Bureau International des Poids et Mesures (BIPM),

Activity measurements of the radionuclides 18 F and 99m Tc for the NMISA, South Africa in the ongoing comparisons BIPM.RI(II)-K4.F-18 and BIPM.RI(II)-K4.Tc-99m (with erratum) C. Michotte 1, M. Nonis 1, M. W. Van Rooy 2, M. J. Van Staden 2, J. Lubbe 2. 1 Bureau International des Poids et Mesures (BIPM), 2 National Metrology Institute of South Africa (NMISA) Abstract In 2015, comparisons of activity measurements of 18 F and 99m Tc using the Transfer Instrument of the International Reference System (SIRTI) took place at the National Metrology Institute of South Africa (NMISA, South Africa). Ampoules containing about 25 kbq of 18 F and 99m Tc solutions were measured in the SIRTI for more than two half-lives. The NMISA standardized the activity in the ampoules by ionization chamber measurements traceable to 4 (LS) coincidence measurements. The comparisons, identifiers BIPM.RI(II)-K4.F-18 and BIPM.RI(II)-K4.Tc- 99m, are linked to the corresponding BIPM.RI(II)-K1.F-18 and BIPM.RI(II)-K1.Tc-99m comparisons and degrees of equivalence with the respective key comparison reference values have been evaluated. 1. Introduction Radionuclides are essential for nuclear medicine where very short-lived (much less than one day) radionuclides are used, particularly for imaging. The use of nuclear medicine is increasing with the accessibility of these radionuclides which are consequently of great interest to the National Metrology Institutes (NMIs) in terms of the standardization and SI traceability. However, sending ampoules of short-lived radioactive material to the Bureau International des Poids et Mesures (BIPM) for measurement in the International Reference System (SIR) [1] is only practicable for the NMIs that are based in Europe. Consequently, to extend the utility of the SIR and enable other NMIs to participate, a transfer instrument (SIRTI) has been developed at the BIPM with the support of the Consultative Committee for Ionizing Radiation CCRI(II) Transfer Instrument Working Group [2]. The BIPM ongoing K4 comparisons of activity measurements of 18 F (half-life T 1/2 = 1.8288 h; u = 0.0003 h [3]) 1 and of 99m Tc (half-life of 6.0067(10) h [3]) are based on the SIRTI, a welltype NaI(Tl) crystal calibrated against the SIR, which is moved to each participating laboratory. The stability of the system is monitored using a 94 Nb reference source (half-life of 20 300(1 600) a [4]) from the Institute for Reference Materials and Measurements (IRMM, 1 Hereafter, the last digits of the standard uncertainties are given in parenthesis. 1/22

Geel), which also contains the 93m Nb isotope. The 18 F or 99m Tc count rate above a low-energy threshold, defined by the 93m Nb x-ray peak at 16.6 kev, is measured relative to the 94 Nb count rate above the same threshold. Once the threshold is set, a brass liner is placed in the well to suppress the 93m Nb contribution to the 94 Nb stability measurements. It should be noted that the uncertainty associated with the 94 Nb decay correction is negligible. The 99m Tc SIR ampoule is placed in the detector well with the brass liner to suppress the 99m Tc x-ray peaks from the counts. The 18 F SIR ampoule is placed in the detector well in a PVC liner to stop the + particles while minimizing the production of bremsstrahlung. No extrapolation to zero energy is carried out as all the measurements are made with the same threshold setting. The live-time technique using the MTR2 module from the Laboratoire National d Essais Laboratoire National Henri Becquerel, France (LNE-LNHB) [5] is used to correct for dead-time losses, taking into account the width of the oscillator pulses. The standard uncertainty associated with the live-time correction, due to the effect of finite frequency of the oscillator, is negligible. Similarly to the SIR, a SIRTI equivalent activity A E is deduced from the 18 F or 99m Tc and the 94 Nb counting results and the 18 F or 99m Tc activity measured by the NMI: A E corresponds to the inverse of a detection efficiency, i.e. A E is the activity of the source measured by the participant divided by the 18 F or 99m Tc count rate in the SIRTI expressed relatively to the 94 Nb count rate. The possible presence of impurity in the solution should be accounted for using - spectrometry measurements carried out by the NMI. The present K4 comparisons are linked to the corresponding BIPM.RI(II)-K1 comparisons through the calibration of the SIRTI against the SIR at the BIPM and consequently degrees of equivalence with the K1 key comparison reference value (KCRV) can be evaluated. The K4 99m Tc comparison results based on primary measurements carried out by the NMI, or ionization chamber measurements traceable to primary 99m Tc measurements made within one year prior to the K4 comparison, are eligible to be included in the KCRV. The protocol [6] and previous comparison results for the BIPM.RI(II)-K4 comparisons are available in the key comparison database of the CIPM Mutual Recognition Arrangement [7]. Publications concerning the details of the SIRTI and its calibration against the SIR can be found elsewhere [8, 9]. 2. Participants As detailed in the protocol, participation in the BIPM.RI(II)-K4 comparisons mainly concerns member states that are located geographically far from the BIPM and that have developed a primary measurement method for the radionuclide of concern. However, at the time of the comparison the National Metrology Institute (NMI) may decide for convenience to use a secondary method, for example a calibrated ionization chamber. In this case, the traceability of the calibration needs to be clearly identified. The present comparisons took place at the National Metrology Institute of South Africa (NMISA), South Africa, in November 2015, who used an ionization chamber calibrated by the 4 (LS) coincidence method one and two months prior to the comparison to standardize the 18 F and 99m Tc solutions, respectively. 2/22

