Are we ready to combat the ecotoxicity of COVID-19 pharmaceuticals? An in silico aquatic risk assessment

Abstract

To fight COVID-19 with uncountable medications and bioproducts throughout the world has taken us to another challenge of ecotoxicity. The indiscriminate usage followed by improper disposal of unused antibacterials, antivirals, antimalarials, immunomodulators, angiotensin II receptor blockers, corticosteroids, anthelmintics, anticoagulants etc. can lead us to an unimaginable ecotoxicity in the long run. A series of studies already identified active pharmaceutical ingredients (APIs) of the mentioned therapeutic classes and their metabolites in aquatic bodies as well as in wastewater treatment plants. Therefore, an initial ecotoxicity assessment of the majorly used pharmaceuticals is utmost requirement of the present time. The present in silico risk assessment study is focused on the aquatic toxicity prediction of 81 pharmaceuticals where 77 are most-used pharmaceuticals for COVID-19 throughout the world based on the literature along with one drug nirmatrelvir [PF-07321332] approved for emergency use by US-FDA and three other molecules under clinical trial. The ecotoxicity of the studied compounds were predicted based on the three aquatic species fish, algae and crustaceans employing the highest quality QSAR models available from the literature as well as using ECOSAR and QSAR Toolbox. To compare the toxicity thresholds, we have also used 4 control pharmaceuticals based on the worldwide occurrence from river, lake, STP, WWTPs, influent and effluent followed by high reported aquatic toxicity over the years as per the literature. Based on the statistical comparison, we have proposed top 3 pharmaceuticals used for the COVID-19 most toxic to the aquatic environment. The study will provide confident predictions of aquatic ecotoxicity data related to abundant use of COVID-19 drugs. The major aim of the study is to fill up the aquatic ecotoxicity data gap of major medications used for COVID-19.

1. Introduction

The Hubei province of Wuhan city in China is attributed to the first reported coronavirus disease-2019 (popular as COVID-19) in late 2019 (Wu et al., 2020). The current outbreak of COVID-19 pandemic caused by 2019-nCoV (novel coronavirus) shares ∼80% genome similarity with already known variant SARS-CoV (severe acute respiratory syndrome-related coronavirus) that was initially reported in the year 2002-2003 in China (Zhou et al., 2020). The pandemic has affected every known corner of the planet and has infected over 666 million people across the globe and has caused the death of nearly 6.7 million people till January 4, 2023. Out of the total infected 659 million individuals, nearly 638 million survived with reports of several post-COVID-19 related symptoms (Woldometer COVID-19 Coronavirus Pandemic tracker, 2023). The recent resurgence of COVID-19 from the beginning of December 2022 makes this work more timely as we may witness another outbreak of COVID-19 throughout the world. The Omicron subvariants XBB.1.5 and BQ.1.1 appear to be more transmissible and one of the dominant strain at the present time in the U.S. along with other major subvariants including BQ.1, BA.5, BN.1, BF.7 and XBB (CDC COVID Data Tracker, 2023). As per the recent report from Center for Disease Control and Prevention (CDC) on December 29, 2022, they quote the following “This week's national ensemble predicts that the number of newly reported COVID-19 deaths will likely increase over the next 4 weeks, with 1,600 to 5,900 new deaths likely reported in the week ending January 21, 2023. The national ensemble predicts that a total of 1,095,000 to 1,109,000 COVID-19 deaths will be reported by this date” (CDC COVID-19, 2023).

Most of the current COVID-19 treatments are based on the symptomatic alleviation of clinical manifestation using approved pharmaceuticals in combination with immunity boosters such as nutraceuticals, supplements, etc. (Fantini et al., 2020; Gautret et al., 2020; Wang et al., 2020; Cao et al., 2020). Throughout the pandemic, most commonly used medications over the world are antivirals (remdesivir, darunavir, oseltamivir, umifenovir, favipiravir, ribavirin, lopinavir), antibacterials (teicoplanin and azithromycin), antimalarial agents (chloroquine/hydroxychloroquine), immunomodulators (bamlanivimab, bevacizumab, sarilumab, thalidomide, and tocilizumab), angiotensin II receptor blockers (losartan), bradykinin b2 receptor antagonists (icatibant), corticosteroids (hydrocortisone, and dexamethasone), anthelmintics (ivermectin, nitazoxanide), and anticoagulants (heparin), etc. (Ojha et al., 2021; Zhang et al., 2020; Tarighi et al., 2021; Yang et al., 2020).

The ingested medicines are eventually released into the environment through bodily discharge substances and end up in larger water bodies either as original drugs or as their metabolites. Another important route of drug exposure to the environment is improper disposal of unused drugs which ultimately end up in the aquatic and soil compartments (Kar and Leszczynski, 2020; Ghosh et al., 2020; Roy and Kar, 2016). Therefore, in the long term, these COVD-19 pharmaceuticals can be a major source of bio-hazards to the environment. Interestingly, according to the multiple reports, these excreted substances are only partially detected or removed from the conventional WWTPs (wastewater treatment plants) (Joss et al., 2005; Nannou et al., 2020). The ecotoxicological concerns related to pharmaceuticals and more importantly, potential contaminants of emerging concerns among them are highlighted in several studies (Al Aukidy et al., 2012; Godoy and Kummrow, 2017; Kar et al., 2020). The current pandemic has caused an enormous release of several anti-virals, antimalarials, antibacterials, immunomodulators, and corticosteroids drugs and their metabolites into aquatic and terrestrial environments, and is expected to pose an alarming risk to the ecosystem, mainly aquatic ones (Jain et al., 2013; Nippes et al., 2021). Hence, unraveling the fate and occurrences of COVID-19 associated pharmaceuticals in a close proximal environment becomes an utmost necessity under this unprecedented pandemic (Essid et al., 2020; Tarazona et al., 2021; Kuroda et al., 2021; Farias et al., 2020).

