Seizure 1992; 1: 275-279

Therapeutic drug monitoring improves seizure

control and reduces anticonvulsant side-effects in
patients with refractory epilepsy

P.J.W. MCKEE, I. PERCY-ROBB & M.J. BRODIE

Epilepsy Research Unit, University Department of Medicine and Therapeutics and the Department of Pathological
Biochemistry, Western Infirmary, Glasgow, UK

Correspondence to Dr M.J. Brodie, Epilepsy Research Unit, Department of Medicine and Therapeutics, Western Infirmary,
Glasgow Gil 6NT, UK. Tel: +41-339 8822 Ext 4572, Fax: +41-334 9329

The benefit to patients of an on-site antiepileptic drug monitoring service, in which physicians’ decisions were
recorded before and after the concentration became available, have been assessed retrospectively. In 109
patients with refractory epilepsy monitored on three (n = 61) or four (n = 48) occasions between 1986 and 1990,
one or more assay result was rated as ‘unexpected’ by the consulting doctor. In 51 of these, revision of the dose
was undertaken as a consequence (active group). Their clinical outcomes were compared with those of the other
54 patients in whom the initial management plan was left unchanged (passive group). Median seizure fre-
quency decreased in both groups, but this was statistically significant only for the actively treated patients
[active group: first visit 1 (0 to 26) seizures/month, last visit 0 (0 to 19) seizures/month, P < 0.01,95% CI -2 to
-0.5; passive group: first visit 2.5 (0 to 57) seizures/month, last visit 1.5 (0 to 44) seizures/month, NS, 95% CI
- 1 to + 101. The number of patients reporting one or more episodes of drug-related toxicity also favoured the
actively managed patients (active group 25%, passive group 56%, P < 0.01). Within this selected group of
patients, the anticonvulsant concentration was a better guide to clinical management than the physician’s
assessment.

Key worok: Antiepileptic drugs; epilepsy; pharmacokinetics; therapeutic drug monitoring.

INTRODUCTION the Western Infirmary in Glasgow since 1984.

The clinicians’ clinical decisions have been
recorded before and after the drug concen-
The potential for reducing the variation in
tration was made availableg. Once experience
inter-individual drug response using the circu-
was gained with the service, a reduced number
lating concentration to adjust the dose has long
of visits were monitored and more assay
been recognized. With the introduction of
results were observed within the target ranges
cheaper and more convenient assay tech-
for the monitored drugs”. We now ask
niques, therapeutic drug monitoring has
whether this implied benefit can be translated
become widely accepted into clinical practice’.
into an improvement in patient management
The pharmacological characteristics of the
by examining seizure frequency and neuro-
major antiepileptic drugs make them suitable
toxic side-effects in a group of epileptic
candidates for monitoring2. However, the
patients monitored between the years 1986
unthinking measurement of anticonvulsant
and 1990.
concentrations without taking into consider-
ation the clinical status of the patient has
attracted criticism3s4 and calls have been made METHODS
for prospective trials to determine the value of
Patients
routine monitoring5-8.
An anticonvulsant drug monitoring service An average of 26 patients attended the epi-
has operated on-site at the epilepsy clinic in lepsy clinic weekly throughout 1986 to 1990.