3. The SIRTI at the NMISA The reproducibility and stability of the SIRTI at the NMISA were checked by measuring the count rate produced by the reference 94 Nb source No. 3, the threshold position (defined by the 93m Nb x-ray peak), the background count rate, the frequency of the oscillator No. 1 for the live-time correction and the room temperature as shown in Figure 1. The plots shown in the Figure represent the differences from the values indicated in the figure caption, using the appropriate units, as given, for each quantity measured. 50 45 40 Nb-94 Threshold Temperature Background Oscillator freq. 35 30 25 20 15 99m Tc 18 F 10 5 0 17/11/2015 19/11/2015 21/11/2015 23/11/2015 25/11/2015 Date Figure 1: Fluctuation of the SIRTI at the NMISA. Black squares: 94 Nb No.1 count rate / s 1 above 7600 s 1 ; circle: threshold position / channel above 90 channels; stars: room temperature / C above 15 C; open squares: background count rate / s 1 above 70 s 1 ; triangles: frequency of the oscillator No.1 / Hz above 999 960 Hz. The SIRTI was very stable during the comparison, except maybe a slight increase in the 94 Nb count rate on the last days and a corresponding slight decrease of the threshold position. However, the 99m Tc measurements are almost insensitive to the threshold position [8]. Consequently, it was decided to use the mean 94 Nb count rate to normalize the 99m Tc measurements (instead of individual 94 Nb results for the 18 F comparison) so that the small fluctuations observed in the 94 Nb results do not influence the 99m Tc comparison results. The mean 94 Nb No. 3 count rate, corrected for live-time, background and decay, measured at the NMISA is 7632.8(5) s 1 which is in agreement with the weighted mean since the set-up of the system in March 2007, 7632.14(25) s 1. Finally, the 94 Nb count rate was checked on the return of the SIRTI to the BIPM after the NMISA comparison, giving a value of 7631.8(13) s 1 in agreement within standard uncertainty with the mean since 2007. 3/22

