Production of scandium radionuclides for theranostic applications: towards standardization of quality requirements

In the frame of “precision medicine”, the scandium radionuclides have recently received considerable interest, providing personalised adjustment of radiation characteristics to optimize the efficiency of medical care or therapeutic benefit for particular groups of patients. Radionuclides of scandium, namely scandium-43 and scandium-44 (43/44Sc) as positron emitters and scandium-47 (47Sc), beta-radiation emitter, seem to fit ideally into the concept of theranostic pair. This paper aims to review the work on scandium isotopes production, coordination chemistry, radiolabeling, preclinical studies and the very first clinical studies. Finally, standardized procedures for scandium-based radiopharmaceuticals have been proposed as a basis to pave the way for elaboration of the Ph.Eur. monographs for perspective scandium radionuclides.


Introduction
In the quest for new radionuclides providing personalised adjustment of radiation characteristics to optimize the efficiency of medical care or therapeutic benefit for particular groups of patients (the so called "precision medicine"), the scandium radionuclides have recently received considerable interest. In the last two decades, several new radionuclides for diagnostic imaging and therapy have been successfully introduced to clinical practice. Additionally, using the same targeting vectors and the combination of positron emitter gallium-68 ( 68 Ga) for diagnostic imaging and matching therapeutic counterpart betaemitting radionuclides such as lutetium-177 ( 177 Lu) and yttrium-90 ( 90 Y) or, recently, alpha emitters bismuth-213 ( 213 Bi) and actinium-225 ( 225 Ac), for therapeutic use, has been recognized as a clear advantage over currently available treatment options. From this perspective, radionuclides of scandium, namely scandium-43 and scandium-44 ( 43/44 Sc) as positron emitters and scandium-47 ( 47 Sc), beta-radiation emitter, seem to fit ideally into the concept of theranostic pair (Mausner and Srivastava 1993;Huclier-Markai et al. 2018). Scandium-47 is also a low energy γ-emitter and allows SPECT and planar imaging. Therapeutic potential of 47 Sc was studied at the Brookhaven National Laboratory already in the 1990s (Mausner and Srivastava 1993;Pietrelli et al. 1992). However, the rapid growth of the scandium radionuclides applications started only after the introduction of 68 Ga-labelled compounds for PET diagnosis in the early 2000s. Scandium-44 was proposed as a potential alternative to 68 Ga for clinical applications in 2010 by the group of Rösch in Mainz, Germany (Roesch 2012;Pruszynski et al. 2010). Similarities in chemistry between 68 Ga and 44 Sc and their different physical properties have opened a wider avenue for applications of other Sc radionuclides.
Scandium radioisotopes can be produced in accelerators and in nuclear reactors (Huclier-Markai et al. 2018;Muller et al. 2018;Mikolajczak et al. 2019). In particular, due to their large number, biomedical cyclotrons accelerating protons up to 20 MeV are expected to provide 43 Sc and 44g Sc in quantities that would allow their wider use in diagnostic imaging (Roesch 2012;Synowiecki et al. 2018). Similarly, cyclotrons accelerating protons to higher energies, nuclear reactors and electron linear accelerators (linacs) might be a source of 47 Sc for therapy (Domnanich et al. 2017a;Jalilian et al. 2020;Qaim 2019). However, due to several production routes possible, the unification of quality parameters of Sc radionuclides has not been yet attempted.

Scandium isotopes with medical potential
Natural scandium has only one stable isotope, scandium-45 ( 45 Sc). Several other scandium isotopes can be produced artificially, however, most of them are short-lived, in seconds range (http://kcvs.ca/isotopesmatter/iupacMaterials/javascript/Interactive%2 0Periodic%20Table%20of%20the%20Isotopes/HTML5/pdf-elements/scandium.pdf). Those with atomic mass smaller than 45 decay by emission of positrons, while those with atomic mass greater than 45 emit electrons. Characteristics of Sc radioisotopes with longer half-lives are given in Table 1 (adapted from Pawlak et al. 2019). Among them, 43 Sc,44,44m Sc and 47 Sc are suitable for medical applications.

Nuclear data and production methodologies
Production routes for 43 Sc,44g,44m Sc and 47 Sc use either calcium, titanium or vanadium targets, typically enriched for the desired isotope. A summary of nuclear reactions, decay data, cross-sections, and targetry is given in Table 2, together with the chemical form, purity, enrichment and cost assessment of the target material (where available). For each of the radionuclides, we also report the production mode (i.e. accelerator, nuclear reactor, radionuclide generator) and examples of production facilities that have already implemented their production. 43 Sc production 43 Sc ðT 1=2 ¼ 3:89 h; E β þ ¼ 476 keV 88:1% ð Þ; branching ratio β þ : 88%; Eγ ¼ 372 keV 22:5% ð Þ Several production routes are possible for 43 Sc, involving proton, deuteron or alpha beams (Chaple and Lapi 2018;Braccini 2016). These differents production routes are described in more detail below.

Production of 43 Sc using alpha beams
Reasonable activities of 43 Sc can be produced by irradiating a nat Ca target. It is obtained both directly throught 40 Ca(α,p) 43 Sc and 40 Ca(α,n) 43 Ti (T 1/2 = 509 ms) → 43 Sc nuclear reactions, using accelerators. Effective production of 43 Sc using alpha beam was presented by Walczak et al. (Walczak et al. 2015). Szkliniarz et al. (Szkliniarz et al. 2016) irradiated natural calcium as calcium carbonate with 20 MeV alpha particle beam. Resulting 43 Ti decays to 43 Sc during the irradiation and no chemistry is needed to separate 43 Sc from 43 Ti due to its very short half-life. Impurities 44m Sc and 44g Sc are formed due to the 42 Ca content (0.65%) in the natural calcium target as well as 46 Sc and 47 Sc due to the 44 Ca content (2% of natural calcium). The 42 Ca(α,np + pn) 46 Sc nuclear reaction has a maximum cross-section around 30 MeV, thus its formation can be decreased by using lower beam energy. Conversely, 47 Sc formation cannot be avoided, as it has a similar maximum (17 MeV) as the 40 Ca(α,p) 43 Sc reaction (14 MeV). The radionuclidic purity of the produced 43 Sc was 99.95% at the end of bombardment (EOB) and 98.9% at 20 h after EOB. The radionuclidic purity decreases with time as 43 Sc half life is lower than that of 44m Sc, 46 Sc and 47 Sc. The use of enriched 40 CaCO 3 target material decreased dramatically the level of radionuclidic impurities to 1.5 × 10 − 5 %. Thick target yield for irradiation of calcium metal was found to be 240 MBq/ μAh, from which 15 GBq production yield was extrapolated for 4 h irradiation with a 25 μA alpha particle beam. Based on the experimental results, 1 GBq yield of 44m Sc is It is worth noting that most of the sources provide the half-life of 44 Sc T 1/2 = 3.97 h. This value has been recently redetermined by Garcia-Tarrano et al. (Garcia-Torano et al. 2016) to be T 1/2 = 4.042 h, about 2% higher than the earlier recommended value Mikolajczak et al. EJNMMI Radiopharmacy and Chemistry (2021) (Koning, 2016;Howard, 1974) nat Ca(α,p) 43 Sc (Synowiecki et al. 2018;Domnanich et al. 2017a) 570 mb (Howard, 1974) (Szkliniarz et al. 2016). Using alpha beam irradiation, Minegishi et al. (Minegishi et al. 2016) reported a remote method for 43 Sc production on an unsolidified, powder calcium oxide (CaO) target. The powdery CaO target material was dissolved in situ in HCl in the target box and remotely recovered as a radio-Sc solution as it is done when using liquid targets. The yield of 43 Sc following isolation via a precipitation method with a typical 0.22 μm sterile filter was 54.8 MBq/μAh at the end of separation (approximately 1.5 h from the EOB).