The European Chemicals Agency (ECHA), and the United States Environmental Protection Agency (US EPA) monitor the release of chemical contaminants across Europe and the US, respectively. Several legislations that are part of ECHA and the “Code of Federal Regulations (40 CFR)” aka “Title 40” of US EPA monitor the release of various groups hazardous substances, including pharmaceutical products in Europe and the US, respectively. Both of these regulatory bodies recommend using alternative testing strategies (ATS), thus avoiding unnecessary animal sacrifice and promoting the use of in silico tools, mainly read-across and quantitative structure-activity relationships (QSAR), for regulatory testing. The development of OECD QSAR Toolbox (The OECD QSAR Toolbox, 2023) by ECHA and ECOSAR software (Mayo-Bean et al., 2012) by US EPA was an attempt to shift conventional (eco)toxicity testing more towards alternative testing strategies (ATS). For EU regulations and US EPA directive details, kindly refer to the resources [(ECHA (European Chemical Agency) 2023; Code of federal regulations, 2023)]. Tarazona et al. (2021) estimated PNECs (Predicted No Effect Concentrations) for several COVID-19 associated pharmaceuticals (dexamethasone, chloroquine, ivermectin, and azithromycin) using QSAR models incorporated in VEGA platform (VEGAHUB Software Platform). Additionally, Desgens-Martin and Keller employed the ChemFate model to predict environmental concentrations of several pharmaceuticals including remdesivir and dexamethasone (Desgens-Martin and Arturo, 2021; Tarighi et al., 2021). The other applications of ATS in ecotoxicity assessment can be traced from (Kuroda et al., 2021; Ali et al., 2021; Kumari and Kumar, 2021).

The present work is an attempt for the aquatic ecotoxicological risk assessment of pharmaceuticals that are currently used (including some newly developed drug molecules) in the treatment of COVID-19 for both mild as well as serious symptoms using in silico models. Additionally, the work aims to promote the use of ATS in the ecotoxicity assessment of pharmaceuticals as major environment contaminants. For the present study, we have compiled a list of 85 pharmaceuticals (77 repurposed approved drug molecules, 1 approved drug for emergency use by US-FDA and 3 molecules under trials, and 4 control molecules to compare the ecotoxicity threshold) from the literature and web platforms (Ojha et al., 2021; Hu et al., 2021; Two COVID-19 antiviral pills advance to late-stage trials 2023; DrugBank online 2023; Tarighi et al., 2021; Pfizer Announces Submission of New Drug Application to the U.S. FDA for PAXLOVID™, 2023) ensuring the inclusion of maximum drug candidates employed directly or indirectly in the treatment of COVID-19 disease (the detailed list of studied pharmaceuticals with their therapeutics class, SMILES, InChI, IUPAC names, MW and MLOGP is given in Table S1 in Supplementary Materials). Out of four molceules under clinical trials, nirmatrelvir [PF-07321332] authorized for emergency use in both high-risk adults and high-risk pediatric patients 12 years of age and older weighing at least 40 kg in combination with ritonavir tablets using brand name of PAXLOVID™ (Pfizer Announces Submission of New Drug Application to the U.S. FDA for PAXLOVID™ 2023). Four control pharmaceuticals (17β-Estradiol, Atenolol, Carbamazepine and Diclofenac) are selected based on worldwide occurrence from river, lake, STP, WWTPs, influent and effluent followed by high aquatic toxicity (Kar et al., 2020). For the evaluation of ecotoxicity, three standard test organisms namely algae (Pseudokirchneriella subcapitata), crustacean (Daphnia magna), and fish (Oncorhynchus mykiss, Pimephales promelas, and Danio rerio) were used as recommended by the OECD guidelines (Kar et al., 2020). The present manuscript is an attempt to identify probable risk and fill up the environmental toxicity data gap for the COVD-19 related pharmaceuticals.

2. Materials and methods

2.1. Studied pharmaceuticals

The aim of the study is to analyze the ecotoxicity potential of different pharmaceuticals employed in the treatment of COVID-19 either directly as a drug or indirectly as a symptom curing agent. We have compiled a list of 81 pharmaceuticals (77 are repruposed small drug molecules and one drug nirmatrelvir [PF-07321332] approved for emergency use by US-FDA and 3 molecules are under clinical trials of Pfizer) from various sources of the literature and the DrugBank database based on the prescription and usage since the inception of COVID-19 (Ojha et al., 2021; Hu et al., 2021; Two COVID-19 antiviral pills advance to Late-Stage Trials, 2023; DrugBank online 2023; Tarighi et al., 2021; Pfizer Announces Submission of New Drug Application to the U.S. FDA for PAXLOVID™, 2023). It is important to mention that we have only considered small drug molecules, no monoclonal antibody or biologicals are considered for the prediction. The studied pharmaceuticals represented almost major therapeutical classes known today, some prominent ones among them are antivirals, antineoplastics, antiplatelets, antimalarials, corticosteroids, etc. (see Tables 2 and 3 for the detailed list of studied pharmaceuticals). For the comparison of ecotoxicity predictions, four popular pharmaceuticals (17β-Estradiol, Atenolol, Carbamazepine, Diclofenac) which have proven record of aquatic toxicity were taken as control pharmaceuticals (Kar et al., 2020).

Table 2.

The predicted ecotoxicity values (pEC50, concentration in mol/L) from selected QSAR models.