4. The 18 F and 99m Tc solutions standardized at the NMISA The 18 F and 99m Tc solutions measured in the SIRTI consisted in saline solutions diluted with de-ionized water. Two SIR ampoules were prepared from each diluted solution. Details are shown in Table 1, including any impurities, when present, as identified by the laboratory. The density and volume of the solutions in the ampoules conformed to the K4 protocol requirements. Drops (max. 3 mm diameter) of solution were observed in the cylindrical part of the ampoules (max. 1 cm above the solution). For the 18 F ampoules, many bubbles were observed in the solution and condensation in the upper part of the ampoule was noted (see Figure 2). The total volume of the bubbles was estimated in the range of 10 mm 3, corresponding to an effective solution density of 0.997 g cm 3. For the condensation, a hypothetic 500 nm thick layer would correspond to a solution volume in the range of 0.1 mm 3. Monte-Carlo simulations indicate that the drops, bubbles and condensation observed should have a negligible effect on the SIRTI measurement result. The 99m Tc ampoules were not sealed, but closed with Parafilm. The 18 F and 99m Tc activities in the SIRTI ampoules were deduced from the measurement at the NMISA of each master solution in a Vinten Isocal IV well-type ionization chamber (IC) and a dilution factor of 1/38.424 78 and 1/41.117 13, respectively. The IC had been calibrated for 18 F and 99m Tc by the NMISA one and two month, respectively, prior to the K4 comparison by the 4 (LS)β- coincidence method. The measurement results are summarized in Tables 2 and 3 while the uncertainty budgets of the NMISA primary measurements are given in appendix 2. Table 1: Characteristics of the solutions measured in the SIRTI Radionuclide 18 F (FDG # ) Density at Solvent Carrier / mol dm 3 / g g 1 20 C Ampoule / g cm 3 number de-ionized water Mass / g Impurity* 1.0 2 3.543 71 3 3.541 34 99m Tc de-ionized water 1.0 1 3.626 18 2 3.594 28 * Ratio of the impurity activity to the main radionuclide activity at the reference date # Fluorodesoxyglucose 4/22

Figure 2: 18 F ampoule where drops, bubbles and condensation are noted Table 2: The 18 F and 99m Tc standardizations by the NMISA Radionuclide 18 F Measurement method ACRONYM* IC calibrated 4P-IC-GR-00-00-00 in Oct. 2015 by 4 (LS)β- coinc. 4P-LS-BP-NA-GR-CO Activity conc. / kbq g 1 Standard uncert. / kbq g 1 6.669 0.041 Reference date YYYY-MM-DD 2015-11-24 12:00 UTC Half-life used by the NMI / h 1.828 90(23) IC calibrated 4P-IC-GR-00-00-00 99m Tc in Sept. 2015 by 4 (LS)β- coinc. 4P-LS-BP-NA-GR-CO * See appendix 1 7.80 0.11 2015-11-23 10:00 UTC 6.0067(10) 5/22

Table 3: The NMISA uncertainty budgets for the activity measurement of the 18 F and 99m Tc ampoules (Nov. 2015) Uncertainty contributions due to Evaluation method Relative standard uncert. 10 4 18 F Comments Relative sensitivity factors Relative standard uncert. 10 4 99m Tc Comments Relative sensitivity factors Counting statistics A 5 Weighing B 1 Statist. analysis of 306 values Mass for source prep. 0.82 7 Statist. analysis of 153 values 1 1 Mass for source prep. 0.1 1 Background A 0.2 Statist. analysis of 201 values Time accuracy of PC clock B 0.1 Accuracy of time stamping Decay correction B 1.5 From literature value uncertainty for the half-life Calibration factor of IC (see appendix 2) B 35 From primary standardization IC cyclical variability B 50 IC long term cyclical behaviour Impurities B 0.0 HPGe measurements 0.0003 2 Statist. analysis of 201 values 1 0.1 Accuracy of time stamping 0.76 0.09 From literature value uncertainty for the half-life 1 134 From primary standardization 1 50 IC long term cyclical behaviour 0.0 HPGe measurements 0.003 1 0.05 1 1 Relative combined standard uncertainty 61 143 6/22

