Br. J. clin. Pharmac. (1985), 19, 693-697 Audit of a monitoring service for free phenytoin G. M. PETERSON, S. McLEAN, 2R. J. von WITT & 'K. S. MILLINGEN School of Pharmacy and 'Department of Medicine, University of Tasmania, Hobart, and 2Department of Clinical Chemistry, Royal Hobart Hospital, Tasmania, Australia This study assessed the value of introducing the measurement of free phenytoin levels in a public hospital. After publicising the availability and purpose of the assay, free phenytoin levels were determined either (a) on the doctor's request or (b) when the total level was requested and the patient's record showed evidence of factors predisposing to an elevated unbound fraction. Total phenytoin was measured by EMIT, and the unbound fraction by ultrafiltration at 37 C using [14C]-phenytoin as a tracer. During a 9 month period, 70 free level determinations were performed on 46 patients. These comprised 20% of all phenytoin assays. The median free phenytoin fraction was 13.6% (range 9.3-28.6%). While total phenytoin levels were below the normal optimum range in 61% cases, free levels were probably therapeutic or above in 70% cases. Dosage adjustments were recommended on the basis of the free level, and were followed more often when the doctor had requested the free level assay (P < 0.05). The results suggest that a free phenytoin level assay can improve the usefulness of therapeutic drug monitoring, particularly when the doctor understands the purpose of the assay. Keywords phenytoin monitoring service Introduction While total plasma levels of phenytoin are routinely used to measure dosage-effectiveness, they only provide a reliable predictor of the anticonvulsant effect if the variation between patients in the unbound (free) plasma fraction of phenytoin is small. The variation is certainly small in vitro using plasma from healthy subjects (Yacobi et al., 1977) and in otherwise healthy epileptic patients, who infrequently receive displacing drugs (Rimmer et al., 1984). As noted by Perucca (1984), however, in practice a considerable number of patients receiving phenytoin have an impairment in plasma protein binding of the drug for various reasons: renal or hepatic disease, old age, late pregnancy, any condition associated with hypoalbuminaemia, and concomitant interacting drugs including sodium valproate, phenylbutazone, sulphonamides, and salicylates. We have previously shown that about 20% of epileptic patients receiving phenytoin have elevated unbound plasma fractions of the drug because of the factors listed above (Peterson et al., 1982). Other workers have also reported considerable variability in the unbound phenytoin fraction in epileptic patients (Walther & Meyer, 1979; DeMonaco & Lawless, 1983; Kilpatrick et al., 1984). Hence, phenytoin fulfills the three criteria proposed by Levy & Moreland (1984) for the consideration of routine measurement of free drug levels: the usefulness of total plasma level monitoring has been Correspondence: Dr G. M. Peterson, School of Pharmacy, University of Tasmania, PO Box 252C, Hobart 7001, Tasmania, Australia 693
694 G. M. Peterson et al. established, the drug is highly (about 90%) bound to plasma proteins, and it exhibits a variable unbound fraction. This study sought to test the value of introducing a free phenytoin level monitoring service at the Royal Hobart Hospital, a 600-bed public hospital. Methods The availability and purpose of the free phenytoin level assay were firstly publicised verbally and in a newsletter which is regularly distributed throughout the Royal Hobart Hospital. Unbound phenytoin fractions in plasma were then determined when either: (a) determination of the free phenytoin level was specifically requested by a doctor, or (b) determination of the total phenytoin level was requested, and the request form or patient's clinical chemistry record indicated factors which could cause an elevated unbound fraction of the drug. The unbound fraction of phenytoin was determined by ultrafiltration with the Amicon MPS-1 micropartition system using YMT ultrafiltration membranes (Kilpatrick et al., 1984), at 37 C. Radioactive 5,5-diphenyl-4-[14C]- hydantoin] of > 98% purity (New England Nuclear) was added to the plasma sample as a tracer. The unbound fraction was calculated by dividing the disintegrations per minute (d/min) in 0.1 ml ultrafiltrate by the d/min in the same volume of plasma before ultrafiltration. The coefficient of variation of the method for repeated measurements on fresh plasma containing 60 pumol/l phenytoin was 4.9% (n = 6). Total phenytoin concentrations in plasma were determined by EMIT (Syva, Palo Alto, Ca, USA), and free phenytoin concentrations were calculated from the total concentration and unbound fraction. Each patient's anticipated optimum range of total phenytoin in plasma was estimated by multiplying the normal optimum range (40-80,mol/l) by the normal value for unbound fraction of