scieee AI-readable full text Open interactive document viewer

Drugs used during the COVID-19 first wave in Vitoria-Gasteiz (Spain) and their presence in the environment

Domingo Echaburu, Saioa,Irazola Duñabeitia, Mireia,Prieto Sobrino, Ailette,Roncano, Bryan,Lopez de Torre-Querejazu, Amaia,Quintana, Ainhoa,Orive Arroyo, Gorka,Lertxundi, Unai

Abstract

This study was funded by the Council of Vitoria-Gasteiz and Fundación Vital, AMVISA, the Basque Government through the financial support as consolidated group of the Basque Research System (IT1213-19), and the Agencia Estatal de Investigación (AEI) of Spain, the 2020 call for the generation of knowledge and scientific and technological strengthening of the R&D&i system and for the R&D&i focused on society's challenges, through project PID2020-117686RB-C31.

Full text

1 ABSTRACT 1 The city of Vitoria-Gasteiz was one of the probable first entrances of the SARS-CoV2 in Spain, 2 one of the worst affected countries in the world during the first COVID 19 wave. Driven by the 3 urgency of the situation, multiple drugs with antiviral activity were used off label. Sadly, most 4 of these treatments were of little or no benefit and thus, the number of patients suffering from 5 COVID-19 attended in intensive care units (ICUs) multiplied. After being administered to 6 patients, a variable proportion of these drugs reach the environment where they may have 7 detrimental effects, although this aspect is usually ignored by healthcare professionals. In this 8 study we measured the patterns of hospital drug use in the city of Vitoria-Gasteiz (Spain) during 9 the first COVID-19 wave pandemic, focusing on those with antiviral activity and those used in 10 the ICUs. Subsequently, we measured concentrations of selected drugs in the city´s wastewater 11 treatment plant influent and effluent and estimated the potential risk for the environment. The 12 hospital use of certain antivirals and drugs used for sedo-analgesia were dramatically increased 13 during the first wave (cisatracurium was multiplied by 25 and lopinavir/ritonavir by 20). A mean 14 of 1.632 daily defined doses of hydroxychloroquine were used during the period of February-15 May 2020. In this study we report the first positive detection of hydroxychloroquine ever in the 16 environment. We also show the second positive report of lopinavir. Low risk was estimated for 17 hydroxychloroquine, lopinavir and ritonavir (Risk quotients (RQ) <1), and medium risk for 18 azithromycin (RQ 0f 0.146). 19 20 KEYWORDS 21 COVID-19 pandemic; drug pollution; One Health; pharmacoepidemiology; LC-q-Orbitrap; target 22 analysis 23 24 25 Drugs used during the COVID-19 first wave in Vitoria-Gasteiz (Spain) and their presence in the environment S. Domingo-Echaburu, M. Irazola, A. Prieto, B. Rocano, A. Lopez de Torre-Querejazu, A. Quintana, G. Orive, U. Lertxundi This is the accepted manuscript of the article that appeared in final form in Science of The Total Environment 820 : (2022) // Article ID 153122, which has been published in final form at https://doi.org/10.1016/j.scitotenv.2022.153122. © 2022 Elsevier under CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/) 2 1. INTRODUCTION 26 Spain was hit hard by the COVID-19 first wave, being considered by many as 27 one of the worst affected countries in the world (Garcia-Basteiro et al., 2021). 28 Phylo-geographic analysis have shown that the city of Vitoria-Gasteiz was one 29 of the probable first entrances of the SARS-CoV2 in the country (Gomez30 Carballa et al, 2021). With 216% intensive care unit (ICU) capacity expanded 31 by March 18th, 2020, our city was one of the hardest hit among one of the 32 most dramatically affected regions in the world (Barrasa et al., 2020). 