Evaluating infusion methods and simplified quantification of synaptic density in vivo with [¹¹C]UCB-J and [¹⁸F]SynVesT-1 PET

Abstract

For some positron emission tomography studies, radiotracer is administered as bolus plus continuous infusion (B/I) to achieve a state of equilibrium. This approach can reduce scanning time and simplify data analysis; however, the method must be validated and optimized for each tracer. This study aimed to validate a B/I method for in vivo quantification of synaptic density using radiotracers which target the synaptic vesicle glycoprotein 2 A: [¹¹C]UCB-J and [¹⁸F]SynVesT-1. Observed mean standardized uptake values (SUV) in target tissue relative to that in plasma (C_T/C_P) or a reference tissue (SUVR-1) were calculated for 30-minute intervals across 120 or 150-minute dynamic scans and compared against one-tissue compartment (1TC) model estimates of volume of distribution (V_T) and binding potential (BP_ND), respectively. We were unable to reliably achieve a state of equilibrium with [¹¹C]UCB-J, and all 30-minute windows yielded overly large bias and/or variability for C_T/C_P and SUVR-1. With [¹⁸F]SynVesT-1, a 30-minute scan 90–120 minutes post-injection yielded C_T/C_P and SUVR-1 values that estimated their respective kinetic parameter with sufficient accuracy and precision (within 7 $\pm$ 6%) . This B/I approach allows a clinically feasible scan at equilibrium with potentially better accuracy than a static scan SUVR following a bolus injection.

Introduction

The synaptic vesicle glycoprotein 2 A (SV2A) is a highly conserved protein expressed ubiquitously and homogeneously in presynaptic vesicles throughout the brain. ¹ Given this expression profile, along with the critical role SV2A plays in mediating synaptic vesicle trafficking, neurotransmitter release, and overall synaptic function, ² positron emission tomography (PET) using radioligands that bind to SV2A provide in vivo measures of synaptic vesicle density, and by proxy, a surrogate for presynaptic terminal density. The radioligands [¹¹C]UCB-J and [¹⁸F]SynVesT-1 bind with high affinity and specificity to the synaptic vesicle glycoprotein 2 A (SV2A),^{3 –5} and have proven to be invaluable tools for investigating relationships between synaptic density and brain function within the context of neuropsychiatric and neurodegenerative disease,^{6 –9} as well as synaptic changes associated with typical development and aging.^10,11

Currently, PET studies utilizing [¹¹C]UCB-J and [¹⁸F]SynVesT-1 feature a bolus injection of the radiotracer followed by 60 minutes of dynamic data acquisition and require arterial input function measurements for full dynamic modeling using the one-tissue compartment (1TC) model in order to quantify volume of distribution (V_T).^{3,12 –14} However, this “gold standard” method of quantification for both radiotracers necessitates a relatively long scan duration, which may be prohibitive for some potential study participants or entire populations, and the arterial blood sampling with corresponding metabolite analyses makes this method less feasible and reproducible across study sites.^3,14

As a possible alternative study design, the radiotracer could be delivered as a bolus plus continuous infusion (B/I) to experimentally establish radiotracer equilibrium between tissue and arterial plasma.^15,16 The B/I approach requires the a priori determination of the infusion schedule, which is described by the parameter K_bol, which defines the magnitude of the bolus in units of minutes of infusion. For example, a K_bol value of 60 means the bolus component is equivalent to 60 minutes of infusion. Once it has been established that the B/I paradigm reliably produces a state of equilibrium, the total scan time can be greatly reduced and V_T can be directly calculated as ratio of the radioligand concentration in tissue target region (C_T) to that in arterial plasma (C_P), since V_T =C_T/C_P at equilibrium . Further, if an appropriate reference region is available, specific, or non-displaceable binding potential (BP_ND) can be estimated by the target tissue to reference tissue ratio -1 (SUVR-1).

