[¹⁸F]PSMA‐1007 renal uptake parameters: Reproducibility and relationship to estimated glomerular filtration rate

Abstract

Background

Scintigraphy with technetium‐99m‐labelled dimercaptosuccinic acid ([^99mTc]Tc‐DMSA) is widely used for renal cortical imaging. Uptake of [^99mTc]Tc‐DMSA has been shown to correlate with glomerular filtration rate (GFR). Prostate‐specific membrane antigen (PSMA) radiopharmaceuticals used for positron emission tomography (PET) show high renal uptake and are being investigated for renal imaging. [⁶⁸Ga]Ga‐PSMA‐11 PET parameters have been shown to correlate with estimated GFR (eGFR). The aim of this study was to investigate the relationship between renal [¹⁸F]PSMA‐1007 uptake and eGFR.

Methods

Hundred and eighty‐five patients underwent PET imaging at 1 and 2 h after injection of 4.0 ± 0.2 MBq [¹⁸F]PSMA‐1007. Serum creatinine levels were measured and GFR estimated using the Chronic Kidney Disease Epidemiology Collaboration (CKD‐EPI) and Modification of Diet in Renal Disease (MDRD) equations. Fifteen patients were excluded due to missing or incorrect data. Thus, data from 170 patients were analyzed. Kidneys were segmented in the PET images using a convolutional neural network with manual correction. For each kidney, mean standardized uptake value (SUV_mean) and segmentation volume in millilitres were measured. Linear regression analyses were performed with eGFR as the outcome variable.

Results

Variation in the eGFR values was explained to a significant degree by SUV_mean and renal segmentation volume in both the 1 and 2 h images. This correlation was stronger for CKD‐EPI eGFR (1 h R ² = 0.64; 2 h R ² = 0.64) than for MDRD eGFR (1 h R ² = 0.47; 2 h R ² = 0.45).

Conclusion

Renal [¹⁸F]PSMA‐1007 uptake parameters correlate with eGFR and are indicative of renal cortical function.

1. INTRODUCTION

Scintigraphy with technetium‐99m‐labelled dimercaptosuccinic acid ([^99mTc]Tc‐DMSA) is a widely used method for imaging renal cortical function. After injection, the radiopharmaceutical is taken up in renal proximal tubules (Willis et al., 1977). The resulting images are read to identify localized uptake reductions or defects, which indicate focal impairment in renal function. Additionally, the contribution of each kidney to the total renal function is often calculated as each kidney's percentage of the total renal tracer uptake. [^99mTc]Tc‐DMSA scintigraphy is primarily used to detect cortical impairment or damage after pyelonephritis, predominantly in children. It can also be used to examine structural anomalies of the kidneys, such as duplicated collecting systems or horseshoe kidney (Mandell et al., 1997; Piepsz et al., 1999, 2009; Vali et al., 2022). Renal [^99mTc]Tc‐DMSA uptake has been shown to correlate with serum creatinine level, creatinine clearance, and glomerular filtration rate (GFR) (Goodgold et al., 1996; Groshar et al., 1991; Kawamura et al., 1978; Taylor et al., 1986; Yee, et al., 1981), although it is not used to quantify glomerular filtration.

Estimation of GFR based on serum creatinine level is perhaps the most widely used method of quantifying renal function. Several equations exist for this purpose, taking into account age, sex, and in some cases race in addition to serum creatinine. The Modification of Diet in Renal Disease (MDRD) equation (Levey et al., 2006) is widely established while the Chronic Kidney Disease Epidemiology Collaboration (CKD‐EPI) equation (Inker et al., 2021) was developed to overcome limitations of the MDRD equation (Stevens et al., 2010), and has better performance when estimating higher GFRs (Earley et al., 2012). In a diverse population, the MDRD and CKD‐EPI equations have similar performance, resulting in estimates within ±30% of measured GFR in 80.4% and 84% of cases, respectively (Levey et al., 2020).

Positron emission tomography (PET), generally performed in combination with computed tomography (CT) as hybrid PET‐CT studies, sees widespread use primarily in oncology. Compared to conventional imaging (CT and bone scan), prostate‐specific membrane antigen (PSMA)‐targeted PET‐CT has brought improved diagnostic performance to prostate cancer (Hofman et al., 2020; Rowe et al., 2016), where it is used for primary staging as well as for detection of recurrence. In many cases, PSMA PET‐CT is performed with intravenous contrast media, requiring estimation of GFR.