33 In those times of scientific uncertainty, and driven by the urgency of the 34 situation, many different drugs with antiviral activity were desperately 35 administered to save patients´ lives. Sadly, many of this off-label treatments 36 showed little or no benefit (Boulware et al., 2021, Skipper et al, 2020, 37 Calvacanti et al 2020, RECOVERY, 2020) and some of them were even harmful 38 for patients (Eftekhar et al., 2021). Moreover, as the number of patients 39 suffering from COVID-19 attended in ICUs multiplied, the amount of sedo40 analgesics and neuromuscular blocking agents also incremented significantly 41 (Corregidor-Luna et al, 2020). 42 One aspect that has been vastly ignored by healthcare providers is the 43 environmental impact of many of the drugs used during the pandemic. After 44 being administered to patients, since a variable proportion of these drugs 45 reach the environment where they may have detrimental effects (Tarazona et 46 al., 2021, Farias et al., 2021, Elsaid et al., 2021). The ecotoxicity of some 47 pharmaceuticals used for COVID-19, e. g. azithromycin and ivermectin is 48 reasonably well addressed. Azithromycin is particularly toxic for 49 cyanobacteria, and ivermectin shows a moderate toxicity for fish and algae and 50 an extremely high toxicity for invertebrates (Janus Info, 2021a). 51 3 In this study we aimed to measure the patterns of hospital drug use in Vitoria52 Gasteiz during the first COVID-19 wave pandemic, focusing on those with 53 antiviral activity and those used in the ICUs. Subsequently, we tried to measure 54 these drugs concentrations in the city´s wastewater treatment plant influent 55 and effluent, to assess their potential ecotoxicological effects. 56 57 2. METHODS 58 2.1 Hospital drug consumption during the COVID-19 first wave in Vitoria59 Gasteiz 60 Vitoria-Gasteiz is the capital city of the Basque Country, an autonomous region 61 located in Northern Spain, where two public acute-care hospitals pertaining to 62 the Araba Integrated Healthcare organization (Txagorritxu and Santiago) 63 attend a population of 248.087 (EUSTAT, 2021). Drug consumption data was 64 obtained from SAP program (an enterprise application software), which is 65 available for Osakidetza, the public health service provider since 1998. This 66 database contains information about all drug consumption and cost for all 67 public healthcare hospitals in the autonomous region. Drug consumption data 68 was obtained for the first wave period (February-May 2020), and compared 69 with the same period form the two previous years, i.e.: 2018 and 2019. 70 Studied drugs were those that were directly used to treat COVID-19 because of 71 their antiviral activity, including hydroxychloroquine and lopinavir/ritonavir, 72 or their immunomodulatory properties, like the antibiotic azithromycin, 73 baricitinib, tocilizumab, methylprednisolone and dexamethasone. Other 74 included drugs were: loperamide, because of its wide use to treat 75 lopinavir/ritonavir provoked diarrhea; neuromuscular blocking agents used in 76 ICUs like cisatracurium and rocuronium; sedo-analgesic drugs like intravenous 77 4 midazolam, fentanyl, remifentanil and propofol; antibiotics like levofloxacin 78 and ceftriaxone; bronchodilators like ipratropium and salbutamol; and low79 molecular-weight heparin enoxaparin. 80 After obtaining raw consumption data, the number of daily defined doses 81 (DDD) for individual drugs were calculated (WHO ATC, 2021). In the case of 82 cisatracurium, no DDD is available, so consumption was arbitrarily normalized 83 considering 20 mg. 84 85 2.2 Waste-water treatment plant influent and effluent concentrations 86 One liter 24h composite samples (200 mL every hour) of both influent and 87 effluent waste-water from the municipal WWTP of the city of Vitoria-Gasteiz 88 (Crispijana) were collected from April 28th to July 13th in polypropylene bottles. 89 Data on chemical oxygen demand, biological oxygen demand, total nitrogen, 90 total phosphorus and daily flow in the influent of the WWTP and the date of 91 each sample are available as supplementary material (Table S1). 