Enhancing the value delivered by the statistician throughout drug discovery and development: putting statistical science into regulated pharmaceutical innovation.

With the dawn of the 21st century, the pharmaceutical industry faces a dramatically different constellation of business and scientific predictors of success than those of just a few years ago. Significant advances in science at the genetic, molecular and cellular levels, combined with progress demonstrated around the globe with drug regulations, have increased business and competitive opportunities. This has occurred in search of better and cheaper medicines that reach patients with unmet medical needs as quickly as possible. Herein lie new opportunities for those who can help business and regulatory leaders make good decisions about drug development and market authorization as quickly and efficiently as possible in the presence of uncertainty. The statistician is uniquely trained and qualified to render such value. We show how the statistician can contribute to the process of drug innovation from the very early stages of drug discovery until patients, payers and regulators are satisfied. Indeed, the very nature of regulated innovation demands that efficient and effective processes are implemented which yield the right information for good decision making. The statistician can take the lead in setting a strategy that directs such processes in the direction of greatest value. This demands skills that enable one to identify important sources of variability and uncertainty and then leverage those skills to make decisions. If such decisions call for more information, then the statistician can render experimental designs which generate the right information needed to make the decision in an efficient, timely manner. To add value to the enterprise, statisticians will have to become more intimately associated with business and regulatory decisions by building on their traditional roles (for example, numerical analyst, tactician) and unique skill sets (for example, analysis, computation, logical thought and work process, precision, accuracy). Business and regulatory savvy, coupled with excellent communication and interpersonal skills, will allow statisticians to help create the knowledge needed to drive success in the future.

A framework establishing clear decision criteria for the assessment of drug efficacy.

Much has been published on various aspects of data analysis and reporting from clinical trials within the biopharmaceutical environment. This ranges from regulatory guidelines on the format and content of registration dossiers to recommendations on data presentation and the statistical methodologies that are appropriate for the diverse types of data one observes in clinical trials. Little has been written about designing a clinical trial analysis and reporting package that focuses on the decisions that must be made throughout the drug development process. Pharmaceutical companies today are under enormous pressure to develop drugs quickly and (cost-) efficiently. Because of this, drugs often move into the later phases of drug development before evidence from prior phases is completely understood. This provides a challenge to clinical trialists to design and execute a clinical trial programme which can expedite drug development. The statistician, as a clinical trialist, must strive to determine the optimum analytical methodology that facilitates decision making for this clinical trial programme. This paper proposes a new framework for the assessment of efficacy in drug development called the 'one programme, one p-value' framework. This framework will accelerate drug development by providing clear criteria for the decisions which must be made along the way. The 'one programme, one p-value' framework is based on the notion that the clinical trial programme comprises exploratory and confirmatory phases. The use of the likelihood function in the exploratory phase facilitates the decision whether (or when) to move into the confirmatory phase. The confirmatory phase consists of one confirmatory trial with a single hypothesis test of the drug's efficacy; hence 'one p-value'. Sponsor interaction with regulatory agencies is necessary at each decision point. Finally, the paper considers how analysis and reporting of efficacy data can be accomplished from a clinical trial programme as described.

Defining, monitoring and combining safety information in clinical trials.

Assessment of clinical trial safety data for industry, regulatory agencies, medical practitioners and patients requires definition and measurement, monitoring, and overall analysis. Prospective 'safety' definitions and reliable measurement tools reduce inefficient data collection and improve the validity of resulting analyses. Statistical tools can help investigators monitor safety data from controlled clinical trials and help improve post-marketing surveillance. Also, when evaluating overall safety, one needs to assess all available information by combining information from many trials and other sources. Planning for this combined assessment, incorporating flexibility to assess unanticipated yet important nuances in individual trials, may be more important than the actual statistical analysis method used. A keen awareness of the future needs of consumers of this information is quite important. Some current proposals to combine safety information will be discussed.

Absence of a relationship between adverse events and suicidality during pharmacotherapy for depression.

This study tested the hypothesis that some patients treated with an antidepressant who develop adverse events (e.g., activation, akathisia) experience emergent suicidality specifically associated with such events. Seventeen double-blind, controlled clinical trials conducted in the United States and Canada with 3,065 patients with major depression were evaluated for treatment-emergent adverse events (events that first occurred or worsened during therapy) and suicidality (a suicidal act or emergence of substantial suicidal ideation or both) with fluoxetine, placebo, and tricyclic antidepressants. Nine relevant adverse event clusters were evaluated: activation, sedation, activation and sedation, decreased libido, mania, psychosis, psychosis and mania, acute brain syndrome, and violence. Incidence rates were determined for suicidality that was and was not temporally associated with an adverse event cluster and were analyzed within and across treatments (incidence difference method). Most patients experienced neither a cluster event nor suicidality. Where suicidality was reported, it generally was not in temporal association with an adverse event cluster. In no cluster was the incidence of suicidality statistically significantly higher when reported in temporal association with an event than when not. Suicidality was associated infrequently with treatment-emergent activation and at comparable rates across treatments. No increased risk of suicidality associated with an adverse event cluster was observed between the treatment groups (fluoxetine versus tricyclic anti-depressants; fluoxetine versus placebo). These results from double-blind, placebo- and comparator-controlled fluoxetine clinical trials in patients with major depression do not suggest a relationship between a treatment-emergent adverse event pattern and suicidality in this population.

