- Email: [email protected]
Radioassays and experimental evaluation of dose calibrator settings for 18F
Applied Radiation and Isotopes 54 (2001) 113±122
www.elsevier.com/locate/apradiso
Radioassays and experimental evaluation of dose calibrator settings for 18F p B.E. Zimmerman a,*, G.J. Kubicek a, J.T. Cessna a, P.S. Plascjak b, W.C. Eckelman b a
Physics Laboratory, National Institute of Standards and Technology, Gaithersburg, MD, USA b PET Department, National Institutes of Health, Bethesda, MD, USA Received 22 October 1999; accepted 25 October 1999
Abstract The positron emitter 18F continues to be one of the most important imaging radionuclides in diagnostic nuclear medicine. Assays of radiopharmaceuticals containing this nuclide are often performed in the clinic using commercial reentrant ionization chambers, or ``dose calibrators''. Meaningful quantitative clinical studies require accurate knowledge of the injected activity which requires proper calibration of these instruments. Radioassays were performed at the National Institute of Standards and Technology (NIST) on a solution of 18F produced at the National Institutes of Health (NIH) using 4pb liquid scintillation (fS) counting with 3H-standard eciency tracing. Cocktails containing water fractions of approximately 0.9 and 9% (both as saline) were used. The massic activity values were measured to be 2.5220.06 and 2.5020.03 MBq gÿ1, respectively, for the 0.9 and 9% water cocktails as of the reference time. The uncertainties on the activity measurements are expanded k 2 uncertainties. The largest uncertainty component was found to be the repeatability on a single LS source, with the cocktails containing 0.9% water fraction exhibiting a larger variability by nearly a factor of two. Reproducibility between LS cocktails with the same water fraction was also found to be a large uncertainty component, but with a value less than half that due to measurement repeatability. Radionuclidic impurities consisted of 48V and 46Sc, at levels of 0.11 2 0.08% (expanded uncertainties) and approximately 2 10ÿ3% (upper limit) relative to the activity of the 18F, as of the reference time. Dose calibrator dial settings for measuring solutions of 18F were experimentally determined for Capintec CRC-12 and CRC-35R dose calibrators in three measurement geometries: a 5-ml standard NIST ampoule (two ampoules measured), a 12-ml plastic syringe containing 9 ml of solution and a 10-ml Mallinckrodt molded dose vial ®lled with 5 ml of solution. The experimental dial settings (and the corresponding expanded uncertainties) for these geometries were found to be 477 2 7, 4742 6, 4822 6 and 4632 7 for the two ampoules, the syringe and the dose vial, respectively, in the CRC-12. The dial settings determined for the CRC-35R were 47227, 47027, 464 2 6 and 456 2 6 for the two ampoules, the syringe, and the dose vial, respectively. The uncertainties in the dial settings are expanded uncertainties. Comparisons between the empirically determined dial settings and the manufacturer's recommended setting of ``439'' indicate that use of the manufacturer's setting overestimates the activity by between 3 and 6%, depending upon the geometry used. Published by Elsevier Science Ltd.
Disclaimer: Certain commercial equipment, instruments or materials are identi®ed in this paper to foster understanding. Such identi®cation does not imply recommendation by the National Institute of Standards and Technology, nor does it imply that the materials or equipment identi®ed are necessarily the best available for the purpose. * Corresponding author. Tel.: +1-301-975-5191; fax: +1-301-926-7416. E-mail address: [email protected] (B.E. Zimmerman). p
0969-8043/00/$ - see front matter Published by Elsevier Science Ltd. PII: S 0 9 6 9 - 8 0 4 3 ( 9 9 ) 0 0 2 6 0 - 2
114
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
1. Introduction The positron emitting radionuclide 18F t1=2 109:77 min continues to be one of the most widely used imaging nuclides in diagnostic nuclear medicine, primarily as the radiopharmaceutical 2-[18F]¯uoro-2deoxyglucose (18FDG). Early uses of this agent focused on brain metabolism imaging (Phelps et al., 1979), but recent studies indicate that 18FDG can be eectively employed in applications such as tumor imaging (Conti et al., 1996) and cardiac viability studies (Sandler et al., 1998). Assays of this radiopharmaceutical are normally carried out in the clinic using commercially available reentrant ionization chambers, also known as ``dose calibrators''. In order to obtain accurate results, however, the correct calibration factor, or ``dial setting'', must be employed. Manufacturers of these devices often publish tables recommending a particular dial setting for a radionuclide. In most cases, however, these dial settings are valid only for a single geometry, which is not usually one in which measurements are carried out in the clinic. In the case of one of the most popular dose calibrators, those manufactured by Capintec, Inc., the reference geometry is a 5-ml, thin-walled, borosilicate ampoule used as a standard geometry at the National Institute of Standards and Technology (NIST). Attenuation of the radiation emanating from the source can be greatly aected by changes in the solution density, thickness and composition of the sample container and ®lling volume of the container. Thus, in order to perform accurate assays of a radioactive solution in a container other than this 5-ml ampoule, it is necessary to determine experimentally the dial setting for the particular geometry that will be used. The purpose of this particular study was to con®rm experimentally the manufacturer's recommended setting for 18F in the NIST 5-ml ampoule and to determine the dose calibrator dial settings in two clinically useful geometries: a 12-ml plastic Monoject syringe ®lled with 9 ml of solution and a 10-ml commercially available molded dose vial ®lled with 5 ml of solution. In collaboration with the Positron Emission Tomography (PET) Center of the National Institutes of Health (NIH), the solutions calibrated at NIST were sent to NIH for calibration of their instrumentation. A similar study of this type was previously performed by a NIST (then the National Bureau of Standards (NBS))/NIH collaboration (Coursey et al., 1983), but the emphasis at that time was the development of a standard measurement protocol that could be used for other assays of 18F, primarily by NBS/ NIST. Since that time, a greater emphasis has been placed on the development of ``transfer standards'' Ð protocols that can be applied in a clinical environment
that are based on measurements made at NIST. Because of the widespread use of dose calibrators, this device has become the de facto transfer standard instrument used in nuclear medicine.
2. Experimental A stock solution of anhydrous [18F]¯uoride ion was prepared from aqueous ¯uoride prepared at the NIH PET Center by the 13-MeV proton irradiation of 96% enriched [18O]H2O in a titanium target. The resolubilization procedure consisted of adding 300±500 ml of activity (1.1±1.8 GBq) to a 1-ml v-vial containing 3 mmol of potassium carbonate in 15 ml of water and 6 mmol of Krypto®x-2.2.2. in 30 ml of acetonitrile. The water was removed with argon ¯ow three times using anhydrous acetonitrile on a 1058C heating block. This ¯uoride solution was then used to prepare FDG by a standard procedure (Adams et al., 1995). The ®nal solution of [18F]FDG contained 20 ml of 0.009 mmol lÿ1 phosphate buer in isotonic saline, pH 6.25. Approximately 15 ml of the solution was then transferred to a 20-ml crimp-seal dose vial and driven to NIST, where it arrived approximately 45 min later. The approximate massic activity upon arrival at NIST was 500 MBq gÿ1. The preparation scheme for the various sources used to perform the radioassays in this study are shown in Fig. 1. Two standard NIST 5-ml glass ampoules (labeled ``ampoule 1'' and ``ampoule 2'', respectively) and a 10-ml Mallinckrodt molded dose vial were weighed prior to the start of the experiments. Using an automated dispenser, nominally 5 g of 18F were transferred to each of the containers, which were then immediately weighed to determine the total dispensed mass. The ampoules were then ¯amesealed and the dose vial was crimp-sealed. Response curves of ``apparent activity'' versus dial setting were obtained for each of the three sample geometries by varying the dial setting with the source in the dose calibrator chamber and noting the output activity reading, as well as the time, at each dial setting. Curves were measured for two Capintec dose calibrators maintained at NIST: a Capintec CRC-12 and, for comparison, a Capintec CRC-35R. Details of this technique have been published previously (Zimmerman and Cessna, 1999, 2000; Zimmerman et al., 1999). The typical number of data points collected for the curves were 16 for the CRC-12 and 4 for the CRC-35R. The contents of the dose vial were then gravimetrically transferred to a 12-ml plastic Monoject syringe, the bottom of which had been sealed with epoxy to prevent leakage. Nominally 4 g of normal (0.15 mol lÿ1) saline solution was gravimetrically added to the syringe to bring the total volume to 9 ml. The syringe
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
piston was reinserted into the syringe barrel and the contents agitated for 1 min to ensure complete mixing. Response curves similar to those developed for the other source geometries were obtained for the two dose calibrators. The solution-®lled syringe and ampoule 2 were then transferred back to NIH to be used for calibrating their Capintec CRC-7, PET phantoms and g-ray cells. The total time for source preparation and measurement in the dose calibrators was less than 3 h. Liquid scintillation cocktails for assaying the 18F were prepared from the solution remaining in ampoule 1. Nominally 0.02 g of the 18F solution was gravimetrically dispensed into each of four glass LS vials containing 10 ml of Packard UltimaGold AB scintillator and 1 ml of normal saline solution. These will be referred to as the ``normal water'' cocktails. An additional set of four vials was prepared by addition of nominally 0.02 g of the 18F to vials containing 11 ml of Packard UltimaGold AB scintillator and 0.1 ml of saline solution. These will be referred to as the ``low water'' cocktails. Two background blanks were prepared by dispensing 10 ml of Packard UltimaGold AB scintillator and 1 ml of normal saline into one LS vial, while 11 ml of UltimaGold and 0.1 ml of normal saline
115
were added to the other. The vials were then ®tted with polyethylene v-cone caps and agitated for several minutes to ensure complete mixing of the cocktails. Because of the high massic activity of the 18F solution, the LS count rates were too high to count immediately. Therefore, the cocktails were allowed to decay for about 12 h (six half-lives). The vials were then counted sequentially for 10 cycles of 10 min for each vial in a Beckman LS7800 LS counter equipped with two Hamamatsu R331-05 phototubes operated in coincidence mode with logarithmic signal ampli®cation. The count rates for these cocktails during the ®rst counting cycle ranged from about 15.5 103 sÿ1 for the ®rst cocktail to about 8 103 sÿ1 for the last vial due to decay during the counting cycle. Determinations of the massic activity of the 18F sources were carried out using the CIEMAT/NIST eciency tracing method (Coursey et al. 1986a,b; Zimmerman and ColleÂ, 1997). This method is a protocol by which the LS counting eciency for a cocktail of interest under known, varying quenching conditions is obtained by following the eciency of a closelymatched (in terms of cocktail composition) standard. Liquid scintillation cocktails containing a gravimetric dilution of NIST tritiated water standard, SRM 4927E
Fig. 1. Preparation scheme for sources of [18F]FDG calibrated in this study. The dial settings (DS) shown for each geometry are from the present study.