5. The 18 F and 99m Tc measurements in the SIRTI at the NMISA The maximum count rate corrected for live-time in the NaI(Tl) was 17 000 s 1 which conforms with the limit of 20 000 s 1 set in the protocol [6]. In addition a relative standard uncertainty of 2 10 4 and 3 10 4 for 18 F and 99m Tc respectively, was added to take account of a possible drift in the SIRTI at high count rate [8]. The time of each TI measurement was obtained from the synchronization of the laptop with a local NTP time server. In principle, the live-time correction should be modified to take into account the decaying count rate [10]. In the present experiments, the duration of the measurements made at high rate has been limited to 300 s and 1500 s for 18 F and 99m Tc respectively, so that the relative effect of decay on the live-time correction is less than one part in 10 4. Two ampoules of each of the 18 F and 99m Tc solutions were measured alternatively for more than two half-lives and the results are shown in Figures 3a and 3b. The reduced chi-squared values evaluated for these series of measurements are 0.83 and 0.71 for 18 F and 99m Tc, respectively, showing that the data are consistent. The absence of significant trend confirms the stability of the SIRTI and the absence of significant impurity in the solutions. The uncertainty budgets for the SIRTI measurements of the 18 F and 99m Tc ampoules are given in Table 4a and 4b. Further details are given in reference [8]. 6. Comparison results and degrees of equivalence The weighted mean and uncertainty of all the measured A E values is calculated taking into account correlations. In the case of 18 F, the last 10 measurements shown in Figure 3a were not considered as the count rate in the SIRTI was lower than the limit of 2000 s 1 set in the protocol. The standard uncertainty u(a E ) is obtained by adding quadratically the SIRTI combined uncertainty from Tables 4a and 4b and the uncertainty stated by the participant for the 18 F and 99m Tc measurements (see Table 2). The correlation between the NMISA and the BIPM due to the use of the same 99m Tc half-life is negligible in view of the small contribution of this half-life to the combined uncertainty of the measurements. The K4 comparison results are given in Table 5 as well as the linked results A e in the corresponding BIPM.RI(II)-K1 comparisons which were obtained by multiplying A E by the linking factors L = 1495.1(18) for 18 F and 12 165 (23) for 99m Tc. The linking factors were obtained through the measurement of 18 F and 99m Tc ampoules from the LNE-LNHB, the NPL and a commercial company in both the SIRTI and the SIR [9]. 7/22

Figure 3a: The 18 F measurement results in the SIRTI at the NMISA. The uncertainty of the 18 F activity concentration and of the 94 Nb mean count rate, which are both constant over all the measurements, are not included in the uncertainty bars shown on the graph. Figure 3b: As for Figure 3a, but for the 99m Tc. 8/22

Table 4a: Uncertainty budgets for the SIRTI measurement of the 18 F ampoules Uncertainty contributions due to 18 F meas. including livetime, background, decay corrections and threshold setting 94 Nb measurement including live-time, background and threshold setting Long-term stability of the SIRTI Nb reference source No.3 instead of No.1 Effect of decay on the live-time correction SIRTI drift at high count rate Ampoule dimensions Ampoule filling height Solution density Unseen droplet on the walls of the ampoule top Comments Standard uncertainty of the weighted mean of 30 measurements, taking into account the correlation due to the 18 F half-life Standard deviation of 10 measurements and sensitivity to threshold setting Weighted standard deviation of 77 series, each series consisting of 10 measurements Weighted standard deviation of 53 series, each series consisting of 10 measurements Maximum measurement duration evaluated from [11] Mean possible drift over all 18 F measurements at the NMISA. From the IRMM report [12] and sensitivity coefficients from Monte-Carlo simulations Solution volume is 3.6(1) cm 3 ; sensitivity coefficients from Monte-Carlo simulations Between 1 g/cm 3 and 1.01 g/cm 3 as requested in the protocol; sensitivity coefficients from Monte-Carlo simulations Evaluation method Relative standard uncert. 10 4 A 1.9 A/B 1.8 A 0.3 A 0.3 B < 1 B 2 B 2* B 2 B 0.7 Evaluated by Monte-Carlo simulation B 1 Relative combined standard uncertainty 4.2 * Included in the type A uncertainty of the measurements of 2 ampoules of 18 F 9/22