Production of 43 Sc using protons
The number of cyclotrons providing regular and intense alpha beams is limited, therefore, production methods using medical cyclotrons (proton beams) are gaining more interest. Domnanich et al. (Domnanich et al. 2017b) demonstrated that 43 Sc can be produced at a medical cyclotron via proton irradiation of enriched 43 Ca or 46 Ti oxide target. The production via the 46 Ti(p,α) 43 Sc nuclear reaction yielded a 43 Sc activity around 200 MBq and high radionuclidic purity (> 98%) at the end of a 7 h irradiation when using 97% enriched 46 Ti. The production via the 43 Ca(p,n) 43 Sc nuclear reaction resulted in higher quantities of 43 Sc, but the product consisted of a mixture of 43 Sc and 44g Sc and an activity ratio of 2:1 when using 57.9% enriched 43 CaCO 3 . This may be increased with higher enrichment of the 43 Ca target (max available is 90% at the moment). However, the question remains whether 44g Sc is a real problem as it decays with almost the same half-life by emitting a positron, as 43 Sc.
Production of 43 Sc using deuterons 43 Sc can be also produced by deuteron bombardment of an enriched 42 Ca oxide target, although the method has not been yet practically utilized. Recently, measurements of the 43 Sc production cross-section in the reaction 42 Ca(d,n) 43 Sc with a deuteron beam have been reported by Carzaniga and Braccini 2019). The authors also studied practical aspects of producing 43 Sc via this route using commercially available targets. Yet, the limited number of medical cyclotrons currently in operation offering deuteron beams prevents the wider application of this method. In summary, 43 Sc can be produced with proton and deuteron beams that are easily available worldwide, the use of enriched (and expensive) target material, 43 Ca or 44 Ca, is mandatory due to their low natural abundance. When proton irradiation route is preferred, the issue of co-produced 44g Sc in 43 Sc needs to be addressed ). On the other hand, irradiation with an alpha particle beam allows to use directly natural calcium as the nuclear reaction of interest involves 40 Ca (natural abundance is 96.94%). However, such a kind of beam is only available in a few places in the world. Though, higher radionuclidic purity can be obtained using enriched 40 (Roesch 2012;Filosofov et al. 2010;Radchenko et al. 2016). Although the generator method has been extensively investigated, only a small number of facilities worldwide use these generators. The parent radionuclide, titanium-44 (T 1/2 = 60 y) is produced through the nuclear reaction 45 Sc(p,2n) 44 Ti. Production of mCi amount of 44 Ti is difficult due to its long half-life and the low cross section (probability) of this nuclear reaction. As an example, irradiation of 1.5 g of Sc target material produced about 185 MBq (5 mCi) (Chaple and Lapi 2018).

Production of 44 Sc using protons
It is thus easier to use an accelerator to produce 44g Sc. Medical cyclotron can be used for proton irradiation of a target containing either natural (Severin et al. 2012) or 44 Ca enriched target . Enriched targets are preferred to avoid the production of contaminants such as 46,48 Sc (Domnanich et al. 2017a). The higher the 44 Ca enrichment, the higher the final radionuclidic purity. However, during the irradiation, 44m Sc is co-produced which may be either an advantage or a disadvantage depending on the application. Indeed, its long half life allows 44m Sc (T 1/2 = 58.6 h) to be used to image slow biological processes. This is possible since 44m Sc/ 44g Sc can act as an in-vivo generator (Alliot et al., 2015a). The production of 44g Sc via the 44 Ca(p,n) 44g Sc nuclear reaction has been implemented at the research cyclotron at Paul Scherrer Institute, Zurich, providing this radionuclide with high radionuclidic purity (> 99%) and at high activities (> 2 GBq) (van der Meulen et al. 2015). This would be the method of choice when one wants to produce cost effectively 44g Sc. Other targets or projectiles have also been used, presenting some advantages for dedicated cases: higher contribution of 44m Sc for example or lower cost when using nat Ca. The use of natural calcium metal provides cost-efficient access to 44g Sc for preclinical experiments. It can be pressed into the cavity of an appropriate target holder (coin or shuttle) and irradiated with high currents, as it has good heat conductivity. However, contaminating radiometals prevent the use of this production route for human application. Exotic production routes have been also studied using 47 Ti in a 47 Ti (p,α) 44 Sc nuclear reaction (Loveless et al. 2021).

Production of 44 Sc using deuterons
Deuteron beams can also be used for production of 44 Sc in a nuclear reaction 44 Ca(d, 2n) 44 Sc/ 44m Sc (Alliot et al. 2015a). A proof of principle has been demonstrated at low beam current (0.3 μA). An activity of 90 MBq (4 h after EOB) can be obtained with a 3 h bombardment on a 500 μm thick 44 CaCO 3 target at 17 MeV (Alliot et al. 2015b). This study showed also that the use of deuterons allows to increase the production of 44m Sc with respect to proton irradiation keeping 44g Sc at the same level as for protons (Duchemin et al. 2015;Duchemin et al. 2016). However, to favor 44m Sc production the use of alpha beam is recommended (Szkliniarz et al. 2016 Scandium-47 can be produced via several different nuclear reactions using a nuclear reactor, a cyclotron, or a linac (Srivastava 2013).

Production of 47 Sc using neutrons
In a nuclear reactor, practically n.c.a 47 Sc is produced via 47 Ti(n,p) 47 Sc nuclear reaction by irradiation of 47 Ti target with fast neutrons (energy greater than 1 MeV). The 47 Ti(n, p) 47 Sc route also requires an enriched target, though 47 Ti oxide is available with very high enrichment and at a reasonable cost. Thus, using the 47 Ti(n,p) 47 Sc nuclear reaction could provide quantities sufficient for therapy. For example, in a HFIR reactor (Oak Ridge National Laboratory) a 3.35 day (one half-life of 47 Sc) irradiation of 10 g target could produce approximately 2800 GBq of 47 Sc at EOB (Kolsky et al. 1998;. However, along with 47 Sc, 46 Sc is co-produced (Domnanich et al. 2017a;Bokhari et al. 2009;Bartoś et al. 2012). This method suffers from the limited number of facilities delivering fast neutrons.
Alternatively, 47 Sc can be produced with thermal neutrons (Mausner 1998;Deilami-Nezhad et al. 2016) that are more widely available. Neutron capture on 46 Ca produces 47 Ca (T 1/2 = 4.5 d), which decays into 47 Sc by β − emission: 46 Ca(n,γ) 47 Ca → 47 Sc, and the obtained 47 Ca can be further exploited as the 47 Ca/ 47 Sc generator system (Mausner 1998). This method suffers mainly from the low natural abundance of 46 Ca (0.004%). Still, 47 Sc production from neutron irradiated natural Ca target was shown to be feasible . Nontheless, to obtain significant activity of 47 Ca the target enriched in 46 Ca must be used (presently 46 Ca is available with a maximum 30% enrichment) which is rather expensive (Chakravarty et al. 2017) but recycling would allow to drastically reduce the cost. For comparison, when 0,97 mg of 46 Ca (48.5 mg of 5% enriched [ 46 Ca]CaCO 3 ) was irradiated in a thermal neutron flux 1.2 × 10 14 ns − 1 cm − 2 for 6 days, around 700 MBq of 47 Ca and 350 MBq of 47 Sc were produced at EOB (Pawlak et al. 2019). The produced 47 Ca decays to 47 Sc with a half-life of 4.5 days, which is longer than the half-life of 47 Sc, enabling multiple separations of in-grown 47 Sc in the generator-like system. Using this approach, Domnanich et al. (Domnanich et al. 2017a) demonstrated that up to 2 GBq 47 Sc can be produced by thermal neutron irradiation of enriched 46 Ca targets. The optimized chemical isolation of 47 Sc from the target material allowed the formulation of up to 1.5 GBq 47 Sc with high radionuclidic purity (> 99.99%) in a small volume (∼700 μL), which was useful for labeling purposes. Three consecutive separations within 1 week were possible by isolating the in-grown 47 Sc (Domnanich et al. 2017a;Pawlak et al. 2019).