92 A total of 16 samples were subsequently transported to the laboratory at the 93 University of the Basque Country (UPV/EHU) and stored at -20°C until their 94 processing. Water was filtered (cellulose filters 0.7 μm, 90 mm, Whatman) and 95 spiked with a deuterated standard mix and processed according to a method 96 previously validated (Gonzalez-Gaya et al., 2021). Briefly, three replicates of 97 250 mL (effluent) and 100 mL (influent) were extracted using in-house made 98 SPE cartridges containing 100 mg of cationic exchange (ZT-WCX), 100 mg of 99 anionic exchange (ZT-WAX) and 300 mg reverse phase (HRX) sorbents from 100 bottom to top. Conditioning was done with 10 mL of MeOH: ethyl acetate (1:1, 101 v/v) and 10 mL Milli-Q water, and after sample loading, the cartridges were 102 eluted with 12 mL of MeOH: ethyl acetate (1:1, v/v) containing 2% ammonia 103 5 and 12 mL of MeOH: ethyl acetate (1:1, v/v) 1.7 % formic acid. Both extracts 104 were combined, evaporated on a Turbovap (Zymark, Hopkinton, USA) at 40 °C 105 under a gentle N2 flow and reconstituted on 250 µL MeOH: Milli-Q water (1:1, 106 v/v). Final extracts were filtered with syringe filters (PP, 0.22 μm, 13 mm, Jasco 107 Analítica, Madrid, Spain) onto amber chromatography vials and analyzed in a 108 Thermo Scientific Dionex UltiMate 3000 UHPLC coupled to a Thermo Scientific 109 Q Exactive Focus quadrupole-Orbitrap mass spectrometer (UHPLC-q-Orbitrap) 110 equipped with a heated ESI source (HESI, Thermo-Fisher Scientific, CA, USA) at 111 the conditions previously reported (González-Gaya et al., 2021). 112 In the validation of the analytical procedure used in this work satisfactory 113 results were obtained and are available as supplementary material (Table S3). 114 Eight calibration levels within 0.1 ng/g and 50 ng/g were injected in triplicate 115 for the determination of instrumental LODs and LOQs. Instrumental LODs were 116 estimated as the lowest concentration detected in the three injection 117 replicates, and in the case of instrumental LOQs, as the lowest concentration 118 detected with a relative standard deviation (RSD) of less than 30% and a 119 closeness to the true concentration values of more than 70%. The procedural 120 LODs and LOQs were stablished as the theoretical concentration measurable 121 and quantifiable in a water sample (a mixture of different wastewater samples) 122 taking into account the instrumental LODs and LOQs, the absolute recoveries 123 and the preconcentration factor of the target compounds. 124 The relative recovery was calculated as the percentage ratio of the compound 125 concentration estimated from the internal calibration to the theoretical 126 concentration, using six Milli-Q water samples spiked with all compounds of 127 interest. 128 6 The RSD values of the three replicates analysed ranged between 1 and 30%, 129 within the standards defined by the European Commission, which indicates an 130 RSD value ≤ 30% as acceptable (EUR-LEX, 2002). 131 132 2.3 Reported environmental concentrations for selected drugs in the 133 literature and ecotoxicity data 134 To find out reported environmental concentrations of the drugs used to treat 135 COVID-19 inpatients in the literature, we used The Pharmaceutical Database 136 published by the German Environment Agency – Umweltbundesamt (UBA, 137 2019). 138 Ecotoxicity data (Predicted no-effect concentration: PNEC) was looked for each 139 individual pharmaceutical. When no data was available, Ecological Structure 140 Activity Relationship tool (ECOSAR v2.0) from the United States Environment 141 Protection Agency was used (ECOSAR, 2021) was used to estimate PNECs. Then, 142 a factor of 25 was used to correct the impact of dilution of the effluent in 143 surface water (Keller et al., 2015). 