The importance of sample size with regard to the robustness of postmortem reference values

The

importance

of

sample

size

with

regard

to

the

robustness

of

postmortem

reference

values

C.

Söderberg

a,b

,

A.

Tillmar

c,d

,

A.

Johansson

c

,

E.

Wernvik

b

,

A.K.

Jönsson

c

,

H.

Druid

a,b,

*

a

DepartmentofOncology-Pathology,KarolinskaInstitutet,Retziusv.3,KICampusSolna,17177Stockholm,Sweden

b

DepartmentofForensicMedicine,NationalBoardofForensicMedicine,Artillerigatan12,58758Linköping,Sweden

c

DepartmentofForensicGeneticsandForensicToxicology,NationalBoardofForensicMedicine,Artillerigatan12,58758Linköping,Sweden

d_{DepartmentofBiomedicalandClinicalSciences,FacultyofHealthandMedicalSciences,LinköpingUniversity,Sandbäcksgatan7,58235Linköping,Sweden}

ARTICLE INFO Articlehistory:

Received16January2020

Receivedinrevisedform1April2020 Accepted3April2020

Availableonline14April2020 Keywords: Postmortemtoxicology Referenceconcentrations Samplesize Power ABSTRACT

Evaluatingpostmortemtoxicologicalresultsisachallengingtaskduetomultiplefactorsaffectingblood concentrationsafterdeath.In ordertoimprovethediagnosticaccuracyincases ofsuspectedfatal intoxication different compilations of postmortem reference drug concentrations are often used. However,itisnotclearwhatconstitutesareliablepostmortemreferencevalue.

Thecurrentstudypresentsreferenceconcentrationsfor13substancesfromsevensubstancegroups accordingtoastandardizedprotocol.Thereferenceconcentrationsweregatheredfrom3767autopsy cases andsubdivided into intoxicationsbyone substanceonly(Group A,n=611),multi-substance intoxications(Group B, n=1355) andpostmortemcontrols, inwhich incapacitationbydrugswere excluded(GroupC,n=1801).Inparticular,thisstudypresentsstatisticalinformationabouttheprecision andconformitychangewithvarioussamplesizes.

Based on the presentdata >10 detectionsare usually needed, for thesubstances examined, to differentiatebetweenintoxicationcasesandcontrols.Repeatedsamplingsshowthatthemedianofsmall samples(N=5)hasahighvariation(normalizedinterquartilerange138–75%)andthatahighnumber ofdetections(N=>20)ineachgroupareneededtoreducethevariation.

©2020TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).

1.Introduction

Fatalintoxications, both intentionaland unintentional,are a global issue. Of all deaths attributed to self-harm worldwide intoxications with pharmaceutical drugs are common in the westernworld,whilepesticideintoxicationspredominateinthe developingworld[1,2].

Identifyingthecauseofdeathinintoxicationsiscomplicatedby detectionsofmultipledrugs,ofwhichseveralmayinisolationorin combinationhavecausedthedeath,orsometimesnot[3–5].One challengeforthepathologistortoxicologististonotonlyidentify potentiallyfatalconcentrationsbutalsotobeabletodeterminethe levels of substances that have not contributed to a fatal intoxication. In the postmortem setting evaluating drug

concentrations presents an additional challengedue to factors such as postmortem redistribution of drugs [6,7], bacterial metabolism [8] and the post mortem interval [9,10]. Forensic investigators have responded to these issues by compiling postmortemreferenceconcentrations.Mostoftenthese compila-tionsaredescriptivepresentationsofdetections[11–13]orspecific casereportswithanoftenlownumberofcases[14].

Small datasets are affectedbyuncertainty in how well the observed concentrations representthe_“true” collectionof con-centrationsforagivensubstanceanditstoxicologicaleffect(e.g. intoxicationandnon-intoxication)onapopulationlevel. Further-more,iftheobservedconcentrationsarescatteredovera broad interval together with overlapbetween intoxications and non-intoxications the uncertainty is even higher. The sample size neededtoprovideareliablepostmortemreferencevalueandthe addedbenefitoflargesamplesizeshavenotbeenexplored.

We have previously published both fatal and non-fatal postmortem reference concentrations based on the Swedish National Forensic Toxicology database using a standardized methodincludingarigorouscase-by-casereview[15–19].

Thepresentstudyprovidespostmortemreference concentra-tionsfor13substances,fromdifferentsubstancegroups,subdivided

* Corresponding author at: Department of Oncology-Pathology, Karolinska Institutet,Retziusv.3,KICampusSolna,17177Stockholm,Sweden.

E-mailaddresses:carl.soderberg@rmv.se(C. Söderberg),

andreas.tillmar@rmv.se(A.Tillmar),anna.johansson@rmv.se(A. Johansson),

emma.wernvik@rmv.se(E. Wernvik),anna.jonsson@rmv.se(A.K. Jönsson),

henrik.druid@ki.se(H. Druid).

http://dx.doi.org/10.1016/j.forsciint.2020.110292

0379-0738/©2020TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).

ContentslistsavailableatScienceDirect

Forensic

Science

International

intosingleintoxications,multi-drugintoxicationsandcontrolcases usingtheSwedishNationalForensicToxicologydatabase.Thissetof reference data was then used to study characteristics such as robustnessandreliabilityfordifferentsamplesizes.

2.Materialandmethods 2.1.Studypopulation

In Sweden all obvious or suspectedunnatural death, deaths withunclearcircumstancesorunclear identityof thedeceased shouldbereportedtothepolice.Thepolicemaythenrequesta forensicautopsyfromtheNationalBoardofForensicMedicine.At autopsyfemoralbloodisroutinelycollected(whenavailable)and willbesubjectedtoabroadtoxicologicalscreeningattheSwedish nationalforensictoxicologylaboratory,andpositivefindingswill beverified.Caseinformation, autopsyfindingsandtoxicological resultsareenteredintoacasemanagementsystemdevelopedto serveasarealtimedatabase[20].

Thepresentstudyconsideredallcasesinwhichaselectionof substances(alprazolam,amitriptyline,carbamazepine,citalopram, nitrazepam, olanzapine,oxazepam, oxycodone, phenytoin, que-tiapine,tramadol,verapamilandzolpidem)hadbeendetectedin femoral blood between 1992 and 2010. Postmortem reference values for olanzapineand quetiapine were generated fromthe same dataset that has in part been presented in a previous publication [19]. The included substances were selected to representdifferentsubstancegroups.

2.2.Samplingprocedures

The samplingprocedure hasbeendescribed in ourprevious publications[15–19].Inbrief,femoralbloodwascollectedfromthe femoralveinatautopsyandpotassiumfluoridewasaddedtoeach sampletoinhibitbacterialmetabolism. Thesampleswerethen transportedtothenationaltoxicologylaboratory,under refriger-atedconditions, where they were stored at 4C until analysis. Duringtransport(1–3days)thesamplesmighthavebeenexposed tohighertemperatures.