Despite its name, PSMA is also expressed in non‐prostatic tissues, such as the proximal tubules of the kidneys (Kinoshita et al., 2006; Silver et al., 1997). Two PSMA radiopharmaceuticals for PET use, ⁶⁸Ga‐PSMA‐11 and ¹⁸F‐PSMA‐1007, display significant renal uptake and are being explored as alternatives to ^99mTc‐DMSA scintigraphy for renal cortical imaging (Ahmadi Bidakhvidi et al., 2021; Sarikaya et al., 2018, 2021; Valind et al., 2021).

A recent study on 25 patients found correlations between [⁶⁸Ga]Ga‐PSMA‐11 PET parameters and estimated GFR (eGFR), with two patients excluded due to high activity in renal calyces (Schierz et al., 2021). [¹⁸F]PSMA‐1007 has significantly lower urinary excretion than [⁶⁸Ga]Ga‐PSMA‐11 (Afshar‐Oromieh et al., 2016; Giesel et al., 2017), making renal uptake easier to distinguish, and thus, potentially making it more suitable for quantification and imaging and of renal cortical function.

For [¹⁸F]PSMA‐1007 to be useful for renal functional imaging, it needs to be established if renal uptake of this radiopharmaceutical is indicative of cortical function. Thus, the aim of this study was to investigate the relationship between renal [¹⁸F]PSMA‐1007 uptake and eGFR.

2. METHODS

2.1. Patients

Between July 2020 and June 2021, 185 patients referred to the Department of Clinical Physiology and Nuclear Medicine, Skåne University Hospital, Lund and Malmö, Sweden for [¹⁸F]PSMA‐1007 PET‐CT underwent scanning at 1 and 2 h after radiopharmaceutical injection, as part of a study on the effect of accumulation time on diagnostic performance. Patients were referred either for staging of newly diagnosed high‐risk prostate cancer or due to biochemical recurrence. High‐risk disease was defined as T3 stage, Gleason sum ≥8, Gleason score at least 4 + 3 in more than half of biopsies, or serum prostate‐specific antigen level (PSA) level ≥20 µg/l. Serum creatinine level was measured in all patients before they underwent PET‐CT. The study was approved by the Regional Ethical Review Board (#2016/417, #2018/117 and #2018/753) and the Swedish Ethical Review Authority (#2020‐00689), and was performed in accordance with the Declaration of Helsinki. All patients provided written informed consent before participating in the study.

2.2. PET‐CT

Participants were injected with 4.0 ± 0.2 MBq/kg (range 3.6–6.0 MBq/kg) of [¹⁸F]PSMA‐1007 and underwent PET‐CT acquisition on a GE Discovery MI (GE Healthcare) system at 1 and 2 h postinjection (p.i.). Images were acquired for 2 min per bed position, and reconstructed using Q.Clear (GE Healthcare), which is a block‐sequential regularization expectation maximization algorithm, using a β‐value of 800 (Trägårdh et al., 2020). Each kidney was segmented in the 1 h p.i. PET images using an in‐house developed convolutional neural network (CNN), trained specifically for this purpose. All segmentations were subsequently reviewed and corrected by a human reader. The CNN failed to segment 2 studies, which were instead segmented manually. For the 2 h p.i. PET images, adjusted 1 h p.i. segmentations were used except in two patients where, due to poor fit for the adjusted segmentations, corrected CNN segmentations of the 2 h p.i. images were used instead. For each segmented kidney, the mean standardized uptake value (SUV_mean) and the segmentation volume (in millilitres) were measured. The SUV_mean for each kidney was multiplied by the segmentation volume into left and right renal uptake (RU) respectively. Left and right RU were then summed to total renal uptake (TRU) for each patient. The left renal uptake percentage (LRU%) was calculated by dividing the left RU by the TRU for each patient.

2.3. Estimated glomerular filtration rate

GFR was estimated based on serum creatinine levels using the CKD‐EPI equation (Inker et al., 2021). A second estimation of GFR was made using the MDRD equation (Levey et al., 2006) to investigate if different estimation methods had any effect on the relationship between eGFR and PET parameters. Both equations result in relative estimations of GFR, assuming a body surface area (BSA) of 1.73 m². The actual BSA for each patient was calculated using the Du Bois formula (Du Bois et al., 1916), and used to determine the absolute eGFR for each estimation method and participant.

