Optimization of parathyroid 11C-choline PET protocol for localization of parathyroid adenomas in patients with primary hyperparathyroidism

Purpose To evaluate the optimal tracer uptake time, the minimal amount of radioactivity and the inter-observer agreement for 11C-choline positron emission tomography/computed tomography (PET/CT) in patients with primary hyperparathyroidism (pHPT). Methods Twenty-one patients with biochemically proven pHPT were retrospectively studied after injection of 6.3 ± 1.2 MBq/kg 11C-choline. PET data of the first nine patients, scanned for up to 60 min, were reconstructed in 10-min frames from 10- to 60-min postinjection (p.i.), mimicking varying 11C-choline uptake times. Parathyroid adenoma to background contrast ratios were calculated and compared, using standardized uptake values (SUVs). Data was reconstructed with varying scan durations (1, 2.5, 5, and 10 min) at 20–30-min p.i. (established optimal uptake time), mimicking less administered radioactivity. To establish the minimal required radioactivity, the SUVs in the shorter scan durations (1, 2.5, and 5 min) were compared to the 10-min scan duration to determine whether increased variability and/or statistical differences were observed. Four observers analyzed the 11C-choline PET/CT in four randomized rounds for all patients. Results SUVpeak of the adenoma decreased from 30 to 40 p.i. onwards. All adenoma/background contrast ratios did not differ from 20- to 30-min p.i. onwards. The SUVs of adenoma in the scan duration of 1, 2.5, and 5 min all differed significantly from the same SUV in the 10-min scan duration (all p = 0.012). However, the difference in absolute SUV adenoma values was well below 10% and therefore not considered clinically significant. The inter-observer analysis showed that the Fleiss’ kappa of the 1-min scan were classified as “moderate,” while these values were classified as “good” in the 2.5-, 5-, and 10-min scan duration. Observers scored lower certainty scores in the 1- and 2.5-min scans compared to the 5- and 10-min scan durations. Conclusion The optimal time to start PET/CT scanning in patients with pHPT is 20 min after mean injection of 6.3 MBq/kg 11C-choline, with a recommended scan duration of at least 5 min. Alternatively, the radioactivity dose can be lowered by 50% while keeping a 10-min scan duration without losing the accuracy of 11C-choline PET/CT interpretation. Electronic supplementary material The online version of this article (10.1186/s13550-019-0534-5) contains supplementary material, which is available to authorized users.


(Continued from previous page)
Conclusion: The optimal time to start PET/CT scanning in patients with pHPT is 20 min after mean injection of 6.3 MBq/kg 11 C-choline, with a recommended scan duration of at least 5 min. Alternatively, the radioactivity dose can be lowered by 50% while keeping a 10-min scan duration without losing the accuracy of 11 C-choline PET/CT interpretation.
Keywords: Primary hyperparathyroidism, 11 C-Choline PET, Inter-observer agreement, Scan protocol Background Primary hyperparathyroidism (pHPT) is a common endocrine disorder, with the highest incidence in elderly females (> 70 years) [1]. Eighty to 90% of pHPT is caused by a single parathyroid adenoma [2]. Surgery, preferably a unilateral minimally invasive parathyroidectomy (MIP), is the only curative treatment.
Lately, new functional imaging techniques using positron emission tomography (PET)/CT, with, e.g., 11 C-methionine [8,9] as radiotracer, have been studied, but no ideal PET tracer for adenomas has emerged so far. Recently, 11 C/ 18 Fcholine was reported for visualization of adenomas and favorable results have been published in literature [10][11][12][13][14][15]. The physiology behind the uptake of 11 C/ 18 F-choline is unknown, but may be based on the elevated concentration of phosphatidylcholine in parathyroid cells [16].
So far, only one study has investigated the clinical performance of 11 C-choline PET/CT in patients with pHPT [13]. Because 11 C-choline is a relatively new technique in patients with pHPT, the optimal uptake time of the tracer and minimal injected radioactivity needs to be further defined. To fulfill the need of routine clinical practice, consistency in the interpretation between different observers is essential.
Therefore, this study aimed to optimize the scan protocol, by assessing the optimal scanning time of 11 Ccholine in combination with the minimal amount of radioactivity needed for clinically acceptable image quality. Also, we studied the inter-observer agreement of 11 C-choline PET/CT for the detection of parathyroid adenomas in patients with pHPT.

