Neither SUVmax nor MTV measured on serial PET-CT scans adequately captured tumor progression over time. Our results did not support the intuitive concept that HNC tumors increase in size and metabolic activity over time. As we move towards risk-adapted therapy, there will be interest to integrate PET-CT as a risk-stratifying biomarker, and incorporate serial FDG PET-CT scans to assess treatment response. Here, we highlight potential limitations in serial PET-CT reproducibility. The temporal variability we observed from a single institution will only be magnified with PET-CT scans obtained from different institutions. Further investigations involving PET-CT requires strict standardization to elucidate its role as a marker of HNC response.
The literature is sparse on serial PET-CT scan reproducibility; however decreases in SUV over time have been described in other sites. In non-small cell lung cancer, upwards of 25% of tumors can have decreased SUVmax
between PET-CT scans (17
), and factors such as tumor necrosis, proximity to lung consolidation, and respiratory motion could account for the SUV reduction.
Several technical or procedural factors have been shown to affect the results of FDG imaging (18
). Technical factors include resolution variability, integrity and stability of different PET scanners, FDG dose injected, timing between injection and scanning, attenuation correction algorithm employed, image analysis software, and region of interest (ROI) determination. At least two guidelines for standardized PET acquisition and analysis have been proposed in attempt to minimize the impact of technical factors (18
). Both guidelines stress the importance of image reconstruction and attenuation consistency and the need to account for artifacts in imaging. Because PET camera specifications can be variable, Shankar et al also recommends that patients should be scanned using the same scanner or the same scanner model, ideally at the same center. This could not be achieved in our study as most patients had staging PET-CT scans upon presentation to our clinic. However, PET-CT scanners at our institution are routinely calibrated and typically differ by an average of 5 – 7%, similar to studies comparing inter-scan variability (20
Our attempts to correct SUV and MTV velocity to account for differences in baseline SUV and the instability in SUVmax failed. While the corrected SUV velocity increased (+0.4/week) compared to the uncorrected SUV velocity (−0.1/week), a significant number of patients demonstrated a negative corrected SUV velocity. Similarly, our attempt to correct MTV fared no better. Our methods to account for technical factors, which likely affected baseline SUV and SUVmax in serial scans, proved insufficient.
Another potential source of error in comparing serial PET-CT scans is the subjectivity of ROI delineation. Shankar et al recommends using a thresholding or edge finding algorithm to eliminate such subjectivity (18
). Likewise, with MTV investigators only have to identify the tumor and involved nodes, and the remainder is computer generated. In fact, we found excellent inter-observer MTV reproducibility, which suggests human bias unlikely accounts for the observed variability.
Several patient-related factors could influence FDG uptake and contribute to our observed results. Impaired glucose metabolism or diabetes can affect FDG uptake and decrease PET sensitivity in detecting smaller lesions (22
). Patients with a large volume of body fat may alter FDG uptake in normal tissues and the tumor (25
). Other factors that can influence FDG bio-distribution include steroids, sedative use to induce muscle relaxation during scanning, and the patient’s hydration status (18
Finally, several tumor-related factors could affect SUV or MTV. Tumor necrosis, cystic nodes, or inflammatory changes in adjacent normal tissue all could influence FDG uptake (27
). These changes could be triggered by tumor biopsy, or dental extractions. Tumor necrosis or cystic nodes could lead to decreased FDG uptake, and thus decreased SUV or MTV. Peri-tumoral inflammation on the other hand could cause increased peri-tumoral SUV, which would lead to MTV overestimation.
Despite the variability in our results, we still observed a significant correlation between primary tumor MTV velocity and disease progression (HR: 2.94, p <0.01), cancer-specific survival (HR=2.43; p=0.004), and overall survival (HR: 1.85, p 0.03). The observation that primary tumor MTV velocity, and not nodal MTV velocity, drives the correlation deserves further attention. While an explanation of this observation remains unclear, it lends support to the hypothesis that primary tumor burden, not nodal tumor burden, predicts overall disease progression. Perhaps the prognostic utility of nodal involvement is better captured with the classic nodal staging system that incorporates size, number, and location of involved nodes. MTV fails to capture number and location of involved lymph nodes. To study this further, we are currently conducting a separate independent study analyzing the prognostic utility of primary tumor MTV versus nodal MTV.
In summary, our data shows significant variability in SUV and MTV from serial PET-CT scans in the same patient at two different time points without any therapeutic intervention. This variability could be mistakenly interpreted as tumor regression or tumor progression in cases where PET-CT is used to assess therapeutic response. This study highlights the challenges of incorporating PET/CT into clinical protocols, and into clinical practice. Prospective studies with standardized protocols on well-calibrated scanners are needed to define the role of serial FDG PET-CT in head-and-neck cancer.