Improvement in the identification and quantification of UV filters and additives in sunscreen cosmetic creams by gas chromatography/mass spectrometry through three-way calibration techniques

Accep ed Manusc ip Imp o emen in he iden i ica ion and quan i ica ion OF UV il e s and addi i es in sunsc een cosme ic c eams by gas ch oma og aphy/mass spec ome y h ough h ee-way calib a ion echniques L. Rubio, L. Val e de-Som, L.A. Sa abia, M.C. O iz PII: S0039-9140(19)30782-9 DOI: h ps://doi.o g/10.1016/j. alan a.2019.120156 A icle Numbe : 120156 Re e ence: TAL 120156 To appea in: Talan a Recei ed Da e: 3 June 2019 Re ised Da e: 10 July 2019 Accep ed Da e: 14 July 2019 Please ci e his a icle as: L. Rubio, L. Val e de-Som, L.A. Sa abia, M.C. O iz, Imp o emen in he iden i ica ion and quan i ica ion OF UV il e s and addi i es in sunsc een cosme ic c eams by gas ch oma og aphy/mass spec ome y h ough h ee-way calib a ion echniques, Talan a (2019), doi: h ps://doi.o g/10.1016/j. alan a.2019.120156. This is a PDF ile o an unedi ed manusc ip ha has been accep ed o publica ion. As a se ice o ou cus ome s we a e p o iding his ea ly e sion o he manusc ip . The manusc ip will unde go copyedi ing, ypese ing, and e iew o he esul ing p oo be o e i is published in i s inal o m. Please no e ha du ing he p oduc ion p ocess e o s may be disco e ed which could a ec he con en , and all legal disclaime s ha apply o he jou nal pe ain. MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 1/31 IMPROVEMENT IN THE IDENTIFICATION AND QUANTIFICATION OF UV FILTERS AND ADDITIVES IN SUNSCREEN COSMETIC CREAMS BY GAS CHROMATOGRAPHY/MASS SPECTROMETRY THROUGH THREE-WAY CALIBRATION TECHNIQUES L. Rubio a , L. Val e de-Som a , L.A. Sa abia b , M.C. O iz a,1 a Depa men o Chemis y, b Depa men o Ma hema ics and Compu a ion Facul y o Sciences, Uni e sidad de Bu gos Plaza Misael Bañuelos s/n, 09001 Bu gos (Spain) Abb e ia ions 2 Abs ac The simul aneous de e mina ion o 2,6-di- e -bu yl-4-me hyl-phenol (BHT), benzophenone (BP), benzophenone-3 (BP3) and diisobu yl ph hala e (DiBP) in se en sunsc een c eams was ca ied ou by gas ch oma og aphy/mass spec ome y (GC/MS) using DiBP-d 4 as in e nal s anda d. The con en o BP3, which is a UV il e , mus no exceed 6% (w/w) in 1 Co esponding au ho . Telephone numbe : +34-947-259571. E-mail add ess: mco [email protected] (M.C. O iz). 2 2,6-di- e -bu yl-4-me hyl-phenol (BHT), benzophenone (BP), benzophenone-3 (BP3), capabili y o de ec ion (CCβ), co e consis ency diagnos ic (CORCONDIA), decision limi (CCα), diisobu yl ph hala e (DiBP), elec on impac (EI), ellip ical join con idence egion (EJCR), gas ch oma og aphy/mass spec ome y (GC/MS), in e nal s anda d (IS), mul i a ia e cu e esolu ion coupled o al e na ing leas squa es (MCR-ALS), pa allel ac o analysis (PARAFAC), p incipal componen analysis (PCA), p obabili y o alse posi i e (α), p obabili y o alse nega i e (β), p og ammed empe a u e apo ize (PTV), single ion moni o ing (SIM), sun p o ec ion ac o (SPF), o al ion ch oma og am (TIC), ul a iole (UV). MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 2/31 cosme ic p oduc s acco ding o Regula ion (EU) 2017/238 and he use o DiBP in cosme ic p oduc s shall be p ohibi ed acco ding o Regula ion (EC) No 1223/2009. The conclusions ob ained wi h he uni a ia e s anda d me hodology in he iden i ica ion o he analy es con ained in he c eams we e w ong. Howe e , a calib a ion based on PARAFAC o PARAFAC2 decomposi ions, whe e he samples o he p edic ion se we e p ojec ed on he model ob ained p e iously wi h he calib a ion se , enabled he unequi ocal iden i ica ion and quan i ica ion o he analy es e en in he p esence o in e e en s no conside ed in he calib a ion model. The PARAFAC2 decomposi ion was used o o e come he shi s in he e en ion ime o BP and BP3. These h ee-way calib a ion echniques a e needed o a oid alse nega i e esul s. The me hod had no p opo ional o cons an bias. The p esence o BHT was de ec ed in he se en sunsc een c eams analysed a an amoun o 6.48 . 10 -2 %, 8.53 . 10 -2 %, 1.70 . 10 -4 %, 1.11 . 10 -4 %, 2.51 . 10 -3 %, 3.20 . 10 -5 % and 6.35 . 10 -3 %. The concen a ions o DiBP ound in ou c eams we e 3.49 . 10 -2 %, 3.19 . 10 -2 %, 3.26 . 10 -2 % and 2.51 . 10 -2 %. On he o he hand, BP was only de ec ed in wo o he cosme ic c eams analysed a an amoun o 7.84 . 10 -3 % and 1.04 . 10 -2 %. In addi ion, BP3 was de ec ed in six o he c eams a an amoun o 4.73%, 3.49%, 4.94 . 10 -3 %, 1.98 . 10 -3 %, 6.62 . 10 -1 % and 1.73%. The e o e, none o he cosme ic c eams con ained BP3 in an amoun highe han 6%. Keywo ds: Benzophenone-3; PTV-GC/MS; PARAFAC; PARAFAC2; UV il e ; sunsc een c eam. 1. In oduc ion The use o sunsc een cosme ic c eams p o ec s he skin om he nega i e e ec s o ul a iole (UV) ays such as sunbu n o skin cance . Un o una ely, some o he addi i es con ained in hese c eams could in e e e in he ho mone le els o he human body [1,2,3]. Regula ion (EC) No 1223/2009 [4] on cosme ic p oduc s es ablishes ules o be complied wi h by any cosme ic p oduc made a ailable on he ma ke , in o de o ensu e he MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 3/31 unc ioning o he in e nal ma ke and a high le el o p o ec ion o human heal h. The con en o oxybenzone o benzophenone-3 (BP3), wi h he chemical name 2-hyd oxy-4- me hoxybenzophenone, was modi ied in Commission Regula ion (EU) 2017/238 [5] and mus no exceed 6% (w/w). In addi ion, he label o he cosme ic p oduc mus include he wo ding "con ains benzophenone-3" when