Clinical and scientific value in the pursuit of quantification of beta cells in the pancreas by PET imaging

To the Editor:

The commentary by Alavi and Werner [1] mentioning our recent paper [2] concludes that ‘it is impossible to visualise and characterise islet cell function and structure with [positron emission tomography] PET’. First, and perhaps it is a matter of misunderstanding, we are not attempting to ‘visualise and characterise islet cell function and structure with PET’. Our objective is limited to determining clinically relevant changes in pancreatic beta cell mass and, in our studies, PET is neither used to visualise islet cell function nor structure. Rather, our use of PET provides an integrated measure of the number of beta cells present in a defined volume of the pancreas. That is, we are measuring beta cell density within the pancreas, which we conclude is clearly an achievable goal even in the face of the biological and technical concerns raised again by Alavi and colleagues, as discussed below:

1. Dispersal of islets throughout the pancreas and relative volume of islets

Dr Alavi has previously raised the concern that PET cannot be used to determine beta cell mass due to the small volume of beta cells within the pancreas and their dispersal throughout the pancreas [1, 3]. We believe that these issues were exceptionally well addressed by Ichise and Harris in a 2011 response [4] to a previous letter to the Editor from Alavi and colleagues [3]. In that response, quantitative imaging of vesicular monoamine transporter type 2 (VMAT2) in beta cells was compared with the measurement of neuroreceptors and transporters in the striatum [4]. Like pancreatic beta cells, dopaminergic terminals occupy a small fraction of the volume of the striatum. Nevertheless, for over 30 years, PET imaging has succeeded to easily image these targets [5]. Further, with the development of higher affinity ligands, which produce a higher contrast, PET has evolved to measure dopamine receptors in cortical regions [6], where the density of dopaminergic terminals (per unit volume) is much lower. Thus, there is clear evidence from 30 years of PET imaging that specific imaging signals representing specific protein targets can be achieved with ligands with appropriately high affinity.

2. Spatial resolution and partial volume effects

It is well known that PET has finite resolution, based on the finite detector size and post-reconstruction filtering. In our study, we minimised these effects in several ways: (1) by using a high-resolution scanner; (2) by using point-spread function reconstruction to minimise the partial volume effect; and (3) by using thin regions-of-interest to minimise resolution effects (see ESM Fig. 1 in [2]).

3. Motion correction

In PET scans, respiratory and bulk body motion can degrade PET images. In our work, we performed dynamic body motion correction to eliminate the latter issue. With regard to respiratory motion correction, while not applied in our study [1], we and others are developing novel methods for event-by-event respiratory motion correction using external devices or with data-driven techniques [7-9]. These methods will become critical for accurate quantification using dynamic PET in the body.

4. Loss of pancreatic volume

Volume loss clearly affects all PET quantification methods, so combined measurement of volume loss using anatomical imaging with functional/pharmacological loss using PET is essential. For example, in type 1 diabetes, the effective assessment of beta cell loss might best be summarised using the combined loss of volume and beta cells per unit volume [10]. However, in our work in type 2 diabetes mellitus, the loss of volume was not statistically significant and is not likely to explain the correlations found between PET measurements and disease-specific indices (see Fig. 4 in [1]).

5. Optimal contrast

With regard to the analysis that deduced that a radioligand contrast ratio of 1000:1 is needed to visualise beta cells in the native pancreas [11], we believe that a more appropriate question is: what degree of change in beta cell mass can be detected given achievable test-retest variability and measured radioligand specificity? In other words, this is a question of signal:noise ratio. In knowing the test-retest coefficient of variation, one can calculate the change in beta cell mass needed to be detectable by PET for a radioligand of known specificity. With a test-retest variability of <10%, which has been obtained by us and others [2, 12], a ~25% change in beta cell mass would be detectable using a radioligand with a specificity of ~35:1, while a ~15% change in beta cell mass would be detectable for a radioligand with a specificity of ~100:1, such as [¹⁸F]fluoropropyl-(+)-dihydrotetrabenazine (¹⁸F-FP-(+)- DTBZ). The clinical value of PET imaging of beta cell mass is not in determining absolute amounts, which are highly variable in both healthy people and those with diabetes [13-15], but in detecting changes in beta cell mass in response to disease progression or therapeutic intervention. We conclude that measuring longitudinal changes in beta cell mass is currently achievable with the proven test-retest variability and the specificity of existing PET and single photon emission computed tomography (SPECT) radioligands.

6. Non-specific binding

With all PET and SPECT imaging agents that bind to molecular targets, the signal not only consists of specific binding to the target, but also non-specific binding to surrounding tissue. We have clearly recognised the importance of defining an accurate reference region and have utilised several independent strategies, including blocking studies and displacement studies in animal models [16-18]. More specifically, we use PET studies of the inactive enantiomer of ¹⁸F-FP-DTBZ to develop and validate a suitable reference method to correct for non-specific binding in the pancreatic tissue [16, 17]. Thus, our use of the spleen as the reference tissue [1] is not based on speculation but has been validated in both nonhuman primates and humans.

Yes, there are hurdles to overcome in achieving the goal of using PET and SPECT (and magnetic resonance imaging [MRI]) imaging as viable research and clinical tools to assess changes in beta cell mass that may occur with type 1 and type 2 diabetes. Faced with these challenges, one could declare the hurdles as insurmountable and declare the task futile, as Alavi and Werner propose [1]. Alternatively, one can take the more productive and forward-looking approach, as we and other independent research teams are doing [12, 14, 19–23], and develop solutions. We have many reasons to take this optimistic approach. Looking back over the 40 year history of PET imaging, there has been huge improvement in PET scanner performance (from single-slice machines with a resolution of ~2 cm to whole body systems with time-of-flight and resolutions better than 5 mm) and constant development of highly specific and sensitive radiopharmaceuticals. In light of this amazing scientific progress, imaging of beta cells with PET is by no means a ‘futile’ endeavour.

Abbreviations

[¹⁸F]-FP-(+)-DTBZ

[¹⁸F]fluoropropyl-(+)-dihydrotetrabenazine

PET

Positron emission tomography

SPECT

Single photon emission computed tomography