Hydralazine dose-response curve analysis

Journal of Pharmacokinetics and Biopharmaceutics, VoI. 18, No. 4, 1990

Hydralazine Dose-Response Curve Analysis

David A. Graves, ~'4 Keith T. Muir, ~ Winston Richards, 2 Barry W. Steiger, ~ lh Chang, 1 and Bababhai Pate l 3

Received January 27, 1989--Final February 20, 1990

A multicenter, parallel, double-blind, 181-patient study compared the safety and antihypertensive efficacy of immediate-release (IR) and extended-release (ER) hydralazine. After 2 to 4 weeks on diuretic, patients were maintained on diuretic and randomized to a hydralazine treatment regimen: IR thrice daily, ER twice daily, or ER once daily. Daily doses of hydralazine were 75, 150, or 300 rag. Although designed as a titration study, important dose-response data were available for analysis with nonlinear mixed effect modeling (NONMEM). Sitting diastolic blood pressure (BP) was selected as the response variable. Several factors were tested for importance, including body weight, time (week) effects, concomitant beta-blocker (BB) therapy, acetylator class, and treatment regimen. All factors were important (p <0.05) except treatment regimen (p .> 0.30). The maximum antihypertensive response (Emax) to hydralazine was 9.4 mm Hg. The daily dose that elicited 50% of the maximum response (Dso) was 0.87 mg/kg for slow acetylators and 1.68 mg/kg for fast acetylators. BP fell 0.52 mm Hg per week independent of other effects, and concomitant BB therapy induced a drop of 6.6 mm Hg in addition to hydralazine, diuretic and week effects. NONMEM's use assisted in evaluations and provided information not obtainable through tradi- tional means.

KEY WORDS: hydralazine; dose response; NONMEM.

I N T R O D U C T I O N

H y d r a l a z i n e has been used for more t han 30 years for the t r ea tmen t o f m o d e r a t e to severe pe r s i s t en t hyper t ens ion . It reduces e leva ted b l o o d p ressu re t h rough p e r i p h e r a l a r t e r io l a r v a s o d i l a t i o n resul t ing f rom di rec t r educ t i on o f m y o g e n i c tone o f the a r t e r io l a r wall. H y d r a l a z i n e p r o d u c e s h e a d a c h e , t a c h y c a r d i a , a n d pa lp i t a t i ons in m a n y pa t ien t s af ter in i t ia l a d m i n i s t r a t i o n and of ten when doses are i nc reased (1,2). Wi th c o n t i n u e d

Presented in part at the 89th Annual Meeting of the American Society for Clinical Pharmacology and Therapeutics, San Diego, California, 1988. t Fisons Pharmaceuticals, P.O. Box 1710 Rochester, New York 14623. 2 Schering Corporation, Kenilworth, New Jersey. 3 Reid-Rowell, Marietta, Georgia. 4 To whom correspondence should be addressed.

279 0090-466X/90/0800-0279506.00/0 �9 1990 Plenum Publishing Corporation

280 Graves etal.

administration, salt and water retention may occur. These adverse effects are sufficiently severe that many patients take hydralazine concomitantly with a diuretic and a beta-blocker.

Double-blind outpatient studies have been performed to investigate the effectiveness of an extended-release (ER) form of hydralazine (Pennkinetic T M system, hydralazine polistirex extended-release capsules), in an effort to increase patient compliance and potentially reduce adverse effects. In these studies (3), ER hydralazine was given once or twice daily and compared with thrice-daily administration of immediate-release (IR) hydralazine. The protocol utilized in these studies was one in which dose titration took place as a part of therapy. Specifically, investigators adjusted each patient's dose on a weekly basis, as required for an arbitrarily defined adequate response without inordinately increasing adverse effects.

Although these studies were not specifically designed to acquire dose- response information, characterization of the dose-response curve in out- patients with moderate hypertension is' desirable since it could help to minimize exposure of patients to doses that are not believed to substantially improve response (4,5). Therefore, this series of analytical investigations attempted to describe the dose-response curve of oral hydralazine.

