Are Certain Diuretics Also Anticonvulsants? Dale C. Hesdorffer, PhD,1–3 James P. Stables, MSA,4 W. Allen Hauser, MD,1–3,5 John F. Annegers, PhD,6† and Gregory Cascino, MD7 A history of diuretic use has been shown to be protective for first unprovoked seizure in adult patients. Recent animal studies suggest that certain diuretics have anticonvulsant activity. We evaluated the potential for the anticonvulsant activity of current diuretic use in a population-based, case–control study in older adults. We also tested chlorthiazide and furosemide for seizure protection in animal models of epilepsy. Concurrent medical prescription of any diuretic was protective for the development of epilepsy [odds ratio (OR) ⴝ 0.62, 95% confidence interval (CI) ⴝ 0.39 – 0.99]. A protective effect for current thiazide use was observed (OR ⴝ 0.53, CI ⴝ 0.31– 0.90), and a protective effect for furosemide was suggested (OR ⴝ 0.44, CI ⴝ 0.1–1.9). In mice, both chlorthiazide and furosemide suppressed the occurrence of maximal electroshock-induced seizures in a dose-dependent manner. Chlorthiazide’s toxic dose for 50% of animals tested (TD50) could not be achieved even with dosing as high as 1,500mg/kg for furosemide; TD50 was 549 mg/kg. Results were similar in rats. Furosemide and chlorthiazide are protective for unprovoked seizures in an epidemiological study and in animal models. Given the potential therapeutic value for seizure control, low toxicity, and low cost, therapeutic efficacy should be explored in clinical studies. Ann Neurol 2001;50:458 – 462 A history of diuretic use is associated with a decreased risk for a first unprovoked seizure in older adults, regardless of therapeutic indication or efficacy.1 Animal studies suggest that furosemide is anticonvulsant in vitro and in vivo,2,3 possibly by the induction of a hyperosmotic environment that interrupts neuronal synchronization.2 If furosemide reduces the occurrence of seizures, its effect should be restricted to the time when it is taken. Considering animal data,2,3 we examined our epidemiological data to determine whether concurrent use of diuretics protected against epilepsy. We also tested the anticonvulsant effect of chlorthiazide and furosemide in laboratory animals. Subjects and Methods Epidemiologic Data STUDY POPULATION. Subjects in this population-based, case–control study were the 145 incident cases of first unprovoked seizure between 1955 and 1984, aged 55 years or older and 290 controls matched for age, gender, and duration of medical follow-up. Patients and controls were identified using the records-linkage system of the Rochester Epidemiology Project.4 Two controls, who were Rochester residents on the day when the seizure disorder of the patient came to medical attention (index date), were selected from the same facility and were matched for age (⫾5 years), gen- From the 1Gertrude H. Sergievsky Center, 2Mailman School of Public Health, and 5Department of Neurology, Columbia University, New York, NY; Departments of 3Health Sciences Research and 7 Neurology, Mayo Clinic and Mayo Foundation, Rochester, MN; 4 Epilepsy Branch, Preclinical Pharmacology Section, National Institute of Neurological Disorders and Stroke, Bethesda, MD; and 6Division of Epidemiology, University of Texas, Houston, TX. †Deceased. 458 © 2001 Wiley-Liss, Inc. der, year of medical evaluation, and duration of medical contact. Patients and controls with a history of any of the following were excluded: clinically detected cerebrovascular disease, central nervous system (CNS) infection, brain trauma with unconsciousness or posttraumatic amnesia of 30 minutes or more, brain surgery, primary or metastatic CNS tumor, mental retardation, or cerebral palsy. Additional details have been reported elsewhere.5 This study relied on data abstraction from existing medical records, and was considered exempt from institutional review board review. Information