he concen a ion is uppe han 0.5% (w/w) and i is no used o p oduc p o ec ion pu poses. This ange does no pose a isk o human heal h, apa om i s con ac which could p oduce alle gy [6]. BP3 is a sunsc een agen used o abso b UV adia ion in plas ics and in pe sonal ca e p oduc s. In addi ion, i is used as a pho o-s abilize o minimize he colou and odou changes o he cosme ic p oduc [7]. The concen a ion o his compound is egula ed because i is ha m ul o human heal h [7] and causes alle gy [8,9]. Benzophenone- ype UV il e s also induce endoc ine dis up ing e ec s [10,11,12]. BP3 pene a es he skin and 1-2% o he sunsc een is abso bed in humans [13,14,15]. BP3 has been de ec ed in blood plasma [16,17], in human b eas milk [18,19] and in u ine since he compound is exc e ed [1,13,14,20]. On he o he hand, mos o he compounds p esen in c eams a e soluble in wa e so nega i e en i onmen al e ec s appea [7] such as ish con amina ions which cause he appea ance o hese compounds in ood [21,22]. In addi ion, he p esence o BP3 in swimming pool wa e could p oduce haza dous p oduc s by eac ion wi h chlo ine [7,23] being a p oblem o human heal h. O he addi i es a e added o sunsc een cosme ic c eams such as benzophenone (BP) and 2,6-di- e -bu yl-4-me hyl-phenol (BHT). BP is ano he UV il e [2,24], whe eas BHT is used as an ioxidan o p e en ancidi y o o inhibi oxida ion in cosme ic o mula ions [2,25]. On he o he hand, he p esence o ph hala es such as diisobu yl ph hala e (DiBP) in hese c eams can be due o manu ac u ing p ocess o o hei mig a ion om packaging when plas ic is used [26]. The con ol and analysis o hese compounds is impo an because hey ha e ha m ul e ec s on heal h oo [27,28]. DiBP has been classi ied as ca cinogenic, mu agenic o oxic subs ance o ep oduc ion (ca ego y 1B) in [29]. The e o e, he use o his MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 4/31 compound in cosme ic p oduc s shall be p ohibi ed as s a ed in [4]. Howe e , he non- in ended p esence o a small amoun o his subs ance shall be pe mi ed i i s p esence is echnically una oidable, ha is, i i comes om impu i ies o na u al o syn he ic ing edien s, he manu ac u ing p ocess, s o age o mig a ion om packaging. Many analysis me hods ha e been de eloped o de ec hese compounds [2,30]. The ex ac ion o he analy es om he cosme ic p oduc can be pe o med wi h o ganic sol en s such as me hanol [24,31], e hanol [32,33], among o he s [26,34]. The use o an ul asonic ba h o a o ex can accele a e he solubilisa ion o hose compounds. Then, he ex ac could be il e ed o cen i uged o ex ac he ac ion o in e es and emo e he insoluble ac ion o he cosme ic ma ix. O he ex ac i e app oaches a e based on solid phase mic oex ac ion [35], o liquid-liquid mic oex ac ion [24]. The analy ical echniques employed o de e mine UV il e s in cosme ics [30] and o he addi i es [2] a e: (i) ch oma og aphic echniques wi h di e en de ec o s; (ii) spec oscopic echniques; and (iii) elec ochemical echniques. The wo la e ha e been less used han ch oma og aphic echniques. The use o liquid ch oma og aphy wi h di e en de ec o s in his de e mina ion has la ely inc eased [32,33,36]. Howe e , gas ch oma og aphy has been less used since a de i a iza ion s ep wi h silyla ing eagen s is some imes equi ed o inc ease he ola ili y and sensi i i y o he compounds [35,37,38]. The de e mina ion o ph hala es and sunsc een agen s in cosme ic p oduc s is ca ied ou using a gas ch oma og aph coupled o a mass spec ome e de ec o wi hou de i a iza ion in [26,39]. Se e al h ee-way algo i hms can be used wi h ch oma og aphic signals. P e ious wo ks [40,41,42] ha e demons a ed he use ulness o h ee-way calib a ions based on he PARAFAC decomposi ion using ch oma og aphic da a ob ained wi h di e en de ec o s ha p o ide mul i a ia e signals (mass spec ome e s o diode a ay de ec o s) [43]. These wo ks highligh he ad an age o using he abundances eco ded a all he ions selec ed (o he abso bance spec um) when quan i ying o iden i ying he analy es acco ding o he equi emen s s a ed in Eu opean egula ions. In addi ion, hese calib a ions a e use ul o he MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 5/31 op imiza ion and e alua ion o he obus ness o analy ical me hods [42] since he second- o de p ope y enables he iden i ica ion o he analy e o in e es as a single ac o independen ly o he change in he ins umen al ac o s, e en in he p esence o unknown in e e en s, which coelu e wi h he analy es. The PARAFAC2 decomposi ion o e comes some de ia ions in he ch oma og aphic signals [43,44]. PARAFAC2 has he second-o de p ope y i he co ela ion be ween he ime p o iles is he same in all he samples, which is a weake condi ion han he equali y o ch oma og aphic p o iles imposed by he PARAFAC model. I he loss o ilinea i y is impo an , hen mul i a ia e cu e esolu ion echniques a e a use ul al e na i e, because hei signal- ela ed equi emen s a e weake han hose demanded by PARAFAC o PARAFAC2. Mul i a ia e cu e esolu ion coupled o al e na ing leas squa es (MCR-ALS) has been widely applied in analy ical chemis y and i s ela ed ields [43,45,46], and i has been used o esol e coelu ed compounds. I s majo limi a ion in iden i ying and quan i ying an analy e is he p esence o o a ional ambigui ies and non-unique solu ions. Howe e , he non-uniqueness p oblem can be alle ia ed o o ally a oided in some cases h ough he in elligen use o he da a s uc u e and app op ia e cons ain s. This p oblem is discussed in dep h in [46]. In his wo k, he simul