The appropriate application of the B/I method requires that scanning and parameter estimation is performed during equilibrium and assumes similar behavior of the radioligand in the plasma across the study population such that equilibrium is established in all study subjects in the selected scan-time window. Therefore, the primary aims of the present study were to evaluate the ability of a B/I paradigm to reliably achieve a state of equilibrium for both [¹¹C]UCB-J and [¹⁸F]SynVesT-1, and to then identify time windows post injection that provided the most reliable parameter estimates by calculating C_T/C_P (i.e., apparent V_T) and SUVR-1 (i.e., apparent BP_ND) and comparing them with their 1TC model parameter counterparts. We additionally tested whether B/I ratios were able to identify group differences in regional SV2A density previously identified using full dynamic compartmental modeling. Overall, we hoped to identify and validate a simplified PET method for in vivo measurements of SV2A density which could aid in the ability to conduct larger (or multi-site) studies.

Materials and methods

Radiotracer synthesis

[¹¹C]UCB-J ¹⁷ and [¹⁸F]SynVesT-1(also referred to as [¹⁸F]SDM-8) ⁵ were synthesized as previously described.

Radiotracer injection and data acquisition

Human subjects

All scanning procedures were approved and overseen by the Yale University Human Investigation Committee and the Yale University Radiation Safety Committee in accordance with the United States federal policy for the protection of human research subjects in Title 45 Part 46 of the Code of Federal Regulations (45 CFR 46). Written informed consent was obtained from all subjects prior to participation after complete explanation of study procedures.

In vivo B/I [¹¹C]UCB-J studies

Data were analyzed from 10 [¹¹C]UCB-J scans acquired from 5 male and 5 female healthy subjects (HS; mean age: 31.4 ± 9.5 years). Doses of 431 ± 176 MBq of [¹¹C]UCB-J with high molar activity (162 ± 122 MBq/nmol) were delivered as B/I. A K_bol value of 150 minutes was selected based on simulations previously performed using the regional TACs from the five healthy control subjects evaluated with PET after bolus injection [¹¹C]UCB-J. ³ Imaging data and arterial blood samples (for metabolite-corrected input function calculation) were collected for 120 minutes following the injection, as previously described. ³

[¹⁸F]SynVesT-1 simulations

Determination of the K_bol value for [¹⁸F]SynVesT-1 was based on the analysis of simulated time-activity curves (TACs) which used arterial input functions and kinetic parameters (K₁ and k₂) from bolus injection studies (n = 14). ¹⁴ In brief, using the equation of time response for B/I protocols ¹⁵ , TACs were simulated for the arterial input function, gray matter regions (frontal, temporal, parietal, and occipital cortices; caudate, putamen, thalamus, hippocampus, and cerebellum) and centrum semiovale (C.S.) by varying K_bol from 60 to 200 minutes in 10 minute increments. The ‘optimal’ K_bol value was determined to minimize the %V_T bias of C_T/C_P at 45–60, 75–90, 90–105, and 105–120 minutes after tracer injection.

In vivo B/I [¹⁸F]SynVesT-1 studies

Data were analyzed from 11 [¹⁸F]SynVesT-1 scans acquired from 9 HS (8 males and 1 female) and 2 males with a diagnosis of major depressive disorder (mean age: 54.1 ± 7.7 years). Doses of 177 ± 12 MBq of [¹⁸F]SynVesT-1 with high molar activity (212 ± 90 MBq/nmol) were delivered as B/I with K_bol = 150 minutes. Imaging data and arterial blood samples (for metabolite-corrected input function calculation) were collected for 150 minutes following the injection. ¹⁴

Equilibrium ratio and 1TC parameter comparisons

For all PET scans, TACs, expressed as the standard uptake volume (SUV = [kBq/mL]/[MBq/kg]), were extracted from 10 regions of interest (ROIs) using subject-specific MRIs nonlinear transformed to the AAL MR template ¹⁸ : frontal, parietal, temporal, and occipital cortices; caudate, putamen, thalamus, hippocampus, amygdala, and cerebellum. The metabolite-corrected arterial input functions were generated as previously described and linearly interpolated to the PET frame times. ³ A TAC was also generated for a 2 mL volume of the C.S., a white matter region proposed as a useful estimate of non-displaceable uptake for radioligands targeting SV2A. ¹⁹