Production of 47 Sc using protons
Scandium-47 can also be produced using high energy proton reaction on 48 Ti (Srivastava 2013; Srivastava and Dadachova 2001). The nuclear reaction on 48 Ti was quite popular as the natural abundance is quite high (73.72%) and the low cross section of the 48 Ti(p,2p) 47 Sc nuclear reaction can be partly compensated by the use of thick targets. However, this method suffers from the co-production of 46 Sc (T 1/2 = 83.79 d) which emits high energy gamma rays abundantly. This impurity is a major concern in the production of 47 Sc both for the risk of unnecessary radiation dose and the regulatory constraints when stored at the therapy wards (Jafari et al. 2019).
Alternative production routes have been explored to try to overcome this issue using 48 Ca, nat V, 44 Ca targets or photonuclear reactions (Mausner and Srivastava 1993;Szkliniarz et al. 2016;Sitarz et al. 2018;Srivastava and Dadachova 2001). The cyclotron production of 47 Sc via the 48 Ca(p,2n) 47 Sc nuclear reaction with a proton energy range of 24 → 17 led to a radionuclidic purity of only around 87%, due to 48 Sc co-production (Misiak et al. 2017). Using enriched 48 Ca for irradiation with 20 MeV protons may be a feasible route for the production of GBq activity levels of 47 Sc, however, the prohibitively high cost of enriched 48 Ca has made it impossible to implement this production route to date (Misiak et al. 2017). Still, the production of 47 Sc in medical cyclotrons providing proton beams at energy range of 15-20 MeV via the 48 Ca(p,2n) 47 Sc nuclear reaction on 48 Ca enriched calcium oxide target could potentially provide wide access to this radionuclide (Braccini 2016). However, the co-produced 48 Sc undermines the 47 Sc purity, and its content strongly depends on the energy of protons impinging the target and on the thickness of the target material, feasibility of this approach has been studied in detail (Carzaniga and Braccini 2019). Another production route proposes the use of natural vanadium targets, since natural vanadium consists of two isotopes: for more than 99.75% it is formed by stable 51 V while the very long-lived 50 V (T 1/2 = 1.4 × 10 17 y) occurs only in 0.25%. Experimental data on nat V can hence be interpreted in broad energy range as coming from a monoisotopic 51 V target and reaction cross-sections can be derived. The cross-sections of the nat V(p,x) 47 Sc nuclear reaction were measured up to 70 MeV proton beam (Jafari et al. 2019;Pupillo et al. 2019;Ditroi et al. 2016), the low reaction yields are the disadvantage of this approach.

Production of 47 Sc using alpha beams
The α-particle irradiation of 44 Ca targets at a cyclotron, inducing the 44 Ca(α,p) 47 Sc nuclear reaction has been considered, though with low yield and radionuclidic purity. The advantage of the 44 Ca(α,p) 47 Sc reaction lies in the short range of α projectiles in Ca target, allowing the use of a relatively small amount of 44 Ca target material for small scale studies with 47 Sc. For example, it is reported that 200 mg of [ 44 Ca]CaO prepared in a diameter of 10 mm would give a yield of approximately 11 MBq at 10 eμA for 2 h irradiation at the end of preparation (approximately 1.5 h from the EOB) in the energy range of 28 → 0 MeV (Minegishi et al. 2016).

Production of 47 Sc in photonuclear reactions
Photonuclear reactions in electron linear accelerators (linacs) using titanium (Jafari et al. 2019;Yagi and Kondo 1977;Mamtimin et al. 2015;Rotsch et al. 2018) and calcium targets Rane et al. 2015) have been explored. However, the cross sections are very small. It may be more convenient to use neutrons and 46 Ca. Mamtimin et al. (Mamtimin et al. 2015) studied the production of 47 Sc via the 48 Ti(γ,p) 47 Sc nuclear reaction by Monte Carlo simulations. Rotsch et al. (Rotsch et al. 2018) evaluated the production yields and purification of the photonuclear-produced 47 Sc from natural titanium oxide targets. In the recent report on photonuclear production Loveless et al. (Loveless et al. 2019a) used eLINAC to produce 47 Sc via 48 Ti(γ, p) 47 Sc reaction. They irradiated a stack of natural titanium foils using bremsstrahlung radiation generated by impinging 22 MeV electrons onto a 0.762 mm thick tungsten radiator. Despite the long irradiation times (10.5-14 h) low activity (approx. 2 MBq) was produced with 90% 47 Sc, 1.2% 46 Sc and 8.3% 48 Sc at EOB. The authors suggested the use of enriched titanium, available in oxide form for low cost. It is expected to improve the attainable radionuclidic purity, but not significantly increase the yield, as the natural abundance of 48 Ti is 73.7%. Importantly, the use of oxide target material requires a special target design, to effectively dissipate the heat during irradiation. Accelerator-based photoproduction of 47 Sc in 48 Ca(γ,n) 47 Ca → 47 Sc nuclear reaction was also reported by Starovoitova et al. (Starovoitova et al. 2015) and Rane et al. (Rane et al. 2015), the latter aiming to develop the 47 Ca/ 47 Sc generator obtained from irradiated 48 Ca target.
To date there is no preferred 47 Sc production route. The presence of radionuclide contaminants is expected in all cases. The choice depends on the availability of the irradiation sites and enriched target materials. For example, higher 46 Ca enrichment may favour its neutron irradiation. At the sites operating linear electron accelerators the photonulear reactions will be preferred (Qaim 2019).