No.	Pharmaceuticals	P. subcapitata (pEC50)	AD	D. magna (pEC50)	AD	O. mykiss (pLC50)	AD	P. promelas (pLC50)	AD	Danio rerio (pEC50)	AD
1	Abivertinib	9.79	Out	6.91	In	6.29	In	7.61	In	7.27	In
2	Acalabrutinib	9.52	Out	4.48	In	5.59	Out	7.15	Out	7.63	In
3	Ampion	4.48	In	3.43	In	4.30	In	3.77	In	4.16	In
4	Aprepitant	9.52	In	4.54	In	5.83	In	7.50	In	13.45	In
5	Apx-115	8.31	In	5.10	In	5.24	In	6.23	In	4.94	In
6	Aspirin	5.73	In	3.89	In	4.36	In	5.15	In	3.77	In
7	At-527 (R07496998)	9.81	Out	6.98	In	6.05	Out	6.76	In	8.66	In
8	Azd1656	8.37	In	3.56	In	5.30	In	6.62	In	6.64	In
9	Azithromycin	10.04	Out	5.06	Out	4.11	In	8.06	In	6.38	In
10	Baricitinib	7.24	In	4.65	In	4.59	In	5.24	In	7.20	In
11	Budesonide	7.82	In	5.63	In	4.79	In	5.91	In	6.88	In
12	Canrenoate potassium	6.20	In	4.19	In	4.64	In	4.75	In	4.44	In
13	Cefditoren	10.65	Out	3.65	In	4.94	In	7.26	In	7.18	In
14	Cenicriviroc	12.14	Out	9.41	In	9.22	In	10.40	In	7.51	In
15	Ciclesonide	9.22	Out	7.40	Out	6.12	In	7.20	In	7.56	In
16	Clarithromycin	9.91	Out	5.33	Out	3.78	In	8.05	In	6.10	In
17	Clevudine	5.83	In	3.03	In	4.42	In	4.20	Out	4.58	In
18	Danoprevir	9.99	Out	4.31	In	5.55	In	7.88	In	8.78	In
19	Darunavir	8.81	Out	5.19	In	5.31	In	7.40	In	6.38	In
20	Deferoxamine	8.67	Out	3.73	In	6.41	Out	8.27	In	5.42	Out
21	Defibrotide	8.86	In	8.55	In	6.45	In	6.29	In	6.42	In
22	Dexamethasone	7.38	In	5.74	In	3.88	In	5.49	In	6.24	In
23	Dipyridamole	9.24	Out	4.51	In	7.06	In	7.18	In	7.58	In
24	Disulfiram	8.64	Out	5.53	In	6.05	In	6.63	In	4.18	Out
25	Doxycycline	7.18	In	4.09	In	3.94	In	5.17	In	6.10	In
26	Duvelisib	10.30	Out	5.64	In	5.19	In	6.83	In	6.47	In
27	Edoxaban	9.78	Out	3.05	In	5.29	In	6.67	In	6.80	In
28	Emtricitabine	6.18	In	4.26	In	4.48	In	4.16	In	4.52	In
29	Enoxaparin	6.39	Out	10.43	Out	9.11	Out	5.48	In	9.07	In
30	Enpatoran	7.41	In	5.31	In	4.62	In	5.82	In	7.82	In
31	Etoposide	7.83	Out	6.36	In	4.25	In	6.55	In	7.44	In
32	Favipiravir	4.30	In	3.44	In	4.25	Out	3.62	In	3.86	In
33	GS-441524	5.90	In	4.47	In	4.29	In	4.36	In	6.63	Out
34	Hydrocortisone	7.16	In	5.39	In	4.61	In	5.30	In	6.08	In
35	Hydroxy-chloroquine	9.02	In	5.46	In	6.07	In	6.67	In	4.76	In
36	Ibrutinib	9.85	Out	6.35	In	5.89	In	7.14	In	6.86	In
37	Ibudilast	6.75	In	5.95	In	5.06	In	5.82	In	4.85	In
38	Ifenprodil	8.41	In	6.32	In	6.17	In	6.79	In	5.17	In
39	Imatinib	9.76	Out	6.00	In	6.47	In	7.73	In	6.67	In
40	Imu-838	8.37	In	5.42	In	5.51	In	6.89	In	5.08	In
41	Isoquercetin	6.07	In	5.47	In	4.60	In	5.70	In	4.97	In
42	LAU-7b	9.77	In	7.67	In	7.11	In	8.00	In	4.74	In
43	Ledipasvir	12.92	Out	9.02	In	8.18	In	9.71	In	13.33	In
44	Leflunomide	7.02	In	4.25	In	4.70	In	5.65	In	4.96	In
45	Levamisole	8.11	In	5.18	In	5.23	In	5.43	In	4.74	In
46	Lopinavir	10.55	Out	6.91	In	6.58	In	9.23	In	7.29	In
47	Losmapimod	8.80	In	4.61	In	5.70	In	6.64	In	5.46	In
48	Melatonin	6.90	In	4.39	In	4.82	In	5.37	In	4.74	In
49	Merimepodib	8.34	In	4.77	In	5.23	In	6.89	In	6.09	In
50	Molnupiravir	6.55	In	3.28	In	4.59	In	4.79	In	5.08	In
51	Naltrexone	7.70	In	5.92	In	5.58	In	5.33	In	7.08	In
52	Niclosamide	8.92	In	4.99	In	5.09	In	6.44	In	5.01	In
53	Nitazoxanide	8.61	In	5.10	In	4.64	In	5.66	In	4.48	In
54	Opaganib	9.88	In	5.80	In	4.50	In	6.62	In	6.46	In
55	Oseltamivir	7.16	In	2.81	In	5.10	In	5.77	In	4.48	In
56	Pentoxifylline	7.04	In	3.22	In	4.93	In	4.90	In	4.34	In
57	PF-00835231	7.94	In	3.92	In	4.76	In	6.70	In	6.39	In
58	PF-06650833	7.72	In	4.54	In	4.67	Out	5.92	In	6.36	In
59	PF-07304814	8.34	Out	6.97	In	6.01	In	6.67	In	6.83	In
60	PF-07321332 (Nirmatrelvir)	8.02	In	3.58	In	4.44	In	5.96	In	10.01	In
61	Piclidenoson	7.77	In	7.02	In	5.75	In	6.92	In	4.97	In
62	Povidone-iodine	5.45	In	4.22	In	4.83	In	4.36	In	3.70	In
63	Prazosin	7.90	In	5.43	In	5.28	In	6.10	In	6.34	In
64	Progesterone	8.00	In	6.38	In	5.97	In	5.87	In	5.91	In
65	Ptc299	10.48	Out	7.57	In	5.06	In	8.11	In	6.90	In
66	Rapamycin	12.05	Out	7.50	Out	6.42	In	11.13	In	7.38	In
67	Remdesivir	9.35	Out	7.62	In	6.58	In	7.46	In	8.86	In
68	Ribavirin	5.37	In	2.69	In	4.68	In	3.45	In	4.52	In
69	Ritonavir	11.72	Out	6.80	In	6.33	In	9.83	In	7.83	In
70	Rosuvastatin	8.77	In	4.09	In	4.99	In	6.97	In	5.31	In
71	Ruxolitinib	7.73	In	5.20	In	5.69	In	5.98	In	6.03	In
72	Selinexor	8.75	In	3.70	In	4.91	In	6.42	In	14.86	In
73	Silymarin	7.13	In	6.98	In	4.95	Out	6.83	In	6.03	In
74	Sofosbuvir	8.65	In	6.97	In	5.91	In	6.46	In	6.76	In
75	Tafenoquine	9.32	In	8.12	In	6.42	In	7.68	In	8.14	In
76	Tamoxifen	9.81	In	7.79	In	6.81	In	8.09	Out	5.35	In
77	Tenofovir alafenamide	8.76	In	7.67	In	6.28	In	6.52	In	6.53	In
78	Tetrandrine	10.69	Out	10.77	Out	7.93	In	8.91	In	8.95	In
79	Thalidomide	6.49	In	3.61	In	4.54	In	5.05	In	4.55	In
80	Umifenovir	9.42	In	7.43	In	6.21	In	7.51	In	6.24	In
81	VERU-111	8.08	In	6.61	In	5.19	In	6.75	In	6.51	In
82*	17β-Estradiol	7.46	In	6.67	In	5.77	In	5.83	In	5.61	In
83*	Atenolol	6.88	In	3.90	In	4.73	In	5.51	In	4.36	In
84*	Carbamazepine	7.48	In	5.84	In	4.75	In	5.99	In	5.22	In
85*	Diclofenac	8.39	In	5.82	In	5.28	In	6.72	In	4.93	In

^⁎Control selected based on worldwide occurrence from river, lake, STP, WWTPs, influent and effluent followed by high aquatic toxicity

Table 3.