Study design and patients
This is a single-center retrospective cohort study of patients with biochemically proven pHPT who underwent preoperative localization of the suspected parathyroid adenoma using 11 C-choline PET/CT in a tertiary referral hospital.
The medical charts of all patients who underwent 11 Ccholine PET/CT between April 2015 and February 2017 were reviewed. Patients eligible for inclusion were those ≥ 18 years, diagnosed with biochemically confirmed pHPT, and who underwent 11 C-choline PET/CT for the localization of the suspected adenoma. In total, 23 patients underwent 11 C-choline PET/CT for the localization of an adenoma. All patients had biochemically confirmed pHPT (calcium and PTH values above the upper limit of normal). However, one patient was known to have multiple endocrine neoplasia type I, and one patient was diagnosed with familial hypocalciuric hypercalcemia (FHH); these two patients were excluded from this analysis. Of the analyzed patients that were operated (67%), the 11 C-choline PET/CT correctly identified the location of the adenoma in all cases. Of the analyzed patients that were not operated (33%), the 11 C-choline PET/CT was positive in all but one patient. This patient was only included in the inter-observer analysis of this study and not in the analysis of the optimal uptake time of the tracer and the radioactivity to be administered (data not shown).
The medical charts were reviewed to determine the injected activity of 11 C-choline in MBq, gender, age, length, weight, preoperative PTH, and corrected calcium (for calculation of corrected calcium refer to Additional file 1).
Data obtained from patient records were anonymously stored using study-specific patient codes in a passwordprotected database. The study was exempt for collection of informed consent after reviewing by the Medical Ethics Committee Groningen (registration number 2016/ 413). 11 C-Choline PET/CT 11 C-choline was produced on site as described in [17]. Further details on production, patient preparation, and PET/CT acquisition can be found in Additional file 2. In the first nine patients, PET/CT images were taken directly after injection of 7.0 ± 0.5 MBq/kg [range 6.1-7.4 MBq/kg] 11 C-choline for up to 40 to 60 min postinjection (p.i.). After an interim analysis of the first nine patients had been performed to determine the most suitable uptake time of the tracer to reduce overall scan duration, all subsequent patients were scanned dynamically 20 min after the injection of mean 5.9 ± 1.4 MBq/kg [range 4.2-8.5 MBq/kg] for a duration of 10 min.

C-Choline PET/CT analysis
Of the first nine patients, the images were reconstructed in time frames of 10 min (0-10 to 50-60-min p.i.) to determine the optimal uptake time of the tracer. Also, these images were reconstructed in scan durations of 5 min (20-25 and 25-30 min), 2.5 min (20-22.5 to 27.5-30 min), and 1 min (20-21 to 29-30 min) to assess image quality as a function of scan duration used as a surrogate for variation in administered radioactivity.
For each of these newly created images, accumulated activity of 11 C-choline in the adenoma was (semi) quantitatively evaluated by placing an automatic volume of interest (VOI) with a threshold of 50% of the maximum tracer uptake in the lesion, using an in-house software (ACCURATE) [18]. In the surrounding normal background tissues (descending aorta, thyroid, shoulder muscle, and the first thoracic vertebrae (T1)), a spherical or cubical VOI, with a fixed size, was positioned. In the thyroid gland, the cubical VOI was positioned in an area on the contralateral side of the adenoma in healthy appearing thyroid tissue. VOI's were copied and transferred to scans of all time frames through an automatic linking feature and manual correction if needed. In each of these newly created images, peak, mean, and maximum standardized uptake values corrected for body weight (SUV peak , SUV mean , and SUV max ) were obtained for adenoma and background organs: descending aorta, thyroid, shoulder muscle, and T1. SUV max represents the maximum tracer uptake seen across all voxels within the volume of interest. SUV peak represents the average uptake in a 1 mL spherical volume of interest positioned such to yield the highest value across all possible locations of the lesion (or organ volume of interest). SUVmean is the mean SUV within the volume of interest. SUV mean was used for analysis of the background tissues, because it provides the most accurate and precise SUV in case of regions with an almost uniform uptake, and SUV peak was used for the adenoma being the preferred SUV metric to quantify tracer uptake in (small) lesions besides SUV max [19]. Moreover, SUV max was used for all as this metric is and has been used frequently and is therefore included to allow comparison with other studies.

