ANNALS O F C LIN IC A L AND LABORATORY SCIEN CE, Vol. 6, No. 5 Copyright © 1976, Institute for Clinical Science

Therapeutic Drug Monitoring: Measurement of Antiepileptic and Drug Levels in Blood by Gas Chromatography with —Selective Detector DEREK P. LEHANE, Ph.D., PAUL MENYHARTH, B.S., GIFFORD LUM, M.D., and ARTHUR L. LEVY, Ph.D.

Chemistry Laboratory St. Vincent’s Hospital and Medical Center of New York, New York, NY 10011

ABSTRACT The nitrogen-specific detector for gas chromatography consists of a non­ volatile rubidium silicate bead, around which nitrogen-containing com­ pounds are pyrolyzed. Speed, sensitivity, specificity, accuracy, small sam­ ple size and minimum sample handling are characteristics of the nitrogen detector that render it superior to conventional gas chromatographic detec­ tors. The detector has been utilized to effect a simple and rapid quantita­ tion of allobarbital, , , heptabarbital, pentabarbital, phénobarbital and , plus the diphenylhydan- toin and . Extraction of the drugs from acidified serum into or­ ganic solvent containing internal standard is followed by oncolumn méthy­ lation with methanolic trimethylphenyl ammonium hydroxide. The drugs, separated on a column of 3 percent OV-lOl on Gas-Chrom Q, 100-120 mesh, are readily quantitated by simple calculations based on peak-height ratios. Therapeutic drug monitoring is discussed in relation to recent con­ cepts of drug-protein binding, drug-drug interactions, drug biotransforma­ tion and problems of multiple drug therapy.

Introduction chromatographic analysis of phosphorus- and nitrogen-containing compounds by Therapeutic drug monitoring is techni­ introduction of a specific alkali flame de­ cally demanding but extremely useful tector. Various commercial modifications because it correlates patient status di­ of this thermionic detector have been rectly with blood levels. Currently, gas- applied to toxicologic assays; recent clin­ chromatographic procedures are being ical applications of the detector include widely applied to the routine monitoring quantitation of blood levels of tricyclic of drug levels in blood. Karmen and Giuf- ,3 analgesics6 and drugs of frida7 increased the sensitivity of gas- abuse.15 4 0 4 THERAPEUTIC DRUG MONITORING 405 Since and the anticonvul­ sants, diphenylhydantoin and primidone, CN* CN each contain two nitrogen atoms, they are ideally suited to gas-chromatographic analysis using the nitrogen-specific de­ Rb Rb* tector. As the detector signal is propor­ surface tional to the nitrogen content of the mol­ of the ecule, the presence of two nitrogen atoms L------JL. bead favors high sensitivity, increased approx­ Rb++e" Rb* imately ten-fold over that of conventional \ flame ionization detectors.15 Recently, the nitrogen detector has been success­ FIGURE 1. Proposed mode of action of the fully applied to the analysis of blood nitrogen-selective detector.8 levels of both barbiturates2 and an­ tiepileptic drugs.5,16 Based on these pro­ cyan radical takes up one electron from cedures, a rapid and reliable assay has the alkali, resulting in a symmetrical been developed for diphenylhydantoin, cyanide ion and a positive alkali which is primidone and eight barbiturate drugs. captured by the bead (figure 1). The cyanide ion migrates to the collector elec­ The Nitrogen-specific Detector trode, but presumably liberates one elec­ tron either by oxidation, resulting in The nitrogen-specific detector is a neutral products, or by reaction with hy­ modification of the standard flame ioniza­ drogen atoms, forming HCN. The elec­ tion detector. A glass bead containing al­ tron thus is collected and is responsible kali as a primarily non-volatile rubidium for the specific nitrogen response. silicate is positioned between the jet and The thermionic nitrogen detector dis­ the collector electrode. The bead is al­ plays significant advantages over conven­ ways at a negative potential of 130v tional flame-ionization or electrón- against the collector electrode. Kolb and capture detectors, not the least being its Bischoff8 have presented a concise ac­ speed, sensitivity, specificity, accuracy, count of the theoretical aspects of the ni­ small sample size, minimum sample trogen detector. The specific sensitivity handling and detection limits indepen­ toward nitrogen-containing compounds dent of the matrix.15 The detection limit is facilitated by a low hydrogen flow rate of the nitrogen detector is 0.1 Pg N, with of 10 ml per minute together with an air­ a 105- fold linearity. flow of approximately 100 ml per minute. The heat generated from such a low Principle volume of hydrogen is insufficient for Internal standard (5-(p-methyl- achieving the necessary temperature for phenyl)-5-phenylhydantoin) in organic alkali emission by the bead; an additional solvent is added to serum. The drugs pass electrical heater is therefore employed. into the organic phase, which is removed The hydrogen is burned around the hot and evaporated to dryness. The residue is bead, thus forming a diluted and cold dissolved in methanolic trimethylphenyl flame zone. If an organic nitrogen- ammonium hydroxide; injection of this containing compound enters this flame solution into the injection port heater of zone around the bead, it is pyrolyzed the gas-chromatograph results in quan­ rather than oxidized, resulting in the pro­ titative methylation of the antiepileptic duction of cyan (C = N) radicals. Such a and barbiturate drugs. 4 0 6 LEHANE, MENYHARTH, LUM AND LEVY

