Interpretation of Postmortem Toxicology Results : Pharmacogenetics and Drug-Alcohol Interaction

INTERPRETATION OF POSTMORTEM TOXICOLOGY RESULTS

Pharmacogenetics and Drug-Alcohol Interaction

Anna Koski

Department of Forensic Medicine University of Helsinki

Finland

ACADEMIC DISSERTATION

To be publicly discussed, with the permission of the Medical Faculty of the University of Helsinki, in the auditorium of the

Department of Forensic Medicine on September 23^rd 2005, at 12 noon.

Helsinki 2005

SUPERVISORS

Professor Antti Sajantila

Department of Forensic Medicine University of Helsinki

Helsinki, Finland Docent Ilkka Ojanperä

Department of Forensic Medicine University of Helsinki

Helsinki, Finland

REVIEWERS

Docent Eero Mervaala Institute of Biomedicine University of Helsinki Helsinki, Finland

Docent Kari Poikolainen

Finnish Foundation for Alcohol Studies

National Research and Development Centre for Welfare and Health Helsinki, Finland

OPPONENT

Professor Jørg Mørland

Division of Forensic Toxicology and Drug Abuse Norwegian Institute of Public Health

Oslo, Norway

ISBN 952-91-9214-2 (paperback) ISBN 952-10-2662-6 (pdf)

http://ethesis.helsinki.fi

Helsinki University Printing House

Helsinki 2005

Obtaining a quantitative result is not the endpoint of the analytical process.

Irving Sunshine

CONTENTS

A

... 6

L

O

P

... 7

A

... 8

I

... 9

R

L

... 10

1 Interpretation of Postmortem Forensic Toxicology Results... 10

1.1 Blood Samples ... 10

1.1.1 Toxicological Tables ... 11

1.2 Other Matrices ... 12

1.3 Postmortem Changes ... 12

1.4 Metabolites... 13

1.5 Alcohol Toxicity ... 13

1.6 Drug Toxicity... 13

1.6.1 Fatal Toxicity Index ... 14

1.6.2 Other Measures of Toxicity... 15

1.6.3 Sources of Bias in Toxicity Indices... 15

2 Pharmacogenetics... 16

2.1 Drug-Metabolizing Enzymes... 16

2.1.1 Cytochrome P450 System ... 16

CYP2D6 ... 17

CYP2C19 ... 19

2.2 Studies on Drug Metabolism ... 19

2.2.1 Tramadol Metabolism ... 19

2.2.2 Amitriptyline Metabolism ... 20

2.3 Clinical Pharmacogenetics... 21

2.4 Postmortem Pharmacogenetics ... 22

3 Drug-Alcohol Interaction ... 23

3.1 Alcohol Effects and Anesthetic Action ... 23

3.1.1 γ-Aminobutyric Acid Receptor Type A ... 24

3.2 Animal Studies... 24

3.3 Postmortem Studies ... 25

A

S

... 26

M

M

... 27

1 Autopsy cases ... 27

1.1 Blood Samples ... 27

1.2 Database ... 27

2 Analysis of Drug Concentrations ... 27

2.1 Screening... 27

2.2 Metabolite Analysis ... 28

3 Genotyping ... 28

3.1 Long Polymerase Chain Reaction... 29

3.2 Restriction Fragment Length Polymorphism Analysis... 29

3.3 Multiplex Single-Base Extension Reaction... 29

4 Case Selection Criteria ... 29

5 Statistical Methods ... 30

R

... 31

1 Pharmacogenetics ... 31

1.1

1.2

1.3

1.4 Allele and Genotype Frequencies (I, II)... 33

2 Fatal Toxicity Indices (IV, V) ... 33

3 Drug-Alcohol Interaction... 34

3.1 Alcohol and Benzodiazepines (III) ... 34

3.2 Alcohol and Other Common Drugs (IV-VI)... 34

D

... 38

1 Methodological Considerations... 38

2 Pharmacogenetics ... 39

3 Drug-Alcohol Interaction... 41

3.1 Alcohol and Benzodiazepines... 41

3.2 Alcohol and Other Common Drugs ... 41

4 Drug Safety... 42

4.1 Newer Antidepressants ... 42

5 Implications for Interpretation... 43

C

... 45

A

... 46

R

... 47

ABBREVIATIONS

ADR adverse drug reaction BAC blood alcohol concentration BDZ benzodiazepine

bp base pair

CI confidence interval

CNS central nervous system

CYP cytochrome P450

CYP2C19 cytochrome P450 enzyme 2C19 CYP2C19 gene encoding CYP2C19 CYP2D6 cytochrome P450 enzyme 2D6 CYP2D6 gene encoding CYP2D6

DDD defined daily dose

DME drug-metabolizing enzyme

EHAT (E)-10-hydroxyamitriptyline EHNT (E)-10-hydroxynortriptyline

EM extensive metabolizer (phenotype)

FTI fatal toxicity index

GC gas chromatography

gEM genetically extensive metabolizer gPM genetically poor metabolizer

gUM genetically ultra-rapid metabolizer ICD-10 International Classification of Diseases, 10^th Revision

IM intermediate metabolizer (phenotype)

kb thousand base pairs

LC/MS-MS liquid chromatography – tandem mass spectrometry

LD50 median lethal dose, the dose required to kill 50% of the given population LIMS laboratory information management system

M1 O-demethyltramadol

M2 N-demethyltramadol, nortramadol M3 N,N-didemethyltramadol

M4 O,N,N-tridemethyltramadol M5 O,N-didemethyltramadol

MR metabolite ratio

MRM multiple reaction monitoring

MS mass spectrometry

NNT N-demethylnortriptyline

p significance level

OPLC overpressured layer chromatography PCR polymerase chain reaction

PM poor metabolizer (phenotype)

RFLP restriction fragment length polymorphism SSRI selective serotonin reuptake inhibitor

TCA tricyclic antidepressant

TLC thin layer chromatography

UM ultra-rapid metabolizer (phenotype) ZHAT (Z)-10-hydroxyamitriptyline ZHNT (Z)-10-hydroxynortriptyline

LIST OF ORIGINAL PUBLICATIONS

This thesis is based on the following articles, which are referred to as I-VI in the text:

I Levo A, Koski A, Ojanperä I, Vuori E, Sajantila A (2003) Post-mortem SNP analysis of CYP2D6 gene reveals correlation between genotype and opioid drug (tramadol) metabolite ratios in blood. Forensic Sci Int 135(1):9-15.

II Koski A, Sistonen J, Ojanperä I, Gergov M, Vuori E, Sajantila A (2005) CYP2D6 and CYP2C19 genotypes and amitriptyline metabolite ratios in a series of medicolegal autopsies.

Forensic Sci Int (in press, published online July 14, DOI: 10.1016/j.forsciint.2005.05.032).

III Koski A, Ojanperä I, Vuori E (2002) Alcohol and benzodiazepines in fatal poisonings. Alcohol Clin Exp Res 26(7):956-9.

IV Koski A, Ojanperä I, Vuori E (2003) Interaction of alcohol and drugs in fatal poisonings. Hum Exp Toxicol 22(5):281-7.

V Koski A, Vuori E, Ojanperä I (2005) Newer antidepressants: evaluation of fatal toxicity index and interaction with alcohol based on Finnish postmortem data. Int J Legal Med (in press, published online March 1, DOI: 10.1007/s00414-005-0528-x).