116
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
(NIST, 1991), were prepared by the gravimetric addition of nominally 0.02 g of the diluted 3H standard to each of 4 LS vials containing 10 ml of Packard UltimaGold AB scintillator and 1 ml of saline solution. As with the 18F solution, an additional set of four LS vials were prepared for the 3H by the gravimetric addition of nominally 0.02 g of the dilute 3H standard to vials containing 11 ml of Packard UltimaGold AB scintillator and 0.1 ml of saline solution. The 3H cocktails were counted in the Beckman LS spectrometer three days previous to preparing the 18F cocktails in order to develop quench correction curves. Appropriate decay corrections to the reference time of 0000 EST 7/17/98 were applied to the LS counting data using a half-life of 109.77 min (NNDC, 1998). This reference time was approximately 14 h after the arrival of the 18F and 2.5 h previous to the start of the LS counting. The total eciency, etot, for 18F was calculated using the following equation: etot bb eb bEC eEC ,
1
where bb and bEC are the respective positron and electron capture (EC) branching ratios and eb and eEC are the calculated eciencies for the respective positron and EC branches. The eciencies for the b + branch were calculated over the observed quenching range using a modi®ed version of the program EFFY4, which is an updated version of EFFY2 (GarcõÂ a-ToranÄo and Grau Malonda, 1985). The modi®cation included an increase in the size of the arrays used in the computation and change in the form of the quenching function Q(E ) (E is the energy of the emitted b-particle). This made the output compatible with other eciencytracing codes used at NIST. The change did not alter the physics of the problem, but changed only the relationship between quenching and the counting eciency. Since quenching is used only as an intermediary variable between the 3H and 18F eciencies, the change has no eect on the ®nal traced eciency for the b + branch. Owing to the extremely low energy of the X-rays and Auger electrons (<1 keV), the eciency for the EC branch was taken to be eEC 0: Input data for the eciency calculations were obtained from the Evaluated Nuclear Structure Data File (ENSDF) at Brookhaven National Laboratory (NNDC, 1998). The actual values employed were: the 3 H maximum beta decay energy Eb,max,H 18:59420:008 keV), the 18F maximum positron decay energy Ebmax,F 633:520:7 keV and the positron branching ratio for 18F bb 0:967320:0004). In order to quantify any possible radionuclidic impurities, two of the LS cocktails were counted again about 5.5, 7.4 and 10.4 d after the reference time for 1
h each. Additionally, the solution remaining in ampoule 1 was analyzed 24 h after the reference time for g-ray emitting radionuclidic impurities using calibrated high-purity germanium (HPGe) detectors. Once the massic activity of the solution was determined, the fraction R CA,Cap =CA,NIST , where CA,Cap and CA,NIST are the respective decay-corrected massic activities of the solution as reported by the dose calibrator and by LS counting, was calculated for each point on each of the response curves. In order to calculate the correct dial setting from each of the calibration curves obtained for each geometry and each dose calibrator, the activity determined from the ``normal'' water fraction LS cocktails was used. Fits of R as a function of the dial setting were then performed for all of the curves and solved for the dial setting value at which R 1: The expanded uncertainty in the dial setting was then determined by solving the ®tting equation for the dial setting values for which R 12U, where U is the expanded uncertainty on the activity. 3. Results and discussion All evaluation of measurement uncertainties throughout this work follow accepted conventions used by the NIST Radioactivity Group and are in concordance with those recommended by the principal metrology standards organizations (Taylor and Kuyatt, 1994). All individual uncertainty components are given as estimated experimental standard deviations (or standard deviations of the mean, if appropriate) or quantities assumed to correspond to standard deviations regardless of the method used to evaluate their magnitude. In accordance with NIST policy, the combined standard uncertainty (calculated by combining the individual uncertainty components in quadrature) is multiplied by a ``coverage factor'' of k 2 to obtain an ``expanded uncertainty'' assumed to give an uncertainty interval having a con®dence level of 90±95%. 3.1. Activity measurements The impurity analysis of the solution using HPGe gray spectrometry indicated the presence of 48V at an activity level of 782 2 21 Bq gÿ1 at the reference time. The uncertainty is a standard uncertainty obtained from combining the standard deviation of ®ve measurements made on the ampoule source and the estimated standard uncertainty in the HPGe eciency. In addition, a small amount of 46Sc was observed at an upper limit of 41 Bq gÿ1 at the reference time. In contrast, the LS data indicated a 48V activity of 50002 200 Bq gÿ1 at the reference time, where the uncertainty
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
is a standard uncertainty calculated from the standard deviation of 8 measurements of the 48V activity. This large discrepancy in the 48V activity between the HPGe and LS data cannot be readily explained. The LS eciency for 48V in the quenching range encountered in these experiments is about 60%, as calculated by EFFY4. Thus, it is certain that the LS value is correct to much better than the factor of 6.4 dierence observed between the two techniques. Since the discrepancy is not readily resolved, the median of the LS and HPGe values, 2900 2 2100 Bq (estimated standard uncertainty) as of the reference time, was adopted as the activity value of the 48V impurity. This value, multiplied by a an eciency factor of 0.6 and decay corrected to the midpoint counting time of each LS cocktail, was subtracted from the counting rates of each of the 18F LS cocktails for all counting cycles to give a net 18F counting rate. The eect on the LS spectrometer deadtime correction of the high counting rates encountered at the beginning of the experiment was investigated by plotting the massic activity calculated for each cycle for each source as a function of the counting rate. If the livetime calculation were in¯uenced by the counting rate, one would expect that the plot would exhibit relatively constant activity values at low count rates, followed by steadily decreasing activity values as the count rate increased. A plot of massic activity versus count rate is presented in Fig. 2 for a typical ``normal'' water fraction LS cocktail. Note that the expected behavior is not
117
observed. The activity values calculated at fairly high count rates are relatively constant, while the values at low count rates are exponentially increasing. At the point where the curve rises steeply (at a count rate of about 104 minÿ1) the activity ratio of 48V to 18F is about 0.3. If the 48V activity is not fully accounted for (which, due to the large uncertainty in the 48V activity, is probably the case, the 18F activity will appear arti®cially high, since the 48V activity is decay-corrected with a much shorter half-life. This eect is almost certainly due to the increasing contribution of the 48V impurity. Regardless of the origin of the eect at low count rate, the calculated activities are fairly ¯at for count rates above 2 104 minÿ1. Experience in this laboratory indicates that the deadtime correction is accurate at count rates as high as 2 105 minÿ1. Thus, there does not appear to be any deleterious eect of the high count rates on the activity calculations. In order to minimize the in¯uence of the uncertainty in the 48V impurity determination, data from only the ®rst 3±4 cycles were used in the calculation of the massic activity of the 18F solution. From the impurity-corrected data, massic activity values of 2.50 2 0.03 and 2.52 2 0.06 MBq gÿ1 were found for the ``normal water'' and ``low water'' LS cocktails, respectively, as of the reference time. The uncertainties on the activities are expanded uncertainties as determined from the uncertainty components detailed in Table 1. When the activities of the 18F and 48V are decay-corrected to the end of bombardment time (700 EST 7/16/98), the ratio of the 48V impurity to the 18F activity is 1.4 10ÿ6.