Reagents Special Apparatus Included are: Necessary reagents include: Reactivials, 1.0 ml.** Extraction solvent: - Rolded filter paper, 12.5 cm diame­ isopropanol-: 94-4-1, v/v/v. ter. tf Hydrochloric acid, 1.0M. Gas-chromatograph model #900, with Methanolic trimethylphenyl am­ heated nitrogen detector. $$ monium hydroxide, 0.1M.* Column: single glass column, 1.83 m x 5-(p-methylphenyl)-5-phenylhydantoin.f 2 mm internal diameter, packed with 3 Diphenylhydantoin. j percent OV-lOl on Gas-Chrom Q, 100- Allobarbital, amobarbital, butabarbital, 120 mesh.§§ pentabarbital, and seco­ Column oven: programmed one min­ .§ ute at 175°, then 24° per minute to 285°, Heptabarbital.11 then four minutes at 285°. Ion-free serum.H Injection port heater: 300° Detector oven: 285° Stock Standard Solutions Gas flow rates: helium 30 ml per min­ ute; hydrogen 10 ml per minute; and air Each drug is dissolved in to 100 ml per minute. give a final concentration ofone g per liter. Before initial use, the column is con­ ditioned overnight at 300°, with a helium Working Standard Solutions flow rate of 30 ml per minute.

Aliquots of each of the stock standard Procedure solutions are mixed and diluted with To a teflon-lined screw-capped tube methanol to give a level for each drug of are added, in sequence, 1 ml serum or 100 mg per 1. From this solution aliquots standard, 0.1 ml 1.0M hydrochloric acid are diluted in ion-free serum to give and 12 ml extraction solvent containing working standard mixtures containing 5 internal standard. The mixture is then mg per 1, 10 mg per 1, 20 mg per 1, 30 mg shaken vigorously for one minute. The per 1, 40 mg per 1, and 50 mg per 1, respec­ organic phase is filtered into a second tively, of each drug. tube and evaporated to dryness at 50° under a stream of nitrogen. The dried res­ Internal Standard Working Solution idues can be stored in stoppered tubes at room temperature prior to chromatog­ The internal standard, 5-(p-methyl- raphy. phenyl)-5-phenylhydantoin, is dissolved The sides of the tube are washed with in extraction solvent to give a concentra­ 0.8 ml extraction solvent. The solution is tion of 1.5 mg per 1. transferred to a Reactivial and is evapo­ rated to dryness under nitrogen.