The 1TC model was applied to dynamic scan data in order to calculate ‘true’ V_T and BP_ND values for [¹¹C]UCB-J³ and [¹⁸F]SynVesT-1, ¹⁴ with CS used as the reference tissue for BP_ND estimations. For each ROI, the apparent V_T was calculated for 30-minute overlapping time windows as the ratio of average (mean) radiotracer concentrations in tissue (C_T) to metabolite-corrected plasma concentrations (C_P). These regional C_T/C_P values were then used to calculate standardized uptake value ratios (SUVR) using C.S. as the reference region:

$$\frac{C_{\text{T}}/C_{\text{P}}}{C_{\text{CS}}/C_{\text{P}}}-1=\text{SUVR-1}$$

As a measurement of lack of achieving equilibrium, the percent difference between C_T/C_P and SUVR-1 values relative to full 1TC model estimates of V_T and BP_ND, respectively, were calculated such that negative values represent an underestimation and positive values indicate an overestimation of the 1TC model parameters:

$$\left(\frac{C_{\mathrm{T}}/C_{\mathrm{P}}}{V_{\mathrm{T}}}-1\right)\times 100=\mathrm{\%}{V}_{\mathrm{T}};\mathrm{ }or\mathrm{ }\left(\frac{\mathrm{SUVR}-1}{{BP}_{\mathrm{ND}}}-1\right)\times 100=\mathrm{\%}{BP}_{\mathrm{ND}}$$

Statistical analyses

To assess whether a state of apparent equilibrium was reliably achieved by the B/I method for each subject, the six scan frames within each 30-minute window of the TAC for the arterial input function, gray matter regions (frontal, temporal, parietal, and occipital cortices; caudate, putamen, thalamus, hippocampus, and cerebellum) and C.S. were fit to a linear model to calculate slope (SUV/hour) and the slope as the percent of the mean SUV for the given 30-minute timeframe (%SUV/hour). The data were checked for statistical outliers using the Grubb’s extreme studentized deviant method. Date from one of the [¹⁸F]SynVesT-1 studies (healthy comparison, Male, age 45 years), specifically the amygdala %BP_ND values starting at the 80–110 min. window, were found to be significant outliers (two-tailed p < 0.05), and thus excluded from the analysis. Descriptive statistics provided throughout are arithmetic mean ± SD.

Results

Bolus/infusion [¹¹C]UCB-J in healthy control subjects

The average TAC for [¹¹C]UCB-J B/I studies conducted in 10 HS is shown in Figure 1. For each subject, we calculated the slope (SUV/hour) and the slope as a percent of the mean SUV (%SUV/hour) for each 30-minute time window (see Statistical analyses, Table 1). Slopes of the arterial input function were negative, with the smallest SUV slope at 90–120 minutes (−9.3 ± 11.2%SUV/hour). For each individual, an average TAC was calculated for gray matter regions (frontal, parietal, temporal, and occipital cortices; caudate, putamen, hippocampus, thalamus, and cerebellum). This “gray matter region” slope was positive at the 20–50 minute window (0.05 ± 9.9%SUV/hour), with negative slopes for the remainder of the scan, ranging from −4.2 ± 9.0%SUV/hour at 30–60 minutes, to −6.8 ± 7.2 slope as %mean at 80–110 minutes. It is further worth noting that by the 40–70 minute window, the %SUV/hour was negative for all individual regions included within the average “gray matter region”, with the exception of amygdala (which had a positive %SUV/hour until the final 90–120 minute window). This suggests that for most gray matter regions, the chosen K_bol was too large, i.e., the initial bolus overshot the equilibrium value. On average, the C.S. slope was smallest at 70–100 minutes (3.5 ± 30.2%SUV/hour); note that the inter-subject variability was very high, with individual subject slopes ranging from −32.5 to +54.0%SUV/hour. At 60–90 minutes, the average of the C.S. slope was nominally greater, but with substantially less inter-subject variability (−6.9 ± 16.8%SUV/hour), with individual values ranging from −27.9 to +17.7%SUV/hour. Note that statistical noise is a major contributor to the variability of C.S. slope values.