1059-131 l/92/040275+05 $08.0010 @ 1992 Bailkre Tindall

276 P.J.W. McKee et al.

Patients were recruited prospectively for anti- patients (n = 7) taking primidone or phenobar-
convulsant monitoring. Criteria for selection bital was too low for analysis and these have
included poor seizure control, dose titration, been excluded.
unwelcome polypharmacy, clinical toxicity,
suspect compliance and the introduction of an
additional anticonvulsant drug. On arrival at Statistics
the clinic, a venous blood sample was with-
drawn and transferred into a lithium heparin Comparisons between the active and passive
tube for immediate analysis. The reason for groups of patients were made using the Chi-
assay, anticonvulsant dose(s), time of last dose squared test for nominal and the Mann Whit-
and of blood sampling were entered on a spe- ney U test for interval data. Differences be-
cially designed form by the consulting doctor. tween the groups for interval data were
After the initial interview with the patient, the explored with the Wilcoxon rank sum test for
clinician recorded his or her management matched pairs. Confidence intervals were cal-
decisions on the same form. When the assay culated to 95% around median values.
result became available, usually around
15min later, the doctor reviewed these while
the patient waited. A number of clinicians, RESULTS
ranging from senior house officer to consultant,
assessed the patients during the years of the Of the 109 patients monitored on three (61
study. patients) or four (48 patients) occasions, who
A total of 40% of 6801 clinic attendances had at least one unexpected assay result, 51
were monitored during the 5 year period, were actively managed. The remaining 54
during which 2857 assays were performed in comprised the passive group. Fifty-seven of
618 patients”. Most patients were monitored these patients were treated with anticonvul-
more than once. When an assay suggested that sant monotherapy, while 48 took two and the
the management plan was inappropriate, it remaining four patients three antiepileptic
was recorded as ‘unexpected’. Those monitored drugs. A total of 377 assays were performed in
on three or four occasions, in whom at least one 336 visits (147 PHT, 138 CBZ, 92 VPA). Four
anticonvulsant concentration had been docu- patients were excluded from analysis because
mented by the doctor as ‘unexpected’, were their seizure data were incomplete.
selected for analysis. The clinical outcomes in The proportion of different seizure types
patients in whom this knowledge had resulted (active group: partial 31%, generalized 69%;
in a revision of drug dose (active group) were passive group: partial 50%, generalized 50%)
compared with those in whom the doctor felt, and the median duration of follow-up [active
on clinical grounds, that the initial therapeutic group: 11.5 (3 to 42) months; passive group: 10
plan should be followed (passive group). (3 to 38) months, 95% CI -2 to +33 did not
The case notes for each patient were exam- differ significantly between the groups.
ined retrospectively. Monthly seizure fre- Median monthly seizure frequencies prior to
quency was taken from the standard Western the first unexpected and last available assay
seizure charts, compiled routinely by all result for both groups are shown in Table 1.
patients, for the 3 months prior to the first Although an improvement in seizure control
unexpected result and the last monitored visit. was observed during the study period in both
Episodes of drug-induced toxicity were diag- sets of patients, this was statistically signifi-
nosed clinically. To be accepted, the event cant (P < 0.05) only for the actively managed
required the patient to complain of symptoms patients.
consistent with the known side-effect profile of The improvement in seizure control among
the drug, which resolved on reducing the dose. the active group was associated with a signifi-
cant increase in drug concentrations (Table 1).
Within the passive group, the concentrations of
Assays all drugs were also observed to rise between
visits, but significantly so only for CBZ. Com-
Carbamazepine (CBZ), phenytoin (PHT) and parisons of seizure frequencies, drug levels and
sodium valproate (VPA) concentrations were times since dosing at first and last clinic visits
measured on-site by enzyme immunoassay revealed no important differences between the
(EMIT, Syva, Palo Alto, USA). The number of groups, except for CBZ. Median CBZ levels
Therapeutic drug monitoring 277

Table 1: Median anticonvulsant concentrations (mg/l) and monthly seizure frequencies for the period between the first
unexpected and last available assay results
First visit Last visit P value 95% Confidence intervals

Carbamazepine concentration active (n = 12) 4.5 7 0.01 +1.5 to +5

passive (n = 28) 8.5 10 0.02 +0.5 to +3
seizures active (n = 12) 1 0 0.001 - 1 to -0.5
passive (n = 28) 1 1.5 NS oto+3
Phenytoin concentration active(n = 39) 10 18 0.01 +5 to +7
passive (n = 13) 14 17 NS -3.5 to +1
seizures active (n = 39) 1 0 0.002 -2 to -0.5
passive (n = 13) 1 0 NS -0.5 to +1.5
Sodium valproate concentration active (IZ = 18) 54 82 0.01 +3.5 to +43
passive (n = 15) 70 77 NS -26 to +12
seizures active (n = 18) 3 1 0.04 -5too
passive (n = 15) 6 2 NS -2 to +5.5