* Eastman Organic Chemicals, Rochester, NY 14650. f Aldrich Chemical Company, Milwaukee, ** Pierce Chemical Company, Rockford, IL WI 53233. 61105. $ Sigma Chemical Company, St. Louis, ff Schleicherand Schuell, Keene, NH 03431. MO 63178. £{ Perkin-Elmer Corporation, Norwalk, CT § Gane’s Chemical Works, Carlstadt, NJ 07072. 06856. 11 Geigy Pharmaceuticals, Ardsley, NY 10502. §§ Applied Science Laboratories, State College, 11 Clinton Laboratories, Santa Monica, CA 90404. PA 16801. THERAPEUTIC DRUG MONITORING 4 0 7 Methanolic trimethylphenyl am­ monium hydroxide (50 ¡A) is added to the vial. The contents are mixed and an aliquot (0.7 /xl)is injected into the gas- chromatograph. Calculations Peak heights are measured for drugs, standards and internal standards. A height ratio for each standard concentra­ tion is calculated: Peak height (standard) _ Peak height (internal standard) R (height ratio) and values of R are plotted graphically against the respective standard concen­ tration. Values of R are similarly calculated for each peak of an analytical sample, and the drug concentration read from the re­ spective standard curve. Results A typical chromatogram of the separa­ tion of the various barbiturate and an­ tiepileptic drugs is shown in figure 2. Application of temperature programming in conjunction with the nitrogen- selective detector achieves maximal sep­ aration of the drugs in the shortest time span, each sample cycle being 12 m in­ utes duration. Calibration of drug levels is performed rapidly and simply by plot­ ting peak height ratios; in this regard, the nitrogen detector is of particular value since the internal standard (peak 10 in figure 2) appears as a sharp spike, while FlGURE 2. Gas chromatogram illustrating the temperature-programmed separation of the barbitu­ with conventional flame ionization detec­ rate and drugs in ion-free serum, tion the internal standard appears as a with relative retention ratios (internal standard = low, broad peak. As shown in figure 2, 1.00). OV-lOl, 3 percent, programmed one minute peaks are well separated from the solvent at 175°, then 24° per minute to 285°, then four min­ utes at 285°. Included are (1) allobarbital, 0.18; (2) front, while peak tailing and overlap are butabarbital, 0.24; (3) amobarbital, 0.29; (4) pen­ minimal. Other major anticonvulsants tobarbital, 0.32; (5) secobarbital, 0.36; (6) phénobarbital, 0.56; (7) heptabarbital, 0.68; (8) (e.g., , ethosuximide, me- primidone, 0.70; (9) diphenylhydantoin, 0.91; and santoin) are separated on this column and (10) 5-(p-methylphenyl)-5-phenylhydantoin (inter­ do not interfere with the analyses. Drug- nal standard). 4 0 8 LEHANE, MENYHARTH, LUM AND LEVY

TABLE I

Comparison of Stated vs. Found Concentrations of Diphenylhydantoin, Phénobarbital and Primidone in Four "Spiked" Human Serum Pools

Pool 1 Pool 2 Pool 3 Pool 4 \xmoles per liter ]imoles per liter ]imoles per liter \imoles per liter Stated Found Stated Found Stated Found Stated Found