2.3.Analyticalmethods

Allanalysisduringthetimeperiodwereconductedusinggas chromatographywithnitrogenphosphorusdetection (GC-NPD). However,whiletheanalyticalmethodstayedthesamethelimitof quantification (LOQ)differedfor certaindrugsduringthestudy period(1992–2010).Tocorrectforthis,thehighestLOQforeach substanceduringthestudyperiodwasusedasacut-off.Fordetails regardingtheusedLOQ,seeTable1.

2.4.Selectionofpostmortemcontrols(groupC)

The postmortem control cases are a subset of the study populationandconsistofcasesinwhichoneofthesubstancesin thisstudywasdetected,butinwhichthedeathwasunrelatedto thesubstance.Hence,causeandmannerofdeathinthecontrol cases were selected to rule out that the deceased had been incapacitatedbythedrugdetected.Casesinwhich themanner and/orcauseofdeathwithpossibledamagetoareservoirorgan (heart,lungs,liver,intestines)wereexcludedtoavoid contamina-tionof theperipheral blood.Thiswas accomplishedbya setof exclusioncriteriabasedonthecauseofdeathcodedeterminedby the forensic pathologist who undertook the autopsy (see supplementalTablesS1andS2).

All cases were thensubject to a manual reviewbymultiple independentreviewers.Casesthatwerehospitalizedandsubjected to care that might affect blood concentrations were excluded. Suicidalsharpforceinjuriestosmallervesselswereexcludedtorule out the effect of slow exsanguination on blood concentrations [21,22].Casesofdrowningwereincludedonlyifthedeathwasa witnessedsuicidaljumpintoabodyofwater.Pulmonaryembolism wasonlyusedasanexclusioncriteriawithregardtoolanzapineand quetiapine asantipsychotics haveshown to increasetherisk of thromboembolism [23]. In unclear cases, or in cases in which unexpectedlyhighconcentrationswerefound,theautopsyreports andpolicereportswerereviewed.Acasewasonlyincludedinthe controlgroupifconsensusbetweenallreviewershadbeenachieved. Fig.1showsageneraloverviewoftheselectionprocess.

2.5.Selectionofintoxicationcases(groupAandB)

Intoxicationcaseswerethosecasesinwhichoneoftheincluded substanceswasfoundandinwhichthecauseofdeathhadbeen attributedtointoxicationbypharmaceuticaldrugs(eitheraloneor incombinationwithalcohol)bytheforensicpathologist responsi-blefortheautopsy.Allcaseswerethensubjecttoamanualreview bymultipleindependentreviewers.Thisreviewservedtoidentify theresponsiblesubstance(s)ineachcase.Thereference concen-trationsfromthecontrolgroup(groupC)wereusedasaguideline for the evaluation. As withthe control group, cases that were hospitalizedandsubjectedtohospitalcarethatmightaffectblood concentrationswereexcluded.Additionally,caseswith concomi-tant gas poisoning and cases where opioid intake could have contributedtothecauseofdeathwereexcluded.Autopsyreports andpolicereportswerereviewedinunclearcases.

Casesofsinglesubstanceintoxicationinwhichthecontribution ofethanoland/orothersubstanceswereexcludedwereincludedin group A. Cases of multi-substance intoxication or cases of substance intoxicationin combination withalcohol (BAC >1

m

) wereincludedingroupB.Fig.1showsageneraloverviewofthe selectionprocess.

The A, B, and C groups are mutually exclusive. A single individualcanonlycontributetoa singlegroup(forexamplea singlecasecannotbothbeclassi_fiedasAandB).However,asingle casecanbeclassifiedasgroupBorgroupCformultiplesubstances, as the same intoxication case might contain multiple key substances and the same control case can involve more than one of the included substances. Therefore, the number of individualsincludedinthestudydoesnotnecessarilymatchthe numberofcasesingroupBandgroupC.

2.6.Statisticalanalysis

TheconcentrationsingroupA,BandCdidnotshowanormal distribution.Thus,onlynon-parametricmetricswereappliedfor allstatisticalanalyses.

Table1

Thehighestlimitofquantification(LOQ)usedforeachsubstanceduringthestudy period(1992–2010).

Substance HighestLOQ(mg/g)

Alprazolam 0.02 Amitriptyline 0.05 Carbamazepine 0.5 Citalopram 0.05 Nitrazepam 0.05 Olanzapine 0.05 Oxazepam 0.1 Oxycodone 0.05 Phenytoin 0.5 Quetiapine 0.05 Tramadol 0.05 Verapamil 0.1 Zolpidem 0.05

Foreachsubstance,the impactofthesizeofthedatasetwas studiedbyrandomlydrawing(withreplacement)Nsamplesfromthe originalsetofconcentrationsineachgroup.Nvariedfrom1upto50. Foreachsamplesize(N),thesamplingprocesswasrepeated1000 times.Thesamplesize(N)waslimitedtolessthanthetotalnumberof casesineach groupforeachsubstance(e.g.a substancewith10 detectionsinagroupwouldonlycontributedatatosamplesizes<10), inordertoreducetheamountofresamplingintherandomdraw.In eachsuchsamplinground,themedianconcentrationwasregistered. Themedianofallmediansandtheinterquartilerange(IQR)ofthe medianswere studied.Further,a normalized IQRwasstudied by dividingIQRwiththemedianconcentrationforeachvalueofN.

Ifasignificantdifferencewasobservedbetweenthedifferent groups inthe originaldata setof concentrations,the powerto

detectthatdifferencewithsmallersamplesizeswasstudied.Since thedatawerenotnormallydistributedtheMann-WhitneyUtest wasappliedtostatisticallycomparetheconcentrationsfortheA,B andCgroupsforeachdrug.Ap-valuebelow0.05wasconsideredto beastatisticallysignificantdifferenceandtheproportionofsuch differenceswasestimatedforeachsamplesize,N,basedonthe 1000replicates,whichrepresentedthepower.

AllstatisticalanalyseswereperformedusingtheRsoftwarev. 3.5.0[24].

2.7.Ethicalconsiderations

ThestudywasapprovedbytheRegionalEthicsReviewBoardin Linköping,Sweden,No2012/343-31.

Fig.1.Schematicworkflowfrominitialdatabasesearchtofinishedreferencevalues. GroupA=single-substanceintoxication.

GroupB=multi-substance(includingalcohol)intoxication. GroupC=controlcases.

Table2

Numberofdetectionsandnumberofincludedandexcludedcases. Substance Totaldetectionsin

database(1992–2010)

Includedcases Excludedcases Detectionsper includedcase Excludedcases (%oftotaldetections) Alprazolam 922 194 728 4.8 79.0 Amitriptyline 1109 459 650 2.4 58.6 Carbamazepine 2526 206 2320 12.3 91.8 Citalopram 4284 1363 2921 3.1 68.2 Nitrazepam 1910 568 1342 3.4 70.3 Olanzapine 473 138 335 3.4 70.8 Oxazepam 823 237 586 3.5 71.2 Oxycodone 188 49 139 3.8 73.9 Phenytoin 622 38 584 16.4 93.9 Quetiapine 133 41 92 3.2 69.2 Tramadol 1737 519 1218 3.3 70.1 Verapamil 398 68 330 5.9 82.9 Zolpidem 1177 337 840 3.5 71.4 Summary Total 16,302 4217 12,085 Mean 5.3 74.7

3.Results 3.1.General

Duringthestudyperiodtoxicologicalanalysisinfemoralblood wasperformedin93623autopsycases,13925(14.9%)ofwhich werepositiveforoneormoreoftheincludedsubstances.