The predicted ecotoxicity values (pEC₅₀, concentration in mol/L) from ECOSAR and OECD QSAR Toolbox.

ID	Pharmaceuticals	ECOSAR			OECD QSAR Toolbox
		Algae	Daphnia	Fish	Algae	Daphnia	Fish
1	Abivertinib	5.88	5.64	4.79	—	—	—
2	Acalabrutinib	5.17	5.45	5.16	—	—	—
3	Ampion	0.52	-1.44	-1.15	—	—	—
4	Aprepitant	6.30	5.96	5.14	—	—	—
5	Apx-115	5.78	5.40	6.00	—	—	—
6	Aspirin	2.32	2.01	2.36	2.32	1.73	2.16
7	At-527 (r07496998)	4.47	4.91	3.78	—	—	—
8	Azd1656	4.54	3.66	3.64	—	—	—
9	Azithromycin	5.60	5.39	4.53	5.60	5.39	4.45
10	Baricitinib	2.87	1.38	1.54	3.42	4.05	1.74
11	Budesonide	3.90	3.72	3.46	3.89	4.45	2.68
12	Canrenoate potassium	3.72	3.90	3.73	—	—	—
13	Cefditoren	5.71	5.49	4.63	5.71	5.49	4.57
14	Cenicriviroc	6.76	8.28	8.80	—	—	—
15	Ciclesonide	6.04	5.46	5.64	6.04	4.40	5.54
16	Clarithromycin	5.56	5.35	4.49	5.56	5.39	4.41
17	Clevudine	7.14	2.69	2.22	—	—	—
18	Danoprevir	5.18	4.53	4.44	—	—	—
19	Darunavir	4.53	4.95	3.89	—	—	3.26
20	Deferoxamine	1.52	1.89	0.73	1.52	1.89	0.35
21	Defibrotide	5.79	5.42	5.79	—	—	—
22	Dexamethasone	3.17	2.92	2.78	3.27	3.90	3.13
23	Dipyridamole	3.97	3.97	3.75	3.97	4.59	3.65
24	Disulfiram	6.55	5.94	6.54	6.55	6.24	6.54
25	Doxycycline	2.35	2.60	1.50	5.05	—	0.64
26	Duvelisib	5.93	4.99	5.59	—	—	—
27	Edoxaban	3.79	3.84	2.85	—	—	—
28	Emtricitabine	1.50	1.87	0.72	—	—	—
29	Enoxaparin	-12.54	-19.22	-17.54	—	—	—
30	Enpatoran	5.09	4.95	4.05	—	—	—
31	Etoposide	2.56	2.32	2.73	2.56	5.95	0.45
32	Favipiravir	2.63	1.06	1.24	3.19	0.77	1.47
33	Gs-441524	3.32	4.02	1.51	—	—	—
34	Hydrocortisone	3.09	2.82	2.70	3.00	3.62
35	Hydroxy-chloroquine	5.45	5.27	4.40	5.45	5.26	4.39
36	Ibrutinib	5.22	5.49	5.25	—	—	—
37	Ibudilast	6.14	5.27	6.05	5.74	4.94	6.24
38	Ifenprodil	6.06	5.79	4.96	—	—	—
39	Imatinib	5.44	5.25	4.38	5.44	5.25	4.30
40	Imu-838	5.54	4.54	5.01	—	—	—
41	Isoquercetin	1.03	0.55	0.77	—	—	—
42	LAU-7b	8.97	7.20	8.01	7.36	7.37	7.27
43	Ledipasvir	7.67	7.28	9.34	—	—	—
44	Leflunomide	4.74	3.93	3.89	5.29	4.65	3.88
45	Levamisole	5.34	5.17	4.28	5.34	5.17	4.29
46	Lopinavir	6.65	6.29	5.50	—	—	—
47	Losmapimod	5.95	5.59	5.42	—	—	—
48	Melatonin	4.25	3.27	3.28	4.81	3.16	3.33
49	Merimepodib	6.14	4.71	4.16	—	—	—
50	Molnupiravir	2.59	2.81	1.74	—	—	—
51	Naltrexone	4.29	4.27	3.32	3.78	—	—
52	Niclosamide	6.64	5.49	5.62	6.62	5.71	5.40
53	Nitazoxanide	3.83	3.47	3.79	4.94	3.30	3.48
54	Opaganib	6.46	6.28	6.05	—	—	—
55	Oseltamivir	3.98	4.01	3.03	3.98	4.00	2.83
56	Pentoxifylline	4.65	3.32	2.85	—	—	—
57	PF-00835231	3.92	2.81	2.86	—	—	—
58	PF-06650833	4.06	3.01	3.04	—	—	—
59	PF-07304814	3.61	2.39	2.47	—	—	—
60	PF-07321332 (Nirmatrelvir)	3.91	2.80	2.85	—	—	—
61	Piclidenoson	5.89	6.13	6.40	4.58	4.97	6.40
62	Povidone-iodine	4.03	2.97	3.00	—	—	—
63	Prazosin	4.33	4.80	3.50	4.58	2.82	3.07
64	Progesterone	4.75	4.67	4.27	4.75	4.35	4.01
65	Ptc299	7.52	7.07	9.00	—	—	—
66	Rapamycin	5.89	5.33	5.51	5.89	5.33	5.41
67	Remdesivir	4.48	4.92	3.80	—	—	—
68	Ribavirin	1.96	0.12	0.41	2.52	2.83	0.71
69	Ritonavir	7.50	8.25	7.07	7.68	8.10	6.61
70	Rosuvastatin	5.06	3.47	3.86	5.33	3.92	3.92
71	Ruxolitinib	5.35	4.24	4.36	4.50	4.59	4.73
72	Selinexor	5.60	5.21	5.68	—	—	—
73	Silymarin	5.02	4.47	4.31	—	—	3.31
74	Sofosbuvir	3.06	2.77	3.15	—	—	—
75	Tafenoquine	7.79	7.27	6.57	—	—	—
76	Tamoxifen	7.76	7.25	6.55	7.76	7.25	6.54
77	Tenofovir alafenamide	4.16	4.67	3.16	—	—	—
78	Tetrandrine	8.96	8.28	7.65	—	—	—
79	Thalidomide	4.52	3.05	3.60	—	—	—
80	Umifenovir	7.08	6.72	5.95	7.12	6.70	5.77
81	Veru-111	5.57	4.52	4.81	—	—	—
82*	17β-Estradiol	3.82	4.02	3.86	2.58	5.38	5.74
83*	Atenolol	6.22	5.17	5.19	3.29	3.41	2.13
84*	Carbamazepine	3.24	3.36	2.33	5.96	4.20	4.17
85*	Diclofenac	6.06	4.32	3.86	5.03	5.16	4.88