2.4. Statistical analysis

Univariable linear regression analysis was performed with eGFR (CKD‐EPI and MDRD) and serum creatinine level as outcome variables, using SUV_mean, segmentation volume, and TRU (at 1 and 2 h p.i.) as explanatory variables. Multivariable linear regression analysis was performed with CKD‐EPI and MDRD eGFR as outcome variables, using SUV_mean and segmentation volume from the same time point (1 or 2 h p.i.) as explanatory variables. To investigate the reproducibility of split renal uptake measurements between measurement time points, LRU% at 1 and 2 h p.i. were tested using Pearson's product moment correlation. The LRU% measurements were also compared using a paired‐samples t‐test to investigate any systematic difference between time points. The relationship between CKD‐EPI and MDRD eGFR was tested with Pearson's product moment correlation. p < 0.05 was considered significant. All statistical calculations were performed using the R statistical software package (Version 4.1.3; The R Foundation for Statistical Computing).

3. RESULTS

Fifteen patients were excluded due to missing or unusable data. Thus, 170 patients were included in the final analysis. All participants were men, and their age ranged from 54 to 83 years (Table 1). Sample PET images and segmentations from a representative patient are presented in Figure 1. Univariable linear regression analysis showed that eGFR and, to a lesser degree, serum creatinine, were correlated to SUV_mean, total renal segmentation size, and TRU (Table 2). The strongest correlations were found between eGFR and TRU (Figure 2). In the 1 h p.i. PET images the resulting equations for CKD‐EPI eGFR and MDRD eGFR, respectively, were as follows:

$$y=62.41+0.0092\ast \mathrm{TRU},$$ (1)

$$y=24.67+0.0145\ast \mathrm{TRU}.$$ (2)

Table 1.

Patient characteristics and PET parameters

Parameter	Mean ± SD (range)
Age (years)	71.0 ± 6.5 (54–83)
Weight (kg)	86.4 ± 13.6 (48–137)
Height (cm)	177.8 ± 6.2 (162–198)
Du Bois BSA (m²)	2.0 ± 0.2 (1.5–2.5)
BMI (kg/m²)	27.3 ± 3.9 (16.6–42.8)
Serum creatinine (µmol/l)	87.3 ± 19.7 (49–207)
CKD‐EPI eGFR (ml/min)	105.8 ± 12.2 (76.6–134.4)
MDRD eGFR (ml/min)	93.1 ± 22.8 (34.6–67.5)
Injected activity (MBq)	345.7 ± 54.0 (191–503)

Parameter	Mean ± SD (range)
	1 h p.i.	2 h p.i.
Renal SUVmean	9.2 ± 1.5 (4.7–13.2)	11.1 ± 2.1 (4.9–16.7)
Segmentation volume (ml)	514.9 ± 86.1 (336.6–819)	537.7 ± 87.9 (340–841.3)
Total renal uptake (SUVmean × ml)	4715.2 ± 1046.0 (2328.4–7551.6)	5957.7 ± 1421.2 (2639–9740)
Left renal uptake %	49.8 ± 5.8 (15.1–100)	49.6 ± 5.9 (14–100)

Abbreviations: BMI, body mass index; BSA, body surface area; CKD‐EPI, Chronic Kidney Disease Epidemiology Collaboration; eGFR, estimated glomerular filtration rate; MDRD, Modification of Diet in Renal Disease; PET, positron emission tomography; p.i., postinjection; SUVmean, mean standardized uptake value; SD, standard deviation.

Figure 1.

Example segmentations from a representative patient. The upper row shows 1 h p.i. segmentations in axial (a), coronal (b), and sagittal (c) planes. The lower row shows the corresponding slices from the 2 h p.i. acquisition, with segmentations in axial (d), coronal (e), and sagittal (f) planes. p.i., postinjection; SUV, standardised uptake value.

Table 2.

Univariable linear models

Outcome variable	Explanatory variable	R² (1 h p.i. data)	R² (2 h p.i. data)
CKD‐EPI eGFR	Renal SUVmean	0.156	0.151
	Renal segmentation volume	0.458	0.463
	Total Renal Uptake	0.621	0.578
MDRD eGFR	Renal SUVmean	0.156	0.089
	Renal segmentation volume	0.352	0.347
	Total renal uptake	0.442	0.385
Serum creatinine	Renal SUVmean	0.108	0.087
	Renal segmentation volume	0.186	0.186
	Total renal uptake	0.285	0.245

Note: All correlations were significant (p < 0.0001).

Abbreviations: CKD‐EPI, Chronic Kidney Disease Epidemiology Collaboration; eGFR, estimated glomerular filtration rate; MDRD, Modification of Diet in Renal Disease; p.i., postinjection; SUVmean, mean standardised uptake value.