Figure 1.

Population [¹¹C]UCB-J time-activity curves. Standardized uptake values (SUV) with units of the measured radioactivity concentration in the tissue or plasma (arterial input function; kBq/mL) relative to the bodyweight-corrected total injected dose (MBq/kg) for 10 healthy subjects injected with a bolus/infusion of [¹¹C]UCB-J (K_bol = 150). Data are presented as the mean $\pm$ SD. Abbreviations: cortex (Ctx), centrum semiovale (C.S.), metabolite-corrected input function (IF).

Table 1.

Population mean (n = 10) standardized uptake value (SUV) slopes for [¹¹C]UCB-J B/I studies.

	Input function		Gray matter a		Centrum semiovale
Time (min)	Slope as %mean	SD of %mean	Slope as %mean	SD of %mean	Slope as %mean	SD of %mean
40–70	−23.3	7.9	−4.9	7.6	−20.7	16.7
50–80	−18.6	7.5	−5.7	6.1	−20.8	27.3
60–90	−15.3	8	−6.1	7.5	−6.9	16.8
70–100	−13	8.8	−5.3	6.3	3.5	30.2
80–110	−11.2	9.7	−6.8	7.2	14.6	70.5
90–120	−9.3	11.2	−6.6	7.7	−18.4	47.3

Note: Standard deviation (SD) refers to across subject variance.

^aGray matter region values represent the within subject mean TAC for frontal, parietal, temporal, and occipital cortices; caudate, putamen, hippocampus, thalamus, and cerebellum.

For each subject, the ratio of radiotracer concentrations in tissue to metabolite-corrected plasma concentrations (C_T/C_P) was calculated for 30-minute overlapping time windows and are displayed in the supplement (Table S1, top). We found that C_T/C_P ratios continued to slowly increase for the total 120-minute scan duration for eight of ten ROIs, including all four cortical regions (Table S1). As time progressed, C_T/C_P overestimated the true V_T value in all regions. This is expected given the overshoot in equilibrium (Table 1) so that TACs were clearing. ¹⁵ The time window with the smallest mean V_T bias across subjects and ROIs was 20–50 minutes (Figure 2(a)), with misestimation ranging from −18.2 $\pm$ 10.3% (amygdala) to +12.6 $\pm$ 8.9% (thalamus). Of note, there was a considerable amount of variability (SD) in the degree of %V_T misestimation across subjects and ROIs, ranging from 9.3% to 12.6% variability at the 20–50 and 90–120 minute windows, respectively.

Figure 2.

Equilibrium ratios versus 1TC model parameters for [¹¹C]UCB-J. Population bias of [a] C_T/C_P values relative to 1TC model estimates of V_T. and, [b] SUVR-1 values relative to 1TC model estimates of BP_ND from [¹¹C]UCB-J B/I studies (K_bol = 150) performed in 10 healthy control subjects. Bias was calculated as the %1TC parameter such that positive values represent overestimations and negative values represent underestimations of the relevant 1TC parameter. Data are presented as the mean±SD. Abbreviations: centrum semiovale (C.S.).