aneous de e mina ion o BHT, BP, BP3 and DiBP in se en sunsc een cosme ic c eams, using DiBP-d 4 as in e nal s anda d (IS), was ca ied ou by means o gas ch oma og aphy/mass spec ome y (GC/MS) wi h a single quad upole mass analyse in selec ed ion moni o ing (SIM) mode. Pa allel ac o analysis (PARAFAC) o PARAFAC2 decomposi ion me hods we e used o disco e i a coeluen ha sha es ions wi h he analy e o in e es is p esen [47,48] and o iden i y unequi ocally he compounds p esen by hei ch oma og aphic and spec al p o iles ollowing he c i e ia es ablished in Decision 2002/657/EC o esidues o e e ina y medicinal p oduc s [49] which a e s ic e han o he o icial egula ions and guidelines [49,50,51]. In his case, a leas a minimum o 3 iden i ica ion poin s is needed o he con i ma ion o each compound; in his wo k 5 ions MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 6/31 we e selec ed. The PARAFAC2 decomposi ion o e come he shi s in he e en ion ime o BP and BP3 in he samples and he analy es we e unequi ocally iden i ied al hough he shi o he e en ion ime is limi ed in egula ed analyses. The esul s o he iden i ica ion we e compa ed wi h he ones ob ained wi h he uni a ia e s anda d me hodology. The de e mina ion o he compounds in he sunsc een c eams analysed was pe o med h ough he p ojec ion o he samples o he p edic ion se on he co esponding PARAFAC o PARAFAC2 model. O he ad an ages o his wo k wi h espec o analy ical me hods p e iously epo ed a e he de e mina ion o di e en compounds, ha a ec human heal h, in sunsc een cosme ic c eams and he de e mina ion o BP3 by GC/MS wi hou a de i a iza ion eac ion, which is no e y usual. 2. Ma e ial and me hods 2.1. Chemicals Benzophenone (CAS no. 119-61-9, pu i ied by sublima ion, ≥ 99% pu i y), 2-hyd oxy-4- me hoxybenzophenone (benzophenone-3, CAS no. 131-57-7, 98% pu i y), 2,6-di- e -bu yl-4- me hyl-phenol (CAS no. 128-37-0, ≥ 99% pu i y), diisobu yl ph hala e (CAS no. 84-69-5, 99% pu i y) and diisobu yl ph hala e-3,4,5,6-d 4 (CAS no. 358730-88-8, analy ical s anda d, 99.7% pu i y) we e pu chased om Sigma-Ald ich (S einheim, Ge many). E hanol (96% ol., CAS no. 64-17-5, HiPe Sol CHROMANORM®, g adien g ade o HPLC) was supplied by VWR In e na ional (Radno , Pennsyl ania, USA). N-hexane (CAS no. 110- 54-3) and ace one (CAS no. 67-64-1) o liquid ch oma og aphy Lich osol ® we e om Me ck KGaA (Da ms ad , Ge many). 2.2. S anda d solu ions MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 7/31 S ock solu ions o BHT a 1 g L -1 , o BP a 5 g L -1 , o DiBP a 2 g L -1 , and o DiBP-d 4 a 0.45 g L -1 we e p epa ed indi idually in hexane. In e media e solu ions we e p epa ed om he o me ones by dilu ion in he same sol en . A s ock solu ion o BP3 a 30.5 g L -1 we e p epa ed in e hanol and in e media e solu ions a concen a ions o 1, 5, 7.5, 10, 15, 20 and 25 g L -1 we e p epa ed om ha s ock solu ion in e hanol. All hose solu ions o BP3 we e dilu ed 25000 imes o p epa e he co esponding calib a ion s anda ds in hexane. The s ock and in e media e solu ions o BP3 we e s able o 15 days. The es o he solu ions o his analy e we e p epa ed daily since his compound was no s able. All he s ock and in e media e solu ions, which weigh was con olled o e i y ha he sol en had no e apo a ed, we e s o ed in c imp ials a 4ºC and p o ec ed om ligh . The labo a o y glasswa e used was ho ough cleaned and plas ic consumables we e a oided as a as possible. The numbe and ype o he samples analysed oge he wi h he concen a ion anges o he sol en s anda ds (analy e s anda ds p epa ed in sol en ) o each analy e in each s age a e collec ed in Table S1 in he Supplemen a y Ma e ial. 2.3. Sunsc een c eam samples Se en di e en sunsc een c eams we e pu chased a local s o es and pha macies (Bu gos, Spain). These cosme ic p oduc s we e: i) c eam 1 (sun p o ec ion ac o (SPF) 50+), ii) c eam 2 (SPF 30), iii) c eam 3 (SPF 50+), i ) c eam 4 (SPF 30), ) c eam 5 (SPF 30), i) c eam 6 (SPF 50+), and ii) c eam 7 (SPF 15). C eam 1 and 2 belong o he same cosme ic b and, whe eas c eam 4 and 7 belong o ano he one. The sunsc een p oduc ’s label o c eam 1, 2 and 7 speci ied ha BP3 was con ained in hei o mula ion and he label o c eam 1, 2, 5 and 7 speci ied ha BHT was one o hei ing edien s. 2.4. Sample p epa a ion me hod MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 14/31 numbe o samples, while he second and hi d ones we e he numbe o scans and ions, espec i ely. 3.1.3.1. Tole ance in e als (s ep ii) To es ima e he ole ance in e als, he e e ence s anda ds measu ed in Sec ion 3.1.1 we e used. In addi ion, h ee e e ence s anda ds ha con ained he IS a h ee concen a ion le els and he analy es a a ixed in e media e concen a ion (see Table S1 in he Supplemen a y Ma e ial, hi d ow) we e also conside ed. The ch oma og ams o all hese samples we e equally agmen ed a ound he e en ion ime o each analy e a e a baseline co ec ion. The esul ing da a ma ices we e a anged in a h ee-way a ay, X 0 , o each analy e excep o DiBP and DiBP-d 4 peaks o which a join a ay was conside ed o bo h compounds. The dimension o he ou h ee-way a ays buil a e gi en in he second ow o Table 1, whe eas he samples included in hose a ays a e de ailed in ows 1 o 3 o Table S1 in he Supplemen a y Ma e ial. The i s dimension co esponds o he numbe o scans conside ed, he second one e e s o he numbe o ions eco ded and he hi d one is he numbe o samples. Then, a PARAFAC decomposi ion was pe o med o each o he a ays. Table 1 ( ows 1-7) also con ains he ea u es o he models