^⁎Control selected based on worldwide occurrence from river, lake, STP, WWTPs, influent and effluent followed by high aquatic toxicity

2.2. Retrieval of structures of pharmaceuticals, their optimization and descriptors calculation

The structures of pharmaceuticals were retrieved from multiple sources including PubChem and Chemical Book search engine. Additionally, few structures were downloaded from DrugBank databases. Lastly, the drugs in the clinical trial were manually drawn using Marvin Sketch software (Version 5.1) (Chemaxon Software, 2023). The molecular structures were compared with all of these sources manually to confirm their correctness and then saved in the SDF format. These processed structures were then utilized for calculating molecular descriptors from the OCHEM platform (Sushko et al., 2011). In addition to SDF files, additional copies of structures in SMILES notations were also prepared, as required by ECOSAR and OECD QSAR Toolbox.

2.3. studied test species for aquatic ecotoxicity assessment

For the ecotoxicity analysis, three standard species algae, daphnia, and fish, which are widely employed in aquatic ecotoxicity estimation and are highly recommended by different regulatory bodies such as OECD, US EPA, etc., were used. The details of the studied species are briefed here.

2.3.1. Algae (pseudokirchneriella subcapitata)

The OECD recommends freshwater algal species (mainly P. subcapitata) along with cyanobacteria (V. fischeri) as the standard test species for assessing the ecotoxicity of environmental pollutants. The OECD Test No. 201 details the specifications required for conducting an ecotoxicity test on algal species. The experimental conditions specified in the test include assessment of algal growth under exposure of the test substance over up to 72 h. The details of the experiments can be found elsewhere (Test No. 201: Freshwater Alga and Cyanobacteria, Growth Inhibition Test, 2023). The implementation of algal species in the ecotoxicity analysis can be traced from several literatures (Gramatica et al., 2012; Roy et al., 2015; Gramatica et al., 2016).

2.3.2. Crustacean (Daphnia magna)

Daphnia magna along with other daphnids represents another group of standard species extensively employed in aquatic ecotoxicity studies of various chemical contaminants. The OECD Test No. 202 details the specifications required for conducting an ecotoxicity test on daphnia species. The test measures immobilization caused in daphnia (young aged <24 hrs) following the administration of test substances at different concentrations (at least five) for 48 h. The complete details of the experiments can be obtained elsewhere (Test No. 202: Daphnia sp. Acute Immobilisation Test, 2023). The implementation of daphnid species in the ecotoxicity analysis can be traced from numerous literatures (Kar and Roy, 2010; Perales et al., 2017; Toropova et al., 2012).

2.3.3. Fish (Oncorhynchus mykiss, Pimephales promelas, and Danio rerio)

Fish is perhaps the most abundantly employed standard test species for ecotoxicity estimation among toxicologists and environmentalists. The OECD Test No. 203 enlists the detailed experimental procedure that is to be followed while carrying out ecotoxicity analysis for the test substances. For this experiment, fish mortalities (50% of the population) are reported after exposure to test chemicals for up to 96 h. The test allows the mixing of one or more species of fish during the course of the experiment. Specific details of the experiments can be obtained elsewhere (Khan et al., 2019). The implementation of fish species in the ecotoxicity analysis can be traced from multiple literatures (Test, 203AD; Khan et al., 2019, 2019).

2.4. QSAR models used for predictions

The QSAR models employed in the ecotoxicity estimation should offer sufficient robustness in terms of the calculated metrics. Additionally, models with a large applicability domain (AD) with minimum prediction errors are considered superior in terms of reliability of predictions. Moreover, QSAR models developed using simple and more interpretable descriptors are accepted and recommended by various data scientists including toxicologists and environmentalists. Lastly, we looked for a model comprising LogP/allied attributes, since the ecotoxicity of organic chemicals is largely due to the lipophilic nature that renders them a potential candidate for PBT (persistence bioaccumulation and toxicity) like substances.

The quest of the model search against various studied species was accomplished using the Scopus search engine. We have downloaded a list of 212, 328, and 1764 publications against algae, Daphnia, and fish endpoints, respectively with the following keywords: ‘Ecotoxicity’ and ‘QSAR’ or ‘Model’ or ‘Validation’. The complete list of the studied papers is reported in Tables S3–S5 in Supplementary Materials. We have selected the best models among the huge list of the literature based on following criteria which are arranged based on priority: (a) Size of the dataset (importance to the global model compared to local ones), (b) validation through training and test sets (to have the essence of prediction capability of the models for the new compounds), (c) acceptable validation metrics values (all models should pass the assigned threshold of internal and external validation metrics), (d) models with 2D descriptors for easy interpretability and reproducibility (if we consider model with 3D and quantum descriptors, the reproducibility of same values can be an issue depending on the nature of algorithm used to prepare the molecules, e) we did not also consider those machine learning models which do not provide any mathematical equation like transparent models The selection of the models is driven with the motivation to achieve reliable, reproducible and interpretative predictions. The models chosen for deriving ecotoxicity are as follows:

The model developed by Khan and Roy (2019) was adopted to derive ecotoxicity predictions of studied pharmaceuticals against algal species i.e. P. subcapitata. The model was developed using a sufficiently large dataset of 334 organic chemicals including pharmaceuticals employing simple and more interpretable 2D descriptors and was extensively validated using various internal and external validation metrics. For calculating ecotoxicity against D. magna, model developed by Khan et al. (2019) was used. The model was derived using a training set of 318 compounds (large AD) and validated using 105 test set compounds. Not only that, the model was also derived using highly interpretable 2D descriptors calculated from Dragon and PaDEL-descriptor calculating software. The models on O. mykiss and Zebrafish (Khan et al., 2019) and the model on P. promelas reported by Khan et al. (2019) were employed for the prediction of ecotoxicity against three fish species. The model equations were converted into a uniform molar unit (mol/L) irrespective of the unit they were reported in the original literature. The detailed specification of different employed models is reported in Table 1 . For model predictions, the DTC Lab Software tools available from http://teqip.jdvu.ac.in/QSAR_Tools/ were used.