Observed SUVR-1 values peaked at 60–90 minutes post-injection for all ROIs except the thalamus, which peaked at 50–80 minutes. The SUVR-1 values for all 30-minute windows and ROIs can be found in the supplement (Table S1, bottom). The time window with the smallest mean bias across ROIs and subjects was at 90–120 minutes (Figure 2(b)), with %BP_ND underestimation ranging from −12.0 $\pm$ 8.8% (amygdala) to −1.5 $\pm$ 7.8% (frontal cortex). This time period is later than the 60–90 min period found for bolus injections. ²⁰ Cerebellum was the only ROI where the 90–120 SUVR-1 overestimated BP_ND (8.2 $\pm$ 10.5%). The overall smaller bias in SUVR-1 was due to partial cancellation of the bias between gray matter regions and the C.S. reference region, as is the case with bolus injections. ²⁰ Values for SUVR-1 demonstrated the least variability across subjects and ROIs at 30–60 minutes, where SD of BP_ND bias ranged from 4.7% (amygdala) down to 2.6% (caudate). The greatest variability of %BP_ND was observed at 90–120 minutes, with an across subject and across ROI average SD of 7.7%. Overall, there was no single time period that provided accurate and precise results for both V_T and BP_ND.

[¹⁸F]SynVesT-1 simulations for k_bol selection

Results of simulations predicting C_T/C_P %V_T bias for various K_bol values in gray matter regions are provided in Fig. S1A, and for the C.S. white matter region in Fig. S1B. Higher K_bol values, i.e., larger bolus fractions, yielded higher TAC values and more positive bias. No K_bol value was ideal for both gray and white matter regions. Ultimately, K_bol = 150 min was selected based on the finding that this K_bol value resulted in the smallest %V_T bias in the gray matter regions (0.5 ± 6.0%) at the earliest (i.e., 45–60 minute) timepoint.

Bolus/infusion [¹⁸F]SynVesT-1 studies

The average regional TACs from 11 B/I [¹⁸F]SynVesT-1 studies are presented in Figure 3. On average, TACs for the arterial input function, gray matter regions, and the C.S. had negative slopes for all timeframes, although these were smaller in magnitude than those of [¹¹C]UCB-J (Table 2). The %mean slope for the arterial input ranged from −6.6 ± 9.3 to −7.7 ± 8.2%SUV/hour. The gray matter region slope never exceeded −4.2%SUV/hour and was less than −2.5%SUV/hour from 80 minutes on. The inter-regional variability (SD) ranged from ±4.4 at 60–90 min. to ±3.3%SUV/hour at 120–150 min. Based on the linear fit to the 6 scan values in each interval, the 70–100 and 120–150 min. windows provided the numerically smallest mean C.S. slope values, but with considerable inter-subject variability (−2.4 ± 13.8 and −2.4 ± 30.2%SUV/hour, respectively). In an effort to reduce noise effects on slope estimation, we also estimated the slope during each period from the fitted 1TC model curve. Indeed, this substantially reduced noise-induced inter-subject variability (Table 2 and Fig. S2), and using this approach, the slope for the same 70–100 min. window was 11.9 ± 5.0%SUV/hour, and for 120–150 min., 6.4 ± 5.1%SUV/hour.

Figure 3.

Population [¹⁸F]SynVesT-1 time-activity curves. Standardized uptake values (SUV) with units of the measured radioactivity concentration in the tissue or plasma (kBq/mL) relative to the bodyweight-corrected total injected dose (MBq/kg) for 11 subjects injected with a bolus/infusion of [¹¹C]UCB-J (K_bol = 150). Data are presented as the mean $\pm$ SD. Abbreviations: cortex (Ctx), centrum semiovale (C.S.), metabolite-corrected input function (IF).

Table 2.

Population mean (n = 11) standardized uptake value (SUV) slopes for [¹⁸F]SynVesT-1 B/I studies.

