Accurate prediction of human drug toxicity: a major challenge in drug development

doi:10.1016/j.cbi.2004.09.008

Chemico-Biological Interactions

Volume 150, Issue 1, 1 November 2004, Pages 3-7

https://doi.org/10.1016/j.cbi.2004.09.008 Get rights and content

Abstract

Over the past decades, a number of drugs have been withdrawn or have required special labeling due to adverse effects observed post-marketing. Species differences in drug toxicity in preclinical safety tests and the lack of sensitive biomarkers and nonrepresentative patient population in clinical trials are probable reasons for the failures in predicting human drug toxicity. It is proposed that toxicology should evolve from an empirical practice to an investigative discipline. Accurate prediction of human drug toxicity requires resources and time to be spent in clearly defining key toxic pathways and corresponding risk factors, which hopefully, will be compensated by the benefits of a lower percentage of clinical failure due to toxicity and a decreased frequency of market withdrawal due to unacceptable adverse drug effects.

Section snippets

Drugs with adverse effects routinely escape preclinical and clinical trials

One of the major challenges in drug development is the accurate assessment of human drug toxicity. A review by Lasser et al. [1] found that in the last four decades, 2.9% of the marketed drugs were withdrawn from the market due to severe adverse drug effects, with seven drugs approved and marketed since 1993 and subsequently withdrawn to be associated with over 1000 deaths. Lasser et al. [1] also estimated that from 1975 to 1999, 10.2% of approved drugs required black box warnings to be added

Reasons for failures in the prediction of human drug toxicity

A clear understanding of reasons for the failure to predict human drug toxicity is the first step to the development of a better approach. A review of the standard approach to drug development yields clues on why human drug toxicity is not always accurately predicted (Table 3). Drug safety is initially evaluated preclinically in laboratory animals, generally in three animal species such as mouse, rat and dog. Preclinical safety study results are submitted to the U.S. Food and Drug

A need for toxicology to be an investigative discipline

Toxicity is a complex biological property that cannot be accurately estimated purely based on dose–response relationships without a sound scientific understanding of the effects. Classical toxicologists rely on Paracelsus’ Principle [7]:

“All things are poison and nothing is without poison. Solely the dose determines that a thing is not a poison.”

This principle of toxicology derived in the 15th century is the cornerstone of today's traditional practice of toxicology. Dose–response relationship

Conclusion

Accurate prediction of human drug toxicity is a major challenge in drug development that needs to be overcome. Failure to predict human drug toxicity is indicated by the occurrence of postmarketing withdrawal and new safety warning labels of marketed drugs. It is necessary to re-evaluate our past practices to determine on areas of weakness. It is proposed here that toxicology should include mechanistic evaluation in additional to the commonly practiced of the determination of a safe dose based

References (7)

J.A. Dimasi et al.
The price of innovation: new estimates of drug development costs
J. Health Econ.
(2003)
J. Easterbrook et al.
A comparison of aroclor 1254-induced and uninduced rat liver microsomes to human liver microsomes in phenytoin O-deethylation, coumarin 7-hydroxylation, tolbutamide 4-hydroxylation, S-mephenytoin 4′-hydroxylation, chloroxazone 6-hydroxylation and testosterone 6-beta-hydroxylation
Chem. Biol. Interact.
(2001)
A.P. Li
A review of the common properties of drugs with idiosyncratic hepatotoxicity and the “multiple determinant hypothesis” for the manifestation of idiosyncratic drug toxicity
Chem. Biol. Interact.
(2002)

There are more references available in the full text version of this article.