Fig. 2. Plot of massic activity (in units of Bq gÿ1) as a function of the LS counting rate (in units of reciprocal minutes) for one of the ``normal'' water (9% water fraction) LS cocktails. Note the log scale on the abscissa axis. The line through the data is intended only to guide the eye.
118
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
These data indicate no dependence of the massic activity measurement on the aqueous composition of the LS cocktail, as is observed when measuring some lowenergy beta emitters (Colle and Zimmerman, 1997; Zimmerman and ColleÂ, 1997). There are, however, relatively large variances due to measurement repeatability and reproducibility. Part of the reason for the large sample repeatability uncertainty stems from the fact that the uncertainty in the 48V impurity determination is also partially embodied in the repeatability component. The dierences in variability between the low and high water fraction cocktails, however, suggest that cocktails containing a small aqueous fraction may have some stability problems, causing poor repeatability and reproducibility. This type of behavior has been observed previously with other radionuclides (Zimmerman and ColleÂ, 1997) and further demonstrates that
reliable results are obtainable only with cocktails having aqueous fractions of about 5% or higher. 3.2. Dose calibrator measurements The experimentally-determined dial settings for the four measurement geometries used in this study (two NIST ampoules, a 12-ml plastic Monoject syringe and a 10-ml Mallinckrodt molded dose vial) in the two NIST Capintec dose calibrators are presented in Table 2. The trend in the results is entirely consistent with that expected from examination of the attenuation characteristics of the containers, with the exception of the apparently too low dial setting on the CRC-35R for the syringe geometry. The highest dial setting DS 482 in the CRC-12), corresponding to the least amount of radiation attenuation, is appropriate for the syringe, while the dose vial has the lowest
Table 1 Uncertainty components evaluated in the determination of the massic activity, CA, of [18F]FDG using 4pb liquid scintillation (LS) counting. The uncertainty components were evaluated separately for the LS cocktails with nominally 9% (by volume, denoted ``normal'') and 0.1% (by volume, denoted ``low'') aqueous fractions Uncertainty component
Uncertainty type (A or B)a, description
s1, LS measurement repeatability A: for a standard deviation on the repeated (degrees of freedom n 9 determination of CA for a single LS sample; mean of four cocktails with the same composition, with an uncertainty in the uncertainty estimator of 34%. s2, LS sample reproducibility A: for a standard deviation on the determination of CA for four LS cocktails with the same composition. u3, Gravimetric determinations B: estimated standard uncertainty on the for LS cocktails determination of 18F mass for a single LS cocktail. u4, 18F decay corrections B: for a standard uncertainty of 0.046% in the 18F half-life over 14 h. u5, Impurity determination; B: for an estimated standard uncertainty of 62% in the determination of activity of the 48V impurity. u6, 18F eciency calculations B: step sizes in EFFY4 calculations u7, Livetime determination B; estimated uncertainty in the correction to the LS counting interval uc, Combined standard quadratic combination of all components P uncertainty of uncertainty; uc i ui 1=2 for i 1±7 with ui si for type A components U k uc , Expanded for k 2, which is assumed to correspond uncertaintya to a con®dence level of about 90±95%
Relative uncertainty (%), normal H2O
Relative uncertainty (%), low H2O
0.56
1.0
0.19
0.30
0.05
0.05
0.05
0.05
0.08 (and PE)b
0.08 (and PE)b
1 10-4 0.02 (and PE)c
1 10-4 0.02 (and PE)c
0.60
1.1
1.2
2.2
a Refer to accompanying text for de®nition of terms. Type A uncertainties generally refer to those uncertainty components that can be evaluated using valid statistical methods, while type B uncertainties are based more on scienti®c judgement and/or practical experience and knowledge. b The relative uncertainty for this component is partially embodied (PE) in the relative standard uncertainties of components s1. c The relative uncertainty for this component is partially embodied (PE) in the relative standard uncertainties of components s1 and s2.
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
dial setting due to the thick glass walls. The results for the 5-ml NIST ampoules, which have thinner glass walls than the dose vial, lie between those obtained for the other two geometries. These results also exhibit very good reproducibility between the two ampoules, with the dial settings diering by only two units for both the CRC-12 and the CRC-35R. This corresponds to a dierence in activity of 0.4%, which is within the uncertainty on the LS activity measurements. Dierences in the output activity values as a function of dial setting for the two dose calibrators in each of the measurement geometries are presented graphically in Fig. 3. The dierence in the syringe dial settings between the two dose calibrators is not entirely surprising. This eect can be explained by the fact that the response of the ionization chamber that comprises the dose calibrator is not uniform along its length. Since the syringe is essentially hanging in the chamber at some height from the chamber ¯oor, dierences in the response of the chamber at that particular height can lead to dierences in the apparent activity. Since the ``sample dippers'' that were used are identical to each other (and thus the syringes should be positioned consistently in the two chambers), the only dierence between the two chambers should be the response at that height. From the data, it appears as though the response of the CRC-35R is signi®cantly lower than that of the CRC-12 at the same height. Experiments are planned that will characterize the response of the various chambers employed at NIST as a function of sample height. The result that is surprising, however, is the apparent variable dierence in output activity values as a function of dial setting from the two dose calibrators for other geometries. In an attempt to reproduce these ®ndings, a 5-ml NIST ampoule containing a solution of approximately 185 MBq of 153Sm was inserted into both dose calibrators and the output activity recorded as a function of dial setting. Samarium-153 was chosen since it was immediately available as a calibrated sol-
119
ution, has relatively high-energy emissions and had sucient activity to provide a reliable response in the dose calibrator. The resulting plot of the ratio of decay-corrected output activity values for the NIST Capintec CRC-12 to the decay-corrected output activity values for the NIST Capintec CRC-35R as a function of dial setting is shown in Fig. 4. Not only is the same type of dependence on the dial setting of the output activity ratio observed, but also the curve in Fig. 4 exhibits the same maximum (around 430) as the data in Fig. 3. Additional studies will need to be performed to test whether or not this is a phenomenon observed in all dose calibrators from this manufacturer or if this is an artifact of the two particular instruments installed in our laboratory. Table 2 presents the percentage dierences between the activities obtained at the experimental settings and the manufacturer's recommended setting of ``439''. As is demonstrated by these data, the activities measured in the dose calibrator using the manufacturer's setting of ''439'' for all geometries are in error by between +7.3 and +4.5% for the CRC-12 and between +6.2 and +2.9% for the CRC-35R. Although the regulatory requirements for accuracy in measuring diagnostic radionuclides such as 18F are not as stringent as those for therapeutic nuclides, good measurement practice dictates that the best possible measurement protocol be employed. This includes using the correct calibration factor for the nuclide being measured in the measurement geometry being used. 3.3. NIH dose calibrator measurements The ampoule sent to NIH (ampoule 2) was measured in two dierent Capintec dose calibrators at the manufacturer's dial setting of ``439'', while the syringe was measured in one dose calibrator at the same setting. The results are tabulated in Table 3. The values obtained are in complete agreement with those obtained using similar dose calibrators at NIST.