Inter-observer agreement
All patients were included in the inter-observer agreement analysis. The acquired image data at 20-min p.i. was post-processed with varying scan duration of 1, 2.5, 5, and 10 min, respectively.
From all patients, two scans with a scan duration of 1 min and 2.5 min, both scans with a duration of 5 min, and the scan with a duration of 10 min were randomly selected.
Four observers visually interpreted the scans in four rounds, with increasing scan duration per round. Observers were asked to identify and localize any abnormally increased 11 C-choline uptake and to localize it in the right upper, left upper, right lower, left lower, or ectopic zones. Next, in case of a focal increased uptake, per location, they had to score how certain they were that the uptake was indeed increased (certainty of increased uptake (CIU)) and how certain they were that the uptake could be attributed to an adenoma (certainty of adenoma (CA)). They completed a standard scoring form per patient, per round including a 5-point scale to score their CIU and CA, with "1" being totally unsure and "5" being totally sure. Further details of the interobserver analysis and assessment can be found in Additional file 3.

Statistical analysis
Patient characteristics are described using means and standard deviations (SD) or medians and ranges for continuous variables (depending on normal distribution). All statistical analyses were performed with IBM SPSS Statistics for Windows, Version 23.0 (Armonk, NY: IBM Corp.). P values of < 0.05 were considered statistically significant. For multiple testing, the Bonferroni correction was applied. Graphics were generated using Graph-Pad Prism, Version 7.02 (La Jolla, CA, USA).

Optimal uptake time of tracer
Of the first nine patients, scanned for up to 40 to 60 min, contrast ratios (adenoma to muscle, T1, aorta, and thyroid tissue) were calculated to determine the optimal uptake time of the tracer. Contrast ratios were calculated by dividing the SUV peak for the adenoma by the SUV mean for the background. The ratios in every time frame were compared with the ratios in the upcoming time frame using the Wilcoxon signed-rank test as normality of the data could not be shown.
The SUV mean for thyroid and SUV peak for adenoma (of the first nine patients) in every uptake time were compared with the same SUVs in the upcoming uptake time using the Wilcoxon signed-rank test as normality of the data could not be shown. Using the Bonferroni correction, p values < 0.01 were considered statistically significant.

Radioactivity to be administered
In the analysis of the radioactivity to be administered, only the first eight patients (scanned for up to 40 to 60 min) were included, since no post-processed images could be made in scan durations of 1, 2.5, and 5 min for one patient, because the original raw data was no longer available. The post-processed images with a scan duration of 10 min was, however, still available.
To assess the effect of injected radioactivity on SUVmax , SUV mean , and SUV peak for adenoma and background tissues, the SUVs in the shorter scan durations (1, 2.5, and 5 min) were compared to the 10-min scan duration (at the determined optimal uptake time) to determine whether statistical differences were observed, using the Wilcoxon signed-rank test as normality of the data could not be shown. For these analyses, the average SUVs of patients per scan duration were calculated, since the 1-, 2.5-, and 5-min scan durations had multiple measurements of SUV (10, 4, and 2, respectively) and the 10min scan duration only had 1 measurement. In this way, we could assess if using different scan durations would result in different SUV data, on average, i.e., if it would result in a systematic bias. Using the Bonferroni correction, p values < 0.017 were considered statistically significant.
Inter-observer agreement The SUV mean for thyroid initially decreased before leveling off from 20-to 30-min p.i. onwards, while the SUVpeak for adenoma was constant until 30-40-min p.i. and decreased afterwards ( Fig. 1) ( Table 2). We found no significant differences between the SUV peak for adenoma for the different uptake times (p > 0.110). There were significant differences between the SUV mean for thyroid in the uptake time of 0-10 versus 10-20 min (p = 0.008) and that of 10-20 versus 20-30 min (p = 0.008). From We focused on the adenoma/thyroid contrast ratio, as the adenoma is usually closest located to the thyroid, refer to Additional file 4 for data on adenoma/aorta, adenoma/muscle and adenoma/T1 contrast ratios.
The adenoma/thyroid ratio became constant from the uptake time of 20-30-min p.i. onwards (Fig. 2). In the uptake time of 0-10 versus 10-20-min p.i., the adenoma/thyroid ratio increased from 1.49 to 1.65 (p = 0.008) ( Table 3). In the uptake time of 10-20 versus 20-30 min, the adenoma/thyroid ratio (increase from 1.65 to 1.85) is just slightly beyond the level of significance (p = 0.028). From the uptake time of 20-30-min p.i. onwards, there were no significant differences with upcoming uptake times for any ratio and p values were higher compared to previous uptake times (p > 0.043).