Table 4b: Uncertainty budgets for the SIRTI measurement of the 99m Tc ampoules Uncertainty contributions due to 99m Tc meas. including live-time, background and decay corrections 94 Nb measurement including live-time, background and threshold setting Long-term stability of the SIRTI Nb reference source No.3 instead of No.1 Effect of decay on the live-time correction SIRTI drift at high count rate Ampoule dimensions Ampoule filling height Solution density Unseen droplet on the walls of the ampoule top Comments Standard uncertainty of the weighted mean of 36 measurements, taking into account the correlation due to the 99m Tc half-life Weighted standard deviation of 7 series, each series consisting of 10 measurements Weighted standard deviation of 77 series, each series consisting of 10 measurements Weighted standard deviation of 53 series, each series consisting of 10 measurements Maximum measurement duration evaluated from [11] Mean possible drift over all 99m Tc measurements at the NMISA. From the IRMM report [12] and sensitivity coefficients from Monte-Carlo simulations Solution volume is 3.6(1) cm 3 ; sensitivity coefficients from Monte-Carlo simulations Between 1 g/cm 3 and 1.01 g/cm 3 as requested in the protocol; sensitivity coefficients from Monte-Carlo simulations Evaluation method Relative standard uncert. 10 4 A 1.7 A 0.7 A 0.3 A 0.3 B < 1 B 3 B 8 B 3* B 0.8 Evaluated by Monte-Carlo simulation B 2 Relative combined standard uncertainty 9.1 * Included in the type A uncertainty of the measurements of 2 ampoules of 99m Tc Table 5: BIPM.RI(II)-K4. comparison results and link to the BIPM.RI(II)-K1 comparisons Radionuclide Measurement method ACRONYM* IC calibrated 18 4P-IC-GR-00-00-00 F in Oct. 2015 by 4 (LS)β- coinc. 4P-LS-BP-NA-GR-CO IC calibrated 99m 4P-IC-GR-00-00-00 Tc in Sept. 2015 by 4 (LS)β- coinc. 4P-LS-BP-NA-GR-CO * See appendix 1 Solution volume (calculated) /cm 3 A E /kbq u(a E ) /kbq Linked A e /kbq u(a e ) /kbq 3.54 10.25 0.06 15 328 96 3.63 and 3.59 12.83 0.18 156 100 2 200 10/22

Every participant in the K4 comparison is entitled to have one result included in the key comparison database (KCDB) as long as the laboratory is a signatory or designated institute listed in the CIPM MRA, and the result is valid (i.e., not older than 20 years). Normally, the most recent result is the one included. Any participant may withdraw its result only if all the participants agree. The KCRV for 18 F has been defined in the frame of the BIPM.RI(II)-K1.F-18 comparison using direct contributions to the SIR, and is equal to 15 276(24) kbq [13]. The key comparison reference value (KCRV) for 99m Tc has been defined in the frame of the BIPM.RI(II)-K1.Tc-99m comparison using direct contributions to the SIR and, as agreed by the CCRI(II) in 2015, the SIRTI results for NIST, KRISS, NIM, LNMRI/IRD and IFIN-HH, giving a value of 153 170(310) kbq as detailed in reference [14]. The degree of equivalence of a particular NMI, i, with the KCRV is expressed as the difference D i with respect to the KCRV D KCRV i Ae (1) i and the expanded uncertainty (k = 2) of this difference, U i, known as the equivalence uncertainty, hence Ui 2u( Di ), (2) taking correlations into account as appropriate [15]. The degree of equivalence between any pair of NMIs, i and j, is expressed as the difference D ij in their results D D D A A (3) and the expanded uncertainty of this difference U ij where U ij 2 ij i j ei e j 2 2 4 ij i j ei ej 2 u ( D ) 4 u u - 2u( A, A ) (4) where any obvious correlations between the NMIs (such as a traceable calibration) are subtracted using the covariance u(a ei, A ej ), as is the correlation coming from the link of the SIRTI to the SIR. The covariance between two participants in the K4 comparison is given by u(a ei, A ej ) = A ei A ej (u L /L) 2 (5) where u L is the standard uncertainty of the linking factor L given above. However, the CCRI decided in 2011 that these pair-wise degrees of equivalence no longer need to be published as long as the methodology is explained. Tables 6a and 6b show the matrices of the degrees of equivalence with the KCRV as they will appear in the KCDB. Only results not older than 20 years are shown. It should be noted that for consistency within the KCDB, a simplified level of nomenclature is used with A ei replaced by x i. The introductory text is that agreed for the comparison. The graph of the degrees of equivalence with respect to the KCRV (identified as x R in the KCDB), is shown in Figure 4a and Figure 4b. The graphical representation indicates in part the degree of equivalence between the NMIs but obviously does not take into account the correlations between the different NMIs. 11/22

Conclusion The BIPM ongoing key comparisons for 18 F (BIPM.RI(II)-K4.F-18) and 99m Tc (BIPM.RI(II)- K4.Tc-99m) currently comprise four and nine results, respectively, and are linked to the corresponding BIPM.RI(II)-K1 comparisons. The last results in these K4 comparisons, for NMISA (South Africa), have been analysed with respect to the KCRVs determined for 18 F and 99m Tc in the frame of the corresponding K1 comparisons. The degrees of equivalence have been approved by the CCRI(II) and are published in the BIPM key comparison database. Other results may be added when other NMIs contribute with 18 F and 99m Tc activity measurements to the K4 or K1 comparisons or take part in other linked Regional Metrology Organization comparisons. It should be noted that the final data in this paper, while correct at the time of publication, will become out-of-date as NMIs make new comparisons. The formal results under the CIPM MRA [7] are those available in the KCDB. 12/22