Table 2 Results of dial setting (DS) determinations for [18F]FDG in the NIST Capintec CRC-12 and CRC-35R dose calibrators in three experimental geometries Geometry
DS, CRC-12
sa
D%, ``439''b
DS, CRC-35R
sa
D%, ``439''b
NIST NIST 12-ml 10-ml
477 474 482 463
7 6 6 7
7.3 6.9 8.3 4.5
472 470 464 456
7 7 6 6
6.2 6.0 3.3 2.9
5-ml ampoule 1 5-ml ampoule 2 Monoject syringe with 9 ml solution Mallinckrodt molded dose vial with 5 ml solution.
a Expanded uncertainties corresponding to the uncertainty in the dial setting, as calculated by propagating the expanded uncertainty in the activity through the respective ®tting equations for each geometry and each dose calibrator. b Dierence in percent of activity measurements made with the manufacturer's recommended dial setting of ``439'' versus the empirical dial setting.
120
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
Fig. 3. Plot of ratios of decay-corrected massic activity readings for 18F from the NIST Capintec CRC-12 (ACRC-12, in MBq gÿ1) and NIST Capintec CRC35R (ACRC-35R, in MBq gÿ1) as a function of dial setting. Data are presented for the NIST 5-ml ampoule (diamonds and squares), a 12-ml Monoject plastic syringe with 9 ml of solution (triangles) and a 10-ml Mallinckrodt molded dose vial with 5 ml of solution (crosses). The estimated standard uncertainty intervals for the data lie within the respective symbols. The lines through the respective data points are intended only to guide the eye.
Fig. 4. Plot of ratios of decay-corrected massic activity readings from the NIST Capintec CRC-12 (ACRC-12, in MBq gÿ1) and NIST Capintec CRC35R (ACRC-35R, in MBq gÿ1) as a function of dial setting. The radiation source was a NIST 5-ml ampoule ®lled with a solution containing 153Sm. The estimated standard uncertainty intervals for the data lie within the respective symbols. The line through the data points is a second-order polynomial ®t, but is intended only to guide the eye.
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
121
Table 3 Results of measurements made on NIST-calibrated [18F]FDG sources using dose calibrators at NIH` Source
Dose calibratora
DA/%b
Ampoule 2 12 ml syringe with 9 ml FDG Ampoule 2
QC Capintec CRC-15R QC Capintec CRC-15R B3 CRC 1245 (no liner)
+8.3% +7.8% +8.2%
a b
Spec®c dose calibrators maintained at NIH. Dierence between the activity reading in the dose calibrator at a dial setting of ``439'' relative to the calibrated activity.
4. Conclusion Radioassays were performed on a solution containing [18F]FDG using 4pb LS counting with 3H-standard eciency tracing. Liquid scintillation cocktails containing approximately 0.9% water (as saline) and 9% water (as saline) were employed. The massic activity values were measured to be 2.5220.06 and 2.5020.03 MBq gÿ1, respectively, for the cocktails containing 0.9 and 9% water as of the reference time. The uncertainties on the activity measurements are expanded k 2 uncertainties. The largest uncertainty component was found to be the repeatability on a single LS source, with the cocktails containing 0.9% water fraction exhibiting a larger variability by nearly a factor of two. Reproducibility between LS cocktails with the same water fraction was also found to be a large uncertainty component, but with a magnitude of less than half that of measurement repeatability. Analysis of radionuclidic impurities indicated the presence of 48V and 46Sc, at levels of 0.1120.08% (1.4 10ÿ4% at end of bombardment) and approximately 2 10ÿ3% (upper limit), respectively, relative to the activity of the 18F as of the reference time. The calibration factors (``dial settings'') for measuring solutions of 18F were experimentally determined for Capintec CRC-12 and CRC-35R dose calibrators in three measurement geometries: a 5-ml standard NIST ampoule (two ampoules measured), a 12-m; plastic syringe containing 9 ml of solution and a 10-ml Mallinckrodt dose vial ®lled with 5 ml of solution. The experimental dial settings for these geometries were found to be 477 2 7, 474 2 6, 482 2 6 and 463 2 7 for the two ampoules, the syringe and the dose vial, respectively, in the CRC-12. The dial settings determined for the CRC-35R were 472 2 7, 470 2 7, 464 2 6 and 45626 for the two ampoules, the syringe and the dose vial, respectively. The uncertainties in the dial settings are expanded uncertainties. Comparison between the empirically determined dial settings and the manufacturer's recommended setting of ``439'' indicate that use of the manufacturer's setting overestimates the activity by between 3 and 6%, depending upon the geometry used. This clearly
demonstrates the need to experimentally determine calibration factors for each radionuclide for each geometry in each individual dose calibrator in order to achieve the most accurate results.
Acknowledgements The authors would like to thank Dr. Francis Schima of the NIST Radioactivity Group for performing the HPGe impurity analyses.
References Adams, H.R., Channing, M.A., Divel, J.E., Dunn, B.B., Kiesewetter, D.O., Plascjak, P., Regdos, S.L., Simpson, N.R., Eckelman, W.C., 1995. Analysis of quality control data. In: Emran, A. (Ed.), Chemists' Views of Imaging Centers. Plenum Press, New York, pp. 175±188. ColleÂ, R., Zimmerman, B.E., 1997. a Compendium on the NIST radionuclidic assays of the massic activity of 63Ni and 55Fe solutions used for an international intercomparison of liquid scintillation techniques. J. Res. Natl. Inst. Stand. Technol. 102, 523±549. Conti, P.S., Lilien, D.L., Hawley, K., Keppler, J., Grafton, S.T., Bading, J.R., 1996. PET and 18F in oncology: a clinical update. Nucl. Med. Biol. 23, 717±736. Coursey, B.M., Hoppes, D.D., Unterweger, M.P., Grau Malonda, A., Margolin, R.A., Kessler, R.M., Manning, R., 1983. Standardization of 18F for use in positron-emission tomography. Int. J. Appl. Radiat. Isot. 34, 1181± 1189. Coursey, B.M., Mann, W.B., Grau Malonda, A., GarcõÁ aToranÄo, E., Los Arcos, J.M., Gibson, J.A.B., Reher, D., 1986a. Standardization of carbon-14 by 4pb liquid scintillation eciency tracing with hydrogen-3. Appl. Radiat. Isot. 37, 403±408. Coursey, B.M., Lucas, L.L., Grau Malonda, A., GarcõÁ aToranÄo, E., 1986b. The standardization of plutonium-241 and nickel-63. Nucl. Instrum. Methods Phys. Res. A 279, 603±610. GarcõÂ a-ToranÄo, E., Grau Malonda, A., 1985. EFFY: a new program to compute the counting eciency of beta particles in liquid scintillators. Comput. Phys. Commun. 36, 307±312.
122
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
National Institute of Standards and Technology (NIST), 1991. Certi®cate, Standard Reference Material 4927E, Radioactivity Standard, Hydrogen-3, Gaithersburg, MD. NNDC, 1998. Data extracted using the NNDC on-line data service from the ENSDF database as of 12/98. M.R. Bhat, Evaluated nuclear structure data ®le (ENSDF). In: Qaim, S.M. (Ed.), Nuclear Data for Science and Technology, Springer-Verlag, Berlin, 1992, p. 817. Phelps, M.E., Huang, S.C., Homan, E.J., Selin, C., Sokolo, L., Kuhl, D.E., 1979. Tomographic measurement of local cerebral glucose metabolic rate in humans with [18F]2¯uoro-2-deoxy-D-glucose: validation of method. Ann. Neurol. 6, 371±388. Sandler, M.P., Bax, J.J., Patton, J.A., Visser, F.C., Martin, W.H., Wijns, W., 1998. Fluorine-18-¯urordeoxygluocose cardiac imaging using a modi®ed scintillation camera. J. Nucl. Med. 39, 2035±2043. Taylor, B.N., Kuyatt, C.E., 1994. Guidelines for evaluating
and expressing the uncertainty of NIST measurement results. NIST Technical Note 1297. Zimmerman, B.E., ColleÂ, R., 1997. Standardization of 63Ni by 4pb liquid scintillation spectrometry with 3H-standard eciency tracing. J. Res. Natl. Inst. Stand. Technol. 102, 455±477. Zimmerman, B.E., Cessna, J.T., 1999. The standardization of 62 Cu and experimental determinations of dose calibrator settings for generator-produced 62CuPTSM. Appl. Radiat. Isot. 51, 515±526. Zimmerman, B.E., Cessna, J.T., 2000. Experimental determinations of commercial dose calibrator settings for nuclides used in nuclear medicine. Appl. Radiat. Isot. 52, 615±619. Zimmerman, B.E., Cessna, J.T., Unterweger, M.P., Li, A.N., Whiting, J.S., Knapp, F.F., 1999. A new experimental determination of the dose calibrator setting for 188Re. J. Nucl. Med. 40, 1508±1516.