Table 1.

The detailed description of models employed in the current manuscript.

Species	Equation	Internal validation for Training set					External validation for Test set				Ref
		NTraining	R2	Q2	MAE95%	MAE100%	NTest	Q2F1	MAE95%	MAE100%
P. subcapitata (Algae)	$\text{pE}{\mathrm{C}}_{50}\left(\text{mol}/\mathrm{L}\right)=3.679+0.029\times \text{CrippenMR}+0.397\times \text{MLOGP}+0.603\times \mathrm{B}06\left[\mathrm{C}-\mathrm{N}\right]+0.786\times \mathrm{B}05\left[\mathrm{C}-\text{Cl}\right]+1.165\times \mathrm{B}02\left[\mathrm{N}-\mathrm{S}\right]-1.467\times \text{nSO}2\text{OH}$	251	0.72	0.70	0.62	0.69	83	0.70	0.60	0.67	54
D. magna (Crustacean)	$\text{pE}{\mathrm{C}}_{50}\left(\text{mol}/\mathrm{L}\right)=3.759+3.021\times B01\left[O-P\right]+0.071\times ALOP2-0.885\times B08\left[O-O\right]+0.279\times nCt-0.107\times minHBint2+0.135\times AMW-0.202\times nHM+0.452\times C-026-0.623\times nArX-14.429\times ETA\_dAlpha\_B$	318	0.65	0.64	0.70	0.84	105	0.65	0.83	0.87	55
Oncorhynchus mykiss (Fish)	$pL{C}_{50}\left(mol/L\right)=5.019-4.973*\text{GD}+0.187*\text{XLogP}+0.041*\text{ALOGP}2+-0.202*\text{SssCH}2+1.41242*\mathrm{B}05\left[\mathrm{C}-\mathrm{P}\right]+0.640*\text{nCrt}-1.8713*\mathrm{B}09\left[\mathrm{O}-\text{Cl}\right]$	83	0.71	0.65	0.63	0.66	36	0.70	0.66	0.74	53
Pimephales promelas (Fish)	$pL{C}_{50}\left(mol/L\right)=-2.616+0.237*\mathrm{S}2\mathrm{K}+2.208*\mathrm{B}02\left[\mathrm{C}-\mathrm{X}\right]+0.053*\mathrm{C}\%+0.265*\text{ALOGP}$	80	0.76	0.74	0.58	0.64	28	0.80	0.47	0.54	56
Danio rerio (Fish)	$pE{C}_{50}\left(mol/L\right)=2.292+0.377*\mathrm{X}5\text{sol}-0.168*\mathrm{F}04\left[\mathrm{O}-\mathrm{O}\right]+1.013*\text{nTB}+0.259*\mathrm{F}07\left[\mathrm{N}-\mathrm{O}\right]-0.354*\mathrm{B}07\left[\mathrm{C}-\mathrm{N}\right]+0.023*\mathrm{T}\left(\mathrm{N}..\mathrm{F}\right)$	35	0.75	0.66	0.25	0.29	15	0.72	0.29	0.35	53

2.5. Ecotoxicity predictions using online expert systems

In addition to the QSAR model-derived predictions, ecotoxicity of pharmaceuticals was also assessed using ECOSAR (version 2.0) (Mayo-Bean et al., 2017) and OECD QSAR Toolbox (version 4.5) (The OECD QSAR Toolbox, 2023), two widely employed online expert systems for the ecotoxicity predictions.

2.5.1. OECD QSAR toolbox based predictions

The molecules in SMILES forms were incorporated into the OECD QSAR Toolbox (version 4.5). For profiling, all available methods including Verhaar, MOA by OASIS, and ECOSAR were given as input. In the next step, experimental results were obtained from resident databases available in QSAR Toolbox 4.5 (The OECD QSAR Toolbox, 2023). The selection of ecotoxicity response from the QSAR toolbox followed a similar protocol as specified for ECOSAR, i.e. taking the lowest available value (worst-case scenario). For compounds having no reported ecotoxicity data in the resident database, a data gap filling tool available in the toolbox was used. For data gap filling, all three available methods i.e. read-across, trend analysis, and QSAR models were used.

2.5.2. ECOSAR based predictions

The individual molecules in SMILES forms were incorporated in the organic module of the ECOSAR software (version 2.0) (Mayo-Bean et al., 2017). The ecotoxicity predictions in ECOSAR are mainly derived from the neutral organics model and/or class-specific local model based on the presence of different functional groups such as amides, anilines, aldehydes, etc. Irrespective of the type of model, the lowest predicted concentration against every studied compound was chosen for comparison purposes to consider the least-case scenario.

Both expert systems predictions were later compared with the model-derived predictions to arrive at a certain conclusion on the ecotoxicity potential and risk assessment of the studied pharmaceuticals. The complete work flow is reported in Fig. 1 .

Fig. 1.

Adapted workflow of the present study.

3. Results

3.1. Predictions using QSAR models

The respective model descriptors were imported into an excel file to predict ecotoxicity using the respective QSAR equations mentioned in Table 1. The AD of the individual pharmaceutical against the respective model was assessed using the standardization approach (Roy et al., 2015). In the standardization approach, the chemical falling outside the range of mean ± 3SD are considered outside the domain of the model. The predicted ecotoxicity by QSAR models against various studied species along with AD information is listed in Table 2 .

3.2. Predictions using online experts

Additionally, predictions were derived using ECOSAR and OECD QSAR Toolbox, two major expert systems. However, the QSAR Toolbox failed to provide ecotoxicity against several studied pharmaceuticals, possibly due to the diverse chemical nature of the studied compounds (a drawback not present in the used QSAR models in the earlier section). For calculating ecotoxicity using the ECOSAR tool and QSAR Toolbox, SMILES notations of individual structures were imported into the software. The ECOSAR tool provides ecotoxicity predictions using LogK_ow calculated from Biobyte's CLogP program. The predicted ecotoxicity values from ECOSAR and OECD QSAR Toolbox against various studied species is listed in Table 3 .