After application of exclusion/inclusion criteria and manual reviewatotalof3767individualswereincludedinthestudy.Of these611(16%)wereincludedingroupA,1355(36%)ingroupB and1801(48%)ingroupC.Themedianagewas54yearsacrossall groups,andwas54yearsingroupA,51yearsingroupBand57 yearsingroupC.Thesexofthedeceasedwasmalein2603cases

(69.1%)acrossallgroups,andtheproportionofmaleswas62.7%in groupA,56.8%ingroupB,and80.6%ingroupCrespectively.

Table 2 presents the total number of detections for each substanceandthenumberofdetectionsincludedinthestudyfor each substance.Notethata singleindividualmaycontributeto morethanonedetectionoftheincludedsubstances.Amajorityof cases(74.7%)inwhichtheincludedsubstanceshadbeendetected wereremovedafterapplicationofexclusioncriteriaandmanual review.Acrossallsubstancesanaverageof5.3detectionsinthe databasewereneededforeachcaseincludedingroupA,BorC. Carbamazepineandphenytoinpresentedahighdegreeofexcluded detections (91.8% and 93.9% respectively) and therefore had a higheramountofdetectionsperincludedcase (12.3and16.4%,

Table3

Femoralbloodconcentrations(mg/g)ofsubstances.

Substance Group N 10th_percentilea _{Median 90}th_percentilea _p-value

Alprazolam A 6 0.17 0.30 0.40 vsB0.995 B 113 0.11 0.20 1.60 VsC<0.001 C 75 0.02 0.05 0.16 VsA<0.001 Amitriptyline A 159 0.90 2.40 6.86 vsB<0.001 B 175 0.60 1.60 5.46 VsC<0.001 C 125 0.10 0.20 0.40 VsA<0.001 Carbamazapine A 24 16.6 36.5 80.0 VsB0.004 B 46 12.0 19.5 79.5 VsC<0.001 C 136 1.05 5.00 11.0 VsA<0.001 Citalopram A 70 1.09 4.45 21.2 VsB<0.001 B 377 0.70 1.40 9.04 VsC<0.001 C 916 0.10 0.30 0.70 VsA<0.001 Nitrazepamb _{A 69 0.40 1.13 3.20 VsB<0.001} B 289 0.30 0.60 1.70 VsC<0.001 C 210 0.06 0.10 0.28 VsA<0.001 Olanzapine A 10 0.29 0.55 2.71 VsB0.138 B 37 0.20 0.40 0.90 VsC<0.001 C 91 0.06 0.10 0.20 VsA<0.001 Oxazepam A 5 2.00 3.50 6.10 VsB0.07 B 109 1.08 1.70 3.64 VsC<0.001 C 123 0.10 0.30 0.90 VsA<0.001 Oxycodone A 19 0.50 0.70 5.22 VsB0.508 B 26 0.40 0.65 2.60 VsC0.007 C 4 0.10 0.10 0.40 VsA0.007 Phenytoin A 2 35.0 38.5 43.0 VsB0.8 B 3 21.0 26.0 80.0 VsC0.018 C 33 1.50 6.00 18.2 VsA0.041 Quetiapine A 9 0.70 2.30 22.0 VsB0.59 B 17 0.96 5.60 42.8 VsC<0.001 C 15 0.09 0.20 0.66 VsA<0.001 Tramadol A 161 2.00 5.10 25.0 VsB0.575 B 184 1.80 4.75 35.7 VsC<0.001 C 174 0.10 0.50 1.80 Vs<0.001 Verapamil A 16 2.10 4.10 12.1 VsB0.066 B 9 0.60 2.80 15.0 VsC<0.001 C 43 0.10 0.20 0.50 VsA<0.001 Zolpidem A 51 0.60 1.50 5.10 VsB0.08 B 205 0.40 1.00 3.36 VsC<0.001 C 81 0.06 0.16 0.50 VsA<0.001

GroupA=single-substanceintoxication.

GroupB=multi-substance(includingalcohol)intoxication. GroupC=controlcases.

a

Ifn<10thenminimumandmaximumvaluesarepresented.Otherwisethe10thand90thpercentilesarepresented.

b

Fornitrazepamtheconcentrationsofnitrazepamanditsmetabolite(7-amino-nitrazepam)wereaddedtogetherforallgroups. 4 C.Söderbergetal./ForensicScienceInternational311(2020)110292

respectively)thanaverage.Incontrast,amitriptylineonlyexcluded 58.6%ofthetotaldetectionsandonlyhad2.4detectionsforeach includedcase.

3.2.Postmortemreferencevalues

Table3liststhefemoralbloodconcentrationsoftheincluded substancessubdividedintogroupsA,BandC.

There were significant concentration differences (p<0.05) betweenintoxications(groupAandgroupB)andcontrols(group C)forallsubstances.However,therewasnosignificantdifference between single intoxications (group A) and multi-substance intoxications (group B) for alprazolam, olanzapine, oxycodone, phenytoin,quetiapine,tramadolandverapamil.

SupplementalTableS3providesexamplesoftheconcentration distributionsacrossgroupsA,BandCforcitalopram, carbamaze-pineandoxazepam.

3.3.Statisticalevaluationofnumberofcasesneededforstable referencevaluesineachgroup

Fig.2displaysthevariationofmedians,whenthemedianshave beensampled1000 timesfromthecasesineachsubstance and group.SupplementalTablesS4andS5presentsubstancespecific variationsofthecorrespondingmedians.TableS4presentsthisdata numericallyina95%interval,andTableS5presentsthenormalized IQR fromthe same data across all substances and groups. The normalizedIQRwithasamplesizeof5or10cases,averagedamong allgroups,washigh(75.3%and47.0% respectively)comparedto20or 30casesineachgroup(32.3%and27.7%respectively).Itisworth notingthattherateofimprovementwasreducedabove20–30cases ineachgroup.Inaddition,evenwith50casesineachgroup,there was stillamoderateamountofvariation (17.5%).Asan example,Fig.3 shows the impact ofthe sample size on the distributionofthe median valuesforatypicalsubstance(citalopram).

Fig.2.Typicalnormalizedinterquartilerange(IQR)ofthemedianvalues,takingallsubstancesandgroupsintoaccount,asafunctionofthesamplesize,N.

Fig.3.Boxplotofthedistributionofmedianvalueswhenperformingreplicativesampling(n=1000)ofsamplesizesN=2,4,6...50forcitalopram(GroupAcases).The referencemedianvalueinthecompletedatasetis4.5.Thedeviationfromtheexpectedmediandecreasesasthesamplesizeincreases.Intheboxplot,thewhiskersare1.5 timestheinterquartilerange(IQR),andoutliersareshownascircles.