Table 6a. Introductory text and table of degrees of equivalence for 18 F Key comparison BIPM.RI(II)-K1.F-18 MEASURAND : Equivalent activity of 18 F Key comparison reference value: the SIR reference value for this radionuclide is x R = 15 276 kbq with a standard uncertainty, u R = 24 kbq (see Final Report). The value x i is the equivalent activity for laboratory i. The degree of equivalence of each laboratory with respect to the reference value is given by a pair of terms: D i = (x i - x R ) and U i, its expanded uncertainty (k = 2), both expressed in MBq, and U i = 2((1-2w i )u i 2 + u R 2 ) 1/2 where w i is the weight of laboratory i contributing to the calculation of x R. Linking BIPM.RI(II)-K4.F-18 to BIPM.RI(II)-K1.F-18 The value x i is the SIRTI equivalent activity for laboratory i participant in BIPM.RI(II)-K4.F-18 multiplied by the linking factor to BIPM.RI(II)-K1.F-18 (see Final report). The degree of equivalence of laboratory i participant in BIPM.RI(II)-K4.F-18 with respect to the key comparison reference value is given by a pair of terms: D i = (x i - x R ) and U i, its expanded uncertainty (k = 2), both expressed in MBq. The approximation U i = 2(u i 2 + u R 2 ) 1/2 is used in the following table as none of these laboratories contributed to the KCRV. Linking CCRI(II)-K3.F-18 to BIPM.RI(II)-K1.F-18 The value x i is the equivalent activity for laboratory i participant in CCRI(II)-K3.F-18 having been normalized to the value of the NPL and the LNE-LNHB (2002) combined as the link. The degree of equivalence of laboratory i participant in CCRI(II)-K3. with respect to the key comparison reference value is given by a pair of terms: D i = (x i - x R ) and U i, its expanded uncertainty (k = 2), both expressed in MBq. The approximation U i = 2(u i 2 + u R 2 ) 1/2 is used in the following table as none of these laboratories contributed to the KCRV. Linking APMP.RI(II)-K3.F-18 to BIPM.RI(II)-K1.F-18 The value x i is the equivalent activity for laboratory i participant in APMP.RI(II)-K3.F-18 having been normalized to the value of the NPL and the LNE-LNHB (2002) combined as the link. 13/22

The degree of equivalence of laboratory i participant in APMP.RI(II)-K3. with respect to the key comparison reference value is given by a pair of terms: D i = (x i - x R ) and U i, its expanded uncertainty (k = 2), both expressed in MBq. The approximation U i = 2(u i 2 + u R 2 ) 1/2 is used in the following table as this laboratory did not contribute to the KCRV. These statements make it possible to extend the BIPM.RI(II)-K1.F-18 matrices of equivalence to all participants in the CCRI(II)-K3.F-18, the APMP.RI(II)-K3.F-18 and the BIPM.RI(II)-K4.F-18 comparisons. Lab i D i U i / MBq IRA 0.04 0.10 BEV 0.11 0.32 CIEMAT -0.06 0.18 PTB 0.04 0.09 LNE-LNHB -0.07 0.11 VNIIM -0.08 0.20 NPL 0.04 0.11 ENEA-INMRI 0.09 0.11 NMISA 0.05 0.20 ANSTO -0.13 0.48 CMI-IIR -0.27 0.24 NIST -0.61 0.32 INER -0.15 0.30 14/22

Figure 4a. Graph of degrees of equivalence with the KCRV for 18 F NOT VALID, See erratum p. 22 (as it appears in Appendix B of the MRA) N.B. The right-hand axis gives approximate relative values only 15/22