VI Koski A, Vuori E, Ojanperä I (2005) Relation of postmortem blood alcohol and drug concentrations in fatal poisonings involving amitriptyline, propoxyphene and promazine. Hum Exp Toxicol 24(8):389-96.

The original publications are reproduced with the permission of the copyright holders.

ABSTRACT

Postmortem forensic toxicology annually reveals more than a thousand fatal poisonings in Finland.

Both alcohol and drugs are found in the vast majority of cases, with certain drugs more often involved than others. Some of the drugs commonly causing fatal poisonings are polymorphically metabolized.

In this thesis, a retrospective, statistical approach was taken to elucidate the role of pharmacogenetics and drug-alcohol interaction in fatal poisonings. More specifically, the objective was to investigate whether certain genetic variants associated with abnormal drug metabolism can be correlated with metabolite ratios in postmortem material, and whether an interaction between alcohol and common toxic drugs is perceptible in fatal poisonings. Methods included genotyping and metabolite analysis using autopsy blood, as well as statistical analysis of the obtained metabolite ratios. Drug safety was evaluated as a function of alcohol and drug concentrations determined in postmortem blood and of fatality rates in relation to the sales of drugs.

Correlations between the CYP2D6 gene dose and the tramadol metabolite ratios were observed in 33 cases involving tramadol. In 195 cases involving amitriptyline, similar correlations were found between the CYP2D6 gene dose and the metabolite ratios related to stereospecific ring hydroxylation and between CYP2C19 gene dose and the metabolite ratios related to N-demethylation. Most importantly, the nonfunctional genotypes were significantly different from the corresponding fully functional genotypes with respect to several of the investigated metabolite ratios. However, no fatal poisonings with accidental or undetermined cause of death were associated with nonfunctional genotypes.

Regarding fatal poisonings involving two common benzodiazepines, blood alcohol concentrations were on average lower in cases involving temazepam than in those involving diazepam or alcohol alone. Diazepam therefore appeared safer in combination with alcohol than temazepam. Among the drugs most commonly causing fatal poisonings, promazine, doxepin, amitriptyline, and propoxyphene were the least safe in combination with alcohol, whereas zopiclone, diltiazem, and the newer antidepressants proved relatively safe. The selective serotonin reuptake inhibitors appeared the safest among the newer antidepressants. Interestingly, the least safe drugs in combination with alcohol were also found to cause more fatal poisonings with respect to their sales than the other drugs included in the study. A similar correlation was also observed within the group of newer antidepressants.

In conclusion, genetic factors seem to play a more dominant role in metabolite ratios than age, gender, or environmental factors, and postmortem genotyping may therefore provide useful information in poisoning cases where the manner of death is unclear. When determining the cause of death, the possibility of a fatal poisoning due to an interaction between alcohol and drugs should be considered seriously, especially when certain toxic drugs are involved. These results have implications not only for the interpretation of postmortem toxicology results but for drug safety in general.

INTRODUCTION

Approximately 50 000 Finns die each year.

Twenty percent of these deaths are investigated in a medicolegal autopsy, and in about half of the autopsy cases samples are taken for toxicological analysis. Forensic toxicology, i.e.

the use of toxicology for legal purposes, today employs a wide variety of analytical methods producing a wealth of data. In postmortem investigation, the main purpose of these data is to help the forensic pathologist to determine the cause of death, but the eventual significance of the results can vary greatly. Besides confirming, revealing, or excluding fatal poisonings, the findings may yield important information on the contributing factors and general circumstances of the death.

According to Act 169/1948, all postmortem forensic toxicology in Finland is centralized to the Laboratory of Toxicology, Department of Forensic Medicine, at the University of Helsinki. Blood, urine, and liver are routinely screened for drugs, alcohols, and drugs of abuse.

Upon request, samples are also analyzed for carbon monoxide, cyanide, and other suspected poisons. Analytical methods include chromato- graphy, spectrometry, and immunological assays. DNA analysis may also be employed for further investigations. Broad drug screens and dedicated analyses are focused on reliable detection, identification, and quantitation of potentially toxic compounds. Quality control is extensively applied to ascertain the integrity of results. The laboratory was accredited by the Finnish Accreditation Service (FINAS) in 1997.

Advances in pharmacology, resulting in new drugs and combinations of drugs, as well as in novel indications for existing drugs, pose a challenge to forensic toxicology. The discipline is further complicated by illegal drugs, designer drugs, and changing local and global trends in drug abuse. To keep current on what is happening ‛on the street’, the analytical methods in the laboratory must be continuously developed. Modern-day forensic toxicology produces a wealth of data, demanding

automation in analysis, reporting, and data management. Forensic toxicology findings are therefore stored in a database such as the Laboratory Information Management System (LIMS) in the Laboratory of Toxicology.

Interpretation of postmortem toxicology results is based on a forensic pathologist’s experience and on previously reported findings regarding toxicity of drugs, alcohols, and other common poisons. In Finland, it is the forensic pathologist who determines the cause of death, although the forensic toxicologist may be consulted on analytical findings. A strength of Finnish forensic medicine is the practice of the investigating pathologist to send a copy of the completed death certificate to the Laboratory of Toxicology. Integrating the information on cause and manner of death to the LIMS creates a nation-wide databank enabling complex research and acquisition of detailed statistics on Finnish fatalities.

There are several confounding factors in the interpretation of postmortem toxicology results. Besides the background information, important issues to be taken into consideration include postmortem redistribution, individual variation, and concomitant findings.

Theoretically, any two compounds that share a mechanism of action or produce a similar response may cause unwanted or pronounced adverse effects. In practice, the most common agent to interact with a drug is alcohol. Alcohol is a frequent finding in forensic toxicology, and being a central nervous system (CNS) depressant, is often deemed to have played a part in causing death. A study was therefore undertaken to elucidate the role of alcohol in fatal poisonings. In addition, the role of genetic factors in drug-related deaths was investigated.

The general purpose of this thesis was to assess the importance of pharmacodynamic drug-alcohol interactions and pharmacogenetic variation affecting drug metabolism in a postmortem context. The findings can be expected to support the interpretation of postmortem toxicology results.

REVIEW OF THE LITERATURE

1 Interpretation of Postmortem Forensic Toxicology Results

A forensic toxicological investigation consists of three main steps: obtaining the case history and suitable specimens, performing toxicological analyses, and interpreting the findings. In a post- mortem case, toxicological specimens are collected at autopsy or external examination and subjected to chemical analysis, which is a process of extraction, detection, identification, and quantitation of the analytes [1,2].

Comprehensive treatises on interpretation of forensic toxicology results have most recently appeared by Jones [3], Richardson [4], and Holmgren [5], but throughout the years the subject has inspired numerous reviews, essays, and book chapters [6-15]. Although the principles are globally applicable, some of the discussion reflects professional experience, personal opinions, and local medicolegal systems of the authors. In the Finnish system, interpretation of postmortem toxicology results involves both the forensic toxicologist, who estimates the relevance of the findings and the need for further analysis, and the forensic patho- logist, who eventually assesses the contribution of the findings to the cause of death.

Several aspects are considered in the interpretation. First of all, the numerical results obtained in the laboratory are meaningful only in context with the individual case and such background information as acute/chronic exposure, emergency treatment, and length of survival [6,8,15]. Besides individual variation, thought should be given to the site of specimen collection, the methods of collection and analysis, findings in the other matrices investigated, autopsy findings, and possible postmortem changes [8,10].

Multiple substances are commonly involved in overdoses, hence the difficulty of attributing the fatality to any single one.