M E T H O D S

Patient Population

Five centers participated in this study. Each center was to enroll 20 to 30 male or female hypertensive patients between the ages of 18 and 70. Patients with a history of persistent essential hypertension with a sitting diastolic blood pressure (BP) equal to or greater than 90 mm Hg who were receiving a diuretic or, if they were previously untreated for their hypertension, had a sitting diastolic blood pressure equal to or greater than 100 mm Hg, entered the study. Patients currently taking a diuretic and a beta-blocker and those taking hydralazine and another antihypertensive drug were also eligible for the study.

Patients were to receive diuretic therapy alone for 2 to 4 weeks after discontinuation of their other antihypertensive medications. Administration of noncardioactive drugs was permitted. If the BP remained equal to or above 90 mm Hg, patients qualified for entry into the hydralazine-treatment phase. Patients whose sitting diastolic blood pressure exceeded 105 mm Hg for two consecutive readings 30 min apart at any visit could enter the study immediately at the discretion of the investigator.

All patients signed IRB-approved informed consent prior to enrollment. Patients were excluded if they had Keith-Wagener grade III- IV retinopathy,

Hydralazlne Dose-Response Curve Analysis 28i

clinically significant congestive heart failure, renal insufficiency (creatinine> 2.5 mg%), or had a history of diabetes mellitus, active liver disease, senility, psychosis, drug or alcohol dependence, noncompliance, a myocardial infarction in the previous 6 months, or angina diagnosed during the previous 2 months. Patients were also excluded if they had a history of secondary hypertension associated with pheochromocytoma, hyper- adrenocorticism, and/or renal artery stenosis. Patients requiring a concomitant medication known to alter BP or having a history of hypersensitivity to hydralazine were excluded from entry. Patients had to have a negative (< 1 : 40) antinuclear antibody test and be willing to be screened to determine acetylator status. Female patients were required to be nonpregnant, could not be nursing, and also needed to be following an accepted method of birth control or be sterile.

Study Design

The majority of patients had their previous antihypertensive medications discontinued for 1 to 2 weeks, with the exception of their previous diuretic (usually hydrochtorothiazide, 50 mg qd). Then, for 2 to 4 weeks (the Diuretic phase), patients were maintained on their diuretic. At the end of this phase, if the patient still qualified (see above), he or she was acetylator-phenotyped using sulfamethazine or dapsone (6,7). The investigator was not informed of the results of this test until the patient had completed the study.

Those who qualified were randomly assigned to one of three treatment groups. Patients were maintained on diuretic throughout the study. The first group (IR tid) took IR hydralazine hydrochloride (Apresoline | three times daily at 8AM, 2PM, and 8 PM. The second group (ERbid) took ER hydralazine twice daily at 8 AM and 8 PM. The third group (ER qd) took ER hydralazine once daily at 8 AM. The study was double-blind, with the blind maintained by providing identical-appearing capsules for all formulations, including supplemental placebo capsules when required. Therefore, all patients were instructed to take medication three times daily, but not all doses contained hydralazine. The treatment phase was intended to be 4 to 7 weeks in length, depending on the duration of the titration period. Importantly, BP was always measured in the clinic prior to that patient's morning dose of medication.

The patient was seen on a weekly basis. The starting dose was equivalent to 75 mg hydralazine HCI daily. At the end of the first week, if the patient's BP was not less than 90 mm Hg, the total daily close was increased to 150 mg. At the end of the second week, the daily dose could be increased again to 300 mg if the investigator felt adequate control was not evident. Titration

282 Graves et al .

was complete after 3 weeks of therapy for the vast majority of patients. After titration, the patient was maintained at that dose for 4 weeks, at which time the patient was considered to have completed the study. Patients were seen approximately once a week during the maintenance phase.

If the patient complained of unacceptable, beta-blocker-responsive adverse effects (headache, palpitations, or tachycardia), the investigator was permitted to give nadolo180 mg once daily to help reduce their incidence and severity. (Three patients had this dose increased to 160 mg.) Once beta-blocker therapy began, it was continued until the patient discontinued the trial.