Targetry
Calcium enriched in 44 Ca, required to reach high radionuclidic purity of 44 Sc, is available only in salt form (oxide, carbonate). The low heat conductivity of the target limits the beam current during irradiation. At high beam currents, the heat accumulation caused burnout and/or cracking of the irradiated target and may lead to gas production through the thermal dissociation of CaCO 3 (Wojdowska 2019). In some application, the issue of low thermal conductivity was handled by pressing the calcium carbonate on top of graphite powder to facilitate heat transfer and to hold the calcium carbonate powder in position (van der Meulen et al. 2015). The improvment of the heat conducitivity is paid by a lower production as part of the projectile will interact with carbon atoms. However, this method yielded up to 2 GBq of 44 Sc when using proton beam energies of near 11 MeV, but can not be automatized easily (see Table 2). Calcium oxide obtianed through thermal decomposition of calcium carbonate, both natural and enriched in 44 Ca, was used for preparation of disk shaped pellets which were encapsulated into aluminum for irradiation. Thus increased density of calcium oxide improved the irradiation yield compared to carbonate, though due to the higroscopicity of CaO to the pellets needed to be protected from moisture during storage and when exposed to irradiation (van der Meulen et al. 2020).
Alternatively, magnesium or aluminium powder were used as an additive for target pellet preparation (Mikolajczak et al. 2018;Stolarz et al. 2018). Nuclear reactions on these atoms under 16 MeV proton energy lead to short lived isotopes mainly, resulting in relatively little radioactive contaminants at the end of irradiation. The magnesium pellet can be selectively dissolved from an aluminium target holder with 3 M hydrochloric acid. Aluminum is also easily dissolved after removal from the holder coin.
Introduction of the powdery target material (calcium oxide or calcium carbonate) directly into the cyclotron represents a high risk of contamination as the material may evaporate during irradiation. If a metal foil is used as a target material cover to prevent evaporation, the choice of metal needs careful consideration as it will also act as a degrader. When the foil is too thin, it may separate from the target material during irradiation and may burn out at higher beam currents, as it was reported at 27 μA 16 MeV on a 12.5 μm aluminum foil by Severin et al. (Severin et al. 2012). Conversely, if a higher degrader thickness is required to adjust optimal energy on the target, it has to be actively cooled (van der Meulen et al. 2015). In medical cyclotrons it is safer to use helium-cooled HAVAR foils to separate the irradiated target material from the vacuum system of the cyclotron (e.g., the ARTMS system). In the target system developed in Debrecen (Mikolajczak et al. 2018), the original parts from a water target were used as an interface to the beam port of the cyclotron (see Fig. 1). This enabled the circulation of helium cooling gas between the two foils. After several months of regular use, the contamination caused by calcium evaporation was clearly visible on the internal surfaces of the target system on the second foil. However, no carbonate migrated through the foils to the cooling cycle, or to the high vacuum side.
The high investment costs and the complicated installation are usually mentioned as major drawbacks of a solid target system. Due to the more complex operation, solid target handling systems will be always more expensive than a simple liquid target but will produce 10 times more radioactivity. Installation of a shuttle type solid target system (commercially available from several companies) requires 4-5 cm diameter passage in the walls of the cyclotron bunker and the hot lab, and some free space in the floor duct between them. However, this can be simply made at one end and nearly impossible at another in already existing facilites.
The challenges related to solid target systems prompted the investigation of 44 Sc production in a liquid target. Hoehr et al. (Hoehr et al. 2014) irradiated approximately 1.5 g/mL natural calcium nitrate solution with 13 MeV proton beam in a relatively small volume liquid target. 28 MBq 44 Sc was produced with a 20 μA beam current for 1 h, which can reach a gigabecquerel level, if enriched target material is used. However, it is questionable, whether the production can be managed in a cost-efficient way given that a high amount of enriched material has to be irradiated (approx. 100 times more, than for a solid target production from 44 CaCO 3 ). The use of liquid targets for radiometal production offers easier installation and lower hardware investment costs, but the presence of corrosive liquids on a cyclotron, which is producing 18 F for daily FDG production is of high concern. IBA (Nirta Ga liquid) and GE (gallium target) also offer liquid targets for radiometal production, however, evidence showing their long-term use is needed. From the regulatory point of view, it might be necessary that a liquid target is used for the production of only one radiometal. In contrast, a solid target system can be used to irradiate targets for several radiometals using dedicated target holders. Gelbart et al. (Gelbart and Johnson 2019) developed a hybrid target system to overcome the installation problems of shuttle type systems. The equipment, used for target dissolution is located in the cyclotron bunker behind the irradiated target, allowing the transport of the produced radioisotope in solution to the hot lab. It was designed for 99m Tc and 68 Ga production, but can be utilized for scandium as well.

Chemical processing
Several separation methods have been described in the literature, each strongly related to the form of the initial target material. Those proposed for separation of scandium comprised mostly solvent extraction (Zhang et al. 1997;Vibhute and Khopkar 1985;Rane and Bhatki 1966;Radhakrishnan and Owens 1972;Kalyanaraman and Khopkar 2002). However, for the radioactive isotopes of scandium extraction chromatography is preferred. Cartridge-or small column-based extraction chromatography methods can be automatized easily, and enable reproducible purification. Considering the further development of scandium radionuclides production, the technical issues associated with irradiation, target handling, dissolution and processing need to be solved. For routine application in clinical trials and future diagnostic use, these steps should be automated in order to facilitate the production and to meet the requirements of GMP and radiation protection. Pourmand et al. (Pourmand and Dauphas 2010) Table 3.
The purity of the resulting batches is of great importance. Most of the published works dealt with 44 Sc. These studies have shown that a high chemical purity of the final 44 Sc fraction is important, since the presence of other metals may interact with the DOTA-chelator (or any other chelator) and, hence, the radiolabeling yield would be compromised. The concentration of common environmental contaminants (Al(III), Cu(II), Pb(II), Zn(II), Fe(III)) in 44 Sc is frequent. The most problematic is Fe 3+ for which the stability with the DOTA ligand is greater than that for Sc 3+ . By contrast, the influence of divalent metal cations (considered as contaminants) is negligible due to the much lower stability of their DOTA complexes. In order to meet the requirements for radiopharmaceutical applications, the obtained final solution containing scandium radionuclide needs to be of high chemical purity and concentrated into a small volume of moderately acidic eluate to facilitate efficient radiolabeling and subsequent in vivo application. Most of the times, authors do not indicate the final chemical purity of the resulting bacthes.

Target recovery
Calcium has six naturally occurring isotopes ( 40 Ca, 42 Ca, 43 Ca, 44 Ca, 46 Ca, and 48 Ca), where 40 Ca is the most abundant, comprising about 97% of naturally occurring calcium. However, calcium doesn't form gaseous compounds near room temperature and atmospheric pressure, hence commonly used enrichment schemes such as gas centrifuge or thermal diffusion are not strictly impossible, but challenging. On the other hand methods based on liquid calcium forms, such as ion exchange chromatography or electrophoresis are difficult to scale up. Hence, enriched calcium materials are very expensive and a recycling process of the target needs to be developed (Jalilian et al., 2020). The choice of method used in the recovery process is related to the form of the initial target material, and it is particularly important when large amounts of the highly enriched target material are used. The recycling will contribute to a significant cost reduction, considering that the target material alone costs around 20-25 € per 1 mge (mg element) of [ 44 Ca]CaCO 3 dependig on the purchased quantity (price for 2020). Additionally, the purchased material needs to be purified before use so the method for recycling applies here as well.
The published data on this topic is quite limited. Few publications deal with the calcium carbonate targets (natural or enriched in 44 Ca) for production of 44 Sc in cyclotrons. One of the early works by Krajewski et al. (Krajewski et al. 2013) reported the use of chelating resin Chelex 100 for separation of 44 Sc and target recycling efficiency of only 60%. In the work by Alliot et al. (Alliot et al. 2015a), using the mixture  (Chakravarty et al. 2017) precipitation Dissolution of the target in 1 M HCl, then alkalized with 25% ammonia. Method takes advantage of the insolubility of Sc(OH) 3 either as a precipitate or coprecipitate, which can be separated from calcium by using microfilters with PTFE membrane (0.22 μm) (Minegishi et al. 2016;Severin et al. 2012;Duval and Kurbatov 1953) bicarbonate/methanol, the rate of solvent evaporation was increased and the solubility of calcium carbonate was lowered (Kan et al. 2002). The recovery yield of enriched calcium of 90 ± 2% was obtained when starting with CaCO 3 material and the HCl solutions from the extraction process. The solution was then loaded on a pre-conditioned AG1 × 8 column to retain all metallic impurities (Cu, Co, Fe…) and enriched 44 Ca was rinsed in 9 M HCl. This solution was evaporated to dryness and the dry residue recovered in a mixture of bicarbonate buffer (pH = 10.33)/methanol, filtered and the residue dried in an oven at 105°C to remove water and methanol. The recycled targets were irradiated, and no significant difference in production yield was observed (Alliot et al. 2015a When starting material is TiO 2 , the enriched titatnium may be recovered and reused. A simple procedure was developed that recovers around 98.5% of the oxide based on precipitation of titanium at basic pH followed by conversion of the oxide using higher temperature (Kolsky et al. 1998). Other processes for recovery of titanium from its acidic solutions use either HF and HNO 3 or converting to nano-sized titanium dioxide (Kolsky et al. 1998). The use of TiO 2 as a starting material and its recycling process have been reported by Loveless et al. (Loveless et al. 2019b). The HCl solution and HNO 3 obtained from the extraction processes were collected and NH 4 OH at pH 8 was added to precipitate TiO 2 . The precipitate was then heated for 4 h at 400°C. Optionally, the enriched titanium oxide can be reduced with calcium or calcium hydride, resulting in titanium pellets with low content of impurities (Lommel et al. 2013).

