH level were not mandatory in our prospectively-maintained database. [26‐28] Nevertheless, these factors were less relevant to the current study as it was not the intention to derive predictive factors for whom CMR could be achieved, rather the key message was to describe the natural history of patients and lesions with CMR and thus their potential treatment implications. Due to the small size of our cohort as a result of infrequent occurrence of CMR, multivariate analysis was not deemed feasible and that subtle correlations between potential prognostic factors and outcomes may not be demonstrated. Follow-up schedule and arrangement of PET/CT in the real-world setting might introduce bias to the estimation of time-to-event endpoints. The unexpected finding that patients on biological therapy were less likely to achieve sustained CMR may be due to selection bias in this small cohort of patients. Variation in tumor loads and difference in tumor biology for those who responded well to chemotherapy alone vs those who required additional biological therapy might be the underlying reasons.

Despite all these, the uniform assessment methodology applied and the evaluation of only patients with CMR should have improved inter- and intra-observer variability in interpreting the PET reports. Although previous study has shown significant variability in quantifying PET parameters, this was thought to be due to differences in method of attenuation correction and variation in imaging protocol [29]. As all patients were scanned using the same protocol and both scanners employed the same attenuation correction methods, we believe the threshold SUV max value calculated in predicting subsequent PMD is reproducible. The mature results so generated with long-term follow-up of 47 months was also reflective of real-world clinical practice. In summary, our results provided additional information on the long-term outcomes and pattern of recurrence in this distinct subgroup of patients with potential treatment implications.

Our study showed that patients who achieved CMR on ¹⁸F-FDG PET-CT have improved clinical outcomes. Although many of them subsequently develop PMD, sustained CMR with systemic therapy was achievable especially with low baseline SUVmax of individual lesion, normal baseline serum CEA as well as RECIST CR at the time of CMR. Discordance was seen between morphological and metabolic treatment response but the two groups had comparable outcomes and we believe PET/CT, especially in those in which aggressive local therapy is contemplated, has a complementary role to cross-sectional imaging in prognostication, treatment monitoring and planning.