were significantly lower at both visits in the and fewer side-effects’* have been reported by
active group (first visit: active group 4.5mg/l, some workers, whereas others have demon-
passive group 8.5 mg/l, P < 0.01, 95% CI +2 to strated no such benefits15*16. Over-reliance on
+6; last visit: active group 7mg/l, passive the drug level without regard for the patients’
group lOmg/l P < 0.05, 95% CI 0 to +4). The clinical circumstances17 and inadequate
median time of sampling since dosing at the knowledge of applied pharmacokineticsls~lg
last visit was shorter in the passive compared may provide an explanation for some of the
with the active group (hours after dosing: negative results. Trials evaluating monitoring
active group 6.5 h, passive group 4 h, P < 0.01, services must take into account the complex
95% CI -4 to -1). interactions between the clinical character-
A total of 53 patients reported neurotoxic istics of the patients and the physician’s ability
side-effects during the course of the study. and experience in interpreting drug concen-
These included sedation (n = 451, dizziness (n. trations, otherwise potential benefit may be
= 43), poor concentration (n = 40) and diplopia overlooked2’. Even among physicians familiar
(n = 10). Fewer patients had an episode of with pharmacokinetic principles, a learning
drug-related toxicity in the active than in the period is required to establish expertise in
passive group (active group: 25% patients, pas- choosing the most suitable patients for moni-
sive group: 56% patients, P < 0.01). The data toringl’.
for each drug are shown in Table 2. Fewer The prospective evaluation of our on-site
monitoring service was designed to explore
Table 2: Number of patients experiencing at least one changes in management decisions brought
episode of drug related toxicity
about by the immediate availability of anticon-
Drug Active Passive P Value vulsant concentrationsg,lO. Improving seizure
Carbamazepine o/12 (0%) 21128 (75%) co.01 control and decreasing the frequency of epi-
Phenytoin 15139 (38%) 6113 (46%) NS sodes of drug-induced toxicity were not specifi-
Sodium valproate 5118 (28%) 6115 (40%) NS cally targeted in the initial protocol. However,
such data were collected and provided the basis
patients treated with CBZ in the active group for the clinical assessment of each patient. The
reported neurotoxic side-effects (P < 0.01). opportunity was, therefore, taken to evaluate
Although more patients taking PHT and VPA retrospectively the impact of anticonvulsant
also experienced side-effects amongst the pas- monitoring on these clinical end-points.
sive group, these differences failed to reach sta- Patients monitored on three or four occasions
tistical significance. were chosen arbitrarily as likely to have suf-
ficient frequency of drug estimations to look for
potential benefit.
DISCUSSION Active and passive patient groups were
chosen for evaluation since they were identi-
Trials of the value of monitoring antiepileptic fied and treated in a similar way. The critical
drugs have been few and have resulted in con- difference between the two was the attending
flicting claims. Savings in medical time11,12, doctor’s decision to alter his management in
decreased number of assays performed10~11,13 response to the drug concentration in one
278 P.J.W. McKee eta/.