3.3. Consensus ecotoxicity predictions

Once the ecotoxicity predictions of all 85 compounds was completed using QSAR models, ECOSAR and OECD QSAR Toolbox, the next challenge was to convert these values into a uniform scale to give equal weightage to each of these test species for comparison. This objective was accomplished using the scaling technique (see the formula in Eq. (1)).

$$Y_{scaled}=\left(Y_{predicted}-Y_{pred\_min}\right)/\left(Y_{pred\_max}-Y_{pred\_min}\right)$$ (1)

The scaling of the predicted response was achieved as per following steps:

Each endpoint was sorted based on the AD characteristic followed by the predicted response. For compounds within the AD, maximum and minimum predicted responses were identified. The scaled value (Y_scaled) of each compound was calculated employing Eq. (1). For compounds falling outside the domain of applicability, the scaled value was kept blank and not included in consensus calculations.

• (1)Step 1 was repeated for all 85 studied pharmaceuticals and species groups i.e. algae, Daphnia, O. mykiss, P. promelas, zebrafish.
• (2)Scaled average toxicity and scaled standard deviation are calculated for all 85 compounds.
• (3)Scaled average toxicity is sorted to find out the predicted top 10 most toxic compounds to the environment.
• (4)The same procedure (steps 1 to 4) is repeated with ECOSAR derived predictions, separately.

The scaling system helps us to make all the prediction value within the same scale of 0 to 1 where 0 defines no toxicity and 1 suggests highes toxicity. For the comparison of ecotoxicity predictions, we have only considered QSAR models and ECOSAR tool based prediction as QSAR Toolbox failed to predict around 53% of studied compounds toxicity. The scaled ecotoxicity values along with the consensus score computed from QSAR models, and the ECOSAR tool are provided in Tables S6 and S7 in Supplementary Materials.

4. Discussion

To compare the toxicity scale, we have predicted toxicity of four control pharmaceuticals along with studied 81 pharmaceuticals used for COVID-19. Analysing Tables S6 and S7 (Supplementary Materials), four control pharmaeuticals reported scaled predicted average toxicity value ranges from 0.38-0.65 [17β-Estradiol (0.4), Atenolol (0.23), Carbamazepine (0.34), and Diclofenac (0.4)] and 0.23-0.4 [17β-Estradiol (0.48), Atenolol (0.60), Carbamazepine (0.38), and Diclofenac (0.58)), respectively as per the ECOSAR and QSAR models. The standard deviation of the predicted toxicity values ranges from 0.04-0.1 and 0.18-0.23, respectively for the ECOSAR and QSAR model based predictions. In case of ECOSAR based predictions, pharmaceuticals having scaled average toxicity value more than 0.6 and standard deviation value less than 0.1 are considered for the further analysis. Again, for the QSAR model based predictions, pharmaceuticals having scaled average toxicity values more than 0.4 and standard deviation value less than 0.23 are considered. Considering the identified toxicity values and standard deviation range from control pharmaceuticals, we have identified top 25 pharmaceuticals for both types of predictions. Based on descending order of average scaled predicted toxicity values, top 25 pharmaceuticals are reported in Fig. 2, Fig. 3 , respectively for ECOSAR and QSAR model based predictions.

Fig. 2.

Top 25 toxic Pharmaceuticals based on ECOSAR for aquatic environment.

Fig. 3.

Top 25 toxic Pharmaceuticals based on QSAR models for aquatic environment.

We have employed the standardization approach (Roy et al., 2015) for the evaluation of applicability domain (AD) which measures the compound's structural dissimilarities in the hypothetical space. Regarding structural coverage and prediction reliability, models showed following statistics: Pseudokirchneriella subcapitata (Algae) (70%), Daphnia magna (Crustacean) (93%), Oncorhynchus mykiss (Fish) (99%), Pimephales promelas (96%) and Danio rerio (Zebra Fish) (80%) [where the number suggests reliable and acceptable predictions of molecules out of 85 studied pharmaceuticals in term of %].

These mentioned top 25 pharmaceuticals showed alarming scaled predicted toxicity as all their values are higher than the proven aquatic toxic pharmaceuticals which are used as control. Therefore, regulatory agencies throughout the world must focus on these pharmaceuticals for the environmental risk assessment followed by preparing protocols for long term effects. For the confident and relaible prediction data, we have checked multiple models and tools as well as applicability domain study is also performed.

To narrow down the comparison, further we have taken top 10 pharmaceuticals based on the scaled average aquatic toxicity employing the ECOSAR and studied QSAR models as illustrated in Table 4 . In the top 10 molecules for both types of predictions, 3 pharmaceuticals are common between ECOSAR and QSAR models’ consensus predictions. Interestingly the order of toxicity (Ledipasvir > Tafenoquine > Lopenavir) is same for both type of predictions. Unfortunately, we could not get any prediction for these three pharmaceuticals employing the QSAR Toolbox.

Table 4.

Comparison of top 10 pharmaceuticals based on higher scaled average aquatic toxicity and lower scaled standard deviation of toxicity employing ECOSAR and studied QSAR models.^*

ID	Pharmaceuticals	Scaled toxicity prediction using ECOSAR		ID	Pharmaceuticals	Scaled toxicity prediction using QSAR models
		Average	Standard			Average	Standard deviation
42	LAU-7b	0.92	0.07	43	Ledipasvir	0.86	0.06
43	Ledipasvir	0.91	0.08	78	Tetrandrine	0.65	0.16
65	Ptc299	0.89	0.07	75	Tafenoquine	0.63	0.22
75	Tafenoquine	0.83	0.08	69	Ritonavir	0.57	0.2
76	Tamoxifen	0.83	0.08	67	Remdesivir	0.56	0.12
80	Umifenovir	0.76	0.08	46	Lopinavir	0.55	0.18
24	Disulfiram	0.74	0.02	77	Tenofovir alafenamide	0.53	0.23
54	Opaganib	0.73	0.06	7	At-527 (r07496998)	0.5	0.12
46	Lopinavir	0.72	0.08	1	Abivertinib	0.49	0.13
61	Piclidenoson	0.71	0.07	74	Sofosbuvir	0.49	0.21

^⁎Bold molecules are common from the prediction results of ECOSAR and QSAR models.

The comparison among three pharmaceuticals based on average scaled predicted toxicity is demonstrated in Fig. 4 . The entire statistical analysis suggested that Ledipasvir is extremely toxic to the aquatic environment followed by Tafenoquine and Lopinavir based on three major aquatic species algae, crustaceans and fish. Therefore, these three molcules along with other pharmaceuticals reported under Table 4 needs to be assessed carefully for the environemntal risk.

Fig. 4.

Top 3 toxic pharmaceuticals based on the scaled average predicted toxicity by QSAR models and ECOSAR.