Diphenylhydantoin 19.8 19.8 39.6 39.6 79.3 80.1 79.3 80.9 Phénobarbital 43.1 42.6 86.1 87.0 172.2 171.4 86.1 87.0 Primidone 22.9 24.3 45.8 44.9 68.7 67.4 free sera, when extracted and chromatog­ matrix, gave the following data (concen­ raphed, show no interfering peaks. Phos­ tration range of added drug in par­ pholipids, which contain nitrogen, do not enthesis): diphenylhydantoin, 95 to 100 display interfering peaks, even when percent, (2.5 to 30.0 mg per 1), primidone, present in serum at supranormal levels. 96 to 98 percent, (2.5 to 20.0 mg per 1), Neutral lipids (e.g., triacylglycerols, and phénobarbital, 97 to 99 percent, (5.0 ) are not detected by the ni­ to 80.0 mg per 1). trogen detector. Therapeutic Ranges Calibration Quantitative estimation of blood bar­ The relationship of standard concentra­ biturate levels is of importance in assess­ tion to peak height ratio in the range 5.0 ing stages of sedation. For example, a to 50.0 mg per 1 is linear for each of the level of 350 /mmoles amobarbital per 1 can nine drugs, the calibration curve in each result in respiratory and/or circulatory case intersecting the origin. Correlation difficulty. However, the concept of sub- coefficients (r2) between mean standard optimal, therapeutic and toxic drug levels concentration and the mean of peak is, in this instance, usually confined to height ratios exceed 0.995. compounds administered as anticonvul­ sants. Kutt and Penry10 have presented Reliability ranges that are generally accepted as Accuracy and reproducibility studies “therapeutic”: diphenylhydantoin, 40 to were performed on four “spiked” human 80 /¿moles per 1, phénobarbital, 65 to 170 serum pools.* Data for accuracy are //.moles per 1, and primidone, 23 to 55 shown in table I. Representative day-to- //.moles per 1. However, there is not univ­ day coefficients of variation at therapeu­ ersal agreement on these ranges; To- tic and toxic levels ranged as follows: seland16 advocates lower levels (12 to 63 diphenylhydantoin, 4.0 to 6.7 percent, /¿moles per I) for diphenylhydantoin, phénobarbital, 5.1 to 6.4 percent, and while Norell, Lilienberg and Gamstorp11 primidone, 6.1 to 7.2 percent (N = 10). in their studies on juvenile epileptics, rec­ Within day coefficients of variation for ommend rather higher levels (48 to 100 these drugs were 1.3 percent, 3.5 percent /¿.moles diphenylhydantoin per 1). As our and 3.5 percent, respectively. understanding of therapeutic drug Recovery studies, performed by addi­ monitoring increases, these discrepan­ tion of increasing amounts of the anticon­ cies will certainly be resolved. vulsants to an analyzed human serum Sources of Error * Syva, Palo Alto, CA 94304 and Lederle Diagnos­ Hemolyzed specimens will yield low tics, Pearl River, NY 10965. diphenylhydantoin levels, owing to un- THERAPEUTIC DRUG MONITORING 409 equal distribution between plasma and small changes in the daily drug intake red blood cells.12 will greatly alter the blood level. Addi­ Prolonged storage of trimethylphenyl tionally, the effect of anticonvulsants in ammonium hydroxide derivatives may increasing hepatic hydroxylases has been lead to destruction of phénobarbital.12 shown to produce abnormalities in liver function and calcium metabolism .1,9 In Discussion several cases, diphenylhydantoin therapy The circulating drugs exist in plasma has caused severe, and sometimes fatal, in free and protein-bound forms; pro­ liver damage.1 tein bound diphenylhydantoin and Drugs (e.g., salicylates, phenyl­ phénobarbital usually account for 90 per­ butazone) displace diphenylhydantoin cent and 50 percent, respectively, of the from plasma binding sites leading to its total circulating concentrations of these rapid biotransformation by the liver. drugs.4 Most analytical procedures mea­ Conversely, saturation of the metaboliz­ sure total drug concentrations in plasma. ing capacity of the liver by other medica­ Since considerable differences in bind­ tions may lead to diphenylhydantoin in­ ing characteristics of plasma proteins toxication. When anticonvulsants are occur in uremia and in newborn infants, monitored, drugs (e.g., disulfiram, sulthi- causing a shift of drug to tissues with a ame or isoniazid) which cause predicta­ drop in plasma drug level,4 therapeutic bly significant elevations of diphenylhy­ drug monitoring is indicated in these dantoin levels in the majority of patients, states. Troupin and Friel17 have sug­ can be administered with relative safety.9 gested that, since the percent of un­ Primidone is in part oxidized to bound drug is more closely related to phénobarbital and, at steady state clinical intoxication than the total level, equilibrium, the ratio of phénobarbital to serum dialysate or salivary drug levels the parent drug in blood approximates might prove of greater value. Serum un­ 2: l .15 Combined therapy with diphenyl­ bound drug levels have correlated well hydantoin and primidone can, however, with dialysate and salivary levels in the result in even greater blood levels of case of diphenylhydantoin, but not in the phénobarbital (the phénobarbital— case of primidone or phénobarbital. primidone ratio approximates 4:l ).13 The metabolic alteration of drugs can Nevertheless, drug interactions may have have profound effects on plasma drug a beneficial clinical impact in that eleva­ levels and the relationship of dosage to tion of a low ineffective level may im­ blood levels. The principal routes of drug prove seizure control. Monitoring blood biotransformation are by hydroxylation, levels of anticonvulsants provides the oxidation, 3-0-methylation and conjuga­ best means to anticipate such interactions tion.4 Richens and Dunlop14 have shown and to regulate doses when multiple that the hydroxylation of diphenylhydan­ medications are clinically indicated9 toin by hepatic microsomal enzymes is a References saturable process; the rate of metabolism 1. Buch Andreasen, P., L y n g b y e , J., and fails to increase in proportion to the T r o l l e , E.: Abnormalities in liver function serum concentration of the drug, leading tests during long-term diphenylhydantoin therapy in epileptic out-patients. Acta Med. to a non-linear relationship between dose Scand. 294:261-264, 1973. and the resulting blood level. The steep­ 2. D v o r c h i k , B. H.: Gas chromatographic ness of this relationship within the method for the microdetermination of barbitu­ rates in blood using a nitrogen-selective flame therapeutic range may result in therapeu­ ionization detector. J. Chromatogr. 105:49-55, tic drug levels which are unstable, since 1975. 4 1 0 LEHANE, MENYHARTH, LUM AND LEVY