Table 6b. Table of degrees of equivalence and introductory text for 99m Tc Key comparison BIPM.RI(II)-K1.Tc-99m MEASURAND : Equivalent activity of 99m Tc Key comparison reference value: the SIR reference value for this radionuclide is x R = 153.17 MBq with a standard uncertainty, u R = 0.31 MBq (see Final Report). The value x i is the equivalent activity for laboratory i. The degree of equivalence of each laboratory with respect to the reference value is given by a pair of terms: D i = (x i - x R ) and U i, its expanded uncertainty (k = 2), both expressed in MBq, and U i = 2((1-2w i )u i 2 + u R 2 ) 1/2 where w i is the weight of laboratory i contributing to the calculation of x R. Linking BIPM.RI(II)-K4.Tc-99m to BIPM.RI(II)-K1.Tc-99m The value x i is the SIRTI equivalent activity for laboratory i participant in BIPM.RI(II)-K4.Tc-99m multiplied by the linking factor to BIPM.RI(II)-K1.Tc-99m (see Final report). The degree of equivalence of laboratory i participant in BIPM.RI(II)-K4.Tc-99m with respect to the key comparison reference value is given by a pair of terms: D i = (x i - x R ) and U i, its expanded uncertainty (k = 2), both expressed in MBq, U i = 2((1-2w i )u i 2 + u R 2 ) 1/2 where w i is the weight of laboratory i contributing to the calculation of x R. These statements make it possible to extend the BIPM.RI(II)-K1.Tc-99m matrices of equivalence to the other participants in BIPM.RI(II)-K4.Tc-99m. 16/22

Lab i D i U i / MBq BEV 2.5 2.7 MKEH 1.2 3.3 PTB -0.5 1.2 LNE-LNHB 0.0 1.5 NPL 0.1 1.6 NIST -0.3 1.4 KRISS 0.9 2.7 NMIJ -0.8 2.3 NIM -0.1 2.3 CNEA 7.0 4.2 LNMRI/IRD 1.5 3.3 IFIN-HH -2.3 2.9 VNIIM 3.4 4.8 ENEA-INMRI -2.4 1.7 NMISA 2.9 4.4 17/22

Figure 4b. Graph of degrees of equivalence with the KCRV for 99m Tc (as it appears in Appendix B of the MRA) N.B. The right-hand axis gives approximate relative values only 18/22

References [1] Ratel G., 2007, The Système International de Référence and its application in key comparisons, Metrologia 44(4), S7-S16. [2] Remit of the CCRI(II) Transfer Instrument Working Group, 2009, CCRI(II) working document CCRI(II)/09-15. [3] Bé M.-M., Chisté V., Dulieu C., Browne E., Chechev V., Kuzmenko N., Helmer R., Nichols A., Schönfeld E., Dersch R., 2004, Table of radionuclides, Monographie BIPM-5, volume 1. [4] NUDAT2.5, National Nuclear Data Center, Brookhaven National Laboratory, based on ENSDF and the Nuclear Wallet Cards. [5] Bouchard J., 2000, Appl. Radiat. Isot. 52, 441-446. [6] SIR Transfer Instrument. Protocol for the ongoing comparisons on site at the NMIs, BIPM.RI(II)-K4. Published on the CIPM MRA KCDB website. [7] CIPM MRA: Mutual recognition of national measurement standards and of calibration and measurement certificates issued by national metrology institutes, International Committee for Weights and Measures, 1999, 45 pp. http://www.bipm.org/pdf/mra.pdf. [8] Michotte C. et al., The SIRTI, a new tool developed at the BIPM for comparing activity measurements of short-lived radionuclides world-wide, Rapport BIPM-2013/02. [9] Michotte C. et al., Calibration of the SIRTI against the SIR and trial comparison of 18 F and 99m Tc at the NPL. In preparation. [10] Baerg A.P. et al., 1976, Live-timed anti-coincidence counting with extending dead-time circuitry, Metrologia 12, 77-80. [11] Fitzgerald R., 2016, Corrections for the combined effects of decay and dead time in livetimed counting of short-lived radionuclides, Appl. Radiat. Isot. 109, 335-340. [12] Sibbens G., 1991, A comparison of NIST/SIR-, NPL-, and CBNM 5 ml ampoules, GE/R/RN/14/91, CEC-JRC Central Bureau for Nuclear Measurements, Belgium. [13] Michotte C. et al., Update of the BIPM comparison BIPM.RI(II)-K1.F-18 of activity measurements of the radionuclide 18 F to include the 2010 result of the LNE-LNHB (France). Metrologia, 2016, 53, Tech. Suppl., 06004. [14] Michotte C., et al., Activity measurements of the radionuclide 99m Tc for the VNIIM, Russia and the ENEA-INMRI, Italy in the ongoing comparison BIPM.RI(II)-K4.Tc- 99m and update of the KCRV of the BIPM.RI(II)-K1.Tc-99m comparison, in preparation. [15] Ratel G., 2005, Evaluation of the uncertainty of the degree of equivalence, Metrologia 42, 140-144. 19/22