Nonlinear Mixed Effect Modeling

Although several models are available for defining dose response, an asymptotic equation is often used. Since a fall in BP was expected during therapy, the following basic equation was used:

Emax * DOSE BP = B A S E L I N E (1)

Dso + D O S E

where B A S E L I N E was the BP at the beginning of the hydralazine treatment phase, Emax was the maximum expected effect due to hydralazine therapy (in mm Hg), D O S E was the daily hydralazine dose (in mg or mg/kg), and Dso was the D O S E that induced a response equal to one-half of Emax. Other factors which could affect BP measurements were also considered during evaluations: Treatment (IR tid, ER bid, or ER qd), acetylator class (fast or slow), the week of the study (duration in weeks), and the use of concomitant beta-blocker therapy (present or absent). The final model is presented in Results.

Nonlinear mixed effects modeling ( N O N M E M ) has appeared in the literature over the past 10 years (8,9). N O N M E M Version 1.4 Level 2 was installed on a VAX 11/785, although the PREDPP modules were not used for the dose-response application. Initial use of simpler models confirmed the need for the asymptotic model, as well as the estimation of interindividual variances on B A S E L I N E , Emax, and Dso. A residual variance (intraindividual) was also estimated which provided an estimate of visit-to- visit variability for BP measurements within patients. The preference for a homoscedastic model was compared during model building to a propor- tional model for both inter- and intraindividual variability. Each N O N M E M execution therefore provided an estimate of (i) dose-response curve parameters, such as Emax and Dso; (ii) interpatient variability in these parameters; and (iii) intrapatient variability.

A stepwise model-building approach was taken starting with a simple "s tep" (yes-no) response followed by a linear dose-response model. The

Hydralazine Dose-Response Curve Analysis 283

asymptotic dose-response function and interpatient variability were then introduced into the model. Each of these factors was introduced into the model to evaluate its effect on BP and dose response. Likelihood ratio tests (with the assumption that both interpatient and intrapatient variation follow a normal distribution) were performed to compare one model versus another and to evaluate the significance level of the effect of a prognostic factor on BP or dose response. The likelihood ratio test statistic can be readily computed by comparing the objective function of each model provided by the N O N M E M program (8). The final dose-response model was therefore developed using N O N M E M in parallel and sequential executions, adding or deleting parameters, keeping those parameters in the model which con- sistently decreased N O N M E M ' s objective function at a level of p < 0.05. The final model and resultant estimates are described in Results.

RESULTS

Of the 181 patients who entered the treatment phase, 7 did not have a BP reading beyond the baseline visit that represented a full week of dosing. These patients were excluded from analysis. Five patients qualified for study without a formal diuretic phase (BP> 105 mm Hg at two consecutive readings 30rain apart). Most patients were seen on at least two occasions, providing at least 90% confidence that their (true) BP was greater than 90m m Hg (10). In the rare (<0 .1%) ins tance that a patient was shown through pill counts to have taken fewer than half of the indicated doses during that week, a dose of 0 mg was used in the data set. A total of 136 patients completed 4 or more weeks of therapy. Some patients were followed for up to 13 weeks by the investigator, and, since the blind was maintained~ this information was included. Three patients were authorized to have their doses increased to 400 mg daily. There were a total of 11 i8 BP observations from 174 patients. A total of 36 patients were prescribed supplemental beta-blockers. Only 9 patients were classified as "intermediate" acetylators, and they were therefore included as "slow" acetylators. Additionally, 6 patients' acetylator classification was unknown due to protocol violations. As the final model included this factor, these patients were removed from final consideration, leaving a total of 1085 BP readings from 168 patients. Due to the size and complexity of the data sets, there was dual data entry and verification. A summary of observations is provided in Table I.