Whenever several drugs are found, the possibility of additive effects, synergism, or antagonism ought to be considered [6].

Especially benzodiazepines (BDZs) and alcohol

are often present in overdoses of other drugs [16,17]. Moreover, tolerance (and cross- tolerance) to drugs or alcohol may have developed prior to death. Since the extent and duration of exposure are seldom precisely or reliably known, concentrations of certain analytes, e.g. lead, barbiturates, BDZs, and several drugs of abuse, are not necessarily very meaningful [6,10,14]. In addition to appreciating the pharmacodynamic properties of drugs, knowledge of pharmacokinetics, of drug bio- transformation in particular, is essential in interpreting the results, especially when the parent compound is either rapidly or poly- morphically metabolized or is not readily detected. Yet another phenomenon to be considered is idiosyncrasy: although most people react to a drug predictably, a few will react differently [6].

What is not found also influences the assessment. Negative findings allow exclusion of many relevant poisons [2], but require comprehension of the limitations of analytical methods. Even the most modern equipment is able to detect only a part of the vast array of pharmacological agents in use today. Due to the great number of possible toxicants, general unknown screening [18] and substance identification [19] are currently key problems in state-of-the-art forensic toxicology. Moreover, once a chemical entity has been detected and identified, quantitation may prove impossible because certified reference materials are unavailable or difficult to obtain [20].

1.1 Blood Samples

Of the matrices available at autopsy, blood is essential for evaluating whether the deceased was under the influence of a drug at the time of death. The specimen of choice for quantitative purposes is femoral venous blood because it is the least susceptible to postmortem changes [3,15,21]. The recommended method of collection is to draw blood from a ligated or severed femoral vein into a plastic tube. A supplementary sample of central blood is often

collected, especially when femoral blood is unavailable or scant, but should be reserved for qualitative purposes only [15]. The drug concentrations in cardiac blood are often higher than in femoral blood [11,22], which may facilitate detection. When time has elapsed between trauma and eventual death, analysis of subdural and epidural blood clots may provide information pertaining to the time of injury.

In toxicological case work, blood drug concentrations are today reliably determined using mass spectrometry (MS) [19], although older tools, such as gas chromatography (GC) coupled with flame ionization detection or with a nitrogen-phosphorus selective detector remain valuable in the analysis of alcohols [23] and nitrogen-containing drugs [24], respectively. To ensure the quality of analytical results and their validity in courts of law, method validation, internal quality assurance, and external proficiency testing are widely employed in forensic toxicology laboratories [25-28].

1.1.1 Toxicological Tables

To aid interpretation of the results, therapeutic, toxic, and lethal concentrations of drugs in human blood, plasma, and serum have been compiled into various handbooks [22,29,30], reviews, and original articles [31-44].

Providing a useful addition to the traditional handbooks and other printed references, the internet can be employed to access many toxicology-related resources, recently reviewed by Goldberger and Polettini [45]. These resources include two extensive compilations of therapeutic and toxic levels of drugs in biological specimens [46,47].

A major problem with using such reference values in postmortem toxicology is that post- mortem drug concentrations are determined in hemolyzed whole blood, whereas clinical studies usually provide information on plasma or serum concentrations [11]. To extrapolate postmortem results to the antemortem state, whole blood/plasma concentration ratios [15]

and whether they stay the same in a decomposing body [3] should be known. The compilations most relevant to forensic toxicology, as discussed by Druid and Holmgren

[48], are therefore those in which the lethal concentrations in postmortem material are cited [33,34,37]. Some of these tables also include statistical information on concentration distributions in different types of fatalities [34,37].

However, how concentrations measured postmortem relate to those measured in life [10,13,15] remains obscure. Drawing correlations between these two situations is therefore not straightforward. Firstly, post- mortem blood is not the same as circulating blood in a living person [13]. Secondly, in a living body the pharmacodynamic response to a drug is dictated by the concentration of a free drug, whereas a postmortem result represents the sum of the free and protein-bound drug. Thirdly, peak concentrations of the drug are likely to have been higher than what is seen postmortem, since the peak concentration may cause irreversible damage but not immediate death [49]. It may take hours for the intoxicated person to succumb, with the drug being metabolized in the meanwhile. Blood alcohol concentration (BAC), for instance, decreases 0.10-0.25‰/h in moderate drinkers and even faster in alcoholics [50]. There are also instances where peak drug concentrations have been reported to be higher postmortem; in two cases in which intoxication had led to hospitalization prior to death, higher amitriptyline and propoxyphene concentrations in blood were found postmortem than antemortem [20].

Finally, when a drug has one or more chiral centers, the pharmacodynamic and pharmaco- kinetic behavior usually differs between the iso- mers. In the hopes of minimizing the expensive production of ineffective isomers, avoiding the side-effects caused by harmful isomers (e.g. (–)- thalidomide), and splitting the drug load on an individual in general, drug development is now aimed at enantiospecific drugs [51]. When a drug may be present in toxicological samples either as a racemate or as an enantiomer, as in the cases of citalopram and amphetamine, interpretation of blood concentrations can be considered confounded by yet another factor, unless enantioselective analysis has been applied to the forensic samples.

1.2 Other Matrices

Other common matrices of interest in postmortem toxicology include urine, liver, gastric contents, bile, blood clots, and vitreous humor [15,52]. Muscle tissue, bone marrow, and hair can be used in severely putrefied cases, and even larvae feeding on the corpse may be examined [8,52,53]. In all, alternative matrices can constitute valuable specimens for postmortem toxicology, as recently reviewed by Drummer and Gerostamoulos [52] and Skopp [15].

Due to site and temporal dependence of drug concentrations [54,55], the toxicological results obtained in other matrices differ in significance from those determined in femoral blood. A positive finding in urine, for instance, shows that the detected substance was present in the body some time before death, but the physiological effects exerted by the compound on the body may not be readily deduced from the concentration in urine [8,15].

Liver is a highly valuable specimen since many substances are present in higher concentrations in the liver and are thus more easily detected than in femoral blood [11,22].

Utilization of hair samples has been recently recapitulated by Kintz [53]. Hair is of exceptional value in exhumed bodies, but can also be used to detect exposure to drugs over a period of several months prior to death [15].

Detection of a compound in gastric contents does not necessarily indicate recent uptake, for drugs can be re-excreted into gastric juices [3].

Vitreous humor is another valuable specimen in postmortem toxicology. It is a relatively well-isolated, sterile compartment protected from trauma and putrefaction, with drug concentrations typically following the concentrations in blood with a certain delay [15]. Glucose, lactate, and potassium are conveniently determined in vitreous humor [56].

Alcohol concentrations in vitreous humor closely follow BACs, when the difference in water content is taken into account [57]. Thus vitreous humor provides the only other matrix, besides peripheral blood, capable of yielding a meaningful quantitative result [11]. Its use as an alternative specimen is, however, limited by the small sample size of only 3-6 ml [3].

1.3 Postmortem Changes

The quality of a postmortem specimen is often poor; it can be watery, putrefied, degraded, or burned. The stability of drugs in postmortem samples is another concern [15,52,58]. Changes in concentration are generally not sufficiently large to affect interpretation, especially when femoral blood is used [58], but such drugs as nitrobenzodiazepines and cocaine may disappear gradually due to bacterial action [59] and hydrolysis [60], respectively. It has also been suggested that, due to reformation of the parent drug from metabolites [58], e.g. by hydrolysis of conjugated entities [15], parent drug concentrations may even increase.