Chelators for Sc
Scandium with its ionic radius (r i ) 74.5 pm (CN = 6) is chemically similar to Y 3+ and the heaviest lanthanides. Similarly to them, scandium is almost exclusively present in its compounds in the trivalent state. Therefore, ligands developed for these cations should be also suitable for chelating Sc. However, the chemistry of trivalent scandium has some differences compared to lanthanides; it is smaller (thus, harder and has a higher preference for hard oxygen donor ligands), and prefers donor numbers from six to eight (Kerdjoudj et al. 2016).
The multi-dentate ligands, which were already used in Gd(III)-based MRI contrast agents as well as for radiolanthanides, that is derivatives of DTPA or DOTA, were also the first choice for scandium chelation for medical applications (Mausner et al. 1995). The stability constant of Sc-DOTA complex was comparable to that for Lu 3+ and heavier lanthanides but higher than those for In 3+ and Ga 3+ . The 13 C NMR studies have shown that Sc(DOTA) similarly to Lu(DOTA) forms in solution complexes with eight-coordination geometry (Majkowska-Pilip and Bilewicz 2011) . The stability constants of scandium(III) complexes DTPA and DOTA (log K ScL 27.43 and 30.79 respectively) were determined from potentiometric and 45 Sc NMR spectroscopic data. Both complexes were fully formed even below pH 2. Complexation of DOTA with the Sc 3+ ion was much faster than with trivalent lanthanides. Proton-assisted decomplexation of the [Sc(dota)] − complex (τ 1/2 = 45 h; 1 M aq. HCl, 25°C) was much slower than that for [Ln(dota)] − complexes (Pniok et al. 2014). Therefore, DOTA and its derivatives were assumed to be very suitable ligands for scandium radioisotopes (Pniok et al. 2014;Huclier-Markai et al. 2015).
Thermodynamic data for scandium(III) complexes with polyamino-polycarboxylic ligands, such as NOTA, EDTA or TETA have been determined using potentiometric titration and free ion selective radiotracer extraction (FISRE) method and the values of stability constants were found to be in the order TETA<NOTA<EDTA<DTPA<DOTA. (Huclier-Markai et al. 2011) DOTA derivatives with phosphinic/methylphosphonic acid pendant arms (i.e. DO 3 A; DO 3 AP PrA , DO 3 AP ABn ) were also investigated showing that the stability constants of the monophosphinate analogues were somewhat lower than that of the Sc(DOTA) complex. (Kerdjoudj et al. 2016) The thermodynamic stability constant of recently developed chelating agent AAZTA, Sc(AAZTA) was reported to be lower than that of Sc(DOTA) but the striking difference was observed on the radiochemical yield at 25°C indicating that AAZTA quickly incorporated 44 Sc (Nagy et al. 2017a). The AAZTA derivative AAZTA 5 (1,4-bis (carboxymethyl)-6-[bis (carboxymethyl)]amino-6-[pentanoic-acid]perhydro-1.4-diazepine) was synthesized representing a bifunctional version with a pentanoic acid at the carbon-6 atom. [ 44 Sc]Sc-AAZTA 5 complexes as well as [ 44 Sc]Sc-AAZTA 5 -TOC were formed at room temperature within 5 min in the pH range 4 to 5.5 and were very stable (Sinnes et al. 2019). Another chelator, H 4 pypa (N 5 O 4 ) has been shown to exhibit high complexation constant with Sc (log K = 27) as well as with In, Lu, Y and La, as determined by potentiometric titration and UV spectrometry (Li et al. 2019a;Li et al. 2019b). Its radiolabeling could be performed at room temperature and in a quite wide range of pH values within 10 min. Also, it has been conjugated to a Glu-urea-Lys based PSMA (prostate-specific membrane antigen, e.g. PSMA-617). When labelled with 44 Sc (pH = 4.5, 30 min RT) this conjugate showed specific tumor uptake (4.86% ID/g from ex vivo biodistribution, 8 h p.i.) without significant off-target uptake, except in the kidney. (Li et al. 2019b) Further search on chelators enabling Sc chelation at room temperature led to the development of the smallcavity triaza-macrocycle-based, picolinate-functionalized chelator H 3 mpatcn. Spectroscopic and radiochemical studies established the [Sc(mpatcn)] complex as kinetically inert and appropriate for biological applications. As a proof-of-concept bifunctional conjugate targeting the prostate-specific membrane antigen (PSMA), picaga-DUPA, chelated 44 Sc to form { 44 Sc}Sc(picaga)-DUPA at room temperature with an apparent molar activity of 60 MBq μmol − 1 and formation of inert RRR-Λ and SSS-Δ-twist isomers. Sc(picaga)-DUPA exhibited a K i of 1.6 nM for PSMA and the 44 Sc labelled Sc(picaga)-DUPA revealed high-quality images in prostate cancer-bearing animals. H 3 mpatcn and its bifunctional analogue picaga represent new additions to the chelator toolbox for the emerging 44/47 Sc theragnostic isotope pair as reported by Vaughn et al. (Vaughn et al. 2020a). Nonetheless, the molar activity calculated from their experimental data was estimated to be 0.03 MBq/nmol and thus the metal to ligand ratio was ranging between 1: 485,500 to 1: 1,084,600. Hence, picagaDUPA seems exhibiting complexation/radiolabeling at RT but complexation constant must be established and the potential of this ligand to provide higher molar activity must be evaluated with sources of 44 Sc of high purity.
Scandium also displays favorable physical properties and chemistry for conjugation to mAb-chelate systems. Since it is chemically similar to 90 Y and close to 177 Lu, the same ligands developed for 90 Y or 177 Lu can be used for chelating 47 Sc. Table 4 provides a summary of the main chelates that have been evaluated, with the corresponding formula and the measured value of the complexation constant with scandium. The table reports values reported in the original manuscripts, accuracy of which is not homogeneous across the data. For simplicity, the uncertainties (available only for selected studies) are not reported.

Radiolabeling studies
Labeling protocols should allow high labeling yields, radiochemical purity and molar activity (Fani and Maecke 2012). Labelling efficiency of ligands is usually tested at various pH, temperature and ligand concentration (or more correctly, at different radiometal-to-ligand molar ratios) and is monitored as a function of time in order to optimize the radiolabeling. Herein, we discuss the development of labelling methods that were developed for scandium radionuclides. Neverthelles, the reported data are often difficult to compare. To overcome this problem, the recommendations on consensus nomenclature rules for radiopharmaceutical chemistry should be followed (Coenen et al. 2017).
Most of the published radiolabeling works have been done with 44 Sc. All have highlighted the importance of reaching a high chemical purity of the final 44 Sc fraction before radiolabeling. The presence of other metals may interact with the DOTAchelator (or any other chelator) that can affect the radiolabeling yield. The content of environmental contaminants (i.e. Al(III), Cu(II), Pb(II), Zn(II), Fe(III)) in 44 Sc is frequent. The most problematic is Fe 3+ , for which the stability with the DOTA ligand is greater than that for Sc 3+ . By contrast, the influence of divalent metal cations (considered as contaminants) is negligible due to the lower stability of DOTA-divalent metal complexes. Although DOTA forms very stable complexes, it exhibits slow formation kinetics at room temperature, that could be increased by heating. However, elevated temperatures remain an important obstacle for efficient labelling of heat-sensitive molecules such as antibodies. Click-chemistry or the new ligands permitting the formation of scandium complexes with faster kinetics, or at much lower temperatures than that required for DOTA, might circumvent this issue. In this context, AAZTA or H4pypa seem to be interesting alternatives (Nagy et al. 2017a;Li et al. 2019a), although their availability is limited. Table 5 provides a summary of the radiolabeling studies published to date. Notably, there are many discrepancies in presentation of results and in most of the cases the quality control results of the batches are not reported. When available, radiochemical yield and molar activity are indicated. This inhomogeneity of the published data makes      (Szkliniarz et al. 2016) the direct comparison of radiotracers efficacy difficult. The data in Table 5 are reported as published in the original papers, pointing to the discrepancies in the measurement units used. Indications on the quality control procedure are also given as well as the conditions of post-radiolabeling purification, wherever reported. The use of suitable buffers, stabilizers/free radical scavengers etc. was addressed.