N O N M E M Results

A total of 31 successful N O N M E M executions (many more were performed in initial, exploratory data analyses) were required to resolve the following

284 Graves et al.

Table I. Frequency of Blood Pressure Observations for Each Dose and Week

Total daily dose (mg)

Week of treatment

0 1 2 3 4 5 6 7 8 >8 Total

0 168 0 3 3 4 1 1 0 0 0 180 75 0 165 68 37 24 27 19 7 0 0 347

150 0 0 95 62 31 31 45 27 6 4 301 300 0 0 0 50 55 21 47 51 14 12 250 400 0 0 0 0 1 2 2 1 1 0 7

Total 168 165 166 152 115 82 114 86 21 16 1085

final model:

02j * DOSE ] BP,j = % - ( ( % j)-UDOSE)J + (05 * DOSE) + BB) + (2)

Olj = O1 + "qlj (3)

02j ~--" 02 -}- ~2j (4)

03j = 03 + r13j or 04j = 04 + r13j (5)

where 0~ was the baseline BP, 02 was the Emax, 0 3 w a s the Ds0 if the patient was a slow acetylator or 04 if the patient was a fast acetylator. 05 was a linear parameter , expressed in mm Hg/week, describing the change in BP that occurred as a result of remaining in the study beyond week 0. 06 was the estimate of the effect of concomitant beta-blocker therapy (BB = 0 if a beta-blocker was not taken prior to the BP measurement, BB = 1 if it was). Daily dose was expressed as mg/kg in the final model. Variance components for patient-to-patient (intersubject) error were included in the final model for baseline (00, Emax (02), and Dso for slow or fast acetylators (03 or 04) and were assumed to be independent of one another (e.g., 3-5). Intersubject and intrasubject errors were modeled homoscedastically (constant error variance).

Results are provided in Table II. Figures 1 and 2 illustrate the typical dose-response curve from this model for fast and slow acetylators, respec- tively. Table I I I indicates variables tested and their level of significance ( importance) to the model.

Importantly, there were no statistically significant differences among treatments ( IR tid, ER bid, or ER qd). During the model-building process,

Hydralaz ine D os e -R esp on se Curve Analysis 285

Table I1. Results from Final NONMEM Model a.

Parameter Estimate 95% C!

Baseline BP (mrn Hg) Emax (mm Hg) Dso for slow acetylators (mg/kg) Dso for fast acetylators (mg/kg) Week slope (mm Hg/week) Beta-blocker effect (rnm Hg) Intersubject SD for baseline (mm Hg) Intersubject SD for Emax (mm Hg) Intersubject SD for Dso (mg/kg) lntrasubject SD (visit-to-visit error) (ram Hg)

98.8 97.7, 99.9 9.37 5.18, 13.6 0.870 0, 1.94 1.68 0.139, 3.22

-0.524 -0.740, -0.308 -6.57 -8.35, -4.79 • 4.58, 6.93 • b , 9.23 • b 1.84 • 4.79, 5.59

~Confidence intervals (95% CI) for this table were considered to be symmetric. Although NONMEM estimated variances, they have been converted into standard deviations (SD).

6Not obtainable; lower confidence limit would require taking the square root of a negative number,

!)8

125

-r E E

~11il ! IL

O

~ n

118

70

O 1 1 3 4 5 $

Dally Do~, mg/kg

Fig. 1. Dose-response curve for fast acetylators. Thick line indicates fit of model without the influence of time and beta-blockers.

286 Graves et al.

I H

I I I

cII . i -

s E

I l l l

'U o

lit

Ill

78

# I l 3 4 S I

Dally Dose, mg/kg

Fig. 2. Dose-response curve for slow acetytators. Thick line indicates fit of model without the influence of time and beta-blockers.

Table IIL Hypothesis Testing with Restricted Models a

Hypothesis Significance

Is the use o f Emax and Dso (curvilinear dose-response) superior to a "s tep" model? Is the use of Emax and Dso superior to a "s lope" (straightline response) model? Do treatment regimens provide different dose-response curves? Is acetylator class an important factor? Is durat ion of therapy an important factor? Is the use of a beta-blocker an important factor? Is total body weight an important factor? ls the use of 3 interindividual variances superior to only 2?