Another type of postmortem change is the production of ethanol from carbohydrates by certain microorganisms in a putrefying body or during storage [61]. To prevent further conversion in the autopsy sample, refrigeration and addition of potassium or sodium fluoride to a final concentration of 1-5% are generally recommended. Microbial activity may produce significant ethanol concentrations, in some cases in excess of 1‰, and a positive BAC should therefore be verified by analysis of urine or vitreous humor whenever possible [62].

The most relevant alteration occurring between death and autopsy is, however, post- mortem redistribution, i.e. the migration of drugs between tissues and blood in a cadaver [63,64]. Literature on postmortem redistribution has recently been summarized in a brief review by Leikin and Watson [11], and the mechanisms involved have been the subject of a more extensive review by Pélissier-Alicot et al. [65].

Drugs that undergo postmortem redistribution are typically lipophilic, weakly basic compounds with a relatively large volume of distribution or preferential binding to the myocardium [21,65,66].

The most important mechanism of redistribution has been estimated to be drug diffusion from the gastrointestinal tract, lungs, and other drug-rich tissues, such as the myocardium, into surrounding tissues and blood [67]. Recent ingestion of a large amount of drugs may result in postmortem diffusion of the unabsorbed drug from the gastric contents to the surrounding organs and vessels [65]. The extent

of redistribution depends on the drug concentration and possible resuscitation attempts, probably shifting cardiac blood towards the periphery [55,65], but also on the length of the delay between death and autopsy and the conditions during the delay [68].

1.4 Metabolites

In intoxications, a large part of the toxic action may derive from the metabolites of the consumed substances. When the metabolite is pharmacologically active (e.g. amitriptyline/

nortriptyline, codeine/morphine/morphine-6- glucuronide, methanol/formic acid), it may contribute to death to the same or even to a greater extent than the parent drug. Knowledge of metabolite levels is therefore of great interpretative value in postmortem toxicology.

Quantitative determination of known metabolites along with the parent drug may further help to determine the time of intake and the type of poisoning (acute vs. chronic or therapeutic exposure [69]), and possibly arouse suspicions of metabolic anomalies [70].

Qualitative identification of metabolites is often used to corroborate the finding of a parent drug.

Furthermore, certain metabolites are present in the body in higher concentrations, thus being easier to detect than the parent drug. Screening for metabolites is particularly important when the suspected parent compound is metabolized extremely fast (e.g. cocaine or heroin [14]), is not excreted in urine, or is not readily detected. Unfortunately, few metabolites are commercially available [71].

1.5 Alcohol Toxicity

Alcohol (ethyl alcohol, ethanol) is a frequent finding in postmortem toxicology, with approximately 400 fatal poisonings attributed to it annually in Finland [72]. Alcohol was detected in 47.7% of Swedish fatal poisonings in 1992-2002 [73], and BACs of 0.50‰ or higher were detected in 50.4% of Finnish fatal drug poisonings in 2000-2001 [72]. Of all of the Finnish postmortem cases analyzed for alcohol in 2000-2004, BACs of 0.20‰ or higher were detected in 45.5% (Vuori et al., unpublished results). The frequent involvement of alcohol in poisonings in general is illustrated by a Finnish

study in which alcohol was found in two-thirds of patients presenting with acute poisoning to an emergency department [74].

The BAC level causing death is often cited as >3.5‰ [42] or ≥4‰ (92mM) [31,33,38], but lower estimates have also been presented [34,75]. Depending on the source, the given value may refer to either a BAC determined postmortem or a peak BAC estimated to have caused an irreversible event leading to death.

However, death may ensue already from lower BACs, especially in alcoholics with a weakened physiological status (heart disease, malnutrition, liver cirrhosis, ketoacidosis), as well as when aspiration of stomach content, postural asphyxia, or hypothermia is involved [75,76]. Old age and concurrent CNS depression from other causes are further factors that may lower the lethal BAC [76]. Tolerance, however, is also an important aspect of alcohol toxicity. People have survived – even driven motor vehicles [77] – at concentrations much higher than 4‰, with a BAC of 15‰ (340mM) probably being the highest reported in a living person [78].

Reported mean and median BACs found in fatal alcohol poisonings usually range from 3‰

to 4‰ [34,49,79-81]. Cumulative frequency distributions have also been published [79-81].

The curves in Figure 1 show the cumulative proportion of fatal alcohol poisonings in which a certain BAC was found.

1.6 Drug Toxicity

Drugs exert their therapeutic effects by various mechanisms. Each drug usually has a specific mechanism of action, which may also mediate the toxic effects produced by higher concentrations, but no single mechanism can be pointed out as the cause of drug toxicity in general. Toxicity is therefore thoroughly investigated in the process of drug development, with each new drug having to pass an extensive series of preclinical and clinical tests before being approved for sale. Even so, virtually all drugs on the market have some degree of toxicity. Postmarketing research on drug toxicity is therefore conducted as well, with the purpose of further improving drug safety. Identification of particularly toxic drugs can lead to restrictions or recommendations intended to

Figure 1. Cumulative frequency distribution of blood alcohol concentrations in fatal alcohol poisonings in Finland in 1983-1985. Modified from a report by Vuori et al. [80].

prevent future adverse events. Restricted prescription of barbiturates since the late 1960s, for instance, was a successful measure in reducing barbiturate poisonings [82], and similarly, dispensing regulations for propoxyphene were tightened in the 1980s, after many reports of propoxyphene-related fatalities [83-85]. The conventional method of assessing acute lethal toxicity, i.e. determining the amount of drug required to kill 50% of a given population (LD₅₀), is obviously not appropriate for people. Various methods have thus been developed for assessing acute drug toxicity in humans in overdose situations, usually by taking an epidemiological approach.

1.6.1 Fatal Toxicity Index

The frequency of poisonings caused by a drug depends to a large extent on its availability and inherent toxicity [86,87]. Controlling for drug availability should thus enable us to compare the degree of inherent drug toxicity for man. From a forensic toxicologist’s point of view, a practical measure of relative drug toxicity is the fatal toxicity index (FTI). It is calculated by relating the number of fatalities attributed to a drug over

a certain time period and area to the consumption of the drug over the same period and area. Consumption can be measured either by number of prescriptions, kilograms, or defined daily doses (DDDs) dispensed, with DDD being the assumed average maintenance dose per day for a drug used for its main indication in adults [88]. This approach has been used to compare both individual drugs and classes of drugs, often in the UK [86,87,89-97].

It must be noted, however, that the FTIs calculated using prescription data are valid for prescription medications only and cannot be applied to over-the-counter drugs such as aspirin, paracetamol (acetaminophen), and ibuprofen. This limitation is not an issue when consumption data in DDDs or kilograms is available.

Death rate per millions of prescriptions was used to demonstrate that nitrazepam is a safer hypnotic than barbiturates [89,91]. By estimating BDZ death rates per diazepam equivalents, temazepam and flurazepam appeared more toxic than average hypnotics, and diazepam more toxic than average anxiolytics [98]. The latter finding, however, was attributed

to the concurrent use of alcohol. Tricyclic drugs as a group had a FTI (expressed as deaths per million prescriptions) higher than the average for all the drugs studied [86,92].

Among antidepressants, tricyclic anti- depressants (TCAs) were associated with a higher FTI and antidepressants introduced after 1973 with a lower FTI than the average for all antidepressants [86,93]. Mianserin, on the other hand, was found early on to have a lower FTI than the TCAs [86,87,94,99], and the selective serotonin reuptake inhibitors (SSRIs) were later shown to cause significantly less fatal intoxications than the TCAs in proportion to their consumption [17,100-102]. Moreover, most SSRI fatalities appear to involve co- ingestion of other substances [16]. Among the TCAs, amitriptyline, desipramine, doxepin, and dothiepin have been estimated to be the most toxic in several FTI studies [17,92,93,99].