Supplemental Table S6 presentspower as a function of the numberofcasesineachgroup.Asanexample,Fig.4presentsthe powertodetectdifferencesbetweengroups(A,BandCcases)for thesubstancecitalopram,asafunctionofthesamplesize,N.In general,separatinggroupAfromgroupBisoftendifficultrequiring eitheralargeamountofcases,ornotbeingpossibleatall.However, separatingtheintoxications(groupsAandB)fromthepostmortem control(groupC)requireslesscases,with10casesprovidingahigh degreeofpower(>0.95)inallbutonecase.Theexceptionbeing separatinggroupBandgroupCforolanzapine,whichrequires17 casestoreachahighdegreeofpower.

4.Discussion

4.1.Casesneededforreliablereferencevalues

In this studywe providereference valuesfor 13 substances usingastandardizedmethodofselectionand evaluation.Akey featureof thecurrentstudyand thepreviousstudiesusingthe samemethod[15–19] is notonly thepresentationofreference concentrationsof“postmortemcontrols”and intoxicationcases, butalsothenumberofdetectionsofeachsubstanceonwhichwe baseourresearch.

It haspreviouslynot beenknownhowmanydetections ofa substancethat are neededto producereliable referencevalues using ourmethod. Indeed, the impactof populationsize with regardtopostmortemreferencevaluesingeneralhasonlybeen discussedbriefly[14].However,itshouldbenotedthatthecurrent international recommendations for clinical laboratories is 120

samplesforeachpartition(suchasageandsex)inordertoboth determinethecentral95%ofthedistributionaswellastoprovide 90% confidencelimits ontheendpoints[25,26]. In theforensic settingitisverydifficulttogeneratetheamountofcasesneeded thatarebothreliable,takesintoaccountnotonlyclinicalaspects butalsoforensicaspectssuchaspostmortemredistributionand thepostmorteminterval.Thepresentstudyprovidesafirststepin discussing sample sizeand reliablereference concentrationsin postmortemtoxicology.

FromTable2it canbeseenthatapproximately74.7% ofthe detectionsforagivensubstanceislostwheninclusion/exclusion criteriaareappliedoruponmanualreview.Oncedetectionshave beenorderedintogroupsA,BandCthereisaneedtoensurethat thenumberofincludedcasesaresufficientforareliablereference value.AscanbeseeninTableS6tencases(n=10)areinalmostall cases sufficient for highpower(>0.95)when separating intox-ications (group A and B) from non-intoxications (group C). Separatingsingleintoxications(groupA)frommulti-intoxications (groupB)ismuchharder,andisnotalwayspossibleevenwith50 casesineachgroup.Thisistobeexpectedsincetheunderlying differenceinantemortemactionbetweentheintoxicationgroups isslight,withthesamemodusoftakingalargeamountofavailable medicationprevalentinbothgroups.

Overmultiplesamplingsfromthelargerpopulationitcanbe seen that the variation of the normalized IQR decreases with increasedsamplesize,andthattherateofimprovementstabilizes aroundasamplesizeof 20–30.Ofspecial noteistheveryhigh variationinlowsamplesizes(<5),indicatingthatforsubstancesin thisgroupmuchcanbegainedbyincreasingthenumberofcases.

Fig.4. Thepowertodetectasignificantdifference,betweengroups(A,BandCcases)forthesubstancecitalopram,asafunctionofthesamplesize,N. 6 C.Söderbergetal./ForensicScienceInternational311(2020)110292

Howmuchvariancethatcanbeacceptedinasamplewithregardto alargerpopulationisdifficulttoaddress,andislikelydependent ontheextentofoverlapbetweennormalandtoxicconcentrations. 4.2.Comparisonwithresultsfromourpreviousstudies

Allsubstancesexceptoxycodoneandtramadolhavepreviously been evaluated using our method in one of our previous publications [15–17,19]. However, except for olanzapine and quetiapine,which in ourprevious study[19] already contained casesfrom1992–2010,thelongerstudyperiodinthepresentstudy hasresultedinanincreasednumberofincludedcases.Comparing thepresentresultswiththeearlieriterationsofreferencevalues thecorrespondingmedianvaluesarethesame,orwithonlyslight variation (10%), for alprazolam (group A and group C), amitriptyline (all groups), citalopram (group B and group C), nitrazepam(all groups),olanzapine (all groups), oxazepam (all groups),verapamil(groupAandgroupC)andzolpidem(groupA andgroupC). Allthesevariationsarewell withintheexpected rangebasedonthesamplesize,asdescribedabove.However,in the case of some substances and groups there are more pronounceddifferences.

Inthecaseofalprazolam(groupB)themedianconcentration has increased by 25.0% (0.16

m

g/g to 0.20

m

g/g) between the previousandthepresent study.Thenumber ofcases increased from67to113.ThemedianofgroupBalprazolamconcentrationsis slightly lower than expected based on the sample size of our previousstudy[16],buttheincreaseto0.20

m

g/ginthepresent studyiswithin theexpectedrange ofvariation.With regardto citalopram, the median of group A hasdecreased 31.5% (from 6.50

m

g/gto4.45

m

g/g)andthenumberofincludedcasesincreased from50to70.However,basedonthesamplesizeofouroriginal study [17], this variation can be expected. Of note is that the standardizedIQRforalprazolamgroupBandcitalopramgroupAat a sample size of 50 was high (40.0% and 36.0% respectively) comparedtotheaverage(17.5%)suggestingalargespreadofthe dataevenatlargesamplesizes.

For carbamazepine themedian of group A decreased 18.9% (from45.0

m

g/gto36.5

m

g/g),themedianofgroupBincreased 39.3%(from14.0

m

g/gto19.5

m

g/g)whencomparedtotheoriginal study[15].Thenumberofcaseshaschangedsignificantlyingroup A(from7to24)andgroupB(from9to46).

Inthecaseofphenytointhereareshiftsinallgroupscompared totheoriginalstudy[15].ThemedianofgroupAdecreased10.5% (from43.0

m

g/gto38.5

m

g/g),themedianofgroupBdecreased 67.5% (from 80

m

g/g to 26

m

g/g) and the median of group C increasedby20%(from5

m

g/gto6

m

g/g).However,thenumberof casesincludedingroupAandgroupBarestillsmall(2and3cases, respectively), which can explain the variability of the median. GroupCshowedamoresignificantincreaseinthenumberofcases comparedtotheoriginalstudy(from14to33).Forverapamilthe groupBmedianconcentrationincreased47.4%(1.9

m

g/gto2.8

m

g/ g) in the present study compared to our previously published values[15].

However,forzolpidemthemedianof groupBincreasedby 11.5%(from0.9

m

g/gto1

m

g/g)andthenumberofcasesincreased from 148 to 205 compared with the previous study [16]. In carbamazepinegroup C themedian of groupC increased11.1% (from 4.5

m

g/g to 5.0

m

g/g) and the number of included cases increasedfrom56to136comparedwiththepreviousstudy[15]. Thesecasesshowthatslightvariationcanbeexpectedevenwith largesamples,howeverchangesonthissmallscalemightalsobe explainedbyinherentmeasurementuncertainty.