on diuretic use was collected blind to the study hypotheses and regardless of hypertensive status. Use of other antihypertensive medications (eg, ␤-blockers) was recorded only for hypertensive individuals. Current diuretic use was defined as a prescription for diuretics at medical contact before the index date. Median time between the index date and the prior visit was 4.6 months for patients and controls (interquartile range, 2.8 –9.4 months). Past diuretic use was defined as a prescription at any medical contact prior to the last contact, but not at the visit before the index date. Diuretics were categorized as thiazides (ie, chlorthiazide, hydrochlorthiazide, bendroflumethiazide, methyclothiazide, hydrochlorthiazide with triamterene, and amiloride HCl hydrochlorthiazide), furosemide, and aldactazide (ie, spirono- DIURETIC USE. Received Mar 14, 2001, and in revised form May 23. Accepted for publication May 24, 2001. Published online Aug 10, 2001; DOI: 10.1002/ana.1136 Address correspondence to Dr Hesdorffer, Gertrude H. Sergievsky Center, Columbia University, 630 West 168th Street, New York, NY 10032. E-mail: email@example.com lactone and hydrochlorthiazide). We excluded 2 patients (current use of an unspecified diuretic and current use of a quinazoline-class diuretic) and 2 controls (past use of an unspecified diuretic and past use of a xanthine-derivative diuretic). We evaluated several possible confounders of the association between diuretics and unprovoked seizure; dementia, congestive heart failure, and a history of hypertension were included in the final model. Patients were considered demented if they met ad hoc criteria similar to those in DSM-III.6 Patients were considered to have congestive heart failure if they met at least three of the following criteria at one point in time: dyspnea on exertion, paroxysmal nocturnal dyspnea, acute pulmonary edema, distended neck veins in a position other than supine and in the absence of venous obstruction, bilateral ankle edema not due to a condition other than cardiac failure, hepatomegaly not due to liver disease, protodiastolic gallop, x-ray evidence of pulmonary congestion with or without cardiomegaly, and arm-to-tongue circulation time greater than 24 seconds. A history of hypertension was defined as at least two blood pressure readings at medical evaluation prior to the index date for nonacute conditions with systolic pressure of 160mm Hg or greater, regardless of diastolic pressure, or a diastolic pressure of 95mm Hg or greater, regardless of systolic pressure. Data were analyzed with SAS7 using conditional logistic regression for matched sets.8 Models were constructed for current or past use of any diuretic and for specific diuretics. We obtained odds ratios (ORs) and 95% confidence intervals (CIs) for diuretic use overall, current use, and past use. We then evaluated the effect of diuretic use in a conditional logistic regression model adjusting for dementia, congestive heart failure, and history of hypertension. Analyses were repeated for specific diuretics. Significance required two-tailed ␣ of 0.05. Because demented patients may have been less likely to take diuretics as prescribed, analysis of current use was repeated with all demented subjects considered to be nonusers of diuretics. STATISTICAL ANALYSIS. Animal Data We tested the anticonvulsant potential of two drugs, chlorthiazide and furosemide, against maximal electroshock seizure (MES) and subcutaneous pentylenetetrazol (Metrazol)-induced seizures. Potential toxicity of these agents was also tested in mice and rats.9,10 The MES test is a model used to evaluate a compound’s protective effects by blocking generalized tonic–clonic seizures. In male Carworth Farms No. 1 mice, an electrical stimulus of 0.2 seconds, 60Hz, 50mA, is administered via corneal electrodes. In Sprague Dawley rats, the stimulus was 150mA and of similar