Radioactivity to be administered
The SD and interquartile range (IQ) of SUV peak for adenoma, SUV mean for background tissue, and SUVmax for all locations decreased with increasing scan duration (Table 4 and Additional file 5). The SUV peak for adenoma and SUV mean for background tissue varied less than SUV max, as witnessed by higher SD and IQ for SUV max for each scan duration (Table 4 and Additional file 5).
In addition, all individual SUVs in the scans with a duration of 1 min seemed more variable than the SUVs in the scan duration of 2.5, 5, and 10 min (Fig. 3, Table 4, and Additional file 5).
There was no significant difference in SUV mean for thyroid between the different scan durations (all p = 0.674), whereas there was a significant difference in SUV max for thyroid between a 1-min versus 10-min scan duration and a 2.5-min versus 10-min scan duration (both p = 0.012). There was no difference in SUV max for thyroid between a 5-min versus 10-min scan duration (p = 0.069). The SUVs for adenoma in the scan duration of 1, 2.5, and 5 min all differed statistically from the same SUV in the 10-min scan duration (all p = 0.012).

Inter-observer agreement
In total, 141 scans were used for this analysis.  SD standard deviation, IQ interquartile range, SUV standardized uptake value Only 7 and 5 patients were scanned until 50 and 60 min, respectively min ( Table 5). Most certainty scores, only representing the certainty score of positive identified lesions, increased with increasing scan duration (Table 6).