With regard to quetiapine and olanzapine there are slight changesinthenumberofcasescomparedtoourpreviousstudy. QuetiapinelostonecaseingroupB(18to17)andgroupC(16to15)

andolanzapinelostthreecasesingroupC(94to91)[19],which hasimpactedthemedianconcentrationsofquetiapineingroupB (from 3.85

m

g/g to 5.60

m

g/g) and group C (from 0.15

m

g/g to 0.20

m

g/g),eventhoughtheevaluatedmaterialisthesame.With regardtoquetiapinegroupAthemedianhasdecreased(from8.1

m

g/gto2.3

m

g/g)eventhoughthe10th/90thpercentilesandnumber ofcasesareunchanged.Thesevariationsareexpectedbasedon samplesize.Thevariationsinnumbersandmediansareprobably due to cases being evaluated differently because of additional reference material (such as updated reference concentrations regardingco-detections)beingavailableforthepresentstudy.

Ingeneral,thelargervariationsofthemedian(_20%),withthe exceptionofsubstancesandgroupswithhighstandardizedIQRat highsamplesizes(alprazolamgroupBandcitalopramgroupA), canbeexplainedbytheinherentvariabilityatthelowersample sizesinthepreviousstudies.However,caremustbetakenevenat large sample sizes due to some substances having a large concentrationspread.

4.3.Comparisonwithotherreviewsandcompilations

Inthescientificliteraturetherearenumerouspublicationsin whichpostmortemconcentrationsoftheincludedsubstancesare presented.Thefollowingsectionwilldiscusspreviouslypublished materialsubdividedbythetypeofstudy.

Wehavechosennottoreviewindividualcasereportsorsmall caseseriessincetheirindividualvaluewithregardtopostmortem referenceconcentrationscanbeconsideredlow[27].Thereview belowfocusesonlargercompilations.

Descriptivecompilations-Thereareafewcompilationsbasedon adescriptiverepresentationofdetectionsofasubstance,suchas the compilations by Launiainen and Ojanperä [12], Ketola and Ojanperä[13]andJones[11],withhighconcentrationpercentiles (90th,95thand97.5th)presentedseparately.Oneadvantageofthis approachisthatthereareoftenahighnumberofdetectionsfor eachsubstanceasnodetectionsareexcluded,whichincreasesthe reliabilityoftheresults.However,inthistypeofpresentationthere isnodistinctionbetweenintoxicationandnon-intoxicationcases, whichisaweaknesssinceitisdifficulttoknowwhereonegroup ends and anotherbegins.These compilations insteadshowifa given concentration can be considered highin relation to the group.Itisimportanttonotethatthisapproachisdependenton theLOQforeachsubstance,ahighLOQcanshiftthepopulation towardsintoxicationcaseswhilealowLOQcandotheopposite. Theconcentrationsseeninthehighpercentilesinallthreestudies alignwellwiththeconcentrationsseeninourintoxicationgroups, withthe90thpercentilein theabovestudies beingwithinthe 10th/90thpercentilerangeofourgroupAcasesforabouthalfthe substances(exceptalprazolam,carbamazepine,nitrazepam, oxaz-epam, oxycodone, phenytoin and quetiapine) and ourgroup B casesformostsubstances(exceptnitrazepam,oxazepam, phenyt-oinandquetiapine).

There are,inaddition,a fewcompilations thatpresentlarge tabulationsofstatedtherapeutic,toxicand_“lethal”concentrations [28,29].Inusingthesecompilationsareviewermustproceedwith caution. While thelist of includedsubstances is extensive, the postmortemreliabilityofthereferencevaluesarenotalwaysclear (asdataregardingsamplesiteandthenumberofdetectionson whichthedataisbasedislacking),whichlimitstheirusefulnessin postmortemcasework.

Evaluated compilations - In forensic practice it is not only importanttoconfirmintoxicationsbutalsotobeabletorulethem out. Thus, there need to bereference concentrations regarding bothpostmortem“normal”(i.e.whichhavenotcontributedtothe causeofdeath)and“toxic”concentrations.Extractingthesecases from a database and reviewing whether or not they are

intoxicationscandothis.Usingacombinationofdetailsavailable in case reports and theextensive databasesearch seen in the databasecompilations,aimstoprovidethebestofbothworlds. Therearesomepublicationsthathavetakensomevariantofthis approach,inwhichacaseserieshasbeensubdividedintonatural deathsandintoxicationgroups,whichwillbediscussedbelow.As withcasereports,itisimportanttorememberthatstudieswitha lownumberofcasesineachgrouphastobeinterpretedcarefully. StudiesbyJonesandHolmgren[30,31]providereferencevalues fornon-intoxicationcausesofdeathforalprazolam(median0.05

m

g/g)andzolpidem(median0.13

m

g/g)thatarewellinlinewithour groupCconcentrations(median0.05

m

g/gand0.16

m

g/g respec-tively).However,theconcentrationsofalprazolamandzolpidem found in intoxication cases (median 0.06

m

g/g and 0.3

m

g/g respectively)are difficulttoevaluate since itis not clearifthe substancesareassessedascontributingtotheintoxicationornot. Theirlowermedianconcentrationscomparedtothepresentstudy (group B median 0.2

m

g/g and 1.0

m

g/g for alprazolam and zolpidem respectively) could be explained by inclusion of intoxicationsinwhichthepresenceofalprazolamandzolpidem aremerelyincidental.

AstudybyWolfetal.[32]concerningalprazolamrelateddeaths between2001and2003compiledreferencevaluesregardingboth combined drug intoxications (mean 0.10

m

g/mL, range 0.01– 1.20

m

g/mL)andnon-intoxicationdeaths(mean0.04

m

g/mL,range <0.002–0.08

m

g/mL) which align well with our own (group B median0.2

m

g/gandgroupCmedian0.05

m

g/g).However,their singleintoxicationsdeathspresentlowerconcentrations(0.01

m

g/ mL and 0.03

m

g/mL respectively) than our single intoxications (median0.3

m

g/g).Thenumberofcasesinthesingleintoxication grouparelow(n=2),andthusmustbeinterpretedcarefully,which alsoappliestoourowndatasincethepresentstudyonlyincludes sixsingleintoxications.Thedeathsintheirnaturalgroup(n=12) andcombineddrugtoxicitygroup(n=87)aremorerobust.

Withregardtocitalopram,astudybyJonassonandSaldeen[33] presentshigherconcentrationsinthecasesinwhichcitalopram had not contributed to death (range: 0.7–1.5

m

g/g, median: 1.08

m

g/g), than those foundin ourgroupC (median0.3

m

g/g). Thisisinterestingsincethestudyusedthesamedatabaseasthe present study. However, they excluded cases in which the concentrationofcitalopramwasbelow0.7

m

g/g,which explains thehigher rangeand median concentrationin theirstudy. The concentrationsin single citalopramintoxications (n=5, median 11

m

g/g)andcontributingcitalopramintoxications(n=9,median 1

m

g/g)doesnotalignwithourgroupAandgroupBconcentrations (median4.45

m

g/g and 1.4

m

g/g respectively) directly.However, thevariationcorrespondstothatseeninlowsamplesizesinthe presentstudy.