Figure 2.

Univariable linear regression models with CKD‐EPI eGFR (a, b) and MDRD eGFR (c, d) as outcome variables, with TRU at 1 h p.i. (a, c) and 2 h p.i. (b, d) as explanatory variables. CKD‐EPI, Chronic Kidney Disease Epidemiology Collaboration; eGFR, estimated glomerular filtration rate; MDRD, Modification of Diet in Renal Disease; p.i., postinjection; TRU, total renal uptake.

For the 2 h p.i. data, the corresponding equations were for CKD‐EPI eGFR and MDRD eGFR were as follows:

$$y=66.87+0.0065\ast \mathrm{TRU}$$ (3)

$$y=33.65+0.0100\ast \mathrm{TRU}$$ (4)

All correlations in the univariable models were significant (p < 0.0001).

Multivariable linear regression with SUV_mean and renal segmentation volume as covariates resulted in slightly higher degrees of explanation (Figure 3). The resulting equations for CKD‐EPI eGFR and MDRD eGFR respectively for the 1 h p.i. data were as follows:

$$y=22.54+0.0988\ast {\mathrm{Volume}}_{\mathrm{segmentation}}+3.5339\ast {\mathrm{SUV}}_{\mathrm{mean}}$$ (5)

$$y=-39.33+0.1616\ast {\mathrm{Volume}}_{\mathrm{segmentation}}+5.3660\ast {\mathrm{SUV}}_{\mathrm{mean}}$$ (6)

Figure 3.

Multivariable linear regression models with CKD‐EPI eGFR (a, b) and MDRD eGFR (c, d) as outcome variables. Segmentation volume and SUV_mean at 1 h p.i. (a, c) and 2 h p.i. (b, d) were used as explanatory variables. CKD‐EPI, Chronic Kidney Disease Epidemiology Collaboration; eGFR, estimated glomerular filtration rate; MDRD, Modification of Diet in Renal Disease; p.i., postinjection; SUV_mean, mean standardised uptake value.

The corresponding equations for the 2 h p.i. data were as follows:

$$y=25.79+0.0972\ast {\mathrm{Volume}}_{\mathrm{segmentation}}+2.4969\ast {\mathrm{SUV}}_{\mathrm{mean}}$$ (7)

$$y=-31.40+0.1569\ast {\mathrm{Volume}}_{\mathrm{segmentation}}+3.6149\ast {\mathrm{SUV}}_{\mathrm{mean}}$$ (8)

All correlations in the multivariable models were significant (p < 0.0001).

The paired‐samples t‐test revealed a statistically significant difference in LRU% between the 1 and 2 h p.i. measurements, with a mean difference of 0.2 percentage points (95% confidence interval [CI] 0.06%–0.03%, p = 0.005). Pearson's correlation coefficient for the LRU% measurements was 0.989 (95% CI 0.985–0.992, p < 0.0001). For 166 out of 170 patients, the difference between measurements did not exceed 2 percentage points (Figure 4). Pearson's correlation coefficient for CKD‐EPI and MDRD eGFR was 0.821 (95% CI 0.765–0.865, p < 0.0001).

Figure 4.

(a) Pearson correlation (r = 0.989, p < 0.0001), and (b) Bland–Altman plot of LRU% measurements from 1 and 2 h p.i. data. Dashed lines indicate mean difference and 95% limits of agreement (mean ± 1.96 SD). LRU%, left renal uptake percentage; p.i., postinjection; SD, standard deviation.

4. DISCUSSION

Renal uptake of [¹⁸F]PSMA‐1007 expressed as SUV_mean, in combination with segmentation volume, correlated moderately but significantly with eGFR in both univariable and multivariable linear regression models. The multivariable models performed slightly better than the univariable ones and had similar performance when using 1 and 2 h p.i. data, reaching at best R ² = 0.64 (CKD‐EPI eGFR, 2 h p.i.). Overall, stronger correlations were found with CKD‐EPI eGFR than with MDRD eGFR. One reason for this could be that the CKD‐EPI equation was developed based on a population with a higher mean GFR (68 ml/min/1.73 m²) than the MDRD population (40 ml/min/1.73 m²) (Earley et al., 2012), more closely matching the participants of our study.