144 3. RESULTS 145 3.1 Hospital drug consumption during the COVID-19 first wave in 146 Vitoria-Gasteiz 147 Figure 1 illustrates the drugs which consumption was more drastically 148 incremented. Interestingly, the average use of cisatracurium was multiplied by 149 25 and lopinavir/ritonavir by 20. Although the hospital use of 150 hydroxychloroquine was not registered before the first wave, a mean of 1.632 151 DDD use during the period of February-May 2020 was recorded. A complete list 152 of the detailed consumption of the selected drugs during the first wave 153 7 pandemic compared to the same period during the two previous years can be 154 consulted as supplementary material (Table S2). 155 -insert figure 1156 157 Waste-water treatment plant influent and effluent concentrations 158 Concentrations of selected drugs detected in the WWTP influent and effluent 159 are available in Figure 2. 160 -Insert figure 2161 3.2 Concentrations and ecotoxicity 162 Table 1 shows a summary of the available information in the literature about 163 the presence in different environmental matrices of some of the most relevant 164 drugs used during the first COVID-19 wave, and ecotoxicity data (PNECs and 165 RQ). 166 -Insert table 1167 4. DISCUSSION 168 Vitoria-Gasteiz was one of the most affected cities early in the first COVID-19 169 wave pandemic. On March 18, 2020, a 216 % expansion in ICUs capacity was 170 required to attend critically ill patients. 171 In this study we have shown that the use of cisatracurium was multiplied by 172 25, and lopinavir/ritonavir by 20, compared with pre-pandemic period. The 173 highest lopinavir/ritonavir/ use was registered on March, with 7.503 DDDs. 174 That would equal to 250 patients taking a daily lopinavir/ritonavir dose during 175 that month (approximately 1/1.000 persons from Vitoria-Gasteiz was 176 hospitalized and taking this antiviral drug on March 2020). 177 8 For most of these drugs, there is scarce information about their potential 178 deleterious effects in the environment. Tarazona et al predicted the potential 179 ecotoxicological consequences of some relevant drugs used during this 180 pandemic (Tarazona et al, 2021). Quantitative structure-activity relationship 181 (QSAR) estimations predicted that hydroxychloroquine (a metabolite of the 182 antimalarial chloroquine) is slightly less toxic than chloroquine for aquatic 183 organisms. So the authors extrapolated the chloroquine a PNEC value (120 184 µg/L) to hydroxychloroquine. An assessment factor (AF) of 100 was used to 185 derive this value from algal toxicity data. Other authors report that the most 186 sensitive organism to this antimalarial drug is the crustacean (Daphnia magna), 187 with a no observed effect concentration (NOEC 21 days, reproduction) of 85.8 188 µg /L (Janus Info, 2021a). The same authors report that this drug is s potentially 189 persistent (0% degradation in 28 days). We found a maximum concentration of 190 0.071 µg/L, which is below the 2 µg/L European Medicines Agency (EMA) 191 default estimation (1% population treated). As far as we are concerned, in this 192 study we provide the first detection of this drug in a WWTP effluent. Expected 193 environmental risk appears to be low, with a RQ <1 (Table 1). 194 There is much more information for the macrolid antibiotic azithromycin, 195 which is included in the European monitoring program under the Water 196 Framework Directive (Gomez-Cortes, 2021). The German Environmental 197 Agency´s Pharmaceutical Database contains more than one hundred reports of 198 positive detections for this drug, which has shown to be particularly toxic to 199 cyanobacteria (Microcystis aeruginosa) with a PNEC of 0.02 μg/L (OECD grow 200 inhibition test, NOEC with an AF of 10) (Tell et al., 2021). This value is below 201 the higher measured concentration in our study, which was 0.073 µg/L 202 indicating a potential risk for the environment (Risk quotient RQ= 0.146). 