Pre-clinical studies
The use of radiolabeled compounds for in vitro and in vivo preclinical animal studies has steadily increased in recent years, becoming a much more widely used method for studying biochemistry, physiology and pharmacology (Kilbourn and Scott 2018). The increasing availability of 44 Sc and its compatibility with numerous chelators, among those DOTA, has led to multiple studies in animals, especially imaging studies. The half-lives of 43 Sc, 44 Sc or 47 Sc are compatible with the pharmacokinetics of a fairly wide range of targeting vectors. Several somatostatin analogs, either DOTA-derivatized (DOTATOC, DOTATATE, DOTANOC) or with NODAGA as a chelator, have been labeled with scandium radionuclides to prove the suitability of the obtained radionuclide solution for radiolabeling (Pawlak et al. 2019;Loveless et al. 2019a) or to compare under the same conditions the in vitro/in vivo imaging/therapeutic potential of scandium radiolabeled peptides to their 68 Ga-or 177 Lu ( 90 Y)radiolabeled counterparts (Walczak et al. 2015;Loveless et al. 2019a;Singh et al. 2017;Pruszynski et al. 2012). The scandium radionuclides labeled bombesin analogs (Koumarianou et al. 2012), RGD peptides (Domnanich et al. 2017c), folate derivatives (Muller et al. 2018;Muller et al. 2014;Siwowska et al. 2019) and PSMA ligands (Umbricht et al. 2017) have been evaluated pre-clinically. Few studies have also reported on scandium radiolabeled antibodies, antibody fragments or affibodies Honarvar et al. 2017; Moghaddam-Banaem 2012) and nanoparticles . A summary is given in Table 6. The early study by Koumarianou et al. (Koumarianou et al. 2011) was conducted on "cold" complexes with natural scandium and gallium. The binding affinity of both nat Sc-DOTATATE and nat Ga-DOTATATE in AR42J cell line was in the sub-    (Ferguson et al. 2020). In more recent studies, the binding affinity of nat Sc-PSMA-617 was evaluated in comparison to nat Ga-PSMA-617, however using different methods and cell lines (PC-3 PIP cells and PSMA+ LNCaP cells, respectively) (Umbricht et al. 2017;Eppard et al. 2017). Though, the binding affinity to the target was found to be in the same molar range. Both [ 44 Sc]Sc-and [ 68 Ga]Ga-PSMA-617 exhibited similar in vivo behavior, with [ 44 Sc]Sc-PSMA-617 displaying higher tumor-to-liver ratios at 15 and 30 min p.i. These values correlated more closely to [ 177 Lu]Lu-PSMA-617 than to [ 68 Ga]Ga-PSMA-617, and therefore were considered useful for pretherapeutic dosimetry (Umbricht et al. 2017).
The comparison of 44 Sc and 68 Ga for imaging of animals with melanocortin-1 receptor (MC1-R), positive tumors using DOTA-NAPamide, at 4 h p.i. revealed significantly higher tumor uptake of 44 Sc (81.7 ± 7.7%ID/g) over 68 Ga (17.3 ± 1.85%ID/g) (Nagy et al. 2017b). DOTA-puromycin was radiolabeled with the generator produced 44 Sc and investigated for the potential imaging of protein synthesis in vivo. In μPET images of tumor-bearing rats significant tumor uptake of [ 44 Sc]Sc-DOTA-puromycin and a clear-cut tumor visualization were demonstrated. In addition, the cellular uptake of [ 44 Sc]Sc-DOTA-puromycin could be suppressed by blocking protein synthesis (Eigner et al. 2013).
In vitro, [ 47 Sc]Sc-folate demonstrated effective reduction of folate receptorpositive ovarian tumor cell viability similar to 177 Lu-folate, but 90 Y-folate was more potent at equal activities due to the higher energy of emitted β − -particles. Comparable tumor growth inhibition was observed in mice that obtained the same estimated absorbed tumor dose (~21 Gy) when treated with [ 47 Sc]Sc-folate (12.5 MBq), [ 177 Lu]Lu-folate (10 MBq), and [ 90 Y]Y-folate (5 MBq), respectively. However, there were no statistically significant differences among the therapeutic effects observed in treated groups (Siwowska et al. 2019).
CHX-A"-DTPA has been successfully used for radiolabeling of the monoclonal antibody (mAb) Cetuximab. ) Another study evaluated DOTA-HPMA (N-(2-hydroxypropyl)methacrylamide) conjugates for labeling efficiency with 68 Ga, 177 Lu and 44 Sc and showed that the 44 Sc labeled polymer allowed for in vivo PET imaging and ex vivo measurements of organ distribution for up to 24 h .

Clinical experiences
The validity, usefulness and advantages of 44 Sc have been demonstrated by studies featuring 44 Sc-radiolabeled targeting vectors, including 44 Sc radiopharmaceuticals in early clinical studies. The first of them has been performed using [ 44 Sc]Sc-DOTA-TOC prepared from 44 Sc produced in a cyclotron (Singh et al. 2017). Two patients were included in this study after being treated by peptide receptor radionuclide therapy due to neuroendocrine neoplasms. The obtained image quality was comparable to that of 68 Ga. This very encouraging proof-of-concept study showed no clinical adverse effects with normal hematology, nor with renal and hepatic profiles.
The other studies were conducted in patients with metastasized castrate-resistant prostate cancer (Eppard et al. 2017;Khawar et al. 2018a;Khawar et al. 2018b). A firstin-human investigation was carried out in a cohort of four patients (mean age 70.0 ± 1.8 years) registered for [ 177 Lu]Lu-PSMA-617 therapy. Physiological tracer uptake was observed in kidneys, liver, spleen, small intestine, urinary bladder, and salivary glands and pathological uptake in both soft and skeletal metastases. SUV values were significantly lower in the kidneys (14.0) compared to [ 68 Ga]Ga-PSMA-11 PET (30.5). All other measured SUV values did not show a statistically significant difference. Tumorto-liver ratios were found to lie between 1.9 and 8.3 for [ 68 Ga]Ga-PSMA-11 and between 2.5 and 8.8 for [ 44 Sc]Sc-PSMA-617 after 120 min. For [ 44 Sc]Sc-PSMA-617 the ratios were higher and no statistically significant differences were observed. Total and % activity were highest in the liver followed by kidneys, spleen, small intestine and salivary glands. Rapid wash-out was seen in liver and spleen, and gradually over time in kidneys. Kidneys received the highest radiation absorbed dose of 0.354 (0.180-0.488) mSv/MBq. No adverse pharmacological effects were observed. Still, the authors concluded that the clinical advantages for individual dosimetry or other applications like intraoperative applications need to be investigated in further studies (Eppard et al. 2017).
The impact of physical properties of 44g Sc on image quality has been studied by Bunka et al. (Bunka et al. 2016) and recently evaluated by Rosar et al. (Rosar et al. 2020) in comparison to 68 Ga with different imaging phantoms. The lower mean positron energy of 44g Sc (0.63 MeV) compared to 68 Ga (0.83 MeV) can result in better spatial image resolutions. However, high-energy γ-rays (1157 keV) are emitted at high rates (99.9%) during 44g Sc decay, which can reduce image quality. Despite the presence of high-energy γ-rays in 44g Sc decay, a higher image resolution of small structures was observed with 44g Sc when compared to 68 Ga. Structures as small as 1 mm could be visualized and analyzed using two different pre-clinical PET scanners. Recently, Lima et al. (Lima et al. 2020) evaluated quantitative capabilities of 44 Sc-PET using a commercial PET scanner and concluded that that clinical 44 Sc-PET imaging has the potential to provide signal recovery in lesions of different sizes comparable to current 18 F-PET standards.
Biodistribution and radiation exposure to normal organs with [ 44 Sc]Sc-PSMA-617 in metastatic castration-resistant prostate carcinoma patients were investigated by Khawar et al. (Khawar et al. 2018a). These authors extrapolated the pharmacokinetics of To date, very limited data is available on radiation doses from 47 Sc labelled radiopharmaceuticals. The absorbed dose in human organs of 47 Sc-EDTMP (ethylene-diamine-tetramethylene-phosphonic acid) has been extrapolated from animal biodistribution using the MIRDOSE formalism. However the animal biodistribution data in this study were collected using EDTMP radiolabeled with the reactor produced 46 Sc of rather low specific activity 148 MBq/mg (Deilami-Nezhad et al. 2017). The radionuclide 47 Sc is a low energy beta emitter, similarly to 177 Lu, and the mean range of its radiation in bone is 0.20 mm, compared to 0.15 mm for 177 Lu (as extrapolated from Bouchet et al. (Bouchet et al. 2000)).