duration. The eyes are pretreated with a solution of 0.5% tetracaine HCl in normal saline. Protection was defined as ablation of the hindlimb tonic extensor component of the seizure.9,10 Average weights for experimental animals used were 23.5 and 120g for mice and rats, respectively. All animals had free access to food and water except during testing periods. The subcutaneous Metrazol test was utilized to evaluate SEIZURE SUPPRESSION. the ability of chlorthiazide and furosemide, administered intraperitoneally, to raise seizure threshold. A dose of 85mg/kg of Metrazol was used in male Carworth Farms no. 1 mice; 70mg/kg of Metrazol was used in Sprague Dawley rats. Animals were observed over a 30 minute interval. This chemical screen produces a behavioral seizure that is clonic in nature and can be potentiated by ␥-aminobutyric acid agonists. The endpoint is elimination of clonic spasms.10 Chlorthiazide and furosemide were evaluated in these models. Effective doses for 50% of animals (ED50) were calculated. In 8 animals per dosage, at a predetermined time of peak effect, dosing for chlorthiazide was taken from 500 to 1,500mg/kg in mice, and from 500 to 1,200mg/kg in rats. Furosemide dosing ranged from 100 to 750mg/kg in mice and from 7.5 to 240mg/kg in rats. Animals received doses of the two compounds until at least 2 points were determined between 0% and 100% protection. Slopes, 95% CI, and standard errors were calculated.11 TOXICITY. For mice, the toxicity of chlorthiazide and furosemide was assessed using the rotorod test.12 Neurological impairment was defined as the inability of the animal to remain on a rotating rod for at least 1 minute in three successive attempts. For rats, a positional sense and gait test12 was performed to identify toxicity. The animals’ hind limb was lowered over an elevated edge to see if it could quickly correct this unnatural limb position. In addition, animals were placed on a surface and observed for any unusual motions, such as circular or zigzag motion, tremor, flattening, stupor, or catalepsy. Toxicity testing was performed at 1,000mg/kg using 7 time points from 15 minutes to 24 hours. All animal experiments received institutional approval, and were carried out under the guidelines of the United States Public Health Service’s Policy on Humane Care and Use of Laboratory Animals. Results Human Epidemiological Data Diuretics were used by 29% of patients and 32% of controls. The most commonly prescribed diuretic was a combination of hydorchlorothiazide and triamterene, prescribed for 45% of current users among patients and 51% among controls. Other hypertensive agents (eg, ␤-blockers) were used by 4% of patients and 8% of controls. Diuretics were prescribed to 70% of individuals with congestive heart failure and 32% of individuals with hypertension. Congestive heart failure associated with valve disease was the main reason for diuretic therapy among normotensive individuals. Diuretic use at any time before the index date (ie, current and past use) was associated with a decreased risk for unprovoked seizure (OR ⫽ 0.68, 95% CI ⫽ 0.44 –1.0) after adjustment (Table 1). This protective effect was greater for current use (OR ⫽ 0.62, 95% CI ⫽ 0.39 – 0.99) than for past use (OR ⫽ 0.85, 95% CI ⫽ 0.46 –1.6). Both current use of thiazide and current use of fu- Hesdorffer et al: Are Certain Diuretics Also Anticonvulsants? 