Among the newer antidepressants, a higher FTI has been reported for venlafaxine than for the other serotoninergic agents [96].

1.6.2 Other Measures of Toxicity

Toxicity indices may also be calculated for ranking purposes by substituting the number of mentions on death certificates for the number of deaths attributed to the drug [103]. Drug toxicity has also been estimated by comparing attempted and completed suicides [104]. Besides fatalities, overdose-related hospital admissions, clinical representation, and outcome have been compared, as have seriousness of side-effects and possibility of drug interactions [105-107].

Yet another measure is related to the therapeutic index; de Jonghe and Swinkels classified antidepressants as ‛safe’ or ‛less safe’ according to whether the amount of drug prescribed for two weeks’ therapy can prove fatal. They considered an antidepressant safe when a two- week supply was not life-threatening in overdose [105,106]. TCAs would therefore be considered ‛less safe’ since significant symptoms can result from ingestion of three to four times the therapeutic daily dose, with a lethal dose only eight to ten times greater [16,108].

Although depressed patients should be allowed only limited access to antidepressant

drugs, as pointed out on several occasions [16,109,110], a reported fatal poisoning caused by citalopram alone involved ingestion of a dose equal to more than a six-month therapeutic supply [111]. In this case, the citalopram concentration determined in femoral blood was approximately 40 times greater than the highest concentration considered therapeutic [37].

In a British study, median fatal concentrations of certain drugs, namely anti- depressants, hypnotics, and volatile anesthetics, were shown to correlate with their aqueous solubility [112]. Inversely, drugs with high FTIs were reported to show more lipophilic character than the least toxic drugs. These drugs were thought to act via a nonspecific mechanism disrupting physiological processes in the lipo- protein membranes of the brain [112]. However, this approach was not applicable to nonnarcotic poisons which exert their lethal effects by very specific mechanisms. Nonspecific membrane- stabilizing activity, also termed a quinidine-like effect, was nevertheless offered as a cause of fatal poisoning and a mechanism of additive interactions [113]. Correlations were also been reported between antidepressant rank orders by FTI and LD50 in mice [95,100,114].

1.6.3 Sources of Bias in Toxicity Indices Toxicity index measures do not necessarily directly represent the inherent toxicity of a drug but can also be related to the indications and manner of use. Antidepressants, for instance, are consumed by people who have suicidal tendencies and who are thus at an elevated risk of death compared with nondepressed individuals [115]. It has also been suggested that prescribing practices may result in biased perceptions of toxicity differences between anti- depressants since dual-action antidepressants, such as venlafaxine, are prescribed to patients already at a relatively high risk of suicide, i.e.

patients whose depression has been resistant to narrow-spectrum serotonergic agents or whose initial symptoms suggest use of something other than a SSRI as a first-line drug [116,117].

Furthermore, some antidepressants may have several indications besides depression, e.g.

obsessive-compulsive behavior, bulimia, and nocturnal enuresis, conditions which generally

are associated with lower risk of drug abuse or self-harm than depression. Yet another consideration is that TCAs may actually be prescribed in subtherapeutic doses because in therapeutic doses, side-effects are common and may lead to noncompliance. For these reasons, TCAs may be ineffective in treating depression, whereas SSRIs, due to their mild side-effect profiles, can be taken in therapeutic amounts and with good compliance, resulting in improvement of the condition, and thus, a lower risk of suicide [109,118].

On the other hand, fatal drug poisonings are not always suicides, but may also occur by accident, and even an intentional overdose can lead to an unintentional death. A Finnish study reported 26% of unintentional deaths among fatal drug poisonings, with 76% of fatal antidepressant poisonings being suicides [102].

Similarly, the proportion of suicides did not exceed 80% in a Danish study on lethal antidepressant intoxications [119]. Another behavioral characteristic affecting poisoning statistics is that some prescription drugs, especially those with euphorigenic or anxiolytic properties, are abused more often than others, while some of the prescribed and acquired medications are never ingested. Fast-acting drugs with short-term effects have been estimated to have a higher abuse potential [120].

These aspects must be kept in mind when compiling and interpreting toxicity index data.

2 Pharmacogenetics

Drug response as well as absorption, disposition, metabolism, and excretion are affected by individual variation. Pharmacogenetics, the study of the heritable component of this variation, is a rapidly expanding field of science;

in the 2000s, hundreds of review articles have appeared on the subject [for recent reviews, see 121-128]. The increasing research interest is explained by the powerful new tools available for DNA analysis and by the observations that even a single nucleotide change in a gene may, due to altered dose-response relationships, lead to clinically significant differences between individuals. More specifically, expression of a functionally altered protein product or an altered

amount of a normal product can be expected to increase the likelihood of adverse drug reactions (ADRs) or of inadequate therapeutic outcome at normal drug dosages. When two or more variants of the same gene locus occur at a frequency of 1% or higher in a population, the gene is termed polymorphic [129]. Poly- morphisms may affect pharmacodynamics, e.g.

the structure of receptors, ion channels, and carrier proteins [121,130], but most of the currently available information concerns pharmacokinetics, especially enzymes involved in drug metabolism.

2.1 Drug-Metabolizing Enzymes

Drug-metabolizing enzymes (DMEs) act as a defense mechanism against foreign compounds.

These enzymes have evolved in animals during the course of interaction with plants [131]. Most exogenous substances enter the body via the gastrointestinal tract, where they are absorbed into the portal circulation, which transports them to the liver. DMEs are predominantly located in the liver, enabling efficient first-pass metabolism of foreign entities and thus constituting an important factor in the bioavailability of ingested drugs. Their body- protecting function comprises rendering a compound more easily excretable; in Phase 1 reactions, DMEs unmask or incorporate a polar, often oxygen-containing function in the compound, thereby creating a site for a Phase 2 reaction, which conjugates the compound with a highly polar agent. Genetic variation in DMEs makes it difficult to predict dosage, efficacy, or safety of a drug. Patients with an abnormal enzymatic status are prone to be predisposed to ADRs (Table 1) [132].

An individual’s response is also affected by several other factors, including age, gender, diet, concurrent medication, general health, lifestyle, and even education and socioeconomic status [133]. In pursuit of personalized medicine, phenotyping panels have been devised for the most common polymorphic DMEs [134-137].

2.1.1 Cytochrome P450 System

The superfamily of cytochrome P450 (CYP) enzymes is the most important metabolic system in Phase 1 [138]. These enzymes are heme-

Table 1. Possible consequences of abnormal enzymatic status depending on the properties of substrate drugs and expected metabolites involved in the reaction.

Enzymatic status

Substrate Expected product Consequence Normal

enzyme in normal quantities

+ Normal dose of drug

Æ Normal metabolite Æ Expected response (toxic effects rare)

+ Active parent drug Æ Inactive metabolite Æ Excessive response + Toxic parent drug Æ Detoxified metabolite Æ Toxic effects Lack of

functional enzyme, inhibited

enzyme + (Pro)drug Æ Active metabolite Æ Lack of response, undertreatment

+ Active parent drug Æ Inactive metabolite Æ Lack of response, undertreatment + Active parent drug Æ Toxic metabolite Æ Toxic effects Excess of

enzyme

+ (Pro)drug Æ Active metabolite Æ Excessive response

containing proteins that show a characteristic absorption maximum at 450 nm in reduced microsomes treated with carbon monoxide. CYP enzymes are located in the endoplasmic reticulum and expressed mainly in the liver, but also in extra-hepatic tissues such as the intestine, brain, and lung [128,139]. There are 57 sequenced human CYP genes and 58 pseudo- genes, the latter having mutated to such an extent that all variants have lost the ability to produce functional enzymes [140].