	Input function		Gray matter a		Centrum semiovale		Centrum semiovale
					(6-point linear fit)		(1TC fit)
Time (min)	Slope as %mean	SD of %mean	Slope as %mean	SD of %mean	Slope as %mean	SD of %mean	Slope as %mean	SD of %mean
									40–70	−6.6	9.3	−3.3	3.5	−24.7	13.0	−20.8	7.6
50–80	−7.4	8.6	−3.4	4.3	−32.4	8.9	−17.1	6.0
60–90	−7.7	8.2	−3.9	4.4	−9.6	19.4	−14.2	5.3
70–100	−7.7	8.0	−4.2	4.1	−2.4	13.8	−11.9	5.0
80–110	−7.6	7.9	−2.5	3.4	−7.3	16.5	−10.2	4.8
90–120	−7.4	8.0	−0.8	3.4	−6.6	12.4	−8.9	4.8
100–130	−7.3	8.2	−1.3	3.5	−13.3	26.6	−7.9	4.9
110–140	−7.1	8.6	−0.6	4.0	−4.5	23.2	−7.0	5.0
120–150	−6.9	9.0	0.9	3.3	−2.4	30.2	−6.4	5.1

Note: Standard deviation (SD) refers to across subject variance.

^aGray matter region values represent the within subject mean TAC for frontal, parietal, temporal, and occipital cortices; caudate, putamen, hippocampus, thalamus, and cerebellum.

As was done for [¹¹C]UCB-J studies, C_T/C_P was calculated for 30-minute overlapping time windows for each subject (see Table S2, top). Notably, C_T/C_P ratios continued to slowly increase for the total 150-minute scan duration for all ten ROIs (Table S2). By the 40–70 minute time window, C_T/C_P ratios overestimated V_T in the majority of ROIs (Figure 4(a)), consistent with the simulation results (Fig. S1A), although the % overestimation was much lower than that of [¹¹C]UCB-J. The 40–70 minute window produced values with the smallest magnitude bias of the V_T estimate when averaged across all ROI’s (−0.3 $\pm$ 6.1%), with misestimation ranging from −14.5 $\pm$ 2.7% (amygdala) to +9.6 $\pm$ 4.9% (thalamus). While the mean bias was numerically greater at 90–120 minutes, it was substantially less variable across ROIs and the mean bias was still less than 4% (3.8 $\pm$ 3.6%): in amygdala, V_T was underestimated (−4.8 $\pm$ 9.4%); In all other ROIs, V_T was overestimated, ranging from 2.2 $\pm$ 5.7% (temporal cortex) to 6.8 $\pm$ 5.0% (thalamus).

Figure 4.

Equilibrium ratios versus 1TC model parameters for [¹⁸F]SynVesT-1. Population bias of [a] C_T/C_P values relative to 1TC model estimates of V_T. and, [b] SUVR-1 values relative to 1TC model estimates of BP_ND from [¹⁸F]SynVesT-1 B/I studies (K_bol = 150) performed in 11 subjects. Bias was calculated as the %1TC parameter such that positive values represent overestimations and negative values represent underestimations of the relevant 1TC parameter. Data are presented as the mean±SD. Abbreviations: centrum semiovale (C.S.).

Observed SUVR-1 values continued to slowly rise thorough the entire 150-minute scan duration (for SUVR-1 values at each 30-minute for all ROIs, see Table S2, bottom). For the majority of the scan durations, SUVR-1 consistently underestimated BP_ND (Figure 4(b)). The 30-minute time window with the smallest average %BP_ND bias across ROIs was the final 30 minutes, that is 120–150 minutes post-injection (+0.8 $\pm$ 2.7%). Here, SUVR-1 overestimated BP_ND in cerebellum (+2.8 $\pm$ 4.4%), thalamus (+1.4 $\pm$ 3.6%), frontal cortex (+0.2 $\pm$ 4.3%), and hippocampus (+0.02 $\pm$ 5.8%). Underestimation of BP_ND in the remaining brain regions ranged from −7.3 $\pm$ 6.4% (amygdala) to −0.2 $\pm$ 5.2 (occipital cortex).

Notably, starting at the 90–120 minute window and continuing through the remainder of the scan, in all regions except amygdala, the %BP_ND misestimation by SUVR-1 was less than 7% with an SD of less than $\pm$ 5%. Thus, a 30-minute scan duration 90–120 minutes post [¹⁸F]SynVestT-1 injection produced C_T/C_P and SUVR-1 values that estimated their relevant kinetic parameters (V_T and BP_ND, respectively) within 7 $\pm$ 6% across almost nearly all brain regions (the only exception being the SUVR-1 estimate of amygdala BP_ND).