Discussion
In this study, we evaluated the scan protocol and interobserver agreement of 11 C-choline PET/CT in patients with biochemically confirmed pHPT. We found that the optimal uptake time of 11 C-choline PET/CT scanning is 20 min, since from 20-min p.i. onwards the adenoma/ background contrast ratios and SUV mean for thyroid became constant.
In addition, we quantitatively analyzed various scan durations as a surrogate for injecting different amounts of radioactivity. Although there was a significant difference in (average) SUVs at shorter scan durations compared to the 10-min scan, the differences for the 2.5and 5-min scan durations were small, well below 10%, and clinically not relevant given the spread in the observed SUV values. However, the observed increased spread in SUV values for the 1-min scan duration was wider compared with the other scan durations. Therefore, we advise for quantitative analysis not to lower the scan duration beyond 2.5 min.
To evaluate the clinical relevance of images with different quality, we performed an inter-observer study regarding lesion detection and localization. We concluded that a 1-min scan is too short for accurate visual interpretation of 11 C-choline PET/CT images. The certainty scores (CIU and CA) were generally lower in the 1-min and 2.5-min scan durations compared to the 5-and 10min scan durations. In clinical routine, it may occur that   SD standard deviation, IQ interquartile range, SUV standardized uptake value the expected dose of 11 C-choline is not fully administered to the patient, e.g., due to a lower production at the radiochemistry lab or to paravasal injection. Also, adenomas can be very small structures. Therefore, combining the results of the quantitative and inter-observer data, we recommend to establish a safety margin for 11 C-choline PET/CT scanning with 6.3 MBq/kg and scan for at least 5 min. Alternatively, the radioactivity dose can be lowered by 50% while maintaining a 10-min scan duration.
As can be expected, we found that the SUV peak is less variable than the SUV max . This is consistent with findings for FDG PET/CT in oncology, as shown by Lodge et al. and Makris et al. [21,22]. In these studies, it was observed that SUV max is more sensitive to noise, showing increased variability and upward bias as compared to SUV peak . The latter can be understood as SUV max represents the uptake in a single voxel, while SUV peak is derived from a 1 mL spherical volume of interest thereby mitigating the effects of image noise. Therefore, if the radioactivity in the adenoma needs to be quantified, it is best to use the SUV peak for parathyroid tracer uptake quantification as it shows a better precision than SUV max .
SUVs in the blood pool are known to be relatively low, and thus, these regions suffer from poor count statistics. SUV max for the aorta in the 1-min scan duration was higher than those seen at longer scan durations. This can be expected because SUV max is more sensitive to noise and thus showing upward biases with increased noise levels [21,22].
Quantitative research on the scan protocol for parathyroid imaging with PET/CT is very limited. The optimal scan time for 18 F-choline PET/CT in pHPT has been investigated once before by Rep et al. 18 F has a considerable longer half-life of 110 min, compared to 20 min for 11 C. Rep et al. only focused on an uptake time of 5 min, 60 min, or 2 h. They found that the optimal scan time of 18 F-choline PET/CT for localization of enlarged parathyroid tissue is 1 h after administration [23]. In the current study with 11 C-choline, we focused on the interval between 5 and 60 min. Prabhu et al. also investigated the utility of an early dynamic PET/CT in detecting parathyroid lesions, but they performed it for the tracer 18 F-choline. They found that early dynamic 18 Fcholine PET/CT imaging could suffice, without the need for a delayed image after 45 min [24]. This is more in line with the optimal uptake time of 20 min found in our study for 11 C-choline.
Our retrospective study has limitations. We assumed that all lesions that were identified on the 11 C-choline PET/CT were parathyroid adenomas in this group of patients with biochemically proven pHPT, although we only have surgical and pathological confirmation in two thirds of the patients (data not shown). Since our study focuses on the technical aspects of lesion detectability, and because determining the sensitivity of this type of PET/CT scan for correct localization of an adenoma was a b Fig. 3 Scatter plot of SUV mean , SUV peak , and SUV max of adenoma and thyroid in the different scan durations (1, 2.5, 5, and 10 min) in n = 8 patients. Scatter plot representing median values and interquartile ranges of a SUV peak for adenoma and SUV mean for thyroid in the different scan durations. b SUV max for adenoma and thyroid in the different scan durations. SUV standardized uptake value, *significant difference (p < 0.017) not the aim of our analysis, we believe for our study objective final surgical and pathological outcome is less crucial.
In our retrospective study, we chose to use SUV corrected for body weight, as this is the SUV type most used. No efforts were made to relate optimal dose to length, weight, or BMI. The latter would require a pharmacokinetic study including arterial sampling to assess which SUV normalization corresponds best with full quantitative pharmacokinetic results. This was beyond the scope of this study, and therefore, the most standard metric was chosen of SUV corrected for body weight. Yet, when analyzing the data using lean body mass (data not shown), results were very comparable and did not affect conclusions. Moreover, the sample size is too small to fully assess the relationship and the interactions between these parameters. Larger datasets are required for further refinement or dose optimization. Yet, at present, our study demonstrated that we can safely reduce scan duration or activity dose with a factor of two without compromising visual image interpretations. The authors acknowledge, however, that investigation of further dose refinements is of interest.
Always the same researcher quantified the 11 C-choline PET/CT scans by placing the VOIs in the adenoma and background. Another researcher would probably find slightly different SUVs. To better estimate this standard error, we tested whether the SUV mean was different if we moved the square VOI of the thyroid 2 voxels up or down. In the 10-min scan durations, the median difference was 3.32% with a standard deviation of 8.33%, which we consider acceptable low (data not shown).
Other limitations were encountered in the interobserver part of the study. The working experience for the observers differed, and two of the observers had limited experience with 11 C-choline PET/CT for the  (Table 1). Note the physiological uptake in the salivary glands, and remaining activity in the vessel used for injection of the tracer. Also, slight uptake is seen in a thyroid nodule in the left thyroid lobe identification of adenomas. Despite these limitations, the inter-observer agreement for the scan duration of 2.5, 5, and 10 min qualified as "good." technical assistance and help with conducting the interim analysis and with Fig. 4.

Funding
Not applicable.

Availability of data and materials
The datasets used during the current study are available from the corresponding author on reasonable request.
Authors' contributions MEN contributed to the conception and design of the study, acquiring the data, analyzing and interpreting the data, and drafting the manuscript. SK contributed to the conception and design of the study, analyzing and interpreting the data, and revising the manuscript. WN contributed to acquiring the data, analyzing and interpreting the data, and revising the manuscript. EDT contributed to acquiring the data, analyzing and interpreting the data, and revising the manuscript. DVG contributed to the conception and design of the study, analyzing and interpreting the data, and revising the manuscript. MTM contributed to analyzing and interpreting the data and revising the manuscript. MO contributed to analyzing and interpreting the data and revising the manuscript. AHD contributed to analyzing and interpreting data and revising the manuscript. GL contributed to analyzing and interpreting the data and revising the manuscript. RAJOD contributed to analyzing and interpreting the data and revising the manuscript. MEM contributed to the conception and design of the study, analyzing and interpreting the data, and revising the manuscript. RB contributed to the conception and design of the study, analyzing and interpreting the data, and revising the manuscript. AHB contributed to the conception and design of the study, acquiring the data, analyzing and interpreting the data, and revising the manuscript. All authors read and approved the final manuscript. CIU certainty increased uptake, CA certainty adenoma, Obs observer, SD standard deviation