Ledipasvir is a common antiviral drug used for the treatment of hepatitis C and it acts as a NS5A potein inhibitor. A single 90 mg oral dose of [14C]-ledipasvir reported around 87% of total recovery of ledipasvir in feces and urine where feces accounted for approximately 86%. Only 2.2% of the dose went through oxidation to generate oxidative metabolite M19 and 70% of the administered dose excreted in unchanged form of ledipasvir through feces. Another important aspect is the median terminal half-life of ledipasvir is extremely high i.e. 47 h (FDA Approved Drug Products: Harvoni (Ledipasvir and Sofosbuvir) tablets or pellets for oral use, 2023). Therefore, once ledipasvir is excreted from the patient in the unchanged form, it can exist in the environment for good amount of time to exert its toxic effects to the ecosystem.

Tafenoquine is an antiparasitic agent used for the treatment of malaria. Once the drug is taken by the patients, degradation occurs through various metabolic pathways where tafenoquine is slowly excreted from the body through feces and renal route (Rajapakse et al., 2015). Tafenoquine is reported to have a long half-life of around 14 days and can be reason of hemolysis for the patients who have glucose-6-phosphate dehydrogenase deficiency. Although preclinical studies showed that Tafenoquine is not carcinogenic and lacks mutagenic potential, fertility studies resulted reduced number of viable fetuses followed by implantation sites and corpora lutea (Rajapakse et al., 2015; Ebstie et al., 2016).

Lopinavir is human immunodeficiency virus (HIV)-1 protease inhibitor used for the treatment of HIV infection, commonly with fixed-dose combination with Ritonavir under the brand name of Kaletra (FDA Approved Drug Products: Kaletra (lopinavir/ritonavir) for oral use, 2023). The primary excretion pathway is feces where unaltered parent drug reported for 2.2% and 19.8% of the administered dose in urine and feces, respectively (Health Canada Product Monograph: Kaletra (lopinavir/ritonavir) for Oral Use, 2023). As the oral Kaletra solution is highly concentrated, there is a high risk of overdose, and the risk increases manifold in the pediatric patients as Kaletra contains approximately 42% (v/v) ethanol (FDA Approved Drug Products: Kaletra (Lopinavir/Ritonavir) for Oral Use, 2023). Lopinavir is >98% protein-bound in plasma, especially to both albumin and alpha-1-acid glycoprotein and [(FDA Approved Drug Products: Kaletra (lopinavir/ritonavir) for oral use 2023, Health Canada Product Monograph: Kaletra (lopinavir/Ritonavir) for Oral Use, 2023)]. Therefore, overdose of Lopinavir is deadly as it is highly protein-bound and no antidote available ill today. Considering these facts, Lopinavir toxicity to the environment, especially the aquatic ecosystem, can be deadly in the long run.

All three studied trial small molecules ((PF-00835231, PF-06650833, PF-07304814) as well as nirmatrelvir [PF-07321332] developed by Pfizer, are relatively safe in terms of aquatic toxicity as not a single one fall under the top 25 toxic compounds. PF-00835231 inhibits SARS-CoV-2 in A549⁺ACE2 cells, and it is one of the first anti 3CL^pro regimen currently tested clinically by Pfizer (de Vries et al., 2021). The other two small molecules Pfizer PF-07304814 (a prodrug that targets the 3CL^pro protease) (Pfizer Takes Antiviral Drug PF-07304814 into Phase 1 Clinical Trial Targeting COVID-19, 2023) and PF-06650833 inhibited human primary cell inflammatory responses to physiologically relevant stimuli generated with rheumatoid arthritis (RA) and Systemic lupus erythematosus (SLE) patient plasma (Winkler et al., 2021). The last molecule nirmatrelvir [PF-07321332] (PF-332) (a potent inhibitor of the viral 3CL^pro protease) is authorized by USFDA for the emergency use which exerts equipotent in vitro activity against the four SARS-CoV-2 variants of concerns (VoC) and it can fully arrest replication of the alpha variant in primary human airway epithelial cells (Abdelnabi et al., 2022). As per ECOSAR, PF-00835231, PF-06650833, PF-07304814 and nirmatrelvir [PF-07321332] showed scaled average toxicity values of 0.41, 0.43, 0.37, and 0.41 respectively which are much below than the control pharmaceutical's toxicity values. All four molecules ranked within the range of 65 to 74 in terms of toxicity out of 81 studied molecules. A similar trend is also observed based on the QSAR models based predictions for these trial molecules where nirmatrelvir [PF-07321332], PF-07304814, PF-06650833 and PF-00835231 reported scaled average toxicity values of 0.36, 0.44, 0.36 and 0.44, respectively. This is an exciting outcome considering that nirmatrelvir [PF-07321332] which is already approved by USFDA for emergency use and other threetrial molecules can emerge as possible medication of COVID-19 in the upcoming days and they are expected to have relatively lower aquatic toxicity.

5. Conclusion

The major aim of the study has been to fill up the aquatic toxicity data gaps against three major aquatic species algae, crustacean, and fish for the pharmaceuticals which are used uncontrollably throughout the world for the fight against COVID-19. The scaled predicted toxicity values considering the consensus of all studied species offer a more clear picture about the possible toxicity values as well as risk of the pharmaceuticals. Prediction comparison between results from developed single QSAR models and the results from the tools like ECOSAR and QSAR Toolbox helped us to decide the most toxic set of COVID-19 pharmaceuticals for aquatic environment. Based on the scaled average predicted toxicity, ECOSAR and studied QSAR models identified top three common pharmaceuticals (Ledipasvir, Tafenoquine, and Lopenavir) which appeared to be toxic for the aquatic environment for the studied species. Considering the high average predicted toxicity values and low standard deviation values, environmental regulatory agencies should prepare risk assessment and management protocols for long term effects of all three drugs. All four studied trial drugs (PF-00835231, PF-06650833, PF-07304814, PF-07321332) showed lower to moderate scaled predicted toxicity, and not a single scaled average toxicity value crossed 0.5 value considering scale of 0 as nontoxic and 1 as toxic. The obtained results can be considered as the first level of aquatic environmental risk assessment which can help to decide the next level of risk study of the concerning pharmaceuticals followed by preparation of risk management protocol.

Declarations

Ethics approval

This is an original article that did not use other information, which requires ethical approval.

Consent to participate

All authors participated in this article.

Consent for publication

All authors have given consent to the publication of this article.

Author contribution

SK and KR contributed to the study conception and design. KK and SK performed material preparation, data collection, and analysis. SK and KR review and edited the final draft. The manuscript was written, edited, and approved by all authors.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

• Supporting data are available in the section of the Supplementary materials. Additional data supporting this study's findings are available on request from the corresponding author.