Our findings show that renal uptake of [¹⁸F]PSMA‐1007, similar to uptake of [^99mTc]Tc‐DMSA, is indicative of renal cortical function. Groshar and colleagues demonstrated a moderate correlation (r = 0.76) between [^99mTc]Tc‐DMSA uptake and creatinine clearance as a measurement of GFR (Groshar et al., 1991). Their findings show that [^99mTc]Tc‐DMSA uptake in itself is insufficient for determination of GFR. The correlations we found between renal uptake of [¹⁸F]PSMA‐1007 and eGFR were comparatively weaker. The is not unexpected as [¹⁸F]PSMA‐1007 is taken up in other organs than the kidneys to a larger extent than [^99mTc]Tc‐DMSA. Consistent with the likely tubular rather than glomerular uptake, the link between [¹⁸F]PSMA‐1007 PET parameters and eGFR is not strong enough to allow estimation of GFR based on PET alone. However, like [^99mTc]Tc‐DMSA, [¹⁸F]PSMA‐1007 seems to be suitable for imaging functioning renal cortical mass. The reproducibility of our split renal uptake measurements is similar to what has been shown for [^99mTc]Tc‐DMSA (Dissing et al., 2008; Tondeur et al., 2000). The statistically significant mean difference of 0.2 percentage points between 1 and 2 h p.i. measurements is not large enough to be clinically significant.

Schierz and colleagues found that renal segmentation volume, as well as the renal SUV_mean‐to‐blood pool SUV_mean ratio, and the renal SUV_mean‐to‐liver SUV_mean ratio, correlated with MDRD eGFR in a study of 25 patients undergoing [⁶⁸Ga]Ga‐PSMA‐11 PET (Schierz et al., 2021). To the best of our knowledge, the present study is the first to demonstrate a correlation between [¹⁸F]PSMA‐1007 PET parameters and eGFR, and it included a considerably larger study population. Interestingly, neither renal SUV_mean nor TRU (referred to as ‘sum TLG’) correlated significantly with eGFR in the study by Schierz et al. Similarly, Conen and colleagues found no significant correlation between renal [⁶⁸Ga]Ga‐PSMA PET SUV_mean and GFR (Conen et al., 2022) in a group of 34 men. This disparity could be caused by differences in population size, segmentation strategy or urinary radiopharmaceutical excretion. Furthermore, differences in eGFR (mean 78.6 ml/min using MDRD in the study by Schierz et al. compared to mean 93.1 ml/min with MDRD in this study) could affect the results.

Rosar and colleagues compared split renal function, measured with technetium‐99m‐labelled mercaptotriglycene ([^99mTc]Tc ‐MAG3) renography, to split renal uptake of [⁶⁸Ga]Ga‐PSMA‐11 in 97 patients and found good agreement (r = 0.91) (Rosar et al., 2020). This supports the use of PSMA PET for measuring split renal function. It remains to be explored if split renal uptake of [¹⁸F]PSMA‐1007 corresponds to split renal function measured with a radiopharmaceutical validated for this purpose. We plan to investigate this in a direct comparison between [¹⁸F]PSMA‐1007 PET and [^99mTc]Tc ‐DMSA scintigraphy.

4.1. Limitations

One limitation of this study is the homogeneity of the study participants, who were all men over 50 years of age. The participants had relatively healthy kidneys. Most of them had an eGFR above 90 ml/min, and no participant had an eGFR below 30 ml/min, regardless of estimation method used. Another limitation is that estimation of GFR itself introduces a source of imprecision, which is underlined by non‐perfect correlation (r = 0.82) between the MDRD and CKD‐EPI estimates in this study population. It would have been preferrable to instead measure GFR as clearance of iohexol or ⁵¹Cr‐labelled ethylenediaminetetraacetic acid ([⁵¹Cr]EDTA). Such a study would be needed to determine if [¹⁸F]PSMA‐1007 PET parameters correlate to measured GFR in addition to eGFR. This would however been a significantly larger endeavour, and is beyond the scope of the current study.

4.2. Conclusion

Renal [¹⁸F]PSMA‐1007 uptake parameters correlate with eGFR and are indicative of renal cortical function. Split renal uptake measurements using this radiopharmaceutical have solid reproducibility. Further studies are needed to establish if [¹⁸F]PSMA‐1007 can replace [^99mTc]Tc ‐DMSA for renal cortical imaging,

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Valind, K. , Jögi, J. , Minarik, D. & Trägårdh, E. (2023) [¹⁸F]PSMA‐1007 renal uptake parameters: reproducibility and relationship to estimated glomerular filtration rate. Clinical Physiology and Functional Imaging, 43, 128–135. 10.1111/cpf.12801

DATA AVAILABILITY STATEMENT

The data used in this study are available from the corresponding author on reasonable request.