www.elsevier.com/locate/apradiso
Radioassays and experimental evaluation of dose calibrator settings for 18F p B.E. Zimmerman a,*, G.J. Kubicek a, J.T. Cessna a, P.S. Plascjak b, W.C. Eckelman b a
Physics Laboratory, National Institute of Standards and Technology, Gaithersburg, MD, USA b PET Department, National Institutes of Health, Bethesda, MD, USA Received 22 October 1999; accepted 25 October 1999
Abstract The positron emitter 18F continues to be one of the most important imaging radionuclides in diagnostic nuclear medicine. Assays of radiopharmaceuticals containing this nuclide are often performed in the clinic using commercial reentrant ionization chambers, or ``dose calibrators''. Meaningful quantitative clinical studies require accurate knowledge of the injected activity which requires proper calibration of these instruments. Radioassays were performed at the National Institute of Standards and Technology (NIST) on a solution of 18F produced at the National Institutes of Health (NIH) using 4pb liquid scintillation (fS) counting with 3H-standard eciency tracing. Cocktails containing water fractions of approximately 0.9 and 9% (both as saline) were used. The massic activity values were measured to be 2.5220.06 and 2.5020.03 MBq gÿ1, respectively, for the 0.9 and 9% water cocktails as of the reference time. The uncertainties on the activity measurements are expanded k 2 uncertainties. The largest uncertainty component was found to be the repeatability on a single LS source, with the cocktails containing 0.9% water fraction exhibiting a larger variability by nearly a factor of two. Reproducibility between LS cocktails with the same water fraction was also found to be a large uncertainty component, but with a value less than half that due to measurement repeatability. Radionuclidic impurities consisted of 48V and 46Sc, at levels of 0.11 2 0.08% (expanded uncertainties) and approximately 2 10ÿ3% (upper limit) relative to the activity of the 18F, as of the reference time. Dose calibrator dial settings for measuring solutions of 18F were experimentally determined for Capintec CRC-12 and CRC-35R dose calibrators in three measurement geometries: a 5-ml standard NIST ampoule (two ampoules measured), a 12-ml plastic syringe containing 9 ml of solution and a 10-ml Mallinckrodt molded dose vial ®lled with 5 ml of solution. The experimental dial settings (and the corresponding expanded uncertainties) for these geometries were found to be 477 2 7, 4742 6, 4822 6 and 4632 7 for the two ampoules, the syringe and the dose vial, respectively, in the CRC-12. The dial settings determined for the CRC-35R were 47227, 47027, 464 2 6 and 456 2 6 for the two ampoules, the syringe, and the dose vial, respectively. The uncertainties in the dial settings are expanded uncertainties. Comparisons between the empirically determined dial settings and the manufacturer's recommended setting of ``439'' indicate that use of the manufacturer's setting overestimates the activity by between 3 and 6%, depending upon the geometry used. Published by Elsevier Science Ltd.
Disclaimer: Certain commercial equipment, instruments or materials are identi®ed in this paper to foster understanding. Such identi®cation does not imply recommendation by the National Institute of Standards and Technology, nor does it imply that the materials or equipment identi®ed are necessarily the best available for the purpose. * Corresponding author. Tel.: +1-301-975-5191; fax: +1-301-926-7416. E-mail address: [email protected] (B.E. Zimmerman). p
0969-8043/00/$ - see front matter Published by Elsevier Science Ltd. PII: S 0 9 6 9 - 8 0 4 3 ( 9 9 ) 0 0 2 6 0 - 2
114
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
1. Introduction The positron emitting radionuclide 18F t1=2 109:77 min continues to be one of the most widely used imaging nuclides in diagnostic nuclear medicine, primarily as the radiopharmaceutical 2-[18F]¯uoro-2deoxyglucose (18FDG). Early uses of this agent focused on brain metabolism imaging (Phelps et al., 1979), but recent studies indicate that 18FDG can be eectively employed in applications such as tumor imaging (Conti et al., 1996) and cardiac viability studies (Sandler et al., 1998). Assays of this radiopharmaceutical are normally carried out in the clinic using commercially available reentrant ionization chambers, also known as ``dose calibrators''. In order to obtain accurate results, however, the correct calibration factor, or ``dial setting'', must be employed. Manufacturers of these devices often publish tables recommending a particular dial setting for a radionuclide. In most cases, however, these dial settings are valid only for a single geometry, which is not usually one in which measurements are carried out in the clinic. In the case of one of the most popular dose calibrators, those manufactured by Capintec, Inc., the reference geometry is a 5-ml, thin-walled, borosilicate ampoule used as a standard geometry at the National Institute of Standards and Technology (NIST). Attenuation of the radiation emanating from the source can be greatly aected by changes in the solution density, thickness and composition of the sample container and ®lling volume of the container. Thus, in order to perform accurate assays of a radioactive solution in a container other than this 5-ml ampoule, it is necessary to determine experimentally the dial setting for the particular geometry that will be used. The purpose of this particular study was to con®rm experimentally the manufacturer's recommended setting for 18F in the NIST 5-ml ampoule and to determine the dose calibrator dial settings in two clinically useful geometries: a 12-ml plastic Monoject syringe ®lled with 9 ml of solution and a 10-ml commercially available molded dose vial ®lled with 5 ml of solution. In collaboration with the Positron Emission Tomography (PET) Center of the National Institutes of Health (NIH), the solutions calibrated at NIST were sent to NIH for calibration of their instrumentation. A similar study of this type was previously performed by a NIST (then the National Bureau of Standards (NBS))/NIH collaboration (Coursey et al., 1983), but the emphasis at that time was the development of a standard measurement protocol that could be used for other assays of 18F, primarily by NBS/ NIST. Since that time, a greater emphasis has been placed on the development of ``transfer standards'' Ð protocols that can be applied in a clinical environment
that are based on measurements made at NIST. Because of the widespread use of dose calibrators, this device has become the de facto transfer standard instrument used in nuclear medicine.
2. Experimental A stock solution of anhydrous [18F]¯uoride ion was prepared from aqueous ¯uoride prepared at the NIH PET Center by the 13-MeV proton irradiation of 96% enriched [18O]H2O in a titanium target. The resolubilization procedure consisted of adding 300±500 ml of activity (1.1±1.8 GBq) to a 1-ml v-vial containing 3 mmol of potassium carbonate in 15 ml of water and 6 mmol of Krypto®x-2.2.2. in 30 ml of acetonitrile. The water was removed with argon ¯ow three times using anhydrous acetonitrile on a 1058C heating block. This ¯uoride solution was then used to prepare FDG by a standard procedure (Adams et al., 1995). The ®nal solution of [18F]FDG contained 20 ml of 0.009 mmol lÿ1 phosphate buer in isotonic saline, pH 6.25. Approximately 15 ml of the solution was then transferred to a 20-ml crimp-seal dose vial and driven to NIST, where it arrived approximately 45 min later. The approximate massic activity upon arrival at NIST was 500 MBq gÿ1. The preparation scheme for the various sources used to perform the radioassays in this study are shown in Fig. 1. Two standard NIST 5-ml glass ampoules (labeled ``ampoule 1'' and ``ampoule 2'', respectively) and a 10-ml Mallinckrodt molded dose vial were weighed prior to the start of the experiments. Using an automated dispenser, nominally 5 g of 18F were transferred to each of the containers, which were then immediately weighed to determine the total dispensed mass. The ampoules were then ¯amesealed and the dose vial was crimp-sealed. Response curves of ``apparent activity'' versus dial setting were obtained for each of the three sample geometries by varying the dial setting with the source in the dose calibrator chamber and noting the output activity reading, as well as the time, at each dial setting. Curves were measured for two Capintec dose calibrators maintained at NIST: a Capintec CRC-12 and, for comparison, a Capintec CRC-35R. Details of this technique have been published previously (Zimmerman and Cessna, 1999, 2000; Zimmerman et al., 1999). The typical number of data points collected for the curves were 16 for the CRC-12 and 4 for the CRC-35R. The contents of the dose vial were then gravimetrically transferred to a 12-ml plastic Monoject syringe, the bottom of which had been sealed with epoxy to prevent leakage. Nominally 4 g of normal (0.15 mol lÿ1) saline solution was gravimetrically added to the syringe to bring the total volume to 9 ml. The syringe
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
piston was reinserted into the syringe barrel and the contents agitated for 1 min to ensure complete mixing. Response curves similar to those developed for the other source geometries were obtained for the two dose calibrators. The solution-®lled syringe and ampoule 2 were then transferred back to NIH to be used for calibrating their Capintec CRC-7, PET phantoms and g-ray cells. The total time for source preparation and measurement in the dose calibrators was less than 3 h. Liquid scintillation cocktails for assaying the 18F were prepared from the solution remaining in ampoule 1. Nominally 0.02 g of the 18F solution was gravimetrically dispensed into each of four glass LS vials containing 10 ml of Packard UltimaGold AB scintillator and 1 ml of normal saline solution. These will be referred to as the ``normal water'' cocktails. An additional set of four vials was prepared by addition of nominally 0.02 g of the 18F to vials containing 11 ml of Packard UltimaGold AB scintillator and 0.1 ml of saline solution. These will be referred to as the ``low water'' cocktails. Two background blanks were prepared by dispensing 10 ml of Packard UltimaGold AB scintillator and 1 ml of normal saline into one LS vial, while 11 ml of UltimaGold and 0.1 ml of normal saline
115
were added to the other. The vials were then ®tted with polyethylene v-cone caps and agitated for several minutes to ensure complete mixing of the cocktails. Because of the high massic activity of the 18F solution, the LS count rates were too high to count immediately. Therefore, the cocktails were allowed to decay for about 12 h (six half-lives). The vials were then counted sequentially for 10 cycles of 10 min for each vial in a Beckman LS7800 LS counter equipped with two Hamamatsu R331-05 phototubes operated in coincidence mode with logarithmic signal ampli®cation. The count rates for these cocktails during the ®rst counting cycle ranged from about 15.5 103 sÿ1 for the ®rst cocktail to about 8 103 sÿ1 for the last vial due to decay during the counting cycle. Determinations of the massic activity of the 18F sources were carried out using the CIEMAT/NIST eciency tracing method (Coursey et al. 1986a,b; Zimmerman and ColleÂ, 1997). This method is a protocol by which the LS counting eciency for a cocktail of interest under known, varying quenching conditions is obtained by following the eciency of a closelymatched (in terms of cocktail composition) standard. Liquid scintillation cocktails containing a gravimetric dilution of NIST tritiated water standard, SRM 4927E
Fig. 1. Preparation scheme for sources of [18F]FDG calibrated in this study. The dial settings (DS) shown for each geometry are from the present study.