Quality specifications
The chemically identical radiopharmaceuticals for diagnosis and therapy using the matched pair 43 Sc or 44 Sc with 47 Sc are very appealing. Although several groups are now producing scandium radionuclides locally and a few centers demonstrated the potential of Sc radionuclides for medical applications, the benefit of their use in the clinical setting is yet to be shown. Given the amount of work and documentation needed to obtain the approval for a clinical trial of a new radiopharmaceutical, this review aims to present the production options for Sc radionuclides and related quality constraints. This information might be useful when implementing Sc radionuclides production technology in the existing or planned production facilities. 43 Sc (β + = 88.1%, E β+ = 476 keV) and 44g Sc (β + = 94.3%, E β+ = 632 keV) have half-lives of 3.89 h and 3.97 h, respectively (Singh and Chen 2015). Although properties of these two isotopes are very similar, the main difference is the high-energy γ-radiation (Eγ = 1157 keV; Iγ = 100%) associated with 44g Sc. The high energy gamma radiation, similarly to that of 89 Zr (Eγ909 keV; Iγ = 99%), may not hamper its clinical use. It is considerd advantageous for 44g Sc as the radionuclide of choice for 3-γ photons imaging, a new modality of imaging (Sitarz 2020). Additionally, 44m Sc, which is co-produced with 44g Sc, has a half-life of 58.6 h and has been suggested for use as an in vivo PET generator for 44g Sc (Alliot et al. 2015a). However, the longer half-life coupled with the high-energy γray emitted from 44g Sc may lead to the unfavorable dosimetry for shorter biodistribution studies (Alliot et al., 2015a). High radionuclidic purity 44g Sc production with proton irradiation necessitates the use of enriched target material and a recycling procedure to keep the cost reasonable. The determination of acceptable limits for each radionuclidic impurity requires careful consideration, as it can have an adverse effect on the implementation of new production methods. Potential contamination with 43 Sc can be argued to have no impact on patient safety and quality of images since it is a positron emitter with no high energy gamma radiation component, in contrast to the expected radiation burden due to 44g Sc contamination in 43 Sc. 44m Sc decays to 44 Sc with low energy gamma (271 keV) emission, which does not cause any harm to the patient. The limit for 44m Sc content should be determined to avoid unnecessary radiation doses to the patient, because of the longer effective half-life of the 44m Sc/ 44 Sc isotope pair. However, the biological half-life and critical organs should be determined for unchelated scandium isotopes before using worst-case dosimetry calculations for the determination of the limit of longer half-life impurities, as discussed earlier for 68 Ga labelled peptides and influence of 68 Ge breakthrough on dosimetry (Velikyan et al. 2013).
From that perspective, 43 Sc has the most favorable radiation characteristics for conventional PET, however it can be produced efficiently only using alpha particle beams. 44 Sc is in the most advanced state and has the highest potential for broad applications.
Medical cyclotrons that currently supply 18 F to hospitals can be used to produce 44g Sc or 43 Sc, with a solid target system. The use of liquid target might not be reasonable as the amount of enriched material nedded is high. With the near to 4 h half-life and reasonable production cross-section, there is a potential for regional distribution following mass production at a single cyclotron unit with solid target. With the increasing numbers of installed cyclotrons, the availability of 43/44 Sc may increase. Aiming for the production of high activities, a solid target is needed. This requires the addition of specialized hardware on the cyclotron and in the processing hot-cell, which is available from cyclotron providers but not frequently installed on 18 F production machines. The co-emission of a high-energy γ-ray similar to 89 Zr has to be taken into consideration when planning for radiation protection. If not controlled, it may increase the radiation dose to the patient and staff. 44 Ti/ 44 Sc generator currently has a high production cost and requires a regular and efficient use of the generator over long periods. 44m Sc half-life enables transportation of 44m Sc-labeled radiopharmaceuticals to hospitals that are located quite far away from the radiopharmaceutical production site (centralized production). This is based on the in-vivo generator principle.
Both positron emitters, 43 Sc and 44g Sc can be used for theranostic studies with 177 Lu or other lanthanides. Scandium has chemistry close to 177 Lu and possibility of centralized production at the regional level, which makes it more desirable than 68 Ga (local production only) for its use in a theranostic pair. Application of 47 Sc, as an alternative radionuclide to 177 Lu, was proposed in earlier works (Srivastava 2013). The advantage of 47 Sc production compared to that of n.c.a. 177 Lu, is the relatively easy isolation of the radionuclide from the target, but the disadvantage is the smaller cross-section of the nuclear reaction compared to 177 Lu production. Nonetheless, the main concern lies with the co-produced impurities. Another limitation concerns the availability of the starting enriched material needed for the set-up of efficient separation chemistry. At the time of writing this article, we are aware of several research groups developing ways of overcoming these limitations, as presented by Jalilian et al. (Jalilian et al. 2020).
When it comes to the scandium-based radiopharmaceuticals, unlike with 68 Ga, the scientific community involved is unfortunately far from following a standardized procedure. From Table 5, it is inferred that neither the metallic impurity contents nor the radionuclidic purity of the resulting radiolabeled compound is always given. There is also a lack of homogeneity in the expression of the molar activity of the resulting radiolabeled vector. Though, some compromise could be found on the radiolabeling protocol and quality control. This is certainly a good basis to pave the way for a monograph for the European Pharmacopeia, together with the use of the recently approved nomenclature guidelines for radiopharmaceuticals (Coenen et al. 2017).
Next to the radionuclide generators and kits, the European Union (EU) directive 2001/83 (The European Parliament and the Council of European Union 2001) dealing with radiopharmaceuticals, defines also the "radionuclide precursors" (EU Commission 2004). The radionuclide precursor is defined as "any other radionuclide produced for the radio-labeling of another substance prior to administration" and as such has to hold a marketing authorization when introduced to the market. Even if used in a hospital radiopharmacy with a cyclotron on-site, and having a status of starting material, as postulated in a position paper by Neels et al. (Neels et al. 2019), the radionuclide precursor needs to be controlled by performing several quality control tests (identity, purity, assay, etc.). To assure safe use in humans, these tests and limits should comply with certain quality standards, and be supported by a suitable quality management system. It is generally accepted that such quality standards are established in Pharmacopoeia monographs.
The expected final form of the radionuclide, regardless of the production method used, is the form of solution for radiolabeling. The Ph.Eur. General monograph on Radiopharmaceutical preparations (0125) can serve as the reference for establishing quality specifications for the final product solution of 43 Sc, 44 Sc or 47 Sc (EDQM 2020a). One should bear in mind that the Ph.Eur. monographs on Lutetium ( 177 Lu) solution for radiolabelling (mon. 2798) (EDQM 2020b), and Yttrium ( 90 Y) chloride solution for radiolabelling (mon. 2803) (EDQM 2020c), can be also very informative, similarly as the monograph of Gallium-68 ( 68 Ga) chloride solution for radiolabelling (mon. 2464) (EDQM 2020d).
One needs to remember, however, that the specified parameters will depend on the production route (target material, nuclear reaction, accompanying nuclear reactions, chemical processing etc.). For example, when using 44 Sc obtained from the 44 Ti/ 44 Sc radionuclide generator, the potential breakthrough of 44 Ti will be of concern as well as its impact on the quality (and safety) of the final radiopharmaceutical. The detailed specifications for quality control of generator eluate as well as for the quality assessement of 44 Sc radiolabeled PSMA-617 prior to its administration to patients were developed by Eppard (Eppard 2018).
The Group 14 (radioactive compounds) of European Pharmacopoeia elaborated the Guide for the elaboration of monographs on radiopharmaceutical preparations, European Pharmacopoeia, EDQM Edition 2018, which should help to prepare the monographs for new radiopharmaceuticals (https://www.edqm.eu/sites/default/files/ guide_-_guide_for_the_elaboration_of_monographs_on_radio-pharmaceutical_ preparations_-_october_2018.pdf). Based on this Ph. Eur. Guide, exemplary general list of tested parameters and the involved methods for the potential scandium radionuclide precursors are briefly discussed below.