Yes; p < 0.001

Yes; p <0.001

No; p > 0.30 Yes; p < 0.025 Yes; p <0.001 Yes; p <0.001 Yes; p <0.001 Yes; p < 0.05

a l f a test was not significant (p > 0.05), that parameterization was excluded from the model.


this treatment factor was tested for different times by estimating a different Emax and Dso for each treatment. In all cases, the addition of this factor was clearly nonsignificant (p > 0.30). The three treatment regimens therefore appeared to have indistinguishable dose-response curves.

Fast acetylators metabolize hydralazine faster and typically require larger doses for any given response (11). Therefore, one would expect that the two acetylator classes have different Dso values. This is, to a great extent, confirmed by the data. The Dso estimate for fast acetylators is about twice of that for slow acetylators and the difference is statistically significant (p <0.025 when this issue was considered individually during model building). There was no statistical evidence of difference in Emax between the two classes. Figures 1 and 2 show the actual dose response of each individual and the fitted curves based on the model.

The effect of simply staying on hydralazine therapy was estimated as -0.524 mm Hg/week, indicating that during a 7-week study period, BP felt about 4ram Hg. Clearly, attentuation of hydralazine's effect was not observed during treatment. The time effect is estimable because a substantial number of patients remained on one dose level for five weeks or more (see Table I). However, an assumption was made that the drop in BP due to time effect was steady from week to week throughout the study period regardless of treatment or dose level. The hydralazine dose-response curve described in the final model is in addition to this week effect. Similarly, the use of a concomitant beta-blocker along with hydralazine and diuretic therapies resulted in a fall of 6.57 mm Hg in addition to that caused by the other effects. Both week and beta-blocker effects were very well estimated. The lack of placebo in this study prevents a precise estimation of the time effect.

Intrasubject (visit-to-visit or within-subject variability) was estimated as + 5.21 mm Hg. Thus, within-patient variability was nearly as large as the between-patient variability estimated at baseline, • mm Hg. These variabilities are somewhat less than previously reported (10), and reflect the homogeneous group entered into this study given specific entrance criteria.

DISCUSSION

This series of analyses indicate that there is a dose-response relationship with hydralazine. This relationship is curvilinear and asymptotic within the dose range used clinically. Only predose measurements of BP were made, so dose-response information about postdose BP measurements is not known.

288 Graves et al.

More stringent significance levels have been used in the literature for hypothesis testing during model building (8,12). This approach provides protection against increased probability of falsely concluding certain factors to be significant (Type I error) due to multiple testings. However, at the same time, it would increase the probability of excluding possibly important factors from the model (Type II error). Typically, a p value is chosen to provide more protection against the error which is of more severe con- sequence. From a model-building point a view, the goal is to find a model that adequately describes the data. Excluding factors that are important but not necessarily significant at a certain level could produce a biased model and widen the confidence limits of parameter estimates. In the similar situation of stepwise linear regression, there has not been an established "best" p value and the convention does favor using 0.05 as an initial choice (13). The need to adjust for multiple testings is appropriate after an adequate model is found and when final inference is to be drawn. One might want to adjust the p values for various factors at this stage using Bonferroni inequality. In this study, the treatment difference in dose response was clearly nonsignificant. The acetylator class was an important factor in describing the data but the difference between the two classes might not be as statistically significant as the value 0.025 indicated when multiple testing is taken into account.

The approximate 95% confidence intervals of Dso for both classes are rather wide and in the case of slow acetylators, the interval appears to include zero. However, a model without the Dso parameters (a step model) cannot adequately describe the data (see Table III). In other words, the data suggest a graded (as well as asymptotic) dose response. The confidence intervals for Dso would be narrowed if more doses would have been given in this study below the Dso.