Inseveralpublicationsregardingoxycodone[34–37]thereare casesinwhichoxycodonehascontributedtointoxicationdeaths witha concentration below ourgroup B and groupA concen-trations.InthestudybyConeetal.[35]theypresentmediansin combinedtoxicitygroups(0.19–0.58

m

g/mL)whichislowerthan ourowndata(groupBmedian0.65

m

g/g),howeverthebloodin theirstudy was collected fromvarious anatomical sites which impactsinterpretation.InthestudybyWolf[37]therangeinsingle oxycodone intoxications (0.21–4.71mg/L) and combined drug toxicity (0.025–17.5mg/L) include cases more in line with our group C median (0.1

m

g/g). Baker and Jenkins [34] present a medianconcentrationinintoxicationdeaths(0.56mg/L)whichare anagreementwithourintoxicationcases(groupBmedian0.65

m

g/gand 10thpercentilegroupA0.5

m

g/g).However,theirrange (0.01–36.54mg/L)includescaseswithlowconcentrationscloserto ourgroupC(median0.1

m

g/g)wherecircumstantialinformation probablyhasplayedaroleforthediagnosis.BakerandJenkinsalso presentalargebodyofnon-intoxicationsdeaths(n=135),which

greatlyexceedsourown(n=4).Inlightofourresults regarding variationin smallsamplestheirmedian (0.26mg/L) isprobably closertothe“truth”thanourown(0.1

m

g/g).Similarlythestudyby Ogle[36]presentslargegroupsofdeathswhereoxycodonewasa contributingfactor(n=117,mean0.48mg/L)andincidentalfinding (n=38,mean0.16mg/L),whichshouldbetakenintoaccountwhen evaluatingthemedianandpercentilesinourcases.Asmentioned above,therearecasesinwhichoxycodonehasbeenclassifiedas contributing which corresponds toourgroupC concentrations, whichisprobablyduetocircumstantialfactorsintheindividual case.

A study by Pilgrim et al. [38] regarding quetiapine reports concentrations in quetiapineassociated deaths(n=114, median 0.7mg/L,range0.02–110mg/L),whichiswellbelowourgroupB concentrations(median5.6

m

g/g).However,theirmaterialisnot onlycomprisedofintoxicationsinwhichquetiapinehasplayeda keyrole,butalsoincludementionsofquetiapineinautopsy-or policereportsaswellasincoronersfindingswhichbroadenstheir materialandcouldexplaintheirlowermedianconcentration.With regard to natural deaths their data (n=60, median 0.25mg/L) alignswellwithourown(median=0.2

m

g/g)andthevariationis withinexpectedlimits.

Numerous publications fromDenmark[39–42] usea similar classificationsystem asours,withslightdifferencesin howthe controlgroupis selected.Inastudyconcerningolanzapine[39] their control group median is lower than ours (0.064

m

g/g comparedto0.1

m

g/g)andslightlyoutsidethevariationwewould expectbasedonourdata,whichisprobablyduetoourLOQof0.05

m

g/g pushingour median upwards. Withregard to multi-drug intoxicationsourresultsalignwell,andwhiletheirstudyincludes fewsingledrugintoxicationcases(n=2)theirresults(1.8

m

g/gand 3.6

m

g/g)arecomparabletothe10th/90thpercentilerangeofour groupAcases(0.29–2.71

m

g/g).Inthestudyconcerningcitalopram [40]theirresultsalignwellwithourownwithregardtocontrols (0.31

m

g/gand0.3

m

g/grespectively)andmixedintoxications(1.3

m

g/gand1.4

m

g/grespectively),andtheirsingledrugintoxication case(3.0

m

g/g)iswellwithinthe10th/90thpercentilerangeofour groupAcases(1.09–21.23

m

g/g).Thevariationsbeingwithinthe expectedrangebasedonthesamplesizedataofthepresentstudy. InthestudybySkovetal.concerningquetiapine[41]themedianof their control groupis lower than ours (0.15

m

g/g compared to 0.2

m

g/g), which, again,is probably becauseof our higher LOQ (0.05

m

g/g).Themedianoftheirmixedintoxicationgroup(n=5, 3.19

m

g/g) is well below ours (n=17, 5.6

m

g/g) but within the variationbasedonsamplesizeweseeinourpopulationwitha similarsamplesize.Theirsingleintoxicationcase (8.99

m

g/g)is wellwithinthe10th/90thpercentilerangeofoursingle intoxica-tion group(0.7–22.0

m

g/g). In a second studyfeaturing, among others,quetiapine[42]anoteworthydetailisthatthemedianin themixedintoxicationgroup(n=7,1.3

m

g/g)islowerthaninthe previousstudy[41] anddivergingmorethanexpectedfromthe oneinthepresentstudybasedonthesamplesize.

In general, this type of approach provides reference values which can both aid in con_firming and excluding intoxications. However, it is important to know how intoxications and non-intoxicationshavebeenclassifiedwhenusingthereferencedata andcomparingitwithotherstudies.Also,asmentionedabove,the numberofcasesineachgroupcanlimittheapplicabilityofthe referencevaluetobegeneralized.Asalways,referencevaluesmust beusedwithcautionandindividualcasecircumstancesmustbe takenintoaccount.

4.4.Theimportanceofcontrolcases

Thestrengthsofourstrategyisthatitproducesnotonlytoxic concentrationsbutalso“normal”(i.e.whichhavenotcontributed

tothecauseofdeath)postmortemreferenceconcentrations.Inour experience having a reference of which concentrations can be considered“normal”greatlyaidsinrulingoutintoxications.Our controlgroup(groupC)comprises,asmentionedinsection2.4, caseswerethedeathwasunrelatedtointakeofdrugs.Inpractice this means cases were the cause of death cannot have been influencedbydrugsandthatthedeceasedshouldhavebeenalive andfullycapabletoact,hadnotsomethingexternal(such asa suicidalhangingorgunshot)occurred.Thecontrolgroupisfurther selectedbyremovingcasesinwhichdamagetoadrugreservoir organ[7,43]mightcontaminatethefemoralbloodusedintesting. Ashasbeenmentionedinsection4.1theselectionprocessused inthisstudydiscardsalargeamountofcases.Thus,groupCisnot merelycaseswithanothercauseofdeaththanintoxication,but ratherahighlyselectedgroupinwhichasmanyuncertaintiesas possiblehasbeenruledout.Theadvantageofthisapproachisthat theresultinggroupCcanbeusedwithhighconfidence. 4.5.Statisticalconsiderations

Inthisstudyweusedseveralstatisticalapproachesinorderto analyzethereliabilityofobtainedreferencevaluesinregardstothe samplesize.Themainobjectivewastobeabletodetecttrends among the studied substances, for example how well a small sample size corresponds to the “true” distribution of a given substanceandcasetype.Byperformingreplicativesamplingitwas possibleto studyhow the concentrationsin small sample sets variedbetweendifferentsamples.Basedonthisitwaspossibleto analyzehowrepresentativeasmallersamplesetisinrelationto thefulldataset,andtoseepossibledifferencesamongdifferent classesofsubstances.Itshouldbenotedthatweonlyused non-parametricstatisticalmethods(duetothenon-normaldistribution oftheconcentrationdata),whichhavethegeneraldisadvantageof a lowerpower compared withparametricstatistical tests. This shortagewas observed for the estimated power when N=1–3 (Fig.4).Infuturestudiesitwouldbevaluabletoperformmore“in depth”studiesforspecificsubstances.Forexamplebyanalyzing andfittingmoredetaileddistributionsinordertoapplyalternative statisticalteststofurtherincreasethespecificity.