ob ained in each case. The co e consis ency diagnos ic (CORCONDIA) [60] measu es he ilinea i y deg ee o he expe imen al h ee-way a ay when he e a e mo e han 2 ac o s in he model. I he a ay is ilinea , hen he maximum CORCONDIA alue o 100 is ound. Fo BP and BP3, a PARAFAC2 decomposi ion was ca ied ou since he e we e shi s in he e en ion ime o hose analy es in he samples. In addi ion, he a iance explained by hese PARAFAC2 models was highe han he one ob ained wi h a PARAFAC model. The PARAFAC model o DiBP and DiBP-d 4 equi ed h ee ac o s whe e he hi d one was ela ed o an in e e en (cha ac e is ic m/z a io: 223) ha elu ed be o e DiBP-d 4 . MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 15/31 The ole ance in e als o he ela i e e en ion ime (see Table 2A, ou h column) and o he ela i e ion abundances (see Table 2B, i h column) we e es ima ed om he loadings o he ch oma og aphic and spec al p o iles, espec i ely, o he PARAFAC o PARAFAC2 models. These wo in e als we e buil ollowing he equi emen s es ablished in [49] and we e used as e e ence o he unequi ocal iden i ica ion o he analy es. PARAFAC decomposi ions p o ide a unique ch oma og aphic p o ile o each compound ha is common o all he samples, whe eas PARAFAC2 decomposi ions p o ide a ch oma og aphic p o ile o each compound o e e y sample. When a PARAFAC2 decomposi ion was conside ed, he median o he e en ion imes o he analy e ob ained om he ch oma og aphic p o ile was used o calcula e he ela i e e en ion ime ( he a io o he ch oma og aphic e en ion ime o he analy e o ha o he in e nal s anda d). Table 2A con ains he e en ion imes (second column o his able) and ela i e e en ion imes ( hi d column o his able) o each analy e ob ained om he models es ima ed wi h he a ay ha con ained he e e ence s anda ds. The ole ance in e als o he ela i e e en ion ime we e buil wi h a ole ance ma gin o ± 0.5% as [49] es ablished. I has also been checked ha he ela i e e en ion imes o BP and BP3 in all he samples ob ained h ough he PARAFAC2 decomposi ion we e wi hin he ole ance in e als es ima ed wi h he e en ion imes o hese analy es in each sample. The e o e, he use o he median o es ima e a ep esen a i e e en ion ime o BP and BP3 was adequa e. On he o he hand, PARAFAC and PARAFAC2 decomposi ions p o ide a unique spec al p o ile o each compound common o all he samples. The ela i e ion abundances o each m/z a io used o de e mine he ole ance in e als acco ding o e . [49] we e calcula ed wi h he co esponding spec al loading (see Table 2B, hi d column) wi h espec o he one o he co esponding base peak o he analy e. 3.1.3.2. Calib a ion models and p ojec ion (s eps iii o iii) MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 16/31 Then, a h ee-way da a a ay was buil o he calib a ion se (X 1 ) and ano he one o he p edic ion se (X 2 ) o each analy e, excep o DiBP and DiBP-d 4 peaks o which a join a ay was conside ed again in bo h cases. The samples ha made up he calib a ion se we e: ou sys em blanks (emp y ials wi hou sol en ), he i s sol en blank measu ed, he i s calib a ion ba ch oge he wi h he wo eplica es injec ed a e i and wo calib a ion s anda ds o he second and o he hi d calib a ion ba ches. The p edic ion se was cons i u ed wi h he es o he samples ( he es o he calib a ion s anda ds, sol en blanks, blank ex ac s and he dilu ed ex ac s o he c eams). Some calib a ion s anda ds we e included in he p edic ion se o gua an ee he easibili y o he p ojec ion o he samples. The dimension o X 1 and X 2 is included in Table 1 ( ows 9 and 15, espec i ely). Then, PARAFAC decomposi ions (o PARAFAC2 decomposi ions in he case o BP and BP3) we e pe o med wi h he h ee-way a ay ha con ained he calib a ion se . When a PARAFAC decomposi ion was ca ied ou , a change in he o de o he dimension o he h ee-way a ays o he calib a ion and p edic ion se s was ca ied ou (see Table 1). In he case o he a ay o DiBP and DiBP-d 4 , some addi ional sol en s anda ds we e added as can be seen in Table 1 since bo h compounds we e comple ely o e lapped and he abundance o DiBP-d 4 was much lowe han he abundance o DiBP, so PARAFAC needs a g ea e a ia ion o DiBP-d 4 . The e o e, a s anda d con aining all he analy es and a highe amoun o DiBP-d 4 (100 µg L −1 ) oge he wi h wo s anda ds ha only con ained 100 µg L −1 o DiBP-d 4 we e added o ha h ee-way a ay. The cha ac e is ics o he PARAFAC and PARAFAC2 models ob ained in each case a e lis ed in Table 1 ( ows 8-15, columns 2-5). An uncons ained one- ac o model was needed o BHT, BP and BP3, whe eas he decomposi ion o he common a ay o DiBP and DiBP- d 4 needed h ee ac o s (CORCONDIA index o 100%). Once hese models we e ob ained, he samples o he p edic ion se we e p ojec ed on he co esponding model. The ex ac ob ained om c eam 6 dilu ed 10 imes exceeded he h eshold alue o he Q and Ho elling’s T 2 s a is ics a he 95% con idence le el when i was MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 17/31 p ojec ed on he PARAFAC2 model o BP3. The e o e, i was conside ed an ou lie and emo ed om he h ee-way a ay ha con ained he samples o he p edic ion se (see he dimension o his a ay in Table 1). Fig. 3 shows he ch oma og aphic, spec al and sample p o iles o he one- ac o PARAFAC2 model ob ained o BP3. As can be seen in Fig. 3 (c), he sample loadings o BP3 we e ze o in he sol en blanks measu ed a e he injec ion o each ex ac o c eam (samples numbe 23, 25, 27, 30, 33, 36, 39, 49, 51, 53, 56, 59, 62 and 65) so he cleanliness o he GC/MS sys em was gua an eed. By way o