The finding that the ER and IR forms of hydralazine have similar dose-response curves is somewhat surprising. Recent pharmacokinetic evaluations indicated that single doses of the ER capsules are absorbed only about 50% compared to an IR reference (14,15). If this was important, the Dso estimates for the ER treatments should have been much greater. Similar observations have been made with ER versus IR propranolol formulations in which reduced apparent bioavailability of the drug was not associated with a reduction in antihypertensive response (16). The selective binding of hydralazine to arteriolar tissues after chronic dosing may in fact be a better predictor of pharmacologic effect than blood concentrations (17). In any case, the single-dose pharmacokinetics studies of the ER capsule (14,15) may not be relevant to chronic dosing in the hypertensive patient, as has been seen with minoxidil (18). Treatment regimen was tested at four different occasions during the model-building process and was clearly less


important in estimating the antihypertensive dose-response curve than the other effects studied, i.e., baseline BP, week effects, acetylator class, and concomitant beta-blocker use.

Hydralazine is not a new antihypertensive agent. Despite more than 30 years of clinical use, little has been published regarding its dose response at various doses during chronic therapy. This common occurrence has prompted many to advocate a systematic evaluation of a drug's dose response prior to marketing (19). Our experience with this analytical approach has been positive, with some minor exceptions. First, there are some patients (i.e., slow acetylators) who typically will not be expected to benefit greatly from increasing doses of hydralazine above 75-100 mg daily. Second, repeated monitoring BP over this time is known to help decision making on outpatient basis, and the large degree of intrapatient (week-to- week) variability confirms this clinical observation. Third, this method (NONMEM, among others) was needed in order to acquire unbiased estimates, permitting the first two conclusions to be drawn. Fourth, the large intrasubject variability observed in many ways confirms the well-recongn- ized need for the size of this study. Fifth, NONMEM required a great deal of CPU time (10 hr) to run all of the intermediate models required during the model-building process, as well as a considerable investment of time on the part of the analysts.

NONMEM has not yet been used extensively to quantify or estimate dose-response curves, although its use has been proposed (19). The basic model in Eq. (1), however, is structually identical to the Michaelis-Menten relationship, previously investigated with real and simulated data (20). Based upon the theoretical advantages apparent from these previous works, NONMEM should provide better estimates of dose-response parameters compared with other approaches.

Two specific limitations should be mentioned, however, which are problems with limitations of dose selection in the clinic, not NONMEM. First, in this trial, we were surprised to find that, for slow acetylators, the Dso was so low. For this group, Dso for a 70-kg patient would be expected to be about 61 mg/day--below the initial dose used in this titration study. Thus, the Dso for slow acetylators was poorly estimated (compared with the fast acetylator Dso) due to lack of data on the ascending portion of the dose-response curve. Second, although not a problem in our study, it is conceivable that, in some therapeutic classes, it will not be possible to give large doses, giving few data on the plateau portion of the dose-response curve. In these cases, Emax wilt be poorly estimated instead.

Results of this study suggest that an additional trial be conducted~ using lower doses and including a placebo treatment group, although estimation of the "week slope" would seem reasonable based on the number

290 Graves eta/.

of observa t ions taken at a cons tan t dose (see Table I; some pat ients r ema ined at 75 m g / d a y for 6 weeks). The "dose esca la t ion" s tudy would be a reason-

able a l ternat ive (19), with the impor t an t feature of a small n u m b e r of

pat ients who rema in on p lacebo th roughou t the trial.

In summary , the use of N O N M E M enab led us to establish a model

o f hydra laz ine ' s the rapeu t ic effect on BP dur ing a t i t ra t ion s tudy which, in m a n y ways, mimicked rou t ine cl inical care. Al though no p lacebo group

was s tudied, key factors in hydra laz ine an t ihyper tens ive response were ident if ied and their magn i t ude (and therefore, cl inical utility) est imated.

The use of N O N M E M permi t ted this analysis which could no t have been

u n d e r t a k e n by t rad i t iona l means .

A C K N O W L E D G M E N T

The authors grateful ly acknowledge the Data H a n d l i n g Sect ion of

Fisons Corpo ra t i on for p repar ing and m a n a g i n g the large data sets that this

analysis required.

R E F E R E N C E S

1. Current Labeling for Apresoline | (hydralazine hydrochloride). 2. Veterans Administration Cooperative Study Group of antihypertensive agents. J. Am. Med.