Cited by (69)

Toxicity prediction and classification of Gunqile-7 with small sample based on transfer learning method
2024, Computers in Biology and Medicine
Drug-induced diseases are the most important component of iatrogenic disease. It is the duty of doctors to provide a reasonable and safe dose of medication. Gunqile-7 is a Mongolian medicine with analgesic and anti-inflammatory effects. As a foreign substance in the body, even with reasonable medication, it may produce varying degrees of adverse reactions or toxic side effects. Since the cost of collecting Gunqile-7 for pharmacological animal trials is high and the data sample is small, this paper employs transfer learning and data augmentation methods to study the toxicity of Gunqile-7. More specifically, to reduce the necessary number of training samples, the data augmentation approach is employed to extend the data set. Then, the transfer learning method and one-dimensional convolutional neural network are utilized to train the network. In addition, we use the support vector machine-recursive feature elimination method for feature selection to reduce features that have adverse effects on model predictions. Furthermore, due to the important role of the pre-trained model of transfer learning, we select a quantitative toxicity prediction model as the pre-trained model, which is consistent with the purpose of this paper. Lastly, the experimental results demonstrate the efficiency of the proposed method. Our method can improve accuracy by up to 9 percentage points compared to the method without transfer learning on a small sample set.
Molecular modelling and de novo fragment-based design of potential inhibitors of beta-tubulin gene of Necator americanus from natural products
2021, Informatics in Medicine Unlocked
The emergence of drug resistance against the known hookworm drugs namely albendazole and mebendazole and their reduced efficacies necessitate the need for new drugs. Chemically diverse natural products present plausible templates to augment hookworm drug discovery. The present work utilized pharmacoinformatics techniques to predict African natural compounds ZINC95486082, ZINC95486052 and euphohelionon as potential inhibitory molecules of the hookworm Necator americanus β tubulin gene. A library of 3390 compounds was screened against a homology-modelled structure of β tubulin. The docking results obtained from AutoDock Vina was validated with an acceptable area under the curve (AUC) of 0.714 computed from the receiver operating characteristic (ROC) curve. The three selected compounds had favourable binding affinities and were predicted to form no interactions with the resistance-associated mutations Phe167, Glu198 and Phe200. The compounds were predicted as anthelmintics using a Bayesian-based technique and were pharmacologically profiled to be druglike. Further molecular dynamics simulations and MM-PBSA calculations showed the compounds as promising anthelmintic drug leads. Novel critical residues comprising Leu246, Asn247 and Asn256 were also predicted for binding. Euphohelionon was selected as a template for the de novo fragment-based design of five compounds labelled A1, A2, A3, A4 and A5; with four of them having SAscore values below 6, denoting easy synthesis. All the five de novo molecules docked firmly in the binding pocket of the β tubulin with no binding interactions with the three known resistance mutation residues. Binding energies of −8.2, −7.6, −7.3, −7.2 and −6.8 kcal/mol were obtained for A1, A2, A3, A4 and A5, respectively. The identified compounds can serve as treasure troves from which future potent anthelmintics can be designed. The current study strives to assuage the hookworm disease burden, especially making available molecules with the potential to circumvent the chemoresistance.
Cardiovascular disease models: A game changing paradigm in drug discovery and screening
2019, Biomaterials
Citation Excerpt :
Even though a step closer toward representing the human population than preclinical studies, clinical trials are also challenging. Clinical trials have relatively small sample sizes, narrow subgroups of patient demographics, and require non-invasive endpoints for study monitoring [8]. As a result, rare events may not be detected until after clinical trials, when the drug is already in the clinic.