Title, definition and production sections
In general, the definition states that the monograph applies to the substance obtained by a certain route of production and in the case of a radionuclide precursor the name of the substance is completed by "for radiolabeling". Therefore, in case of the new monographs, e.g. for gallium-68 and technetium-99 m, the title of the monograph itself contains a reference to the production method, like: Gallium ( 68 Ga) Chloride (Accelerator-Produced) Solution For Radiolabelling (mon. 3109) (EDQM 2020e) or Sodium Pertechnetate ( 99m Tc) Injection (Accelerator-Produced (mon. 2891) (EDQM 2020f).
Such titles and definitions are intended to make the recipient/reader aware that other methods of obtaining radionuclides may and, in fact do, result in a completely different profile of potential contaminants, mainly radionuclidic. These constraints apply to the radionuclides of scandium, therefore it is not possible to develop a single monograph even for one of the scandium radionuclides, but it would be necessary to develop several monographs for this radionuclide depending on the way the radionuclide is produced. In the following, a monograph-like characteristics are proposed.
Identification: (characteristic of type of radiation -spectra, approximate half-life, pH,
The most prominent gamma photon of scandium-47 has an energy of 0.159 MeV.
To test the radiochemical purity of obtained 47 Sc when used for radiolabeling of selected common chelators such as DOTA, DTPA or EGTA, or peptide conjugates such as DOTATATE or DOTA-bombesin, with further TLC development to assess the radiochemical purity of 47 Sc-labelled compound. Usually, the radiochemical purity should be not less than 98-99% to confirm the quality of 47 Sc.
This approach can also serve as the indirect test of 47 Sc suitability for radiolabeling, and the experiments can be designed to check the effective specific activity of obtained 47 Sc solutions. Such tests provide very useful information during the process development, especially when instrumental methods for determination of chemical purity are not available on-site.
Bacterial endotoxins General applicable limit for the presence of bacterial endotoxins in radiopharmaceutical preparations is less than 175 IU/V, V being the maximum volume to be used for the preparation of a single patient dose, if intended for use in the manufacture of parenteral preparations without a further appropriate procedure for the removal of bacterial endotoxins. Whereas, however, it should be borne in mind that the limit of 175 IU/V applies to the final injection preparation, therefore the determination of the limit of bacterial endotoxins for the radionuclide precursor should take into account the additional effect of the labelled substance and should therefore be established on the basis of actual results obtained for several batches of the radionuclide precursor intended for radiolabeling.
Sterility The test applies if intended for use in the manufacture of parenteral preparations without a further appropriate sterilization procedure.
From the regulatory point of view, labeling efficiency of ligands is usually tested at different pH, temperature and ligand concentrations (or more precisely, at different radiometal-to-ligand molar ratios), and monitored as a function of time to optimize the radiolabeling.
The preparation of 43/44/47 Sc-radiolabeled peptides for patient administration is designed to be performed on a modular system in which the final radiotracer is typically purified on a cartridge and diluted in a physiological vehicle before injection. Using the same commercially-available modular entity as is commonly used for the preparation of 68 Ga-radiopharmaceuticals has the advantage to allow the preparation of 43/44/47 Scradiopharmaceuticals for clinics without further evaluation and GMP validation of the system.
All these elements combined with a larger availability of the isotope production sites and broader distribution network are expected to allow more clinical trials with scandium radionuclides labelled radiopharmaceuticals (Domnanich et al. 2017c). These radionuclides have attracted considerable interest in the last decade and their broader availability opens new avenues for investigations.

Conclusions
The 44 Sc/ 47 Sc or 43 Sc/ 47 Sc pairs are appealing true theranostic radionuclides for Nuclear Medicine. Reliable methodologies for the production of all discussed herein medical scandium radionuclides exist and are increasingly reliable. With the increasing availability of scandium PET isotopes, some preclinical studies have been conducted, but remain limited and mostly performed with peptides. Since scandium exhibits suitable conjugation chemistry to be coupled with MAbs, and 44m Sc half life is adapted to long biodistribution time, it paves the way for bringing new developments in this area. Scandium-based vectors from diagnosis to therapy offer a great opportunity for dosimetric calculations and the development of personalized medicine.
Among scandium radionuclides, the 44 Sc seems to be in the most advanced state. Several production and purification methods were developed and some hold the promise of their relatively easy adaptation to the locally available infrastructure. At present, the primary importance of 44 Sc lies in its potential broad-scale availability based on the production in medical cyclotrons and the possibility to use it as the diagnostic match with 177 Lu or 47 Sc. However, efficient production of 47 Sc is still not developed though the half life of 47 Sc allows its production in a centralized facility, which might provide this radionuclide to several countries. Its widespread use will be questionable if one considers the current availability of 177 Lu and the problems associated with the use of a therapeutic agent without centralised marketing authorisation. The shorter half-life of 44 Sc can be beneficial considering the lower radiation dose for the patients. The main issue is the co production of 46 Sc when Ti target is used, but that is not a problem when produced through Ca target.
The 44 Sc is an alternative to 68 Ga, a well-established radiometal, with a similar field of application. The availability of 68 Ga from generator gave a boost of its widespread use, but it is clear now, that after the initial success in the introduction of new diagnostics (somatostatin analogues and PSMA), production capacity and cost efficiency will be the main factors determining the future of these examinations, and the interest of centralized production. Additionally, the short half-life of 68 Ga may limit its use for some pharmakokinetics studies. These factors favor fluorine labeling over radiometals. 44 Sc can find its place in the palette of medical cyclotrons, which could be capable of producing scandium before the daily 18 F irradiation. Using the [ 18 F]FDG transportation channels for shipping 44 Sc solution for radiolabeling or the GMP produced 44 Sc tracers to the nuclear medicine departments facilitates the use of new specialized tracers, currently relying on the 68 Ge/ 68 Ga generator supply.
The authors trust that the medical applications of scandium radionuclides will be growing, however standardization of their quality is still needed. The question remains whether there is enough experience already to draft the pharmacopoeia monograph for any of scandium radionuclides as radipharmaceutical precursors and which of the scandium radionuclides becomes economically feasible.