Appendix 1. Acronyms used to identify different measurement methods Each acronym has six components, geometry-detector (1)-radiation (1)-detector (2)-radiation (2)-mode. When a component is unknown,?? is used and when it is not applicable 00 is used. Geometry acronym Detector acronym 4 4P proportional counter PC defined solid angle SA press. prop. counter PP 2 2P liquid scintillation counting LS undefined solid angle UA NaI(Tl) NA Ge(HP) Ge(Li) Si(Li) CsI(Tl) ionization chamber grid ionization chamber Cerenkov light detector calorimeter solid plastic scintillator PIPS detector Radiation acronym Mode acronym positron PO efficiency tracing ET beta particle BP internal gas counting IG Auger electron AE CIEMAT/NIST CN conversion electron CE sum counting SC mixed electrons ME coincidence CO bremsstrahlung BS anti-coincidence AC gamma rays GR coincidence counting with efficiency tracing X - rays XR anti-coincidence counting with efficiency tracing photons (x + ) PH triple-to-double coincidence ratio counting alpha - particle AP selective sampling SS mixture of various radiations GH GL SL CS IC GC LC CA SP PS CT AT TD MX high efficiency HE Examples method 4 (PC) -coincidence counting 4 (PPC) -coincidence counting eff. trac. defined solid angle -particle counting with a PIPS detector 4 (PPC)AX- (Ge(HP))-anticoincidence counting 4 CsI-,AX, counting calibrated IC internal gas counting acronym 4P-PC-BP-NA-GR-CO 4P-PP-MX-NA-GR-CT SA-PS-AP-00-00-00 4P-PP-MX-GH-GR-AC 4P-CS-MX-00-00-HE 4P-IC-GR-00-00-00 4P-PC-BP-00-00-IG 20/22

Appendix 2. Uncertainty budgets for the NMISA primary measurements of 18 F (Oct. 2015) and 99m Tc (Sept. 2015) 4P-LS-BP-NA-GR-CO Uncertainty contributions due to Evaluation method 18 F Relative standard uncert. 10 4 Relative sensitivity factors 99m Tc Relative standard uncert. 10 4 Relative sensitivity factors Comments Counting statistics A 2 0.18 23 0.23 Statist. analysis of 20 values Weighing B 5 1 5 1 Mass of sources Dead time B 2 0.005 5 0.014 D ± 0.05 s Background A 10 0.0013 6 0.0014 Background square root statist. applied Coincidence resolving time B 1 0.005 5 0.025 R ± 0.01 s Counting time B 0.1 1 0.1 1 Calibration of timer Adsorption B 9 1 0.0 1 18 F : (1) / 99m Tc : (2) Decay correction B 2 1.6 7 4.3 From literature value uncert. for the half-life Extrapolation of effic. curve B 20 1 130 1 Alternative fits to data Afterpulse correction B 10 0.00012 20 0.02 Beta-plus branching ratio: 0.9686 (19) B 20 1 From literature value uncertainty in Betaplus branching ratio Impurities B 0.0 0.0 HPGe measurements Relative combined standard uncertainty 34 134 (1) Count rates after multiple rinsings relative to the expected count rates if rinsings were not done (2) Count rate measurements after multiple rinsings comparable to background count rates 21/22

Erratum In the present paper, degrees of equivalence for NMISA, South Africa, have been included in the comparison of 18 F activity measurements BIPM.RI(II)-K4.F-18. The corresponding Table 6a is correct. However an error in the Graph 4a happened and the uncertainty bars for the BIPM.RI(II)-K1.F-18 comparison results in red are wrong. The correct graph is shown below. Corrected Figure 4a. Graph of degrees of equivalence with the KCRV for 18 F It should be noted that no error took place in the graph appearing on the Key Comparison Data Base on line. 22/22