459 Table 1. Odds Ratios and 95% Confidence Intervals for Diuretic Use and Idiopathic/Cryptogenic Seizures in Older Adults Variables No. Cases Exposed No. Controls Exposed Crude ORa 95% CI Adjusted ORb 95% CI 42 27 15 105 92 75 17 200 0.92 0.77 1.40 1.00 0.64–1.3 0.51–1.2 0.80–2.4 Referent 0.68 0.62 0.85 1.00 0.44–1.00 0.39–0.99 0.46–1.60 Referent Any diuretic use Current diuretic usec Past diuretic use onlyd No diuretic use a Adjusting for the matching variables (age, gender, and duration of medical care in the community). Adjusting for the matching variables, dementia, congestive heart failure, and history of hypertension. c Diuretic use on the visit just before the index date. d Diuretic use before the index date but not on the visit immediately before the index date. b OR ⫽ odds ratio; CI ⫽ confidence interval. rosemide were associated with a decreased unprovoked seizure risk (Table 2), but ORs were not statistically significant when adjusting for matching variables. When adjusted for dementia and congestive heart failure as well as matching variables (see Table 2), patients were 47% less likely than controls to have current thiazide treatment, 56% less likely than controls to have current furosemide treatment, and 10% more likely than controls to have current aldactazide treatment. Only the protective effect of current thiazide use was statistically significant. Results were unchanged when individuals with dementia were considered to be noncompliant (see Table 2). Animal Data: Seizure Suppression Chlorthiazide suppressed seizures in the MES model with a clear dose–response relationship (Table 3) in both rodent species. The mouse ED50 2 hours after administration was 881mg/kg with chlor- CHLORTHIAZIDE. Table 2. Crude and Adjusted Odds Ratios and 95% Confidence Intervals for Use of Specific Diuretics and Idiopathic/Cryptogenic Seizures in Older Adults Variables Whole group Any diuretic usec Thiazided Aldactazide Furosemide No diuretic use Current or past diuretic usec Current thiazided Current aldactazide Current furosemide Past diuretic use No diuretic use Demented cases counted as unexposed to diureticse Current or past diuretic usec Current thiazided Current aldactazide Current furosemide Past diuretic use No diuretic use Crudea Adjustedb No. Cases Exposed No. Controls Exposed OR 95% CI OR 95% CI 32 5 3 103 76 6 8 198 0.87 1.3 0.80 1.0 0.58–1.3 0.54–3.3 0.25–2.5 Referent 0.85 1.4 0.56 1.0 0.47–1.5 0.58–3.6 0.17–1.8 Referent 19 4 2 15 103 64 6 5 15 198 0.67 1.2 0.84 1.5 1.0 0.41–1.1 0.43–3.2 0.21–3.4 0.85–2.5 Referent 0.53 1.1 0.44 0.84 1.0 0.31–0.90 0.41–3.1 0.10–1.9 0.45–1.6 Referent 13 4 1 7 118 61 6 4 12 205 .54 1.2 0.61 1.1 1.0 .30–.96 0.45–3.3 0.08–4.4 0.52–2.4 Referent 0.47 1.1 0.47 0.82 1.0 0.25–0.87 0.40–3.1 0.06–3.5 0.34–2.0 Referent a Adjusting for the matching variables. Adjusting for the matching variables, dementia, congestive heart failure, and history of hypertension. c Excluding 2 cases with current use and 2 controls with past use of unknown (n ⫽ 2) or miscellaneous diuretics (n ⫽ 2 and includes 1 quinazoline and 1 xanthine derivative). Current indicates diuretic use on the visit just before the index date; past indicates any diuretic use before the index date but not on the visit immediately before the index date. d Including thiazides with or without the potassium-sparing agents triamterene or amiloride. e Adjusted analysis controls for the matching variables (age, gender, and duration of medical care in the community), congestive heart failure, and hypertension. b 460 Annals of Neurology Vol 50 No 4 October 2001 Table 3. Effective Dose for 50% of Animals: Values for Chlorthiazide and Furosemide 95% CI Test Protective Index Time (hr) ED50 (mg/kg) Low High Slope (SE) 2.0 2.0 2.0 881 ⬎1,000 ⬎1,500 700 0 0 1,103 0 0 6.6 (1.9) 0 (0) 0 (0) ⬎1.7 0.5 0.5 0.5 423 ⬎600 548 253 0 412 624 0 707 3.8 (1.3) 0 (0) 5.7 (2.0) 1.3 Chlorthiazide MES ScM TD50 Furosemide MES ScM TD50 The ED50 was calculated from dose responses based on the time of peak effect over several time periods for activity and toxicity. CI ⫽ confidence interval; SE ⫽ standard error; MES ⫽ maximum electric shock; ScM ⫽ subcutaneous Metrazol; TD50 ⫽ toxicity for 50% of animals. thiazide, for rats 165mg/kg. The greatest suppression of seizures (3 of 4 mice protected) occurred at 2 hours. A TD50 could not be achieved with dosing as high as 1,500mg/kg in mice and 1,200mg/kg in