example, he loadings o he h ee- ac o PARAFAC model o DiBP and DiBP-d 4 a e shown in Fig. 4. The sample loadings o he ac o co esponding o DiBP Fig. 4 (a)) inc eased wi h he concen a ion o he calib a ion s anda ds as expec ed and he eplica es we e simila . The sample loadings o his analy e in he ex ac s o c eam 3 and 4 dilu ed 10 imes we e ou side he calib a ion ange (samples numbe 80 and 82 no shown in ha igu e) so he amoun o DiBP in hose dilu ions could no be quan i ied. The loadings o he sample p o ile o he ac o associa ed o he in e nal s anda d (Fig. 4 (b)) we e ze o in he sol en and sys em blanks, whe eas hey emained nea ly cons an in he es o he samples excep o he addi ional samples added o he h ee-way a ay whe e he loadings we e highe as expec ed. On he o he hand, he sample loadings o he hi d ac o (Fig. 4 (c)) we e e y high in he ex ac s o c eam 3 and 4 dilu ed 10 imes. Some o he m/z eco ded o DiBP we e sha ed wi h his ac o being 223 i s mos cha ac e is ic m/z (see Fig. 4 (d)). This ac o was a ibu ed o an uniden i ied in e e en elu ing nea he beginning o he DiBP-d 4 peak as can be seen in ed in Fig. 4 (e). The analy es we e unequi ocally iden i ied since he ela i e e en ion imes (see Table 2A, i h column) and he ela i e ion abundances (see Table 2B, six h column) es ima ed om he loadings o he ch oma og aphic and spec al p o iles ob ained in his analysis, espec i ely, we e wi hin hei co esponding ole ance in e als (see Table 2A ( ou h column) and Table 2B ( i h column), espec i ely). As can be seen in Fig. 1, he ela i e abundances es ima ed om he spec al PARAFAC loadings o DiBP ob ained we e wi hin MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 18/31 he ole ance in e als o all he m/z a ios. The e o e, he conclusion ha DiBP was no p esen in he c eams s a ed in Sec ion 3.1.1 was w ong since a coelu ing in e e en (see Fig. 2 (b)) p esen in hose samples sha ed some m/z a ios wi h DiBP. This in e e en did no appea in he PARAFAC model conside ed (see Table 1) since i was no p esen in he samples o he calib a ion se and he ex ac s o he c eams we e p ojec ed on he model ob ained wi h he calib a ion se . The e o e, he e was no p oblem in he iden i ica ion o DiBP using PARAFAC. I is impo an o bea in mind ha one o he ad an ages o he PARAFAC decomposi ion o e he uni a ia e s anda d me hodology is ha his h ee-way echnique p o ides a unique spec al p o ile o each analy e ha is common o all he samples. The esul s o he s eps ix, x and xi o he p ocedu e will be shown in he ollowing sec ion al hough many app oaches can be used o pe o m a calib a ion based on PARAFAC o PARAFAC2 decomposi ions [61]. 3.2. Quan i ica ion using h ee-way echniques The sample loadings o each analy e we e nume ically high since hey came om he i s mode o a PARAFAC decomposi ion o om he hi d mode o a PARAFAC2 decomposi ion and hese modes a e no no malized in he decomposi ion. The e o e, hey we e manually no malized p io o s anda diza ion. Once he sample loadings o each analy e we e s anda dized by di iding each o hem by ha o he in e nal s anda d, calib a ion models “s anda dized sample loading e sus ue concen a ion” using he s anda ds con ained in he calib a ion se we e i ed and alida ed. The pa ame e s o he eg ession models es ima ed o each analy e a e included in Table S2 in he Supplemen a y Ma e ial. A quad a ic eg ession model was conside ed o all he analy es excep o DiBP. One ou lie was de ec ed in he calib a ion models o BP and BP3 since hese da a had a s uden ized esidual g ea e han 3 in absolu e alue. The e o e, hose ou lie s we e emo ed, and a new i ing was ca ied ou wi h he emaining da a in bo h cases (see he calib a ion models MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 19/31 ob ained in Table S2 in he Supplemen a y Ma e ial). The lowes mean o he absolu e alue o he ela i e e o s in calib a ion was 1.97% (n = 11) o BP, whe eas he highes alue was 6.86% (n = 8) o BP3 when he samples wi h p edic ed concen a ion lowe han he co esponding CCβ had been excluded. In addi ion, he mean o he absolu e alue o he ela i e e o s in p edic ion o he second and hi d calib a ion ba ches we e calcula ed. Taking hese alues in o accoun , i was concluded ha BP3 was no s able o e ime since he e o s inc eased om he second calib a ion ba ch o he hi d one. The e o e, he solu ions we e p epa ed daily. Table S2 in he Supplemen a y Ma e ial also con ains he pa ame e s o he co esponding accu acy lines buil wi h he calib a ion s anda ds, ha is he eg essions “p edic ed concen a ion e sus ue concen a ion”. The ellip ical join con idence egion (EJCR) es was compu ed and Fig. S2 in he Supplemen a y Ma e ial shows he con idence ellipses, a a 95% con idence le el, o he slope and he in e cep o he accu acy line es ima ed o each analy e. All he con idence ellipses con ained he poin (0,1). In addi ion, Table S2 in he Supplemen a y Ma e ial con ains he p- alues o his es . These p- alues we e highe han 0.05 (0.973 o BHT, 0.960 o BP, 1.000 o DiBP and 0.123 o BP3) so he in e cep and he slope we e signi ican ly no di e en om 0 and 1, espec i ely. The e o e, he me hod had no cons an o p opo ional bias a a 95% con idence le el. The alues o decision limi (CCα) and capabili y o de ec ion (CCβ) o each analy e wi h he p obabili ies o alse posi i e (α) and alse nega i e (β) ixed a 0.05 a e lis ed in he i s wo ows o Table 3. The blank ex ac con ained DiBP in all i s dilu ions excep o he one dilu ed 8000 and 10000 imes, whe eas he amoun o he es o he analy es