Assoc. 237:2303-2310 (1977). 3. J. Adir, C. L. Curry, R. E. Taylor, C. D. Poku, and K. S. Rotenberg. Comparative efficacy

and safety of immediate-release and controlled-release h.ydralazine in black hypertensive patients. Clin. Ther. 9:640-650 (1987).

4. R. Temple. Government viewpoint of clinical trials. Drug Inform. J. 16:10-17 (1982). 5. D. Shen, K. O'Malley, M. Gibaldi, and J. L. McNay. Pharmacodynamics of minoxidil as

a guide for individualizing dosage regimens in hypertension. Clin. PharmacoL Ther. 17:593- 598 (1987).

6. W. Olson, J. Miceli, and W. Weber. Dose-dependent changes in sulfamethazine kinetics in rapid and slow isoniazid acetylators. Clin. Pharmacol. Ther. 23:204-211 (1978).

7. K. Cart, J. A. Oates, A. S. Nies, and R. L. Woosley, Simultaneous analysis of dapsone and monoacetyldapsone employing high performance liquid chromotography. Br. J. Clin. Pharmacol. 6:421-427 (1978).

8. D. R. Mungall, T. M. Ludden, J. Marshall, D. W. Hawkins, R. L. Talbert, and M. H. Crawford. Population pharmacokinetics of racemic warfarin in adult patients. J. Phar- macokin. Biopharm. 13:213-227 (1985).

9. S. L. Beal and L. B. Sheiner. N O N M E M User's Guide-Part I: User's Basic Guide, Technical Report of the Division of Clinical Pharmacology, University of California, San Francisco, 1980.

10. B. Rosner and B. F. Polk. Predictive values of routine blood pressure measurements in screening for hypertension. Am. J. EpidemioL 117:429-442 (1983).

11. D. E. Drayer and M. M. Reidenberg. Commentary: Clinical consequences of polymorphic acetylation of basic drugs. Clin. PharmacoL Ther. 22:251-258 (1977).

12. J. Grevel, B. Whiting, A. W. Kelman, W. B. Taylor, and D. N. Bateman. Population analysis of the pharmacokinetic variability of high-dose metoclopramide in cancer patients. Clin. Pharmacokin. 14:52-63 (1988).

13. Anonymous. The Stepwise procedure. In SAS User's Guide: Statistics, Version 5, SAS Institute, Cary, NC, 1985, pp. 763-774.


14. T. M. Ludden, A. M. M. Shepherd, K. S. Rotenberg, J. W. Woodworth, and L. K. Ludden. Relative bioavailability of immediate and sustained release hydralazine formulations. J. Pharm. Sci. 77:1026-1032 (1988).

15. A. M. M. Shepherd, S. H. D. Jackson, D. Rogers, T. M. Ludden, J. W. Woodworth, L. K. Ludden, and K. T. Muir. Effect of food on blood hydralazine levels in hypertension. (Abstract) Clin. PharmacoL Ther. 43:147 (1988).

16. M.J. Serlin, L. E. Orme, M. Maclver, G. J. Green, R. G. Sibeon, and A. M. Breckinridge. The pharmacodynamics and pharmacokinetics of conventional and long-acting propranolol in patients with moderate hypertension. Br. ~ Clin. PharmacoI. 15:519-527 (1983).

17. K. O'Malley, J. L. Segal, Z. H. lsraili, M. Boles, J. L. McNay, and P. G. Dayton. Duration of hydralazine action in hypertension. Clin. PharmacoL Ther. 18:581-586 (1975).

18. J. L. MeNay. Dose response studies with vasodilator antihypertensive drugs. Clin. Exp. Hyper. Theory Pract. A4:61-69 (1982).

19. L. B. Sheiner, S. L. Beal, and N. C. Sambol. Study designs for dose-ranging. Clin. Pharmacol. Ther. 46:63-77 (1989).

20. L. B. Sbeiner and S. L. Beal. Evaluation of methods for estimating population pharmacokinetic parameters. I. Michaelis-Menten Model: Routine clinical pharmacokinetic data. J. Pharmacokin. Biopharm. 8:553-571 (1980).

Documents

Hydralazine dose-response curve analysis