Cardiovascular disease is the leading cause of death worldwide. Although investment in drug discovery and development has been sky-rocketing, the number of approved drugs has been declining. Cardiovascular toxicity due to therapeutic drug use claims the highest incidence and severity of adverse drug reactions in late-stage clinical development. Therefore, to address this issue, new, additional, replacement and combinatorial approaches are needed to fill the gap in effective drug discovery and screening. The motivation for developing accurate, predictive models is twofold: first, to study and discover new treatments for cardiac pathologies which are leading in worldwide morbidity and mortality rates; and second, to screen for adverse drug reactions on the heart, a primary risk in drug development. In addition to in vivo animal models, in vitro and in silico models have been recently proposed to mimic the physiological conditions of heart and vasculature. Here, we describe current in vitro, in vivo, and in silico platforms for modelling healthy and pathological cardiac tissues and their advantages and disadvantages for drug screening and discovery applications. We review the pathophysiology and the underlying pathways of different cardiac diseases, as well as the new tools being developed to facilitate their study. We finally suggest a roadmap for employing these non-animal platforms in assessing drug cardiotoxicity and safety.
Metabolism-dependent cytotoxicity of citrinin and ochratoxin A alone and in combination as assessed adopting integrated discrete multiple organ co-culture (IdMOC)
2018, Toxicology in Vitro
Citrinin (CTN) and ochratoxin A (OTA) can be present as co-contaminants in cereals, foods and feed commodities, and can affect human health. Metabolism-dependent toxicity of these two mycotoxins, separately as well as in combination, is not yet understood. To fill this gap we adopted integrated discrete multiple organ co-culture (IdMOC) technique, which obviates animal experiments from the perspectives of species difference as well as animal welfare concerns. IdMOC facilitates co-culture of a metabolically competent cell (HepG2) and a metabolically incompetent cell (3T3) that are physically separated but provides for extracellular product(s) from one cell to interact with the other. After ascertaining that HepG2 is metabolically competent and 3T3 is not, adopting luciferin-IPA metabolism assay, CTN and OTA were tested separately and in combination in the co-culture set-up, when both proved to be metabolism-dependent cytotoxic agents. Hepatocytes metabolize CTN into a diffusible product that is cytotoxic to 3T3 cells but the cytotoxicity of OTA appears to be limited to the hepatocytes, i.e., local acting. As a combination at a concentration of 20% of IC₅₀ of each, CTN forms a reactive metabolite that diffuses out of HepG2 to cause cytotoxicity to 3T3 cells synergistically with OTA parent molecule. The CYP isoenzymes involved in the metabolism OTA and CTN were identified adopting in silico methods which indicated that OTA and CTN can bind CYP proteins at specific sites.
Toxic effect prediction of cefatirizine amidine sodium and its impurities by structure-toxicity relationship of cephalosporins
2018, Toxicology in Vitro
The three-dimensional (3D) structure-toxicity relationship of cephalosporins was explored by computing the most stable conformations of 33 kinds of cephalosporins in aqueous solution and using the teratogenicity and lethality of these compounds obtained in zebrafish embryo toxicity testing to evaluate their toxic effects. The toxic effect of cefatirizine amidine sodium, a novel cephalosporin which has finished preclinical study, was investigated. It is thought that the teratogenic effect of the triazine ring at the C-3 position is the main toxic effect of cefatirizine amidine. In addition, cefatirizine amidine is no more toxic than cefathiamidine and ceftriaxone. The results of the zebrafish embryo toxicity test combined with gene expression microarray technology were consistent with the prediction. The toxic effects of some potential process-related impurities of cefatirizine amidine were also predicted.
Extrapolation of In Vitro Results to Predict Human Toxicity
2018, In Vitro Toxicology
A gold standard for toxicity prediction in a living system is very much needed by the toxicology community. The enormous effort that gets unaccounted for between the numbers of successful drugs reaching the market and the investment during drug discovery needs to be addressed for better success. Classical methods are not sufficient enough anymore to predict the toxicity as a whole by taking into account the species specificity, drug metabolism, drug exposure, acute versus chronic effect, and existing ailment multiple effects occurring at the same time. Routine biochemical measurements typically include enzyme activity and metabolite concentration measurement in a cell-free environment. Observations gathered from one type of in vitro system remotely mimic the whole animal behavior. Moreover, these biochemical measurements are typically the end point of toxicity when the derangement in the biological processes causes terminal damage to the system. Onset as well as progress of toxicity can hardly be accounted for from these “end point”–type observations. A better predictive method need to be able to integrate various forms of alteration measured in multiple cell system under different condition—a steady accumulation of changes eventually culminates into an irreversible outcome. An endeavor to collate various results obtained from different in vitro systems on different measurements together may reveal the holistic picture behind toxicity.

View all citing articles on Scopus

View full text