Discussion

When successfully applied, the advantage of the B/I method is that a state of equilibrium can be achieved. ¹⁵ At equilibrium, parameters (i.e., V_T and BP_ND) can be estimated from plasma and tissue TACs directly, bypassing the need for modeling, with the additional benefit of requiring a shorter scan duration.^15,21 However, different brain regions and each radiotracer displays unique kinetics and will reach equilibrium at different times (or not at all); thus, every new tracer and protocol must be validated. ²² Therefore, the overarching goal of these studies was to evaluate B/I methods of measuring synaptic density in vivo using radiotracers that target the presynaptic SV2A protein: [¹¹C]UCB-J and [¹⁸F]SynVesT-1. To this end, we assessed B/I paradigms and aimed to identify time windows post injection which provided the most reliable parameter estimates using C_T/C_P (apparent V_T) and SUVR-1 (apparent BP_ND).

Here, using a B/I delivery of [¹¹C]UCB-J with K_bol = 150 min, C_T/C_P overestimated V_T with increasing magnitude at later timepoints as a state of equilibrium is approached, although the magnitude of overestimation was smaller than that of bolus injections.^15,20 Although SUVR-1 displayed smaller observed bias of BP_ND relative to C_T/C_P estimation of V_T (as expected, given the cancellation of errors that occurs when dividing the target region by the reference tissue when calculating the SUVR), this B/I paradigm underestimated BP_ND (−3.4 $\pm$ 7.7% at 90–120 minutes) to a comparable extent than our previous bolus studies (−1 $\pm$ 7%). ²⁰ It is additionally worth noting that while 30–60 minute period provided the least variability in BP_ND underestimation, this earlier part of the scan produced estimates with a greater magnitude of bias (−10.9 $\pm$ 3.5%). Therefore, to select the timeframe with the smallest intersubject variability comes at the cost of reduced accuracy, and vice versa. Furthermore, no 30-minute time window provided precise and accurate estimates of both V_T and BP_ND, with a continued interest among investigators for using either or both outcome measures. It is possible that this poor performance was due to a suboptimal choice for K_bol. However, we believe that a more likely cause is the higher noise of ¹¹C-labeled tracers at late time points, For example, we observed a higher degree of inter-subject variability for [¹¹C]UCB-J compared to [¹⁸F]SynVesT-1 in the SUVR-1 estimates of BP_ND at later times post-injection (see error bars in Figures 2(b) and 4(b)). Given these findings, we generally conclude that a B/I method for [¹¹C]UCB-J PET provides no advantage- and introduces disadvantages- relative to bolus injections and is not recommended.