rats. The peak effect of chlorthiazide was examined in 4 mice under the MES model at a dosage of 750mg/kg (Table 4). FUROSEMIDE. Furosemide suppressed seizures in the MES model for both species, showing a dose–response relationship (see Table 3). The ED50 was 423mg/kg in mice and 36mg/kg in rats, and the TD50 was 549mg/kg in mice and 394mg/kg in rats. The peak effect of furosemide was examined in 4 mice under the MES model at a dose of 300mg/kg (see Table 4). Seizures were suppressed in 2 of 4 mice at 30 minutes. The peak effect of furosemide in rats occurred 2 hours after administration. Discussion In previous epidemiological studies, we have shown that diuretics are associated with a decreased risk for unprovoked seizures, regardless of indication or efficacy.1 Our findings suggest that current treatment with thiazide or furosemide is associated with a decreased risk for unprovoked seizure, and is functionally anti- convulsant. Our laboratory findings show both chlorthiazide and furosemide can suppress MES-induced seizures in a dose-dependent manner. These findings argue against a protective effect of diuretics mediated by lowering blood pressure, insofar as animals were not hypertensive. Mouse ED50 for chlorthiazide (881mg/kg) and for furosemide (423mg/kg) exceed doses given to freeliving populations. Our population-based data suggest that human therapeutic doses of these drugs are anticonvulsant. It is possible that diuretics were not taken as prescribed in the epidemiological study. We considered the effect of the most extreme example of this bias by recategorized people with dementia as nonusers (see Table 2). The protective effect of diuretic use remained. It is also possible that information on diuretic therapy was not included in the medical record. The likelihood of this misclassification error should be the same for patients and controls, biasing the estimate of diuretic effect towards the null (ie, no effect). An additional weakness of our epidemiological analyses was the definition of current diuretic use. We were unable to examine diuretic use on the index date, but we have no Table 4. Time to Peak Effect of Chlorthiazide and Furosemide Administered Intraperitoneally to the Carworth Farms No. 1 Mouse Test Chlorthiazide MES Tox Furosemide MES Tox Time (hr) Dose (mg/kg) 0.25 0.50 1.00 2.00 4.00 6.00 24.00 No. Deaths 750 1,000 0/4 0/8 0/4 0/8 0/4 1/8 3/4 2/8 1/4 1/8 0/8 1/8 1 300 750 1/4 2/8 2/4 7/8 1/4 6/8 0/4 6/8 0/4 6/8 6/8 6/8 6 8.00 Results are expressed as the number of mice protected or toxic over the number tested. MES ⫽ maximum electric shock; Tox ⫽ toxicity. Hesdorffer et al: Are Certain Diuretics Also Anticonvulsants? 461 reason to assume that diuretics were stopped in the short interval between these two visits. Misclassification of diuretic use on the index date may have introduced bias, although the nature of this bias is unclear. Selection bias (ie, the tendency for differential ascertainment of patients and controls) was unlikely in our study. We studied all cases of first unprovoked seizure that came to medical attention among residents of Rochester, MN. Controls were drawn from the same population and were matched to patients for age, gender, and duration of medical follow-up in the community. There was no recall bias in this study; patients and controls were not interviewed. Rather, we relied on record abstraction, which was done blind to the study hypotheses. Little is known about the anticonvulsant effect of diuretic agents in humans. The carbonic anhydrase inhibitor acetazolamide is used primarily as a diuretic, but is also used as an anticonvulsant.13,14 This agent is thought to alter seizure threshold through CO2 accumulation in the brain, thereby altering neurophysiological function.15 No patient was prescribed acetazolamide. Our findings in humans and in animal studies demonstrate that furosemide and thiazide use is