Joining the DoTS: new approach to classifying adverse drug reactions

Short abstract

A new classification system for adverse drug reactions based on time course and susceptibility as well as dose responsiveness should improve drug development and management of adverse reactions

The pharmacological classification of adverse drug reactions whose causality has been established currently rests on the perceived dose dependence and predictability of the adverse reaction. It is based on a proposal of Rawlins and Thompson, prefigured by others (see table A on bmj.com), to classify adverse drug reactions into two types¹: type A reactions, dose dependent and predictable from the known pharmacology of the drug; and type B reactions, not dose dependent and unpredictable.² This classification is simple; it helps drug regulation because prelicensing studies can reveal type A reactions,³ and it predicts that dose titration will reduce the risk of some reactions. However, it is sometimes difficult or impossible to assign a reaction to one type. For example, dose dependent (type A) nausea and vomiting due to erythromycin could also be classified as type B because it is not pharmacologically predictable.

Furthermore, other types of adverse reactions are not comfortably classified by the system. For example, osteoporosis from corticosteroids depends not only on dose but also on duration of treatment. And some reactions, such as asthma from β adrenoceptor antagonists, do not occur in all patients. The classification has gradually been extended to other alphabetically labelled types (see table A on bmj.com), including type C (dose and time dependent (chronic) reactions), type D (delayed reactions), type E (withdrawal reactions), and type F (failure of therapy).⁴ These modifications have mitigated some of the difficulties of the classification system but have introduced others.

The current classification is defined only by properties of the drug—its known pharmacology and the dose dependence of its effects. However, other criteria should be taken into account in a comprehensive classification, including properties of the reaction (the time course of its appearance and its severity) and properties of the individual (the genetic, pathological, and other biological differences that confer susceptibility). We therefore propose a three dimensional classification system based on dose relatedness, timing, and patient susceptibility (DoTS).

Figure 1.

Credit: LIANNE PAYNE

Dose relatedness

Traditionally, immunological and certain other adverse drug reactions have been considered not to be dose related. However, effects of drugs involve interactions between chemical entities and are therefore subject to the law of mass action. This implies that all drug effects, beneficial or adverse, are dose related. Examples of immunological reactions that are clearly dose dependent include hay fever in response to high pollen counts⁵; the immunogenic response to hepatitis B vaccine⁶; desensitisation by the use of increasing doses of antigen (for example, cephalosporins)⁷; and type IV hypersensitivity skin reactions.⁸

It is therefore misleading to suggest that type B adverse drug reactions are not dose dependent.² In fact, it is clearer to divide adverse drug reactions into reactions that occur at supratherapeutic doses (toxic effects); reactions that occur at standard therapeutic doses (collateral effects); and reactions that occur at subtherapeutic doses in susceptible patients (hypersusceptibility reactions). We use the term collateral effects for reactions that occur at standard therapeutic doses because the term side effects is often colloquially used to refer to all adverse effects. Collateral effects include those that occur due to a different pharmacological effect from the therapeutic action and those that occur through the therapeutic pharmacological effect but in another tissue.

Time relatedness

Many pharmacological effects depend on both the concentration of the drug at the site of action and the time course of its appearance there. For example, a given dose of furosemide (frusemide) induces a greater diuresis when it is infused than when it is given as a bolus.⁹ And the toxicity of methotrexate is greater when a low dose is given repeatedly than when the same total amount is given as a single dose.¹⁰

We distinguish two patterns of time courses of adverse drug reactions, time dependent and time independent (see table B on bmj.com for details of the classification and its implications).

Time independent reactions

Time independent reactions occur at any time during treatment, independent of the duration of the course. They typically occur either when the concentration ofthe drug at the site of action changes (for example, digoxin toxicity when renal function worsens) or when the pharmacological response is altered without a change in concentration (for example, digoxin toxicity in association with potassium depletion). When such a reaction occurs, its time course may be affected by the kinetics of the drug, but that is not an aspect of its time dependency as defined here.

Time dependent reactions

There are six subtypes of time dependent reactions—rapid, first dose, early, intermediate, late, and delayed.

Rapid reactions occur only when a drug is administered too rapidly—for example, the red man syndrome with vancomycin.¹¹

First dose reactions occur after the first dose of a course of treatment and not necessarily thereafter. Examples include hypotension after the first dose of an angiotensin converting enzyme inhibitor¹² and type I hypersensitivity reactions. In type I hypersensitivity reactions the reaction occurs after the first dose of a course, whether or not previous exposure has been recorded; 30% of those who develop anaphylaxis with penicillin have no such record.¹³ We regard a previous sensitising exposure as causing a change in susceptibility.

Early reactions occur early in treatment then abate with continuing treatment. These are adverse drug reactions to which patients develop tolerance (such as nitrate induced headache).

Intermediate reactions occur after some delay; however, if a reaction has not occurred after a certain time, there is little or no risk that it will occur later. Examples are hypersensitivity reactions of type II (thrombocytopenia due to quinine), type III (interstitial nephritis with penicillins), and type IV (cutaneous hypersensitivity to antihistamines), and the ampicillin/amoxicillin pseudoallergic rash.¹⁴ Non-allergic reactions of this type include the increased risk of neutropenia with carbimazole and of venous thromboembolism with antipsychotic drugs. We believe that intermediate reactions occur in populations of individuals with different susceptibilities. Those at high risk have the reaction and stop taking the drug; those at low risk do not have the reaction and can be regarded as healthy survivors. Thus, after a time the population risk seems to fall.