Our group has recently reported on the development and utility of an ¹⁸F-labeled SV2A radiotracer, [¹⁸F]SynVesT-1 (also known as [¹⁸F]SDM-8), with comparable or improved imaging characteristics to that of [¹¹C]UCB-J.^5,14 There are a number of advantages of opting for an ¹⁸F over a ¹¹C-labeled radiotracer. Their relatively long half-life (109.7 minutes for ¹⁸F vs. 20.4 minutes for ¹¹C) allows for a] central production and multisite radiotracer distribution; b] a greater number of scans can be performed with a single synthesis; c] improved spatio-temporal resolution and reduced noise.^{23 –25} While simulations predicted no K_bol was able to perfectly recapitulate 1TC V_T results, a B/I paradigm with K_bol = 150 was determined to be most likely to result in the smallest V_T bias of C_T/C_P 45 minutes post injection of [¹⁸F]SynVesT-1. Indeed, the 40–70 minute time window produced C_T/C_P values with the smallest V_T bias, with all regions, except amygdala (−14.5%) and C.S. (11.8%), having a bias of less than 10% (Figure 4). Although the magnitude of the average bias across brain regions was larger relative to that observed at 40–70 minutes, regional variability was smaller at later timepoints. Specifically, 90–120 minutes was the time window with the least variability in the V_T bias across ROIs, with the magnitude of the bias still being less than 7%. Furthermore, the B/I paradigm used here yielded SUVR-1 values which estimated 1TC BP_ND with minimal bias: starting at the 90 minutes and continuing through to the end of the 150 minute scan, any 30-minute window could be used to calculate SUVR-1. Across this time, the magnitude and inter-subject variation in %BP_ND was less than 7 $\pm$ 5% for all brain regions (with the exception of amygdala). As a comparator, the degree of bias observed here was within the absolute test-retest variability (aTRV) of 30-minute 1TC V_T (7–13%) and BP_ND (6–9%) estimates, previously reported in bolus studies. ²⁶ Additionally, the B/I approach examined here produce similar BP_ND estimates to that of bolus studies. ¹⁴ However, as expected, the B/I method avoided the overestimation of V_T from late tissue-to-plasma ratios, with percent biases of 5 ± 2% (B/I) vs. 46 ± 20% (bolus, 60–90 minutes, unpublished results). Furthermore, B/I equilibrium ratio estimates of V_T and BP_ND for [¹⁸F]SynVesT-1 outperformed those for [¹¹C]UCB-J, in terms of both the magnitude and variability of the bias.

A notable limitation of this work is the older (mean age of 54 years) and predominately male (90%) sample used for the [¹⁸F]SynVesT-1 studies. Furthermore, two individuals with major depressive disorder were included in our analysis of [¹⁸F]SynVesT-1; it is currently unknown how living with a chronic mood disorder could potentially impact tracer pharmacokinetics and pharmacodynamics in ways that would have biased the present findings. Therefore, it will be necessary to perform confirmation studies to ensure the B/I method is still valid in younger, female, or other clinical populations. Here it is also worth reiterating that SV2A PET acts as a surrogate (albeit, a well-defined and widely accepted surrogate) for historically postmortem measures of synaptic density.^{1 –4} However, an important caveat is that this method assumes that radiotracer affinity for SV2A is constant across the brain and across patient groups; this caveat also applies to virtually all PET tracers that bind to a brain protein. In addition, studies focused on a specific region of interest may want to reconsider the timing of the scan window, as not all brain regions display the same tracer uptake kinetics and therefore achieve a state of equilibrium at different times. Similarly, in cases where either V_T or BP_ND, but not both parameters, is the desired outcome measure, the scan window could be shifted earlier (for C_T/C_P estimates of V_T) or later (for SUVR-1 estimates of BP_ND). As a further consideration, investigators who desire to use V_T as the outcome measure, blood sampling will still be required. However, we have previously demonstrated that less-invasive venous plasma can be used as a reliable estimate of the input function for other ¹⁸F-labled tracers, ²¹ a possibility we are actively perusing for SynVesT-1.

In summary, using a B/I approach with [¹¹C]UCB-J, we were unsuccessful at consistently achieving a state of equilibrium across subjects, thus we were unable to identify a timeframe post injection when we could accurately and precisely estimate both V_T and BP_ND. However, switching to the fluorinated SV2A tracer, [¹⁸F]SynVesT-1 the B/I method employed here could be used to closely estimate 1TC parameters; Specifically, using a 30-minute scan duration 90–120 minutes post [¹⁸F]SynVestT-1 injection produced C_T/C_P values that estimated V_T and SUVR-1 values that estimated BP_ND 7 $\pm$ 6% across most brain regions. Thus, with B/I delivery of [¹⁸F]SynVesT-1, a short scan with minimal to no blood sampling is sufficient for accurate quantification of tracer distribution and specific binding. Additionally, use of an [¹⁸F], relative to a [¹¹C], makes central tracer synthesis and reginal distribution a possibility; In combination with a shorter scan duration and limited computational burden, this B/I approach with [¹⁸F]SynVesT-1 will make SV2A PET accessible to a greater number of institutions, increasing the feasibility of larger-scale, multi-site studies.