group, but not in the other. Among these pliance), the physician should be guided by the
selected patients, the merit of acting upon the drug concentration in choosing the correct
drug level or ignoring it was explored. course of action.
Overall, a significant improvement in seiz- Although this study suggests that the serum
ure frequency was observed among the drug concentration can favourably alter clini-
actively, but not the passively treated patients. cal management, it was undertaken in a select
The associated changes in drug concentrations group of patients. We must proceed now to
between visits are difficult to interpret. explore the value of therapeutic drug monitor-
Increases and decreases in dosage were made ing among the whole clinic population using
in both groups. The drug concentration data, clinical end points in patients randomized to
therefore, represent an overall summary of monitored and unmonitored groups.
these alterations. That the anticonvulsant
levels increased significantly among the active
group for all drugs suggested that they were ACKNOWLEDGEMENT
generally unexpectedly low. This is reflected
also in the lower drug levels in the active group Our grateful thanks go to Mrs Moya Dewar for
at the first visit, although this was only stat- expert secretarial assistance.
istically significant for CBZ. The serum CBZ
concentration also increased between visits in
the passive group. However, since the magni- REFERENCES
tude of the CBZ concentration is proportional
to the time after dosingg, these differences may 1. Brodie, M.J. and Feely, J. Practical clinical pharmaco-
logy. Therapeutic drug monitoring and clinical trials.
be a kinetic artefact.
British Medical Journal 1989; 296: 1110-1114.
Fewer toxic events were recorded in the 2. Thomson, A.H. and Brodie, M.J. Pharmacokinetic
active group. In terms of single anticonvulsant optimisation of anticonvulsant therapy. Clinical Phar-
drugs, this was statistically significant only for macokinetics 1992; 23: 216-230.
CBZ. This was slightly surprising since PHT 3. Brodie, M.J., McIntosh, M.E. and Hallworth, M. Thera-
peutic drug monitoring-the need for audit? Scottish
and VPA are more difficult to use clinically’i. Medical Journal 1985; 30: 75-82.
However, the range of concentrations at the 4. Chadwick, D.W. Overuse of monitoring of blood con-
final visits included a higher maximum value centrations of antiepileptic drugs. British Medical
for PHT and higher minimum level for VPA in Journal 1987; 2: 123-724.
the active compared with the passive group, 5. Hvidberg, E.F. Monitoring drug plasma levels in clini-
cal practice: a procedure for evaluation is needed. Eur-
suggesting more effective drug levels among
opean Journal of Clinical Pharmacology 1980; 17: 317-
the former. 319.
In this study, management decisions that 6. Spector, R., Park, G.D., Johnson, G.F. and Vessell, ES.
ignored the availability of an unexpected anti- Therapeutic drug monitoring. Clinical Pharmacology
convulsant concentration were associated with and Therapeutics 1988; 43: 345-353.
McInnes, G.T. The value of therapeutic drug monitor-
poorer seizure control and an increased inci- ing to the practising physician-an hypothesis in need
dence of concentration-related side-effects. of testing. British Journal of Clinical Pharmacology
During the years 1986 to 1990, 26% of total 1989; 27: 281-284.
concentrations were categorized as ‘unex- Cridland, J.S. The value of therapeutic drug monitor-
pected’ and in 60% of these a change in drug ing. British Journal of Clinical Pharmacology 1990; 28:
278-279.
dose resulted lo . This study has implications, Larkin, J.G., Herrick, A.L., McGuire, G.M., Percy-
therefore, for the management of 10% of moni- Robb, I.W. and Brodie, M.J. Antiepileptic drug moni-
tored patients, in whom no such alteration was toring at the epilepsy clinic: a prospective evaluation.
made in the face of an unexpected assay result. Epilepsia 1991; 32: 89-95.
We have stressed previously the importance of 10. McKee, P.J.W., Larkin, J.G., Brodie, A.F., Percy-Robb,
I.W. and Brodie, M.J. Five years of drug monitoring on
using the anticonvulsant concentration as a site at the epilepsy clinic. Therapeutic Drug Monitor-
guide to dosage alteration in the light of the ing 1992; in press.
clinical status of the patientl*g,lO. Although 11. Patsalos, P.N., Sanders, J.W.A.S., Daley, J. and Las-
this continues to be the correct strategy, an celles, P.T. Immediate anticonvulsant drug monitoring
unexpected assay result should alert the in management of epilepsy. Lancet 1987; ii: 39.
12. Elliot, K., Watson, I.D., Tsintis, P., Gray, J., Stewart,
physician to undertake a thorough revision of M.J., Hadr, H. and Lawson, D.H. The impact of near-
the patient’s clinical circumstances. If no patient testing on the organisation and costs of an
obvious explanation can be identified (e.g. anticonvulsant clinic. Therapeutic Drug Monitoring
timing of sampling, drug interaction, poor com- 1990; 12: 434-437.
Therapeutic drug monitoring 279

13. Wing, D.S. and Duff, H.J. The impact of a therapeutic dose be increased? A randomised prospective study.
drug monitoring programme for phenytoin. Thera- Epilepsia 1988; 29: 129-139.
peutic Drug Monitoring 1989; 11: 32-31. 18. Beardsley, R.S., Freeman, J.M. and Appel, F.A. Anti-
14. Reid, L.D., Horn, J.R. and McKenna, D.A. Therapeutic convulsant serum levels are useful only if the
drug monitoring reduces toxic drug reactions: a meta- physician appropriately uses them: an assessment of
analysis. Therapeutic Drug Monitoring 1990; 12: 72- the impact of providing serum level data to physicians.
78. Epilepsia 1983; 24: 330-335.
15. Froscher, W., Eichelbaum, M., Gugler, R., Hilden- 19. Levine, M., McCollom, R., Chang, T. and Orr, J. Evalu-
brand, G. and Penin, H. A prospective randomised trial ation of serum phenytoin monitoring in an acute care
on the effect of monitoring plasma anticonvulsant setting. Therapeutic Drug Monitoring 1988; 10: 51-57.
levels in epilepsy. Journal of Neurology 1981; 224: 20. Reid, L.D., McKenna, D.A. and Horn, J.R. Effect of
193-201. therapeutic drug monitoring services on the number of
16. Schumacher, G.E., Barr, J.T., Browne, T.R. and Col- serum drug assays ordered for patients: a meta-analy-
lins, J.F. Test performance characteristics of the serum sis. Therapeutic Drug Monitoring 1989; 11: 253-263.
phenytoin concentration @PC): the relationship be- 21. Brodie, M.J. Established anticonvulsants and the
tween SPC and patient response. Therapeutic Drug treatment of refractory epilepsy. Lancet 1990; 336:
Monitoring 1991; 13: 318-324. 350-354.
1’7. Woo, E., Chan, K.M., Yu, Y.L., Chan, Y.W. and
Huang, C.Y. If a well stabilised epileptic patient has a
subtherapeutic antiepileptic drug level, should the