in he blank ex ac was below he co esponding CCα alues. The amoun o each analy e ound in he sunsc een c eams oge he wi h he co esponding 95% con idence in e al a e de ailed in Table 3. The label o c eam 1, 2, 5 and 7 speci ied ha BHT was con ained in hei o mula ion bu BHT was de ec ed in he se en c eams analysed. On he o he hand, BP3 was ound in all he sunsc een c eams excep o c eam 4 since he con idence in e al o MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 20/31 ha analy e in ha c eam con ained ze o and he alues we e below CCα as can be seen in Table 3. The c eams 1, 2, 6 and 7 we e he ones which con ained a concen a ion o BP3 be ween 0.5% and 6%. Howe e , only he p oduc ’s label o c eams 1, 2 and 7 speci ied ha BP3 was one o hei ing edien s as equi ed in [5]. In addi ion, DiBP and BP we e only de ec ed in ou and wo cosme ic c eams, espec i ely. The amoun ound o DiBP (below 0.035%) may come om impu i ies o na u al o syn he ic ing edien s, he manu ac u ing p ocess, s o age o mig a ion om packaging o he cosme ic p oduc . 4. Conclusions The p ojec ion o he samples o he p edic ion se on he co esponding PARAFAC o PARAFAC2 model enabled he de e mina ion o BHT, BP, DiBP and BP3 in he sunsc een cosme ic c eams analysed e en in he p esence o in e e en s no conside ed in he calib a ion model. PARAFAC2 decomposi ion o e came he p oblems due o he shi s in he e en ion ime o BP and BP3. In addi ion, he unequi ocal iden i ica ion and quan i ica ion o each analy e acco ding o he equi emen s es ablished by EU egula ions we e possible using a PARAFAC o PARAFAC2 decomposi ion despi e some o he m/z a ios o a coelu ing in e e en we e sha ed wi h DiBP. In ac , he unequi ocal iden i ica ion o his analy e could ne e ha e been achie ed om he mass spec um eco ded a i s e en ion ime due o he p esence o a coelu ing in e e en p esen in he sunsc een c eams analysed. The p esence o BHT was de ec ed in he se en sunsc een c eams analysed, whe eas BP3, DiBP and BP we e de ec ed in some o hese c eams. None o he cosme ic c eams con ained BP3 in an amoun highe han 6% as es ablished in Regula ion 2017/238 [5]. 5. Acknowledgmen s The au ho s hank he inancial suppo p o ided by Spanish MINECO (AEI/FEDER, UE) h ough p ojec CTQ2017‐88894‐R and by Jun a de Cas illa y León h ough p ojec MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 21/31 BU012P17 (bo h co‐ inanced wi h Eu opean FEDER unds). L. Val e de-Som hanks JCyL o he pos doc o al con ac h ough BU012P17 p ojec . CONFLICT OF INTEREST The au ho s decla e ha hey ha e no known compe ing inancial in e es s o pe sonal ela ionships ha could ha e appea ed o in luence he wo k epo ed in his pape . 6. Re e ences [1] K.N. Jallad, Chemical cha ac e iza ion o sunsc eens composi ion and i s ela ed po en ial ad e se heal h e ec s, J. Cosme . De ma ol-US. 16 (2017) 353-357. h ps://doi.o g/10.1111/jocd.12282. [2] M. Lo es, M. Llompa , G. Al a ez-Ri e a, E. Gue a, M. Vila, M. Celei o, J.P. Lamas, C. Ga cia-Ja es, Posi i e lis o cosme ic ing edien s: Analy ical me hodology o egula o y and sa e y con ols - A e iew, Anal. Chim. Ac a. 915 (2016) 1-26. h ps://doi.o g/10.1016/j.aca.2016.02.033. [3] T. Wong, D. O on, Sunsc een alle gy and i s in es iga ion, Clin. De ma ol. 29 (2011) 306-310. h ps://doi.o g/10.1016/j.clinde ma ol.2010.11.002. [4] Regula ion (EC) No 1223/2009 o he Eu opean Pa liamen and o he Council o 30 No embe 2009 on cosme ic p oduc s, O . J. Eu . Union, L342 (2009) 59-209. [5] Commission Regula ion (EU) 2017/238 o 10 Feb ua y amending Annex VI o Regula ion (EC) No 1223/2009 o he Eu opean Pa liamen and o he Council on cosme ic p oduc s, O . J. Eu . Union, L36 (2017) 37-38. [6] Scien i ic Commi ee on Consume P oduc s, Heal h & Consume P o ec ion Di ec o a e-Gene al, Opinion on benzophenone-3. COLIPA Nº S38, Eu opean Commission, 2008. MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 22/31 [7] J.C. DiNa do, C.A. Downs, De ma ological and en i onmen al oxicological impac o sunsc een ing edien oxybenzone/benzophenone-3, J. Cosme . De ma ol. 17 (2018) 15-19. h ps://doi.o g/10.1111/jocd.12449. [8] A.R. Heu ung, S.I. Raju, E.M. Wa shaw, Benzophenones, De ma i is 25 (2014) 3- 10. h ps://doi.o g/10.1097/DER.0000000000000025. [9] M.D. Palm, M.N. O'Donoghue, Upda e on pho op o ec ion, De ma ol. The . 20 (2007) 360-376. h ps://doi.o g/10.1111/j.1529-8019.2007.00150.x. [10] J. Wang, L. Pan, S. Wu, L. Lu, Y. Xu, Y. Zhu, M. Guo, S. Zhuang, Recen ad ances on endoc ine dis up ing e ec s o UV il e s, In . J. En . Res. Pub. He. 13 (2016) 1- 11. h ps://doi.o g/10.3390/ije ph13080782. [11] M. K ause, A. Kli , M.B. Jensen, T. Søebo g, H. F ede iksen, M. Schlump , W. Lich ens eige , N.E. Skakkebaek, K.T. D zewiecki, Sunsc eens: a e hey bene icial o heal h? An o e iew o endoc ine dis up ing p ope ies, In . J. And ol. 35 (2012) 424-436. h ps://doi.o g/10.1111/j.1365-2605.2012.01280.x. [12] M. Schlump , P. Schmid, S. Du e , M. Conscience, K. Mae kel, M. Hensele , M. G ue e , I. He zog, S. Reolon, R. Cecca elli, O. Faass, E. S u z, H. Ja y, W. Wu ke, W. Lich ens eige , Endoc ine ac i i y and de elopmen al oxici y o cosme ic UV il e s - an upda e, Toxicology 205 (2004) 113-122. h ps://doi.o g/10.1016/j. ox.2004.06.043. [13] H. Gonzalez, A. Fa b o , O. La kö, A.M. Wennbe g, Pe cu aneous abso p ion o he sunsc een benzophenone-3 a e epea ed whole-body applica ions, wi h and wi hou ul a iole i adia ion, B i . J. De ma ol. 154 (2006) 337-340. h ps://doi.o g/10.1111/j.1365-2133.2005.07007.x. MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 23/31 [14] H.G. Gonzalez, A. Fa b o , O. La kö, Pe cu aneous abso p ion o benzophenone-3, a common componen o opical sunsc eens, Clin. Exp. De ma ol. 27 (2002) 691- 694. h ps://doi.o g/10.1046/j.1365-2230.2002.01095.x. [15] C.G Hayden, M.S. Robe s, H.A. Benson, Sys emic abso p ion o sunsc een a e opical applica ion, Lance . 