Human CYP forms are divided into families and subfamilies on the basis of similarities in amino acid sequence. The individual isozymes are very versatile and are often capable of catalyzing several types of oxidative reactions [138]. There is increasing evidence that CYPs are involved in chemical carcinogenesis and chemical-induced toxicity through metabolic activation, i.e. formation of reactive metabolites [139]. Families CYP1, CYP2, and CYP3 participate extensively in drug metabolism, with three of the major isozymes (CYP2C9, CYP2C19, and CYP2D6) being poly- morphic to a clinically significant degree [125].

The following sections focus on research involving the hepatic enzymes CYP2D6 and CYP2C19 and the polymorphic genes CYP2D6 and CYP2C19 encoding them.

CYP2D6

The CYP2D6 enzyme was originally called sparteine hydroxylase or debrisoquine hydroxylase due to two separate clinical trials where some of the subjects experienced ADRs because they were unable to hydroxylate these compounds [141,142]. CYP2D6 has been estimated to participate in the metabolism of more than 70 common drugs and 20-25% of all drugs in clinical use [125,128,138]. Most importantly, CYP2D6 metabolizes many psychoactive substances such as several antidepressants (TCAs, SSRIs, mianserin, mirtazapine, venlafaxine) and various antipsychotics (haloperidol, perphenazine, risperidone, thioridazine). CYP2D6 substrates also include opioids (codeine, dextro- methorphan, ethylmorphine, methadone, oxycodone, tramadol), β-blockers (metoprolol, propranolol, timolol), type 1 antiarrhythmics (flecainide, mexiletine, propafenone), and methylenedioxymethamphetamine (MDMA, i.e.

‛ecstasy’) [138]. The major reaction types catalyzed by CYP2D6 appear to be ring oxidation and O-demethylation. Substrates of CYP2D6 tend to be basic in character, with a protonatable nitrogen atom at a distance of 5-7 Å from the site of the oxidative reaction [143].

Figure 2. Chromosome 22, gene CYP2D6, and the sequence positions of the major CYP2D6 polymorphisms in the nine exons, with adjacent pseudogenes CYP2D7P and CYP2D8P shown.

According to current knowledge, CYP2D6 is the most polymorphic CYP gene [144], with more than 80 allelic variants documented to date [145,146]. In humans, the 4.2-kb region containing the CYP2D6 gene (MIM*124030) resides on the long arm of chromosome 22 (22q13.1), with two pseudogenes, CYP2D7P and CYP2D8P, in close proximity upstream (Figure 2).

In addition to interindividual variation, the CYP genes show interethnic variation.

Approximately 7% of the Caucasian population and 1% of Orientals carry a homozygously defective CYP2D6 genotype (gPM). They produce no active CYP2D6 enzyme, and thus, regarding CYP2D6 substrates, exhibit a poor metabolizer phenotype (PM). The major nonfunctional (null) alleles *3 (frame shift), *4 (splicing defect), and *5 (deletion of the entire gene) are responsible for approximately 90% of gPMs in Europeans [123]. Of the alleles associated with decreased CYP2D6 activity, *9,

*10, and *17 do not contribute significantly to drug metabolism in Caucasians [147], whereas the frequency of newly described allele *41 is approximately 8% [148]. The functional alleles

*1 and *2 are common in European, African, and Asian populations, with a combined allele frequency of ~71%, ~68%, and ~52%, respectively. Allele *4 is relatively frequent in Europeans (20%), while alleles *10 and *17 are

common in East Asian (38-70%) and Black African (24%) populations [147].

Individuals carrying two functional copies of CYP2D6, i.e. genotypically extensive metabolizers (gEMs), are predicted to have an extensive metabolizer phenotype (EM). Since the range of metabolic ratios (MRs) associated with one functional gene generally overlaps with that observed for gEMs [149,150], the carriers of one functional gene are typically also considered EMs [127,128]. A nonfunctional CYP2D6 allele in combination with a functionally deficient allele [128,151,152] is currently considered to predict an intermediate metabolizer phenotype (IM). Furthermore, inhibitors or high-affinity substrates of CYP2D6, such as quinidine, paroxetine, or fluoxetine, may temporarily convert gEMs to IMs or PMs, thus constituting a source of clinically significant drug interactions [128].

Expression of CYP2D6, unlike many other CYP genes, is noninducible, but during human evolution its metabolic capacity has been up- modulated by duplication and multiduplication of the entire gene. Some individuals may therefore carry extra copies of CYP2D6. Three or more copies of CYP2D6, constituting an ultra-rapid metabolizer genotype (gUM), is considered to lead to ultra-extensive production of CYP2D6 protein, thus predicting an ultra- rapid metabolizer (UM) phenotype [153].

However, the sensitivity of genotyping to predict the UM phenotype is low; common estimates of the frequency of duplication in European UMs range from 10% to 30%

[128,132]. For instance, a duplicated copy of CYP2D6 has been found in 4 of 18 UMs (22%) phenotyped with sparteine [150] and in 14 of 64 UMs (23%) phenotyped with debrisoquine [154]. Moreover, the phenotypes exhibited by gUMs and gEMs overlap [128,152].

Probe drugs used in CYP2D6 phenotyping include debrisoquine, sparteine, and dextro- methorphan [124]. In genotyping, the poly- morphic positions of the CYP2D6 gene are not detected directly in one step because of high sequence homology with neighboring pseudo- genes [155]. Therefore, a CYP2D6-specific fragment is first amplified by polymerase chain reaction (PCR) in parallel with two other possible fragments identifying a deleted CYP2D6 gene (*5) and a (multi)duplicated one (*1xN, *2xN, *4xN, *10xN, *35xN). The poly- morphic positions are then identified using such techniques as restriction fragment length poly- morphism (RFLP) analysis [155], multiplex single-base extension reaction (e.g. SNaPshot^TM) [156], real-time fluorometric melting point analysis [157], pyrosequencing [158], and oligo- nucleotide microarray technology (‛gene chips’) [151,159].

CYP2C19

The CYP2C19 polymorphism was originally discovered as a deficiency in (S)-mephenytoin 4'-hydroxylation [160]. In addition to mephenytoin, omeprazole has been used as a probe drug in CYP2C19 phenotyping [161].

Other CYP2C19 substrates include anti- depressants (TCAs, SSRIs, mianserin, moclobemide, venlafaxine), antipsychotics (clozapine, perphenazine), BDZs (diazepam, flunitrazepam, temazepam), β-blockers (metoprolol, propranolol), several proton pump inhibitors, dextromethorphan, phenytoin, and (S)-warfarin [138]. Substrates of CYP2C19 are often weakly basic in character and have two hydrogen bond donor/acceptor atoms. There are typically seven or eight chain atoms between the site of metabolism and the site forming a hydrogen bond. The major reactions catalyzed

by CYP2C19 include dealkylation and ring hydroxylation [143].