116
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
(NIST, 1991), were prepared by the gravimetric addition of nominally 0.02 g of the diluted 3H standard to each of 4 LS vials containing 10 ml of Packard UltimaGold AB scintillator and 1 ml of saline solution. As with the 18F solution, an additional set of four LS vials were prepared for the 3H by the gravimetric addition of nominally 0.02 g of the dilute 3H standard to vials containing 11 ml of Packard UltimaGold AB scintillator and 0.1 ml of saline solution. The 3H cocktails were counted in the Beckman LS spectrometer three days previous to preparing the 18F cocktails in order to develop quench correction curves. Appropriate decay corrections to the reference time of 0000 EST 7/17/98 were applied to the LS counting data using a half-life of 109.77 min (NNDC, 1998). This reference time was approximately 14 h after the arrival of the 18F and 2.5 h previous to the start of the LS counting. The total eciency, etot, for 18F was calculated using the following equation: etot bb eb bEC eEC ,
1
where bb and bEC are the respective positron and electron capture (EC) branching ratios and eb and eEC are the calculated eciencies for the respective positron and EC branches. The eciencies for the b + branch were calculated over the observed quenching range using a modi®ed version of the program EFFY4, which is an updated version of EFFY2 (GarcõÂ a-ToranÄo and Grau Malonda, 1985). The modi®cation included an increase in the size of the arrays used in the computation and change in the form of the quenching function Q(E ) (E is the energy of the emitted b-particle). This made the output compatible with other eciencytracing codes used at NIST. The change did not alter the physics of the problem, but changed only the relationship between quenching and the counting eciency. Since quenching is used only as an intermediary variable between the 3H and 18F eciencies, the change has no eect on the ®nal traced eciency for the b + branch. Owing to the extremely low energy of the X-rays and Auger electrons (<1 keV), the eciency for the EC branch was taken to be eEC 0: Input data for the eciency calculations were obtained from the Evaluated Nuclear Structure Data File (ENSDF) at Brookhaven National Laboratory (NNDC, 1998). The actual values employed were: the 3 H maximum beta decay energy Eb,max,H 18:59420:008 keV), the 18F maximum positron decay energy Ebmax,F 633:520:7 keV and the positron branching ratio for 18F bb 0:967320:0004). In order to quantify any possible radionuclidic impurities, two of the LS cocktails were counted again about 5.5, 7.4 and 10.4 d after the reference time for 1
h each. Additionally, the solution remaining in ampoule 1 was analyzed 24 h after the reference time for g-ray emitting radionuclidic impurities using calibrated high-purity germanium (HPGe) detectors. Once the massic activity of the solution was determined, the fraction R CA,Cap =CA,NIST , where CA,Cap and CA,NIST are the respective decay-corrected massic activities of the solution as reported by the dose calibrator and by LS counting, was calculated for each point on each of the response curves. In order to calculate the correct dial setting from each of the calibration curves obtained for each geometry and each dose calibrator, the activity determined from the ``normal'' water fraction LS cocktails was used. Fits of R as a function of the dial setting were then performed for all of the curves and solved for the dial setting value at which R 1: The expanded uncertainty in the dial setting was then determined by solving the ®tting equation for the dial setting values for which R 12U, where U is the expanded uncertainty on the activity. 3. Results and discussion All evaluation of measurement uncertainties throughout this work follow accepted conventions used by the NIST Radioactivity Group and are in concordance with those recommended by the principal metrology standards organizations (Taylor and Kuyatt, 1994). All individual uncertainty components are given as estimated experimental standard deviations (or standard deviations of the mean, if appropriate) or quantities assumed to correspond to standard deviations regardless of the method used to evaluate their magnitude. In accordance with NIST policy, the combined standard uncertainty (calculated by combining the individual uncertainty components in quadrature) is multiplied by a ``coverage factor'' of k 2 to obtain an ``expanded uncertainty'' assumed to give an uncertainty interval having a con®dence level of 90±95%. 3.1. Activity measurements The impurity analysis of the solution using HPGe gray spectrometry indicated the presence of 48V at an activity level of 782 2 21 Bq gÿ1 at the reference time. The uncertainty is a standard uncertainty obtained from combining the standard deviation of ®ve measurements made on the ampoule source and the estimated standard uncertainty in the HPGe eciency. In addition, a small amount of 46Sc was observed at an upper limit of 41 Bq gÿ1 at the reference time. In contrast, the LS data indicated a 48V activity of 50002 200 Bq gÿ1 at the reference time, where the uncertainty
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
is a standard uncertainty calculated from the standard deviation of 8 measurements of the 48V activity. This large discrepancy in the 48V activity between the HPGe and LS data cannot be readily explained. The LS eciency for 48V in the quenching range encountered in these experiments is about 60%, as calculated by EFFY4. Thus, it is certain that the LS value is correct to much better than the factor of 6.4 dierence observed between the two techniques. Since the discrepancy is not readily resolved, the median of the LS and HPGe values, 2900 2 2100 Bq (estimated standard uncertainty) as of the reference time, was adopted as the activity value of the 48V impurity. This value, multiplied by a an eciency factor of 0.6 and decay corrected to the midpoint counting time of each LS cocktail, was subtracted from the counting rates of each of the 18F LS cocktails for all counting cycles to give a net 18F counting rate. The eect on the LS spectrometer deadtime correction of the high counting rates encountered at the beginning of the experiment was investigated by plotting the massic activity calculated for each cycle for each source as a function of the counting rate. If the livetime calculation were in¯uenced by the counting rate, one would expect that the plot would exhibit relatively constant activity values at low count rates, followed by steadily decreasing activity values as the count rate increased. A plot of massic activity versus count rate is presented in Fig. 2 for a typical ``normal'' water fraction LS cocktail. Note that the expected behavior is not
117
observed. The activity values calculated at fairly high count rates are relatively constant, while the values at low count rates are exponentially increasing. At the point where the curve rises steeply (at a count rate of about 104 minÿ1) the activity ratio of 48V to 18F is about 0.3. If the 48V activity is not fully accounted for (which, due to the large uncertainty in the 48V activity, is probably the case, the 18F activity will appear arti®cially high, since the 48V activity is decay-corrected with a much shorter half-life. This eect is almost certainly due to the increasing contribution of the 48V impurity. Regardless of the origin of the eect at low count rate, the calculated activities are fairly ¯at for count rates above 2 104 minÿ1. Experience in this laboratory indicates that the deadtime correction is accurate at count rates as high as 2 105 minÿ1. Thus, there does not appear to be any deleterious eect of the high count rates on the activity calculations. In order to minimize the in¯uence of the uncertainty in the 48V impurity determination, data from only the ®rst 3±4 cycles were used in the calculation of the massic activity of the 18F solution. From the impurity-corrected data, massic activity values of 2.50 2 0.03 and 2.52 2 0.06 MBq gÿ1 were found for the ``normal water'' and ``low water'' LS cocktails, respectively, as of the reference time. The uncertainties on the activities are expanded uncertainties as determined from the uncertainty components detailed in Table 1. When the activities of the 18F and 48V are decay-corrected to the end of bombardment time (700 EST 7/16/98), the ratio of the 48V impurity to the 18F activity is 1.4 10ÿ6.