4.6.Limitations

Itiswell knownthatcertainsubstances, suchasopioidsand benzodiazepines,aresubjecttothedevelopmentoftoleranceinan individual[44].Ourmethodhasnotdifferentiatedtolerantand non-tolerantindividuals.Inordertoevaluatetoxicologicalfindingswere tolerancemightbeafactorconcurrenthairanalysiscanprovide additional information. Absence ofa drugin theinner segment(s)ofa hair sample suggests a lack of tolerance [45–47], which is of importancewheninterpretingthetoxicologicalresultinblood.

Manysubstanceshaveactivemetabolites(inthepresentstudy thisisthecaseforamitriptyline,olanzapine,oxazepam,oxycodone andtramadol),whichmayhaveadirectpharmacologicaleffector induce other toxicity [48,49]. Chiral structures can impact the effect of a drug [50] and genetic mutations affect both the metabolismandeffectofdrugs[51,52].Inourstudieswehaveonly evaluatedparentcompoundswithoutregardtoactivemetabolites, genetics and chiral structures. Thus, these factors must be evaluatedonacase-by-casebasis andtakenintoaccountwhen usingreferencevalues.

With regard to multi-drug intoxications there is also the complicatedproblemoftowhichextentdrugsinteractwithone anotherinadetrimentalmanner,producingintoxicationgreater thanthesumofitsparts.Thishastobekeptinmindwhenthere aremultipledrugsofthesameclasspresent,ormultipledrugsthat act on the same physiologic system (such as respiration or

consciousness)[44].Inourmaterialitisdifficulttoevaluatethe extentoftheseinteractions.

4.7.Conclusionsandrecommendations

Postmortem toxicology is a complex matter with multiple sources of possible error and often a paucity of information regardinganyintakeofdrugs.Inthisenvironmentitisimportant toprovidethebesttoolsforevaluationpossible,whileatsametime beingmindfulofthepitfallsand limitationsofthepostmortem setting.

Thereferencevaluespresentedinthisstudy,andourprevious publications,alongwiththestatisticalinformationfillan impor-tantroleinthisregard.Table4presentasummaryofnumerical benchmarksforusingourmethod.

In conclusion, while we believewe are beingas diligent as possibleinthecompilationofourresults,thevariationinsmalland mediumsamplesizesurgescautionwhenusingthecurrentand otherreferencecompilations.Casecircumstancesandpostmortem factorsmustthusalwaysbetakenintoaccount.Futureresearch shouldstrivetousedatabaseswithasmanydetectionsaspossible, inordertominimizetheimpactofvariationofsmallsamplesizes. CRediTauthorshipcontributionstatement

C. Söderberg: Conceptualization, Investigation, Writing -originaldraft,Writing-review&editing.A.Tillmar: Conceptuali-zation, Methodology, Formal analysis, Investigation, Writing -review&editing.A.Johansson:Investigation,Writing-review& editing.E.Wernvik:Investigation,Writing-review&editing.A.K. Jönsson:Conceptualization,Investigation,Datacuration,Writing -review & editing. H. Druid: Supervision, Conceptualization, Methodology,Investigation,Fundingacquisition,Writing-review &editing.

DeclarationofCompetingInterest None.

Acknowledgements

Thisstudy wassupportedby theSwedishNationalBoard of ForensicMedicine.

AppendixA.Supplementarydata

Supplementarymaterialrelatedtothisarticlecanbefound,in the online version, at https://doi.org/10.1016/j.forsciint.2020. 110292.

Table4

Keynumericalfactors. Numericalfactors

Averagenumberofdetectionsneededtoprovideasinglecaseingroup A,BorCafterevaluation

5.3 Numberofcasesineachgroupneededtoreachapowerof>0.95when separatingintoxicationsfromcontrols

10a

Normalizedinterquartilerange(IQR)ofmediansacrossallgroupswitha samplesizeof:

5 75.3% 10 47.0% 20 32.3% 30 27.7% 40 21.8% 50 17.5% a

OlanzapinegroupBcomparedtogroupChasbeenexcludedasanoutlier,which require17casestoreachapowerof>0.95.

References

[1]M. Eddleston,Patterns andproblems of deliberate self-poisoningin the developingworld,QJM93(11)(2000)715–731.

[2]V.Ajdacic-Gross,etal.,Methodsofsuicide:internationalsuicidepatterns derivedfromtheWHOmortalitydatabase,Bull.WorldHealthOrgan.86(9) (2008)726–732.

[3]K.W.Simonsen,etal.,Fatalpoisoningindrugaddictsinthenordiccountriesin 2012,ForensicSci.Int.248(2015)172–180.

[4]M.Hickman,etal.,Londonauditofdrug-relatedoverdosedeaths: character-isticsandtypology,andimplicationsforpreventionandmonitoring,Addiction 102(2)(2007)317–323.

[5]E.Groot,etal.,DrugtoxicitydeathsafterreleasefromincarcerationinOntario, 2006-2013:reviewofcoroner’scases,PLoSOne11(7)(2016)e0157512. [6]D.J.Pounder, G.R.Jones,Post-mortemdrugredistribution—atoxicological

nightmare,ForensicSci.Int.45(3)(1990)253–263.

[7]A.L.Pelissier-Alicot,etal.,Mechanismsunderlyingpostmortemredistribution ofdrugs:areview,J.Anal.Toxicol.27(8)(2003)533–544.

[8]G.Skopp,Postmortemtoxicology,ForensicSci.Med.Pathol.6(4)(2010)314– 325.

[9]G.Skopp,Preanalyticaspectsinpostmortemtoxicology,ForensicSci.Int.142 (2-3)(2004)75–100.

[10]E.Saar,etal.,Assessmentofthestabilityof30antipsychoticdrugsinstored bloodspecimens,ForensicSci.Int.215(1-3)(2012)152–158.

[11]A.W. Jones, A.Holmgren, Concentration distributionsof the drugsmost frequentlyidentifiedinpost-mortemfemoralbloodrepresentingallcausesof death,Med.Sci.Law49(4)(2009)257–273.

[12]T.Launiainen,I.Ojanpera,Drugconcentrationsinpost-mortemfemoralblood comparedwiththerapeuticconcentrationsinplasma,DrugTest.Anal.6(4) (2014)308–316.

[13]R.A.Ketola,I.Ojanpera,Summarystatisticsfordrugconcentrationsin post-mortemfemoralbloodrepresentingall causesofdeath,DrugTest.Anal. (2019).

[14]O.H.Drummer,Post-mortemtoxicology,ForensicSci.Int.165(2-3)(2007) 199–203.