203 9 Regarding the antiviral lopinavir, we found a concentration of 0.033 µg/L. The 204 Swedish environmental classification of pharmaceuticals states that “lopinavir 205 has high potential for bioaccumulation” (log Dow = 4.7, which is >4) (Janus 206 Info, 2021b). So far, there is just one additional report of its presence in the 207 environment in the German database from South Africa. The study of Tarazona 208 et al (Tarazona et al., 2021) clearly reflects the scarce available 209 ecotoxicological information on these medicinal products (with the remarkable 210 exception of oseltamivir). They concluded that “despite the uncertainties in the 211 extrapolation of the ecotoxicity data, available information suggests that the 212 predicted concentrations for the antiviral and pharmacokinetic boosters are in 213 the range of the generic PNEC values for antivirals, and that specific attention 214 is required for sublethal effects on fish”. However, we found no experimental 215 data regarding lopinavir ecotoxicity. ECOSAR tool v2.0 predicts a chronic value 216 (geometrical mean of the NOEC and LOEC) of 4.5 µg/L for Daphnids (ECOSAR, 217 2021). Thus, the RQ for this drug is predicted to be low (RQ=2.9 x 10-3). In the 218 case of ritonavir, the same tool predicts a chronic value of 2.5 µg/L for fish 219 (RQ= 4.6 X 10-3). 220 We believe that the potential ecological impact of antivirals on viruses present 221 in the environment has not been sufficiently addressed so far. Very recently 222 Kuroda and co-workers predicted the occurrence, ecotoxicological risk and 223 acquired resistance of antivirals associated with COVID-19 in environmental 224 waters (Kuroda et al., 2021). They suggested that the removal efficiencies at 225 conventional WWTPs will remain low for half of the substances, and that high 226 concentrations might be present in effluents and thus persist in the 227 environment. They also estimated a high ecotoxicological risk in receiving river 228 waters for lopinavir and ritonavir, and medium risk for hydroxychloroquine 229 16  UBA. German Environment Agency, 2019. Umweltbundesamt. Für Mensch und Umwelt. 388 Available at: https://www.umweltbundesamt.de/en/database-pharmaceuticals-inthe389 environment-0 [Accessed March 17, 2021]. 390  WHO Collaborating Centre for Drug Statistics Methodology. ATC/DDD Index. Available at: 391 http://www.whocc.no/ atc_ddd_index/ [Accessed March 17, 2021]. 392  Wood TP, Duvenage CS, Rohwer E. The occurrence of anti-retroviral compounds used for 393 HIV treatment in South African surface water. Environ Pollut. 2015 Apr;199:235-43. doi: 394 10.1016/j.envpol.2015.01.030. 395 Table S1. Physicochemical data about the influent of the municipal WWTP of Vitoria-Gasteiz (Crispijana) during the first wave RAW WASTEWATER Date Flow Total nitrogen Chemical oxigen demand Biological oxigen demand Total phosphorus (m3/day) (mgN/l) (mgO/l) (mgO/l) (mg/l) April 28, 2020 138.300 36,1 372 240 3,79 April 29, 2020 109.900 38,3 362 210 4,01 May 4, 2020 108.900 45,6 475 280 4,75 May 7, 2020 89.400 51,9 514 310 5,03 May 11, 2020 191.600 27,2 330 150 3,45 May 14, 2020 144.600 31,1 260 140 3,09 May 18, 2020 104.160 47,5 410 250 4,51 May 21, 2020 95.808 44,5 462 270 5,15 May 25, 2020 83.328 50,9 522 300 5,34 June 1, 2020 75.744 60,2 666 380 6,13 June 8, 2020 81.120 48,7 457 290 5,02 June 15, 2020 77.760 55,7 523 300 5,43 June 22, 2020 76.224 55,6 614 330 5,80 June 29, 2020 68.928 58,0 645 380 6,05 July 6, 2020 66.912 57,7 738 390 6,93 July 13, 2020 73.440 50,6 628 320 5,19 Table S2. Hospital drug use in Vitoria-Gasteiz. PRE-PANDEMIC FIRST WAVE 2018 2019 Mean 2020 Mean Multiplying factor Drug Use in COVID-19 ATC DDD February March April May February March April May February March April May Hydroxiclhoroquin (mg) Antiviral P01BA02 516 mg - - - - - - - - - - 2.611.200 752.800 6.000 842.500 No pre-pandemic registered consumption Cisatracurium (mg) Neuromuscular blocker M03AC11 Not available 310 960 900 970 250 1.260 1.950 1.200 1.110 990 55.340 46.380 7.700 27.603 24,9 Lopinavir/ritonavir (mg) Antiviral J05AR10 800 mg 271.200 95.400 223.20 0 154.000 223.232 114.80 0 192.00 0 96.200 105.100 67.200 6.002.600 2.223.200 3.200 2.074.050 19,7 Loperamide (mg) Anti-diarrheal A07DA03 10 mg 400 678 498 1.188 726 462 804 1.364 570 920 5.384 2.500 520 2.331 4,1 Baricitinib (mg) Inmunomodulatory L04AA37 4 mg 336 336 672 896 1.976 2.112 2.688 2.016 1.224 4.022 6.448 3.512 1.260 3.811 3,1 Azythromycin (mg) Inmunomodulatory J01FA10 300 mg 115.000 128.10 0 108.10 0 103.800 133.600 103.60 0 118.00 0 117.900 115.850 154.200 899.500 241.600 95.000 347.575 3,0 Midazolam (mg) Sedoanalgesic N05CD08 15 mg 20.488 20.805 22.635 23.193 21.275 23.910 23.388 22.478 22.358 17.446 102.070 80.645 20.870 55.258 2,5 Fentanil iv (mg) Sedoanalgesic N01AH01 Not available 315 228 209 265 279 318 259 348 273 360 1.206 762 337 666 2,4 Ipratropium (mg) Bronchodilator R03BB01 Not available 1.310 1.262 1.246 1.087 1.186 1.219 1.015 1.261 1.240 1649 3551,5 1463,5 725,5 1.847 1,5 Rocuronium (mg) Neuromuscular blocker M03AC09 Not available 19.550 11.750 17.150 22.350 18.200 20.300 14.200 23.050 16.025 19.800 23.850 35.250 17.350 24.063 1,5 Ceftriaxone (g) Antibiotic J01DD04 2 g 2.534 2.640 2.371 2.136 2.385 2.504 2.532 2.997 2.572 2.829 6.316 3.145 2.410 3.675 1,4 Salbutamol (mg) Bronchodilator R03CC02 Not available 11.513 9.645 8.415 7.718 10.610 9.468 7.180 9.825 9.556 8.928 26.943 9.813 5.508 12.798 1,3 Tocilizumab (mg) Inmunomodulatory L04AC07 20 mg 21.196 27.408 32.846 31.704 26.364 31.404 23.996 25.140 29.406 20.796 78.640 38.100 19.024 39.140 1,3 Propofol (g) Sedoanalgesic N01AX10 Not available 983 1.058 917 1.116 1.149 1.177 947 1.162 1.118 1.495 1.854 1.373 821 1.386 1,2 Morphine iv (mg) Sedoanalgesic N02AA01 30 mg 75.830 71.250 97.010 104.740 33.990 54.400 83.270 81.795 62.825 64.300 80.080 76.650 49.960 67.748 1,1 Enoxaparin (g) Thromboprophylaxis B01AB05 20 mg 502 573 527 500 533 543 560 553 558 509 507 703 542 565 1,0 Remifentanil (mg) Sedoanalgesic N01AH06 Not available 1.842 1.892 1.822 2.075 1.659 1.416 1.183 1.712 1.654 1.643 1.500 1.576 1.520 1.560 0,9 Levofloxacin (g) Antibiotic J01MA12 500 mg 1.153 951 776 709 767 772 741 697 861 537 1.105 557 270 617 0,7 Dexamethasone (mg) Inmunomodulatory H02AB02 1,5 mg 36.581 40.420 33.676 33.936 23.297 40.283 25.822 25.004 40.352 28.258 43.184 21.171 16.231 27.211 0,7 Metilprednisolone (g) Inmunomodulatory H02AB04 20 mg 173 195 153 187 178 191 205 189 193 128 143 120 124 129 0,7 Table S3. Surrogates, relative recoveries (R %), instrumental and procedural limits of detection (LOD) and quantification (LOQ), ionisation mode and coefficient of determination (R2) for each compound studied. *CAL: External calibration. These compounds could not be corrected for due to the lack of suitable surrogates for correction. Their recovery, marked with **, is absolute, not relative and has been applied to the sample results. Compound Surrogate R % Instrumental Procedural Ionisation R2 LOD (ng/g) LOQ (ng/g) LOD (ng/L) LOQ (ng/L) Azithromycin *CAL 67** 2.5 8.5 5.2 17.2 Positive 0.9975 Hidroxychloroquine *CAL 57** 1.4 4.6 2.8 9.4 Positive 0.9888 Lopinavir Azoxystrobin- (cianophenoxi-d4) 89 1.0 3.3 2.0 6.8 Positive 0.9948 Ritonavir *CAL 62** 4.8 16.0 9.7 32.5 Positive 0.9859 Table S4. Concentration data for each day (ng/L) 28 Apr 29 Apr 4 Mar 7 Mar 11 Mar 14 Mar 18 Mar 21 Mar 25 Mar 1 Jun 8 Jun 15 Jun 22 Jun 29 Jun 6 Jul 13 Jul 28 Apr 29 Apr 4 Mar 7 Mar 11 Mar 14 Mar 18 Mar 21 Mar 25 Mar 1 Jun 8 Jun 15 Jun 22 Jun 29 Jun 6 Jul 13 Jul INFLUENT EFLUENT Azithromycin 51 28 25 51 9 27 4 6 73 47 69 43 51 42 Hydroxychloro -quine 32 32 57 71 32 Lopinavir 2 11 14 9 7 4 27 33 32 27 29 12 1 12 13 1 12 1 8 7 12 27 5 Ritonavir 22 2 28 22 16 22 24 32 22 2 24 34 16 38 12 13 16 15 8 15 16 14 16 11 13 19 16 23