Fig. 2. Plot of massic activity (in units of Bq gÿ1) as a function of the LS counting rate (in units of reciprocal minutes) for one of the ``normal'' water (9% water fraction) LS cocktails. Note the log scale on the abscissa axis. The line through the data is intended only to guide the eye.
118
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
These data indicate no dependence of the massic activity measurement on the aqueous composition of the LS cocktail, as is observed when measuring some lowenergy beta emitters (Colle and Zimmerman, 1997; Zimmerman and ColleÂ, 1997). There are, however, relatively large variances due to measurement repeatability and reproducibility. Part of the reason for the large sample repeatability uncertainty stems from the fact that the uncertainty in the 48V impurity determination is also partially embodied in the repeatability component. The dierences in variability between the low and high water fraction cocktails, however, suggest that cocktails containing a small aqueous fraction may have some stability problems, causing poor repeatability and reproducibility. This type of behavior has been observed previously with other radionuclides (Zimmerman and ColleÂ, 1997) and further demonstrates that
reliable results are obtainable only with cocktails having aqueous fractions of about 5% or higher. 3.2. Dose calibrator measurements The experimentally-determined dial settings for the four measurement geometries used in this study (two NIST ampoules, a 12-ml plastic Monoject syringe and a 10-ml Mallinckrodt molded dose vial) in the two NIST Capintec dose calibrators are presented in Table 2. The trend in the results is entirely consistent with that expected from examination of the attenuation characteristics of the containers, with the exception of the apparently too low dial setting on the CRC-35R for the syringe geometry. The highest dial setting DS 482 in the CRC-12), corresponding to the least amount of radiation attenuation, is appropriate for the syringe, while the dose vial has the lowest
Table 1 Uncertainty components evaluated in the determination of the massic activity, CA, of [18F]FDG using 4pb liquid scintillation (LS) counting. The uncertainty components were evaluated separately for the LS cocktails with nominally 9% (by volume, denoted ``normal'') and 0.1% (by volume, denoted ``low'') aqueous fractions Uncertainty component
Uncertainty type (A or B)a, description
s1, LS measurement repeatability A: for a standard deviation on the repeated (degrees of freedom n 9 determination of CA for a single LS sample; mean of four cocktails with the same composition, with an uncertainty in the uncertainty estimator of 34%. s2, LS sample reproducibility A: for a standard deviation on the determination of CA for four LS cocktails with the same composition. u3, Gravimetric determinations B: estimated standard uncertainty on the for LS cocktails determination of 18F mass for a single LS cocktail. u4, 18F decay corrections B: for a standard uncertainty of 0.046% in the 18F half-life over 14 h. u5, Impurity determination; B: for an estimated standard uncertainty of 62% in the determination of activity of the 48V impurity. u6, 18F eciency calculations B: step sizes in EFFY4 calculations u7, Livetime determination B; estimated uncertainty in the correction to the LS counting interval uc, Combined standard quadratic combination of all components P uncertainty of uncertainty; uc i ui 1=2 for i 1±7 with ui si for type A components U k uc , Expanded for k 2, which is assumed to correspond uncertaintya to a con®dence level of about 90±95%
Relative uncertainty (%), normal H2O
Relative uncertainty (%), low H2O
0.56
1.0
0.19
0.30
0.05
0.05
0.05
0.05
0.08 (and PE)b
0.08 (and PE)b
1 10-4 0.02 (and PE)c
1 10-4 0.02 (and PE)c
0.60
1.1
1.2
2.2
a Refer to accompanying text for de®nition of terms. Type A uncertainties generally refer to those uncertainty components that can be evaluated using valid statistical methods, while type B uncertainties are based more on scienti®c judgement and/or practical experience and knowledge. b The relative uncertainty for this component is partially embodied (PE) in the relative standard uncertainties of components s1. c The relative uncertainty for this component is partially embodied (PE) in the relative standard uncertainties of components s1 and s2.
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
dial setting due to the thick glass walls. The results for the 5-ml NIST ampoules, which have thinner glass walls than the dose vial, lie between those obtained for the other two geometries. These results also exhibit very good reproducibility between the two ampoules, with the dial settings diering by only two units for both the CRC-12 and the CRC-35R. This corresponds to a dierence in activity of 0.4%, which is within the uncertainty on the LS activity measurements. Dierences in the output activity values as a function of dial setting for the two dose calibrators in each of the measurement geometries are presented graphically in Fig. 3. The dierence in the syringe dial settings between the two dose calibrators is not entirely surprising. This eect can be explained by the fact that the response of the ionization chamber that comprises the dose calibrator is not uniform along its length. Since the syringe is essentially hanging in the chamber at some height from the chamber ¯oor, dierences in the response of the chamber at that particular height can lead to dierences in the apparent activity. Since the ``sample dippers'' that were used are identical to each other (and thus the syringes should be positioned consistently in the two chambers), the only dierence between the two chambers should be the response at that height. From the data, it appears as though the response of the CRC-35R is signi®cantly lower than that of the CRC-12 at the same height. Experiments are planned that will characterize the response of the various chambers employed at NIST as a function of sample height. The result that is surprising, however, is the apparent variable dierence in output activity values as a function of dial setting from the two dose calibrators for other geometries. In an attempt to reproduce these ®ndings, a 5-ml NIST ampoule containing a solution of approximately 185 MBq of 153Sm was inserted into both dose calibrators and the output activity recorded as a function of dial setting. Samarium-153 was chosen since it was immediately available as a calibrated sol-
119
ution, has relatively high-energy emissions and had sucient activity to provide a reliable response in the dose calibrator. The resulting plot of the ratio of decay-corrected output activity values for the NIST Capintec CRC-12 to the decay-corrected output activity values for the NIST Capintec CRC-35R as a function of dial setting is shown in Fig. 4. Not only is the same type of dependence on the dial setting of the output activity ratio observed, but also the curve in Fig. 4 exhibits the same maximum (around 430) as the data in Fig. 3. Additional studies will need to be performed to test whether or not this is a phenomenon observed in all dose calibrators from this manufacturer or if this is an artifact of the two particular instruments installed in our laboratory. Table 2 presents the percentage dierences between the activities obtained at the experimental settings and the manufacturer's recommended setting of ``439''. As is demonstrated by these data, the activities measured in the dose calibrator using the manufacturer's setting of ''439'' for all geometries are in error by between +7.3 and +4.5% for the CRC-12 and between +6.2 and +2.9% for the CRC-35R. Although the regulatory requirements for accuracy in measuring diagnostic radionuclides such as 18F are not as stringent as those for therapeutic nuclides, good measurement practice dictates that the best possible measurement protocol be employed. This includes using the correct calibration factor for the nuclide being measured in the measurement geometry being used. 3.3. NIH dose calibrator measurements The ampoule sent to NIH (ampoule 2) was measured in two dierent Capintec dose calibrators at the manufacturer's dial setting of ``439'', while the syringe was measured in one dose calibrator at the same setting. The results are tabulated in Table 3. The values obtained are in complete agreement with those obtained using similar dose calibrators at NIST.
Table 2 Results of dial setting (DS) determinations for [18F]FDG in the NIST Capintec CRC-12 and CRC-35R dose calibrators in three experimental geometries Geometry
DS, CRC-12
sa
D%, ``439''b
DS, CRC-35R
sa
D%, ``439''b
NIST NIST 12-ml 10-ml
477 474 482 463
7 6 6 7
7.3 6.9 8.3 4.5
472 470 464 456
7 7 6 6
6.2 6.0 3.3 2.9
5-ml ampoule 1 5-ml ampoule 2 Monoject syringe with 9 ml solution Mallinckrodt molded dose vial with 5 ml solution.
a Expanded uncertainties corresponding to the uncertainty in the dial setting, as calculated by propagating the expanded uncertainty in the activity through the respective ®tting equations for each geometry and each dose calibrator. b Dierence in percent of activity measurements made with the manufacturer's recommended dial setting of ``439'' versus the empirical dial setting.