CYP2C19 is a large gene of more than 90 kb, including nine exons, on chromosome 10q24.1-q24.3 (MIM*124020). Approximately 2-3% of the Caucasian population [162] and 14-21% of East Asians [161] are CYP2C19 gPMs. Allele *2 (splicing defect) is the only common defective mutation in Caucasians (15%) and Blacks (17%). In addition to a high frequency (30%) of allele *2 (splicing defect), allele *3 (premature stop codon) is present in the Chinese at a frequency of 5% [161]. The mutations corresponding to these alleles are 681G>A in exon 5 [163] and 636G>A in exon 4 [164], respectively. They are readily detected by first amplifying a fragment covering exons 4 and 5 and then applying one of the various techniques mentioned above (section CYP2D6).

Of the CYP2C19 genotypes commonly observed in Caucasians, *1/*1 is considered to predict an EM, *1/*2 an IM, and *2/*2 a PM of CYP2C19 substrates [126,165].

2.2 Studies on Drug Metabolism

Human liver microsomes have been used extensively in studying metabolic poly- morphisms, but the ‛well-characterized’ human liver microsomes used in in vitro studies may contain enzyme variants that metabolize well the probe drug but not the drug being investigated [166]. Therefore, in vitro studies are not reviewed in detail in the following sections, focusing instead on the role of CYP enzymes in the metabolism of the opioid drug tramadol and TCAs, especially amitriptyline.

2.2.1 Tramadol Metabolism

Tramadol is administered as a racemic mixture of (+)- and (–)-trans-tramadol, i.e. (R,R)- and (S,S)-tramadol, respectively. CYP2D6 has been shown to convert tramadol to O-demethyl- tramadol (M1) in vitro [167,168] and in vivo [169]. The formation of (+)-M1 is important for the hypoalgesic effect because it has a higher affinity for opioid receptors than the parent drug [170]. Demethylation of tramadol in vitro is stereoselective, with (+)-tramadol being preferentially O-demethylated by CYP2D6 and (–)-tramadol N-demethylated by CYP3A4 [167].

NMe₂ O

H

OMe

NHMe O

H

OMe

NHMe O

H

OH NMe₂

O H

OH

NH₂ O

H

OMe

NH₂ O

H

OH

CYP3A4 CYP3A4?

tramadol M2 M3

CYP2D6 CYP2D6? CYP2D6?

CYP3A4? CYP3A4?

M1 M5 M4

Figure 3. Outline of main pathways of tramadol metabolism in Phase 1, starting from (R,R)- tramadol, i.e. (+)-trans-tramadol, the isomer with the highest affinity for the µ-opioid receptor.

Cyclohexyl oxidation is not shown.

Apparently, the N-demethylation product N-demethyltramadol (M2, nortramadol) is further N-demethylated to N,N-didemethyl- tramadol (M3) by CYP3A4 and O-demethylated to O,N-didemethyltramadol (M5) by CYP2D6, possibly followed by formation of O,N,N-tri- demethyltramadol (M4) from M3 via CYP2D6 as well as from M5 via CYP3A4 (Figure 3) [171]. At low tramadol concentrations, in vitro M1 formation predominates, while M2 is the major metabolite at higher concentrations [168].

2.2.2 Amitriptyline Metabolism

Metabolism of TCAs is well known [for reviews, see 123,172], for it was the subject of intensive research even before the discovery of CYP genes. In early studies, aliphatic ring hydroxylation of both amitriptyline [173,174]

and nortriptyline [175-177] in vivo correlated with polymorphic 4-hydroxylation of debrisoquine, whereas N-demethylation of amitriptyline did not [178]. The 2-hydroxylation reactions of imipramine [179] and desipramine [180] have also been suggested to be under the same genetic control as sparteine hydroxylation.

In parallel with amitriptyline metabolism, N-demethylation of imipramine to desipramine does not cosegregate with sparteine polymorphism [179].

Nortriptyline is the N-demethylated metabolite of amitriptyline, but also a drug on its own. Studies on nortriptyline metabolism are of great relevance here because nortriptyline formation is quantitatively the most important pathway in amitriptyline metabolism [173], with (E)-10-hydroxyamitriptyline (EHAT) formed to a lesser extent.

The major metabolite of nortriptyline is (E)-10-hydroxynortriptyline (EHNT) [176], or more precisely, the (–)-enantiomer of EHNT [181]. The enzymes mediating (Z)-10-hydroxy- metabolite formation are not known, but (Z)-10-hydroxyamitriptyline (ZHAT) and (Z)-10-hydroxynortriptyline (ZHNT) have been detected in vitro [182] and in vivo [181].

Nortriptyline is further demethylated to N-demethylnortriptyline (NNT) (Figure 4).

In vivo, the stereoselective formation of (–)-EHAT and (–)-EHNT in particular has been shown to depend on the activity of CYP2D6 [181]. In amitriptyline demethylation to nortriptyline, several enzymes have been implicated, namely CYP2C19, CYP3A4, CYP1A2, and CYP2D6 [183,184]. The results suggest a dominant role of CYP2C19 at therapeutic concentrations and involvement of CYP3A4 at higher concentrations.

HHH H

NMe₂

HHH H

NHMe

NMe₂

O H

NHMe

O H

HH H H

NH₂ CYP2C19

CYP2C19

amitriptyline nortriptyline NNT

CYP2D6 CYP2D6

CYP2C19

EHAT EHNT

Figure 4. Outline of amitriptyline metabolism in Phase 1, with formation of (E)-hydroxy- metabolites shown. The major pathway is indicated with bold arrows. N-oxide formation is not shown.

2.3 Clinical Pharmacogenetics

The discipline of clinical pharmacogenetics is aimed at individualized drug therapy, primarily by identifying patients at risk prior to initiating treatment (improved risk prediction). When the drug to be prescribed is exceptionally toxic or expensive, diagnostic phenotyping or genotyping, in addition to therapeutic drug monitoring, may help to improve safety and efficacy, i.e. to avoid ADRs and therapeutic failure. This will maximize medical and financial benefits and minimize the burden of medication both on the individual and on public health. Furthermore, in nonresponsive patients, phenotyping or genotyping may allow differentiation between ultra-rapid metabolism and noncompliance [185,186].

Phenotyping typically involves measurements in plasma or urine. After giving the patient a single oral dose of a probe drug, the concentrations of the unchanged drug and a relevant metabolite are analyzed, and the obtained MR is compared with a reference value or distribution determined in a large population.

The metabolism of a probe drug cannot, however, accurately represent the metabolism of another drug because several enzymes are typically involved in the metabolism of a drug, and the probe drug and the drug to be

prescribed may eventually behave differently [166].

Genotyping offers several advantages over phenotyping: the patient is not exposed to probe drugs; drawing one blood sample takes little time; genotyping can also be carried out post- mortem, when clinical phenotyping is no longer an option; and genotyping is very specific, with no interference from comedication. On the other hand, the latter is also a disadvantage since interactions arising from comedications are not taken into account.

In evaluating the concordance between tramadol metabolism, dextromethorphan pheno- type, and CYP2D6 genotype, only a modest correlation was found between the tramadol/M1 plasma ratio and the urinary dextromethorphan/

dextrorphan ratio in general, but when the subjects were segregated according to the number of functional CYP2D6 genes, a much stronger relationship was observed in gEMs [187]. The impact of the CYP2D6 genotype on 10-hydroxylation of nortriptyline [188-191] and amitriptyline [192], and the effect of the CYP2C19 genotype on N-demethylation of amitriptyline [192,193] have also been examined in volunteers and psychiatric patients.