[15]H.Druid,P.Holmgren,Acompilationoffatalandcontrolconcentrationsof drugsinpostmortemfemoralblood,J.ForensicSci.42(1)(1997)79–87. [16]A.K.Jonsson,etal.,Sedativeandhypnoticdrugs—fatalandnon-fatalreference

bloodconcentrations,ForensicSci.Int.236(2014)138–145.

[17]M.Reis,etal.,Referenceconcentrationsofantidepressants.Acompilationof postmortemandtherapeuticlevels,J.Anal.Toxicol.31(5)(2007)254–264. [18]C.Soderberg,etal.,Referencevaluesoflithiuminpostmortemfemoralblood,

ForensicSci.Int.277(2017)207–214.

[19]C.Soderberg,etal.,Antipsychotics-postmortemfatalandnon-fatalreference concentrations,ForensicSci.Int.266(2016)91–101.

[20]H.Druid,P.Holmgren,P.Löwenhielm,Computer-assistedsystemsforforensic pathologyandforensictoxicology,J.ForensicSci.41(5)(1996)830–836. [21]F.C. Kugelberg,etal.,Influenceofbloodloss onthepharmacokineticsof

citalopram,ForensicSci.Int.161(2–3)(2006)163–168.

[22]F.C.Kugelberg,P.Holmgren,H.Druid,Codeineandmorphineblood concen-trationsincreaseduringbloodloss,J.ForensicSci.48(3)(2003)664–667. [23]A.K.Jonsson,etal.,Antipsychoticsassociatedwithpulmonaryembolismina

Swedishmedicolegalautopsyseries,Int.Clin.Psychopharmacol.23(5)(2008) 263–268.

[24]R:ALanguageandEnvironmentforStatisticalComputing,RFoundationfor StatisticalComputing,2012.

[25]G.L.Horowitz,Referenceintervals:practicalaspects,EJIFCC19(2)(2008)95– 105.

[26]G.L.Horowitz,etal.,Defining,Establishing,andVerifyingReferenceIntervals in the Clinical Laboratory; Approved Guideline, third edition, Clinical LaboratoryandStandardsInstitute,2010.

[27]F.Musshoff,Casereportsinforensictoxicology,DrugTest.Anal.5(9-10)(2013) 723–724.

[28]C.L.Winek,etal.,Drugandchemicalblood-leveldata2001,ForensicSci.Int. 122(2-3)(2001)107–123.

[29]M.Schulz,etal.,Therapeuticandtoxicbloodconcentrationsofnearly1,000 drugsandotherxenobiotics,Crit.Care16(4)(2012)R136.

[30]A.W.Jones,A.Holmgren,Concentrationsofalprazolaminbloodfromimpaired drivers and forensic autopsies were not much different but showed a highprevalenceofco-ingestedillicitdrugs,J.Psychopharmacol.27(3)(2013) 276–281.

[31]A.W.Jones,A.Holmgren,Concentrationsofzolpidemandzopicloneinvenous bloodsamplesfromimpaireddriverscomparedwithfemoralbloodfrom forensicautopsies,ForensicSci.Int.222(1-3)(2012)118–123.

[32]B.C.Wolf,etal.,Alprazolam-relateddeathsinPalmBeach County,Am.J. ForensicMed.Pathol.26(1)(2005)24–27.

[33]B.Jonasson,T.Saldeen,Citalopraminfatalpoisoningcases,ForensicSci.Int. 126(1)(2002)1–6.

[34]D.D. Baker, A.J. Jenkins, A comparison of methadone, oxycodone, and hydrocodone related deaths in northeast Ohio, J. Anal. Toxicol. 32 (2) (2008)165–171.

[35]E.J.Cone,etal.,Oxycodoneinvolvementindrugabusedeaths.II.Evidencefor toxicmultipledrug-druginteractions,J.Anal.Toxicol.28(7)(2004)616–624. [36]A.Ogle,etal.,Clinicalhistoryandcharacteristicsofpersonswith oxycodone-relateddeathsinHillsboroughCounty,Floridain2009,ForensicSci.Int.223 (1-3)(2012)47–52.

[37]B.C.Wolf, etal., One hundredseventy two deaths involvingtheuse of oxycodoneinPalmBeachCounty,J.ForensicSci.50(1)(2005)192–195. [38]J.L.Pilgrim,O.H.Drummer,Thetoxicologyandcomorbiditiesoffatalcases

involvingquetiapine,ForensicSci.Med.Pathol.9(2)(2013)170–176. [39]M.Nedahl,S.Johansen,K.Linnet,Referencebrainandbloodconcentrationsof

olanzapineinpostmortemcases,J.Anal.Toxicol.(2018).

[40]M.Nedahl,S.S.Johansen,K.Linnet,ReferenceBrain/Bloodconcentrationsof citalopram,duloxetine,mirtazapine andsertraline,J.Anal.Toxicol.42(3) (2018)149–156.

[41]L.Skov,S.S.Johansen,K.Linnet,Postmortemquetiapinereference concen-trationsinbrainandblood,J.Anal.Toxicol.39(7)(2015)557–561. [42]L. Skov, S.S. Johansen, K. Linnet, Postmortem femoral blood reference

concentrations of aripiprazole, chlorprothixene, and quetiapine, J. Anal. Toxicol.39(1)(2015)41–44.

[43]C. Sastre, et al., Post mortem redistribution of drugs: current State of knowledge,Curr.Pharm.Des.23(36)(2017)5530–5541.

[44]T.Richardson,Pitfallsinforensictoxicology,Ann.Clin.Biochem.37(Pt1) (2000)20–44.

[45]H.Druid,etal.,Evaluationoftheroleofabstinenceinheroinoverdosedeaths usingsegmentalhairanalysis,ForensicSci.Int.168(2-3)(2007)223–226. [46]G.Hoiseth,etal.,Ishairanalysisusefulinpostmortemcases?J.Anal.Toxicol.

42(1)(2018)49–54.

[47]G.A.Cooper,etal.,Societyofhairtestingguidelinesfordrugtestinginhair, ForensicSci.Int.218(1-3)(2012)20–24.

[48]D.E.Drayer,Pharmacologicallyactivemetabolitesofdrugsandotherforeign compounds.Clinical,pharmacological,therapeuticandtoxicological consid-erations,Drugs24(6)(1982)519–542.

[49]M.Pirmohamed,N.R.Kitteringham,B.K.Park,Theroleofactivemetabolitesin drugtoxicity,DrugSaf.11(2)(1994)114–144.

[50]S.W.Smith,Chiraltoxicology:it’sthesamething...onlydifferent,Toxicol. Sci.110(1)(2009)4–30.

[51]U.A.Meyer,U.M.Zanger,Molecularmechanismsofgeneticpolymorphismsof drugmetabolism,Annu.Rev.Pharmacol.Toxicol.37(1997)269–296. [52]F.Musshoff,U.M.Stamer,B.Madea,Pharmacogeneticsandforensictoxicology,

ForensicSci.Int.203(1-3)(2010)53–62. 10 C.Söderbergetal./ForensicScienceInternational311(2020)110292