Late reactions occur rarely or not at all at the beginning of treatment, but the risk increases with continued or repeated exposure. Examples include many of the adverse effects of corticosteroids and tardive dyskinesia with dopamine receptor antagonists. Withdrawal reactions are late reactions that occur when a drug is withdrawn or its dose is reduced after prolonged treatment. They include opiate and benzodiazepine withdrawal syndromes, hypertension after withdrawal of clonidine or methyldopa, and acute myocardial infarction after withdrawal of β blockers.

Delayed reactions are observed some time after exposure, even if the drug is withdrawn before the reaction appears. Examples are carcinogenesis (vaginal adenocarcinoma in women who were exposed to diethylstilbestrol in utero) and teratogenesis (phocomelia due to thalidomide).

Susceptibility

The risk of an adverse drug reaction differs among members of an exposed population. In some cases the risk of an adverse reaction will be present in susceptible subjects and absent in others. In other cases susceptibility follows a continuous distribution—for example, increasing susceptibility with increasing impairment of renal function. Although reasons for hypersusceptibility may be unknown, several types are recognised. These include genetic variation, age, sex, physiological variation, exogenous factors, and disease (table 1). More than one susceptibility factor can be present.

Table 1.

Sources of altered susceptibility to adverse drug reactions

Source of susceptibility	Examples	Implications
Genetic	Porphyria	Screen for abnormalities; avoid specific drugs
	Succinylcholine sensitivity
	Malignant hyperthermia
	CYP isozyme polymorphisms
Age	Neonates (chloramphenicol15)	Adjust doses according to age
	Elderly people (hypnotics16)
Sex	Alcohol intoxication	Use different doses in men and women
	Mefloquine, neuropsychiatric effects17
	Angiotensin converting enzyme inhibitors, cough
	Lupus-like syndrome18
Physiology altered	Phenytoin in pregnancy19	Alter dose or avoid
Exogenous factors	Drug interactions	Alter dose or avoid co-administration
	Interactions with food (eg grapefruit juice with drugs cleared by CYP3A420)
Disease	Renal insufficiency (eg lithium21)	Screen for abnormalities; avoid specific drugs; use reduced doses
	Hepatic cirrhosis (eg morphine22)

Using DoTS

To show how the categorical form of the classification works, the box gives examples of three adverse drug reactions that readers of draft versions of this paper have challenged us to classify. Classifying an adverse drug reaction in this way will allow doctors to consider the implications of managing it (see table B on bmj.com).

A more sophisticated probabilistic analysis is also possible, but we shall not discuss it in detail here. It requires an estimate of the probability of an adverse drug reaction at different doses and times after administration, for different degrees of susceptibility. The information can be displayed as a series of three dimensional graphs or equivalent nomograms (figure).

Figure 2.

Graphs showing how probability of adverse drug reaction (y axis) might vary with variations in time after administration (x axis, arbitrary units) and dose (z axis, arbitrary units) in people with high, medium, and low susceptibility having an adverse effect of intermediate type

A one dimensional classification system based on time course could meet all the requirements for a useful classification of adverse drug reactions (table 2). For example, the time course of a reaction is evident in each patient, so that all reactions can be classified by individual observation, supplemented, if necessary, by observations in the population. The association between halothane and hepatitis was first shown by a careful analysis of the time course of the reaction in individual cases.²³ The time course also helps to distinguish similar adverse drug reactions, such as the two forms of heparin induced thrombocytopenia,²⁴ the two forms of chloramphenicol induced anaemia,²⁵ and photoallergic and phototoxic reactions.²⁶ However, time course alone is an unsatisfactory basis for classification because it ignores important information on dose dependence and individual susceptibility. Our proposal for a three dimensional classification should provide important insights for drug development and regulation, for pharmacovigilance, for monitoring patients, and for the prevention, diagnosis, and treatment of adverse drug reactions.

Table 2.

How dose related, time related, and susceptibility related classifications of adverse drug reactions fulfil criteria for satisfactory classification

	Classification
Criterion	Dose related	Time related	Susceptibility related
Allows classification on basis of clinical features	No; dose dependency is not always clear from clinical observations and dose ranging studies are not always available	Yes; the time course of a reaction can be directly observed in individual cases or populations	Sometimes, depending on type of susceptibility
Give insight into mechanism of reaction	No; only implies the range of doses at which it occurs	Yes; different mechanisms have different time courses	Yes; mechanism and susceptibility are often linked
Avoids assigning a reaction to more than one category	No	Yes	No; an adverse drug reaction may be associated with multiple susceptibility factors
Suggests how to monitor adverse reactions	Yes	Yes	Yes
Suggests population strategies for pharmacovigilance	Yes	Yes; also tells the patient when to be alert for an adverse reaction	Yes (can identify patients at high risk or low risk)
Helps in making decisions on treatment or avoiding adverse reactions	Only some types	Yes	Only some types
Guides drug development and regulation	Yes; can help in defining the therapeutic dosage range	Yes; suggests strategies for monitoring during drug development and after marketing	Yes; defines subgroups at high risk or low risk

Examples of DoTS (dose-time-susceptibility) classification

• Osteoporosis due to corticosteroids: Do—collateral effect; T—late; S—age, sex.

• Anaphylaxis due to penicillin: Do—hypersusceptilbility; T—first dose; S—not understood; requires previous sensitisation

• Hepatotoxicity due to isoniazid: Do—collateral effect; T—intermediate; S—genetic (drug metabolism), age, exogenous (alcohol), disease (malnutrition)

Summary points

The current classification of adverse drug reactions based on dose response is inadequate

The time course of the reaction and the susceptibility of the patient also need to be taken into account

A three dimensional approach to adverse drug reactions is proposed based on dose, time, and susceptibility (DoTS)

This approach would improve drug development and patient care