phenytoin (0.10)/ observed value. The results were presented as shown in Figure 1, and communicated to the particular hospital ward or clinic. The action taken on the basis of the results sheet was then determined by examining each patient's medical record, generally 1-2 weeks later. An albumin level (bromocresol green dyebinding method using a Technicon SMA II) and immunospecific albumin level (immunonephelometric method using a Beckman ICS analyser II) were also determined in most plasma samples. Results Over a 9 month period, 70 free phenytoin level assays were performed. These comprised 20% of all phenytoin assays. The 70 assays involved 46 patients (24 female, 22 male) aged 1 to 86 years (median 51 years). Reasons for performing the assay are listed in Table 1. The most common displacing drug was sodium valproate (13 assays). Doctors were apparently unaware of the assay's potential usefulness in elderly or pregnant patients receiving phenytoin. The unbound phenytoin fractions ranged from 9.3 to 28.6%, with a median value of 13.6%. The total phenytoin concentration was within the normal optimum range (40-80,umol/ 1) in only 21 (30%) of the cases, and below this range in 43 (61%) cases. However, if one compares the free levels found with the expected optimum range for free phenytoin (40-80,umol/l x 0.10 = 4-8,umol/l), most results were either within (49%) or above (21%) this range. Hence, use of the total phenytoin assay indicated that most patients needed increases in dosage, whereas the free phenytoin assay level showed that most patients probably required no alteration in dosage, or even a reduction. Follow-up showed that the actual dosage adjustments made were in accordance with the recommendation in all cases (15) when the free phenytoin assay had been specifically requested by the patient's doctor. However, recommendations following laboratory-initiated free level assays were apparently followed in only 41 out of 55 (75%) cases. This difference in the implementation of results ordered by the doctor and those initiated by the laboratory was statistically significant (chi-square = 4.77, 1 df, P < 0.05). These analyses include cases where the dosage adjustment recommendation is the same based on the total and free phenytoin level. To determine whether the total or free phenytoin levels were used when contemplating a dosage adjustment, one has to examine the subset of samples where the dosage adjustment recommendations differ. In practice, this arose in one of two situations: (a) when the free level was within the anticipated optimum range of 4-8,umol/l, but the total level was below the normal optimum range of 40-80,umol/l (recommended no change in dosage), or (b) when the free level was above 8,umol/l, but the total level was within or below the 40-80
Results sheet Short report 695 Name XSXXXXXXXXXXXX.... Number.. X... Date. 8/2184. Phenytoin is normally highly (90%) bound to plasma albumin. Only the unbound or free fraction of phenytoin in plasma has anticonvulsant activity. The commonly stated therapeutic range of phenytoin in plasma (40 to 80 pmol/l refers to the total phenytoin level and assumes that the extent of phenytoin binding to plasma albumin varies little between patients. However, the free fraction of phenytoin in plasma can be greatly increased in certain circumstances (e.g. treatment with interacting drugs such as sodium valproate, aspirin, warfarin, and phenylbutazone; pregnancy; hyperthyroidism; liver disease; and renal failure). When the phenytoin free fraction is elevated, the patient has a lower than normal therapeutic range for phenytoin, and toxicity may appear at total phenytoin levels normally within the therapeutic range. This patient has/is reeeiving... which can increase the free fraction of phenytoin in plasma. We have determined this patient's free phenytoin fraction in plasma, and the results follow. Total phenytoin concentration (omol/l) Therapeutic range, in patients with normal plasma protein binding (pmol/l) Percentage of phenytoin free in plasma (normally about 10%)... 23... 40-80 Free phenytoin concentration (pmol/l) Therapeutic range for free phenytoin (pmol/l) Anticipated therapeutic range for total phenytoin in this patient (pmol/l) Suggested change to phenytoin dosage based on assay results Figure 1 Example of free phenytoin level results sheet. 4-8 16-32... change no iecrease deereaee Table 1 Reasons for performing the free phenytoin level assay Number of assays Assay performed on: Reason Doctor's request Laboratory initiative Renal failure 2 3 Hepatic disease 5 4 Hypoalbuminaemia 4 10 Pregnant/post-partum 0 7 Age over 65 years 0 16 Displacing drug 2 15 Persistently low total 2 0 levels 15 55