350 (1997) 863-864. h ps://doi.o g/10.1016/S0140- 6736(05)62032-6. [16] N.R. Janjua, B. Kongshoj, A.M. Ande soon, H.C. Wul , Sunsc eens in human plasma and u ine a e epea ed whole-body opical applica ion, J. Eu . Acad. De ma ol. 22 (2008) 456-461. h ps://doi.o g/10.1111/j.1468-3083.2007.02492.x. [17] V. Sa eiya, S. Risk, H.A.E. Beson, Liquid ch oma og aphic assay o common sunsc een agen s: applica ion o in i o assessmen o skin pene a ion and sys emic abso p ion in human olun ee s, J. Ch oma og . B 803 (2004) 225-231. h ps://doi.o g/10.1016/j.jch omb.2003.12.022. [18] D. Molins-Delgado, M.M. Olmo-Campos, G. Vale a-Juan, V. Pleguezuelos- He nández, D. Ba celó, M.S. Díaz-C uz, De e mina ion o UV il e s in human b eas milk using u bulen low ch oma og aphy and babies' daily in ake es ima ion, En i on. Res. 161 (2018) 532-539. h ps://doi.o g/10.1016/j.en es.2017.11.033. [19] R. Rod íguez-Gómez, A. Za a-Gómez, N. Do i al-Ga cía, O. Balles e os, A. Na alón, De e mina ion o benzophenone-UV il e s in human milk samples using ul asound-assis ed ex ac ion and clean-up wi h dispe si e so ben s ollowed by UHPLC-MS/MS, Talan a 134 (2015) 657-664. h ps://doi.o g/10.1016/j. alan a.2014.12.004. [20] M. Schlump , B. Co on, M. Conscience, V. Halle , B. S einmann, W. Lich ens eige , In i o and in i o es ogenici y o UV sc eens, En i on. Heal h Pe sp. 109 (2001) 239-244. h ps://doi.o g/10.1289/ehp.01109239. MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 30/31 FIGURE CAPTIONS Fig. 1 Tole ance in e als o he ela i e ion abundances o DiBP es ima ed o each m/z a io wi h h ee e e ence samples ha con ained a di e en concen a ion le el o DiBP: 25 µg L -1 (in blue), 75 µg L -1 (in ed) and 125 µg L -1 (in g een). Rela i e abundance o each m/z a io ob ained wi h: he PARAFAC models ob ained wi h he e e ence samples (ligh blue ci cles) and wi h all he samples o he analysis o he c eams (black c oss), he ex ac ob ained om c eam 1 dilu ed 10000 imes (pink iangle), he ex ac ob ained om c eam 2 dilu ed 10000 imes (pu ple diamond), he ex ac ob ained om c eam 3 dilu ed 10000 imes ( ed squa e) and he ex ac ob ained om c eam 6 dilu ed 8000 imes (ligh g een s a ). (Fo in e p e a ion o he e e ences o colou in his igu e legend, he eade is e e ed o he web e sion o he a icle). Fig. 2 Ch oma og aphic p o iles o he PARAFAC models ob ained wi h he h ee-way a ays ha con ained he calib a ion and p edic ions se s oge he in he same a ay o : (a) BHT, (b) DiBP and DiBP-d 4 , (c) BP3 (a PARAFAC2 decomposi ion o he a ay was ca ied ou in his las case). Fig. 3 One- ac o PARAFAC2 model ob ained wi h he h ee-way a ay ha con ained he calib a ion se o BP3. Loadings o he: (a) ch oma og aphic p o ile, (b) spec al p o ile and (c) sample p o ile. The sample loadings o he 20 samples o he calib a ion se a e ep esen ed by g ey ci cles, whe eas he sample loadings o he 62 samples o he p edic ion se , which ha e been p ojec ed on he PARAFAC2 model, a e ep esen ed by ed diamonds. (Fo in e p e a ion o he e e ences o colou in his igu e legend, he eade is e e ed o he web e sion o he a icle). MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT 31/31 Fig. 4 Loadings o he sample p o ile o : a) ac o 1 (DiBP), b) ac o 2 (DiBP-d 4 ) and c) ac o 3 (unknown in e e en ) o he h ee- ac o PARAFAC model i ed wi h he common h ee-way a ay o DiBP and DiBP-d 4 ha only con ained he calib a ion se . The sample loadings o he 24 samples o he calib a ion se a e ep esen ed by g ey ci cles, whe eas he ones o he 63 samples o he p edic ion se , which ha e been p ojec ed on he model, a e ep esen ed by ed diamonds. d) Loadings o he spec al p o ile and e) loadings o he ch oma og aphic p o ile. In igu es (d-e), ac o 1 (DiBP) is in ligh blue, ac o 2 (DiBP-d 4 ) is in ligh g een, while ac o 3 (in e e en ) is in ed. (Fo in e p e a ion o he e e ences o colou in his igu e legend, he eade is e e ed o he web e sion o he a icle). MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT Table 1 Dimensions (scans × ions × samples) o he h ee-way a ays buil o e e y analy e in each expe imen al s age and cha ac e is ics o he PARAFAC o PARAFAC2 models (numbe o ac o s, cons ain s imposed, explained a iance and CORCONDIA index) ob ained om he decomposi ion o each a ay. Analy ical s age BHT BP DiBP and DiBP-d 4 BP3 Tole ance in e als Dimension o X 0 21 × 5 × 9 49 × 5 × 9 41 × 10 × 9 56 × 5 × 9 Model PARAFAC PARAFAC2 PARAFAC PARAFAC2 # Fac o s 1 1 3 1 Non-nega i i y cons ain s None None In modes 1 and 2 None Expl. Va (%) 99.71 99.88 99.45 99.79 CORCONDIA (%) a --- --- 100 --- Analysis o he sunsc een c eams Dimension o X 1 (calib a ion se ) 20 × 21 × 5 b 49 × 5 × 20 24 × 41 × 10 b 56 × 5 × 20 Model PARAFAC PARAFAC2 PARAFAC PARAFAC2 # Fac o s 1 1 3 1 Non-nega i i y cons ain s None None In he h ee modes None Expl. Va (%) 99.74 99.87 99.08 99.82 CORCONDIA (%) a --- --- 100 --- Dimension o X 2 (p edic ion se ) 63 × 21 × 5 b 49 × 5 × 63 63 × 41 × 10 b 56 × 5 × 62 Expl. Va .: Explained a iance by he model. a The CORCONDIA index canno be calcula ed in he PARAFAC o PARAFAC2 decomposi ion wi h only one ac o . b In his case, he dimension o he