120
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
Fig. 3. Plot of ratios of decay-corrected massic activity readings for 18F from the NIST Capintec CRC-12 (ACRC-12, in MBq gÿ1) and NIST Capintec CRC35R (ACRC-35R, in MBq gÿ1) as a function of dial setting. Data are presented for the NIST 5-ml ampoule (diamonds and squares), a 12-ml Monoject plastic syringe with 9 ml of solution (triangles) and a 10-ml Mallinckrodt molded dose vial with 5 ml of solution (crosses). The estimated standard uncertainty intervals for the data lie within the respective symbols. The lines through the respective data points are intended only to guide the eye.
Fig. 4. Plot of ratios of decay-corrected massic activity readings from the NIST Capintec CRC-12 (ACRC-12, in MBq gÿ1) and NIST Capintec CRC35R (ACRC-35R, in MBq gÿ1) as a function of dial setting. The radiation source was a NIST 5-ml ampoule ®lled with a solution containing 153Sm. The estimated standard uncertainty intervals for the data lie within the respective symbols. The line through the data points is a second-order polynomial ®t, but is intended only to guide the eye.
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
121
Table 3 Results of measurements made on NIST-calibrated [18F]FDG sources using dose calibrators at NIH` Source
Dose calibratora
DA/%b
Ampoule 2 12 ml syringe with 9 ml FDG Ampoule 2
QC Capintec CRC-15R QC Capintec CRC-15R B3 CRC 1245 (no liner)
+8.3% +7.8% +8.2%
a b
Spec®c dose calibrators maintained at NIH. Dierence between the activity reading in the dose calibrator at a dial setting of ``439'' relative to the calibrated activity.
4. Conclusion Radioassays were performed on a solution containing [18F]FDG using 4pb LS counting with 3H-standard eciency tracing. Liquid scintillation cocktails containing approximately 0.9% water (as saline) and 9% water (as saline) were employed. The massic activity values were measured to be 2.5220.06 and 2.5020.03 MBq gÿ1, respectively, for the cocktails containing 0.9 and 9% water as of the reference time. The uncertainties on the activity measurements are expanded k 2 uncertainties. The largest uncertainty component was found to be the repeatability on a single LS source, with the cocktails containing 0.9% water fraction exhibiting a larger variability by nearly a factor of two. Reproducibility between LS cocktails with the same water fraction was also found to be a large uncertainty component, but with a magnitude of less than half that of measurement repeatability. Analysis of radionuclidic impurities indicated the presence of 48V and 46Sc, at levels of 0.1120.08% (1.4 10ÿ4% at end of bombardment) and approximately 2 10ÿ3% (upper limit), respectively, relative to the activity of the 18F as of the reference time. The calibration factors (``dial settings'') for measuring solutions of 18F were experimentally determined for Capintec CRC-12 and CRC-35R dose calibrators in three measurement geometries: a 5-ml standard NIST ampoule (two ampoules measured), a 12-m; plastic syringe containing 9 ml of solution and a 10-ml Mallinckrodt dose vial ®lled with 5 ml of solution. The experimental dial settings for these geometries were found to be 477 2 7, 474 2 6, 482 2 6 and 463 2 7 for the two ampoules, the syringe and the dose vial, respectively, in the CRC-12. The dial settings determined for the CRC-35R were 472 2 7, 470 2 7, 464 2 6 and 45626 for the two ampoules, the syringe and the dose vial, respectively. The uncertainties in the dial settings are expanded uncertainties. Comparison between the empirically determined dial settings and the manufacturer's recommended setting of ``439'' indicate that use of the manufacturer's setting overestimates the activity by between 3 and 6%, depending upon the geometry used. This clearly
demonstrates the need to experimentally determine calibration factors for each radionuclide for each geometry in each individual dose calibrator in order to achieve the most accurate results.
Acknowledgements The authors would like to thank Dr. Francis Schima of the NIST Radioactivity Group for performing the HPGe impurity analyses.
References Adams, H.R., Channing, M.A., Divel, J.E., Dunn, B.B., Kiesewetter, D.O., Plascjak, P., Regdos, S.L., Simpson, N.R., Eckelman, W.C., 1995. Analysis of quality control data. In: Emran, A. (Ed.), Chemists' Views of Imaging Centers. Plenum Press, New York, pp. 175±188. ColleÂ, R., Zimmerman, B.E., 1997. a Compendium on the NIST radionuclidic assays of the massic activity of 63Ni and 55Fe solutions used for an international intercomparison of liquid scintillation techniques. J. Res. Natl. Inst. Stand. Technol. 102, 523±549. Conti, P.S., Lilien, D.L., Hawley, K., Keppler, J., Grafton, S.T., Bading, J.R., 1996. PET and 18F in oncology: a clinical update. Nucl. Med. Biol. 23, 717±736. Coursey, B.M., Hoppes, D.D., Unterweger, M.P., Grau Malonda, A., Margolin, R.A., Kessler, R.M., Manning, R., 1983. Standardization of 18F for use in positron-emission tomography. Int. J. Appl. Radiat. Isot. 34, 1181± 1189. Coursey, B.M., Mann, W.B., Grau Malonda, A., GarcõÁ aToranÄo, E., Los Arcos, J.M., Gibson, J.A.B., Reher, D., 1986a. Standardization of carbon-14 by 4pb liquid scintillation eciency tracing with hydrogen-3. Appl. Radiat. Isot. 37, 403±408. Coursey, B.M., Lucas, L.L., Grau Malonda, A., GarcõÁ aToranÄo, E., 1986b. The standardization of plutonium-241 and nickel-63. Nucl. Instrum. Methods Phys. Res. A 279, 603±610. GarcõÂ a-ToranÄo, E., Grau Malonda, A., 1985. EFFY: a new program to compute the counting eciency of beta particles in liquid scintillators. Comput. Phys. Commun. 36, 307±312.
122
B.E. Zimmerman et al. / Applied Radiation and Isotopes 54 (2001) 113±122
National Institute of Standards and Technology (NIST), 1991. Certi®cate, Standard Reference Material 4927E, Radioactivity Standard, Hydrogen-3, Gaithersburg, MD. NNDC, 1998. Data extracted using the NNDC on-line data service from the ENSDF database as of 12/98. M.R. Bhat, Evaluated nuclear structure data ®le (ENSDF). In: Qaim, S.M. (Ed.), Nuclear Data for Science and Technology, Springer-Verlag, Berlin, 1992, p. 817. Phelps, M.E., Huang, S.C., Homan, E.J., Selin, C., Sokolo, L., Kuhl, D.E., 1979. Tomographic measurement of local cerebral glucose metabolic rate in humans with [18F]2¯uoro-2-deoxy-D-glucose: validation of method. Ann. Neurol. 6, 371±388. Sandler, M.P., Bax, J.J., Patton, J.A., Visser, F.C., Martin, W.H., Wijns, W., 1998. Fluorine-18-¯urordeoxygluocose cardiac imaging using a modi®ed scintillation camera. J. Nucl. Med. 39, 2035±2043. Taylor, B.N., Kuyatt, C.E., 1994. Guidelines for evaluating
and expressing the uncertainty of NIST measurement results. NIST Technical Note 1297. Zimmerman, B.E., ColleÂ, R., 1997. Standardization of 63Ni by 4pb liquid scintillation spectrometry with 3H-standard eciency tracing. J. Res. Natl. Inst. Stand. Technol. 102, 455±477. Zimmerman, B.E., Cessna, J.T., 1999. The standardization of 62 Cu and experimental determinations of dose calibrator settings for generator-produced 62CuPTSM. Appl. Radiat. Isot. 51, 515±526. Zimmerman, B.E., Cessna, J.T., 2000. Experimental determinations of commercial dose calibrator settings for nuclides used in nuclear medicine. Appl. Radiat. Isot. 52, 615±619. Zimmerman, B.E., Cessna, J.T., Unterweger, M.P., Li, A.N., Whiting, J.S., Knapp, F.F., 1999. A new experimental determination of the dose calibrator setting for 188Re. J. Nucl. Med. 40, 1508±1516.