The nortriptyline/10-hydroxynortriptyline ratio was shown to be influenced by CYP2D6 geno-

type and gender in one study [188], whereas in another, age and gender as factors did not reach statistical significance, with only the number of mutated CYP2D6 alleles being significant [191].

Theoretically, PMs are at an elevated risk of developing excessive plasma concentrations of certain drugs, for instance, of nortriptyline or desipramine, when these are given as such or are formed from amitriptyline or imipramine, respectively. Such instances have in fact been documented in several case reports [185,194- 196]. Furthermore, a German study recently found that gPMs were overrepresented among 28 patients reported to suffer from ADRs after taking CYP2D6-dependent antidepressant drugs, with an observed frequency of 29% compared with a frequency of 7% in a random German population. The same study also genotyped 16 nonresponders and found that duplication of CYP2D6 was overrepresented, with an allele frequency of 12.5% vs. 1.8% in a general German population [132].

ADRs are, however, relatively rare, probably because most drugs can be metabolized by several enzymes, so that when one is not fully active, complementary pathways may compensate, thus preventing harmful accumulation [187,192]. For instance, no evidence of an increased ADR rate was found in gPMs treated with fluoxetine or nortriptyline [197]. Nevertheless, on the basis of advanced study settings and calculations, preliminary genotype-based dose recommendations for certain antidepressants have been published [126,165,198]. Furthermore, since CYP2D6- inhibiting comedication may convert EMs to PMs, e.g. when paroxetine is combined with desipramine [199], it may also constitute a favorable interaction by converting an unresponsive UM to a responsive patient.

2.4 Postmortem Pharmacogenetics

Even before the advent of postmortem geno- typing, the possibility of a defective CYP2D6 genotype leading to death was discussed in a case report presenting two fatal poisonings involving imipramine and desipramine. Chronic accumulation of imipramine and desipramine, particularly of desipramine, was suspected since very low imipramine/desipramine ratios were

found. The patients had been comedicated with thioridazine and chlorpromazine, both of which inhibit the CYP2D6 enzyme responsible for desipramine metabolism [138]. However, CYP2D6 is also involved in the metabolism of thioridazine and chlorpromazine, and high concentrations of imipramine and desipramine can competitively inhibit CYP2D6. The possibility of a drug interaction was therefore also considered [200].

In 1999, CYP2D6 genotyping was demonstrated to be feasible using autopsy blood, with 22 suspected overdose fatalities and 24 controls successfully genotyped for CYP2D6 alleles *1, *3, and *4. No gPMs were found among the overdose cases, but CYP2D6 inhibitors were present in eight cases. However, the cases were not preselected according to known CYP2D6-catalyzed reactions, and the relevance of the investigated metabolic reactions to CYP2D6 was not discussed. Furthermore, no MRs allowing comparison between genotypes were calculated [201].

A case report of a toddler, deceased at the age of two years and genotyped for CYP2D6 postmortem, was published in 2000. The cause of death was determined as dextromethorphan poisoning following a therapeutic ingestion of cough medicine. Although the dextro- methorphan/dextrorphan ratio of 2.5 suggested slow O-demethylation, a reaction catalyzed by CYP2D6, the CYP2D6 genotype was that of an EM. No concurrent analytes were found in general drug screening [202].

In another case reported in 2000, the death of a child was investigated in depth when high blood concentrations of both fluoxetine and norfluoxetine were found in postmortem toxicology. The parents were first accused of homicide, but were vindicated by the results of genotyping, with DNA analysis revealing a homozygously defective CYP2D6 genotype. The interpretation of the results was therefore chronic accumulation of fluoxetine and norfluoxetine. In fact, the nine-year-old boy had been prescribed fluoxetine at 100 mg/day, a dose five times the DDD, and his medical history indicated several hospitalizations due to seizure episodes. The parents eventually filed a malpractice suit against the neurologist who had

prescribed fluoxetine at an exceptionally high dosage and yet failed to recognize the symptoms of toxicity in the patient [79,203].

In a Swedish series of 242 fatal drug intoxications, both CYP2D6 and CYP2C19 genotypes were determined and the genotype distributions compared with those in 281 controls (blood donors). The CYP2C19 geno- typing results in the autopsy cases were similar to those in the blood donors, but the prevalence of CYP2D6 gPMs in fatal intoxications was found to be lower (4.7%) than expected from the frequencies of these genotypes in the blood donors (8.5%), leading the authors to suggest that intoxication victims might exhibit a lower frequency of CYP2D6 gPMs than the general population [204]. No explanation has been offered for this observation.

In 15 fatalities involving oxycodone [205]

and in 21 involving methadone [206], CYP2D6 genotyping has been used to aid interpretation of postmortem toxicology results, although without calculating the relevant MRs.

In 53 Swedish autopsy cases involving citalopram, genotyping of CYP2D6 and CYP2C19 was combined with enantioselective analysis of citalopram enantiomers by chiral liquid chromatography. No gPMs were found for CYP2C19 and only 2 gPMs (3.8%) for CYP2D6. The authors suggested that pharmacokinetic interactions are likely to play a more important role than pharmacogenetic deficiencies in drug metabolism [207].

In summary, postmortem pharmaco- genetics is a relatively new area of research; the extent to which it will contribute to medicolegal investigations remains to be seen.

3 Drug-Alcohol Interaction

Drug-alcohol interactions have been widely investigated in animals and in clinical settings, especially with regard to psychomotor performance, but studies on human postmortem material are scarce. Alcohol is, however, a frequent finding in fatal poisonings. For instance, alcohol was detected in 47.7% of Swedish fatal poisonings in 1992-2002 [73], and BACs of 0.50‰ or more were detected in 50.4%

of Finnish fatal drug poisonings in 2000-2001

[72]. An overview of the current knowledge of the mechanisms of drug-alcohol interactions will be provided below and research on postmortem material summarized. Although most studies on drug-alcohol interactions relate to animals or living persons, clinical studies pertaining to psychomotor skills are beyond the scope of this review.

3.1 Alcohol Effects and Anesthetic Action A pharmacologically active drug typically has one or more known mechanisms and sites of action, e.g. binding directly to a specific proteinaceous receptor or enzyme. With regard to alcohol, while its effects are well known, the mechanism is less clear. Knowledge of this mechanism is, however, crucial in elucidating the interaction between drugs and alcohol. Many aspects of ethanol, recently reviewed by Jones [208], differ from medicinal drugs. Ethanol is often ingested in large quantities, has nutritional value, and is evenly distributed throughout the body. The molecular structure of ethanol is small and simple, with several potential physiological targets. In alcohol-related fatalities, anesthesia and CNS depression leading to respiratory failure are considered the mechanisms of major importance [208].

General anesthesia can be produced by a wide variety of chemical entities, including alcohols, alkanes, ketones, ethers, and inert gases, but the mechanism of action remains largely unresolved. An early effort to explain it, the Meyer-Overton hypothesis, based on the independent but similar findings of Meyer in 1899 and Overton in 1901, states that there is a correlation between anesthetic potency and oil solubility (i.e. hydrophobicity) of a compound [209]. Anesthesia was then proposed to occur when a critical drug concentration is achieved in the cell membrane, but the intramembrane volume was later found to be a better parameter than the intramembrane concentration for equal degrees of narcosis produced by different agents [210]. In the 1970s, the anesthetic site of action was concluded to be located within the neuronal membranes [211], and the physiological site of action of general anesthetics was thought to involve proteins rather than the lipid region of the membrane [212]. This proposition was based