3. G if f o r d , L. A., Turner, P., and P a r e , C. M. 11. N o r e l l , E., Lilienberg, G., and G a m s to rp , B.: Sensitive method for the routine determi­ I.: Systematic determination of the serum nation of tricyclic antidepressants in plasma level as an aid in the management of using a specific nitrogen detector. J. children with epilepsy. Europ. Neurol. Chromatogr. 205:107-113, 1975. 23:232-244, 1975. 4. G l a z k o , A. J.: Antiepileptic drugs; bio­ transformation, metabolism, and serum half- 12. Pippenger, C. E. and K u tt, H.: Common er­ life. Epilepsia 26:367-391, 1975. rors in detection of anticonvulsant drugs by gas 5. G o u d ie , J. H. and B u r n e t t , D.: A gas- liquid chromatography. Clin. Chem. 29:666, chromatographic method for simultaneous de­ 1973. termination of phenobarbitone, primidone, 13. R e y n o ld s , E. H., Fenton, G., Fenwick, P., and phenytoin in serum using a nitrogen detec­ J o h n s o n , A. L. and L a u n d y , M.: Interaction of tor. Clin. Chim. Acta 43:423-429, 1973. phenytoin and primidone. Brit. Med. J. 6. Jam es, S. P. and WARING, R. H.: Determina­ 2:594-595, 1975. tion of blood levels of pentazocine by gas R ic h e n s , D u n l o p , chromatography with a nitrogen detector. J. 14. A. and A.: Serum- Chromatogr. 78:417-419, 1973. phenytoin levels in management of epilepsy. 7. K a rm e n , A. and Giuffrida, L.: Enhancement Lancet 2:247-248, 1975. of the response of the hydrogen flame ioniza­ 15. R ie d m a n n , M.: Analytische Methoden des tion detector to compounds containing halo­ Nachweises von Drogen und Giftstoffen. gens and phosphorus. Nature 201:1204-1205, Naturwissenschaften 59:306-310, 1972. 1964. 8. K o lb , B. and BlSCHOFF, J.: A new design of a 16. Toseland, P. A., Albani, M., and G a u c h e l, thermionic nitrogen and phosphorous detector F. D.: Organic nitrogen-selective detector for G.C. J. Chromatogr. Sci. 12:625-629, used in chromatographic determination of 1974. some anticonvulsant and barbiturate drugs in 9. K u tt, H.: Interactions of antiepileptic drugs. plasma and tissues. Clin. Chem. 21:98-103, Epilepsia 26:393-402, 1975. 1975. 10. K u tt, H. and P e n ry , J. K.: Usefulness of blood 17. TrOUPIN, A. S. and Friel, P.: Anticonvulsant levels of antiepileptic drugs. Arch. Neurol. level in saliva, serum and cerebrospinal fluid. 32:283-288, 1974. Epilepsia 26:223-227, 1975.