696 G. M. Peterson et al.,umol/l range (recommended a reduction in dosage). In this subset of samples, the dosage adjustment recommendation based on the free level was followed in all seven cases when the doctor specifically requested the assay, but was followed in only 50% (11 of 22) cases when the assay was initiated by the laboratory. This difference was statistically significant (chi-square = 5.64, 1 df, P < 0.05). There was a moderate correlation between the unbound phenytoin fraction and the plasma albumin level measured by the routine dyebinding method (n = 54, r = -0.70, P < 0.001). A stronger correlation existed with the plasma albumin level measured by the immunospecific method (n = 65, r = -0.80, P < 0.001). There was a reasonably strong correlation between plasma albumin levels measured by the two methods (n = 52, r = 0.87, P < 0.001). The mean albumin level was 37.3 g/l (s.d. 5.7) with the dye-binding method, and 35.8 g/l (s.d. 9.1) with the immunospecific method. Discussion The results of our previous study (Peterson et al., 1982) indicated that the unbound fraction of phenytoin in plasma is elevated (> 14.5% unbound) in as many as one in five epileptic patients. This proportion was doubled in the present study, principally due to the selection criteria used. This study shows that in the subgroup of patients with abnormal phenytoin binding, use of a free phenytoin assay can improve the value of therapeutic drug monitoring as a measure of dosage-effectiveness. This was illustrated by the finding that, amongst the patients selected, total phenytoin levels were below the normal optimum range in 61% cases, while free levels were probably therapeutic or above in 70% cases. In most cases, the free level was apparently used in preference to the total level when considering dosage adjustments of phenytoin. This was more likely when doctors specifically requested the free level assay and, presumably, therefore understood its pharmacological basis, than when doctors received unrequested advice from the monitoring service. Clearly, to rationally apply a new assay's result, the doctor must understand the purpose of the assay. It should be recognised, however, that the statistical analyses did not take into account the clinical status of each patient when considering a dosage adjustment of phenytoin. Furthermore, there was no attempt in this study to determine the influence of monitoring free phenytoin levels on the clinical control of patients. Additional clinical research is obviously required, for instance, to confirm that the optimum range of free phenytoin concentrations in plasma matches that extrapolated from the total concentrations. A free phenytoin level assay performed after the audit period further illustrates the clinical usefulness of the assay. A 55 year old female patient with renal failure due to chemotherapy for multiple myeloma (serum creatinine 444,umol/l; albumin 16 g/l by dye-binding method, 11 g/l by immunospecific method) was receiving 500 mg phenytoin daily. She was in a semiconscious state and having generalized major seizures. Her total phenytoin level was only 57,umol/l, but the unbound fraction was 42% and the free level 24,u mol/l, well above the expected optimum range. Her phenytoin was ceased, but she died 2 days later when peritoneal dialysis was discontinued. Finally, our results indicate that plasma albumin levels measured by an immunospecific assay provide a better indicator of phenytoin binding capacity than albumin levels measured by the traditional and less specific dye-binding method. We suggest that the immunospecific method be routinely used for determining plasma albumin levels in protein binding studies. We thank the staff of the Clinical Chemistry Department at the Royal Hobart Hospital for their assistance, and Miss H. Lawler for typing the manuscript. References DeMonaco, H. J. & Lawless, L. M. (1983). Variability of phenytoin protein binding in epileptic patients. Arch. Neurol., 40,481-483. Kilpatrick, C. J., Wanwimoiruk, S. & Wing, L. M. H. (1984). Plasma concentrations of unbound phenytoin in the management of epilepsy. Br. J. clin. Pharmac., 17, 539-546. Levy, R. H. & Moreland, T. A. (1984). Rationale for monitoring free drug levels. Clin. Pharmacokin., 9 (Suppl. 1), 1-9. Perucca, E. (1984). Free level monitoring of antiepileptic drugs. Clinical usefulness and case studies. Clin. Pharmacokin., 9 (Suppl. 1), 71-78. Peterson, G. M., McLean, S., Aldous, S., von Witt,
R. J. & Millingen, K. S. (1982). Plasma protein binding of phenytoin in 100 epileptic patients. Br. J. clin. Pharmac., 14, 298-300. Rimmer, E. M., Buss, D. C., Routledge, P. A. & Richens, A. (1984). Should we routinely measure free plasma phenytoin concentration? Br. J. clin. Pharmac., 17, 99-102. Walther, H. & Meyer, F. P. (1979). Interindividual differences of protein binding in man. Int. J. clin. Short report 697 Pharmac. Biopharm., 17, 392-395. Yacobi, A., Lampman, T. & Levy, G. (1977). Frequency distribution of free warfarin and free phenytoin fraction values in serum of healthy human adults. Clin. Pharmac. Ther., 21 283-286. (Received August 21, 1984, accepted November 16, 1984)