a ay co esponds o he numbe o samples × scans × ions. MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT Table 2 Tole ance in e als o : A) he ela i e e en ion ime and o B) he ela i e ion abundances es ima ed om he loadings o he ch oma og aphic and spec al p o iles, espec i ely. Iden i ica ion o e e y analy e in he analysis o sunsc een c eams. In he case o BP and BP3, he median o he e en ion imes ob ained in he co esponding PARAFAC2 decomposi ion was conside ed. A) Re en ion ime Analy e R (min) Rela i e R Tole ance in e al Iden i ica ion in he analysis o he c eams Rela i e R BHT 8.229 0.806 (0.802-0.810) 0.806 BP 9.016 0.883 (0.879-0.887) 0.883 DiBP-d 4 10.211 1.000 (0.995-1.005) 1.000 DiBP 10.220 1.001 (0.996-1.006) 1.001 BP3 11.120 1.089 (1.083-1.094) 1.089 B) Diagnos ic ions Analy e m/z a io Spec al loading Rela i e abundance (%) Tole ance in e al (%) b Iden i ica ion in he analysis o he c eams Rela i e abundance (%) BHT 91 6.85 . 10 - 2 7.13 (3.57-10.70) 7.08 145 1.15 . 10 - 1 11.94 (9.55-14.33) 11.94 177 7.67 . 10 - 2 7.98 (3.99-11.97) 8.03 205 a 9.61 . 10 - 1 100 - 100 220 2.29 . 10 - 1 23.81 (20.24-27.38) 23.94 BP 51 1.37 . 10 - 1 17.75 (14.20-21.30) 17.42 77 4.38 . 10 - 1 56.77 (51.09-62.45) 56.63 105 a 7.71 . 10 - 1 100 - 100 152 3.27 . 10 - 2 4.24 (2.12-6.36) 4.34 182 4.40 . 10 - 1 57.05 (51.35-62.76) 57.61 DiBP-d 4 80 5.65 . 10 - 2 5.68 (2.84-8.52) 5.70 153 a 9.94 . 10 - 1 100 - 100 171 2.48 . 10 - 2 2.50 (1.25-3.75) 2.57 209 1.17 . 10 - 2 1.18 (0.59-1.77) 1.23 227 4.95 . 10 - 2 4.98 (2.49-7.47) 5.14 DiBP 104 7.82 . 10 - 2 7.86 (3.93-11.79) 7.66 149 a 9.95 . 10 - 1 100 - 100 167 2.75 . 10 - 2 2.76 (1.38-4.14) 2.75 205 1.37 . 10 - 2 1.38 (0.69-2.07) 1.39 223 5.24 . 10 - 2 5.27 (2.64-7.91) 5.27 BP3 77 1.56 . 10 - 1 22.86 (19.43-26.29) 22.33 105 8.16 . 10 - 2 11.99 (9.59-14.39) 11.74 151 5.73 . 10 - 1 84.19 (75.77-92.61) 83.43 227 a 6.81 . 10 - 1 100 - 100 228 4.21 . 10 - 1 61.86 (55.67-68.05) 61.80 a Base peak. b Acco ding o e . [42], he es ima ion o he ole ance in e al is di e en depending on he alue o he ela i e abundance o he co esponding m/z a io. The ole ance ma gin was ±50% o ela i e in ensi ies lowe o equal o 10%, a ± 20% o ela i e in ensi ies om 10% o 20%, a ± 15% o ela i e in ensi ies om 20% o 50% and ± 10% o ela i e in ensi ies highe han 50%. MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT Table 3 Decision limi (CCα), capabili y o de ec ion (CCβ) a x 0 = 0 and concen a ion o each analy e ound in he sunsc een c eams oge he wi h he co esponding con idence in e als o he p edic ed concen a ion a a 95% con idence le el. BHT BP DiBP BP3 CCα (x 0 = 0) (µg L − 1 ) 4.03 6.32 5.93 142.67 CCβ (x 0 = 0) (µg L − 1 ) a 7.93 12.40 11.65 279.80 Sunsc een c eam 1 Concen a ion (% w/w) 6.48 . 10 - 2 < CCα b 3.49 . 10 - 2 4.73 In e al a a 95% con idence le el (% w/w) (4.62 . 10 - 2 , 8.21 . 10 - 2 ) b (6.75 . 10 - 3 , 6.20 . 10 - 2 ) b (4.17, 5.52) b Sunsc een c eam 2 Concen a ion (% w/w) 8.53 . 10 - 2 < CCα b 3.19 . 10 - 2 3.49 In e al a a 95% con idence le el (% w/w) (6.70 . 10 - 2 , 1.02 . 10 - 1 ) b (3.75 . 10 - 3 , 5.91 . 10 - 2 ) b (2.90, 4.17) b Sunsc een c eam 3 Concen a ion (% w/w) 1.70 . 10 - 4 < CCα b 3.26 . 10 - 2 4.94 . 10 - 3 In e al a a 95% con idence le el (% w/w) (1.46 . 10 - 4 , 1.92 . 10 - 4 ) c (4.50 . 10 - 3 , 5.98 . 10 - 2 ) b (4.17 . 10 - 3 , 5.95 . 10 - 3 ) c Sunsc een c eam 4 Concen a ion (% w/w) 1.11 . 10 - 4 7.84 . 10 - 3 < CCα d 8.41 . 10 - 4 In e al a a 95% con idence le el (% w/w) (8.68 . 10 - 5 , 1.34 . 10 - 4 ) c (3.92 . 10 - 3 , 1.14 . 10 - 2 ) d (-1.56 . 10 - 4 , 1.56 . 10 - 3 ) c Sunsc een c eam 5 Concen a ion (% w/w) 2.51 . 10 - 3 1.04 . 10 - 2 < CCα e 1.98 . 10 - 3 In e al a a 95% con idence le el (% w/w) (5.75 . 10 - 4 , 4.27 . 10 - 3 ) e (7.49 . 10 - 3 , 1.31 . 10 - 2 ) e (1.08 . 10 - 3 , 2.75 . 10 - 3 ) c Sunsc een c eam 6 Concen a ion (% w/w) 3.20 . 10 - 5 < CCα 2.51 . 10 - 2 6.62 . 10 - 1 In e al a a 95% con idence le el (% w/w) (6.87 . 10 - 6 , 5.56 . 10 - 5 ) c (2.40 . 10 - 3 , 4.67 . 10 - 2 ) (3.84 . 10 - 2 , 1.06) Sunsc een c eam 7 Concen a ion (% w/w) 6.35 . 10 - 3 < CCα g < CCα g 1.73 In e al a a 95% con idence le el (% w/w) (5.40 . 10 - 4 , 1.17 . 10 - 2 ) g (1.56, 2.00) g a α = β = 0.05. b The concen a ion alue was calcula ed using he ex ac dilu ed 10000 imes (n = 2). c The concen a ion alue was calcula ed using he ex ac dilu ed 10 imes (n = 1). d The concen a ion alue was calcula ed using he ex ac dilu ed 1300 imes (n = 2). e The concen a ion alue was calcula ed using he ex ac dilu ed 1000 imes (n = 2). The concen a ion alue was calcula ed using he ex ac dilu ed 8000 imes (n = 2). g The concen a ion alue was calcula ed using he ex ac dilu ed 3000 imes (n = 2). MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT Fig. 1 Rela i e abundance (%) MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT Fig. 2 8.20 8.21 8.22 8.23 8.24 8.25 8.26 Time (min) 0 2 4 6 8 10 12 104 (a) BHT In e e en 1 In e e en 2 10.18 10.19 10.20 10.21 10.22 10.23 10.24 10.25 Time (min) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 106 (b) DiBP IS In e e en Ch oma og aphic loading MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT Fig. 3 Ch oma og aphic loading 77 105 151 228 m/z a io 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 (b) 227 10 20 30 40 50 60 70 80 Sample numbe 0 2 4 6 8 10 12 14 104 (c) PREDICTION SET CALIBRATION SET MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT MANUS CRIP T ACCEP TED ACCEPTED MANUSCRIPT Fig. 4