protective for the development of seizures. Furosemide suppresses seizures in vitro and in vivo, possibly by creating a hyperosmotic environment, disrupting synchronization of neuronal activity.2 In another study, 3 loop-diuretics (furosemide, bumetanide, and ethacrynic acid) prevented the occurrence of sound-triggered seizures in audiogenic seizure-prone rats, whereas mannitol had no effect,3 suggesting that the mechanism is not through diuresis. Chlorthiazide and furosemide suppress seizures in humans and in animals and demonstrate a dose–response relationship in animals. These data may be important to clinical researchers evaluating new anticonvulsant agents in patients receiving diuretic therapy. Further clinical studies are needed to determine whether these or other diuretics offer an adjuvant or alternative treatment for epilepsy. This work was funded in part through grants from the National Institute of Neurologic Diseases and Stroke (NS16308); the National Institutes of Health (AR30582, M01RR00645); and the Epilepsy Foundation of America. 462 Annals of Neurology Vol 50 No 4 October 2001 We thank Pat Perkins for her help in data collection. We also thank Drs Hal Wolfe, Steve White, and Jose Woodhead of the University of Utah and Dr Harvey Kupferberg of the Anticonvulsant Screening Project, Epilepsy Branch, NIH, for assistance with the animal data. We are grateful to Dr Syd Shapiro of the Mailman School of Public Health for his time and careful advice. References 1. Hesdorffer DC, Hauser WA, Annegers JF, Rocca WA. Severe uncontrolled hypertension and adult-onset seizures: a case–control study in Rochester, Minneosta. Epilepsia 1996;37: 736 –741. 2. Hochman DW, Baraban SC, Owens JWM, Schwartzkroin PA. Dissociation of synchronization and excitability in furosemide blockade of epileptiform activity. Science 1995;270:99 –102. 3. Reid KH, Guo SZ, Iyer VG. Agents which block potassiumchloride cotransport prevent sound-triggered seizures in postischemic audiogenic seizure-prone rats. Brain Res 2000;864: 134 –137. 4. Melton LJ III. History of the Rochester Epidemiology Project. Mayo Clin Proc 1996;71:266 –274. 5. Hesdorffer DC, Hauser WA, Annegers JF, et al. Dementia and adult-onset unprovoked seizures. Neurology 1996;46:727–730. 6. Kokmen E, Beard CM, O’Brien PC, et al. Is the incidence of dementing illness changing? A 25-year time trend study in Rochester, Minnesota (1960 –1984). Neurology 1993;43: 1887–1892. 7. SAS. Proc PHREG: A procedure for Cox’s proportional hazards regression analysis. Presented by Ying So at the SAS users group international conference, February 17–20, 1991. 8. Breslow NE, Day NE. The analysis of case–control studies. In: Davis W, ed. Statistical methods in cancer research, vol 1. Lyon, France: International Agency for Research on Cancer, 1980:248 –279. 9. White HS, Johnson M, Wolf HH, Kupferberg HJ. The early identification of anticonvulsant activity: role of the maximal electroshock and subcutaneous pentylenetetrazol seizure models. Ital J Sci 1995;16:73–77. 10. White HS, Woodhead JH, Franklin MR. General principles: experimental selection, quantification, and evaluation of antiepileptic drugs. In: Levy RH, Mattson RH, Melrum BS, eds. Antiepileptic drugs, 4th ed. New York: Raven Press, 1995: 99 –110. 11. Finney DJ. Probit analysis, 3rd ed. London: Cambridge University Press, 1971. 12. Dunham MS, Miya TA. A note on a simple apparatus for detecting neurological deficit in rats and mice. J Am Pharm Assoc Sci Ed 1957;46:208 –209. 13. Millichamp JG. Relation of laboratory evaluation to clinical effectiveness of antiepileptic drugs. Epilepsia 1969;10:315–328. 14. Lombroso CT, Forxythe I. A long-term follow-up of acetazolamide (Diamox) in the treatment of epilepsy. Epilepsia 1960; 1:493–500. 15. Woodbury DM, Kemp JW. Some possible mechanisms of action of antiepileptic drugs. Pharmakopsychiatr Neuropsychopharmakol 1970;3:201–226.