Pattern analysis method for assessing successful weight reduction.

Researchers reporting on clinical weight loss programmes have been challenged to provide information about long-term follow-up. To evaluate long-term weight reduction, a methodology was developed for analysis of longitudinal data. Under this method, a pattern of interest is defined a priori, and the proportions of patients attaining the pattern are compared between treatments. The methodology permits objective summarization and statistical analysis of the data. Applications of the method and corresponding results demonstrate its utility for assessing the treatment effect on weight reduction and systolic blood pressure. To show that the method may be further generalized, suggestions for alternative types of patterns are presented.

Fluoxetine: a randomized clinical trial in the treatment of obesity.

Fluoxetine hydrochloride (Lovan, Eli Lilly and Company, Indianapolis, Indiana, USA), a specific serotonin uptake inhibitor, was compared with placebo in 458 obese outpatients in a 52-week double-blind randomized ten-site trial to study its effect on weight reduction. Patients in the fluoxetine and placebo groups were predominantly Caucasian (81% and 85%, respectively) and female (81% and 79%, respectively), with a mean body mass index (BMI) of 36.2 and 35.8 kg/m2, respectively, and a mean age of 43 years (both groups). Fluoxetine therapy (60 mg/day) resulted in statistically significantly (P < or = 0.05) greater mean weight loss than placebo to week 28. Although some patients continued to lose weight throughout the 52-week therapy period, maximum mean weight loss occurred at week 20. There was no treatment difference at 52 weeks. The change in visit frequency (biweekly to week 8, monthly to week 20, then bimonthly to week 52) may have affected results. Patients with higher baseline BMIs (> 40 kg/m2) attained and maintained a greater weight loss than patients with lower baseline BMIs (< 40 kg/m2). Two sites demonstrated greater efficacy than the study as a whole. The use of nutrition counselling at one site and behaviour modification at the other, or other site-to-site differences, may account for the improved efficacy. Fluoxetine was well tolerated and appeared to be safe therapy for the treatment of obesity with efficacy demonstrated for 28 weeks.

A simple stopping rule for declaring treatment ineffectiveness in clinical trials.

We consider the problem of stopping a clinical trial before its scheduled termination due to the apparent ineffectiveness of the experimental therapy, as compared with a control. We propose a simple-to-implement, intuitive decision rule based on the unadjusted attained significance levels from any appropriate statistical test. The proposed procedure may be used at any time during the study as an aid to help determine whether the study of an experimental treatment should be terminated early with the conclusion of treatment ineffectiveness. Much of the power of the usual fixed-sample test is retained while maintaining the nominal test size.

Data monitoring and interim analyses in the pharmaceutical industry: ethical and logistical considerations.

The characteristics of data monitoring and the need for the use of data monitoring committees in clinical trials sponsored by the pharmaceutical industry differ from those of trials sponsored by government. Data monitoring is a continuous process in industry trials due to the regulatory requirements and the need to more thoroughly evaluate safety of new compounds. As part of this process, interim analyses are employed to make decisions about treatment effects. In some cases, such analyses may require the use of an external data monitoring committee to assist in the data review, analysis and decision making. A number of examples of interim analyses, with and without data monitoring committees, are discussed. Issues surrounding the need for external data monitoring committees and recommendations are presented. In particular the issues of sponsor participation in the data monitoring committee and controls of the decision making process are considered.

Effect of a leukotriene antagonist, LY171883, on cold air-induced bronchoconstriction in asthmatics.

Isocapnic hyperpnea (ISH) of cold air induces bronchoconstriction in many asthmatic subjects. Although this response is well described, it is unclear whether this bronchoconstriction is related to the release of bronchoactive mediators. We examined whether pretreatment with LY171883, a competitive antagonist of leukotriene D4 activity via LTD4 receptors, reduced the bronchospastic response to cold air ISH in asthmatics using a randomized, double-blind, two-phase crossover design. In 20 subjects, 2 wk of treatment with either LY171883 600 mg twice a day or placebo did not result in a change in FEV1 (3.45 +/- 0.21 L placebo versus 3.59 +/- 0.20 L LY171883; p greater than 0.05). Nineteen subjects underwent cold-air ISH; LY171883 increased the geometric mean respiratory heat loss required to reduce the FEV1 by 20% (PD20RHE) from 1.00 kcal/min with placebo to 1.24 kcal/min with LY171883 (p less than 0.05). A similar difference was noted when responses were expressed as a function of minute ventilation. LY171883 produced greater shifts in the PD20RHE in more reactive subjects (r = 0.69, p less than 0.002). Among 11 different symptom scores recorded by 18 subjects, there was a significant decrease in daytime chest tightness with LY171883 (p less than 0.03). The increase in PD20RHE while the subjects received LY171883 is consistent with the hypothesis that LTD4 becomes available during cold-air ISH and may mediate bronchoconstriction. The small magnitude of the effect on the PD20RHE may be due to the role of other mediators in cold-air-induced bronchoconstriction or, alternatively, to an inadequate blockade of LTD4 effects.

Monitoring versus interim analysis of clinical trials: a perspective from the pharmaceutical industry.

The definitions of "interim analysis" and "monitoring" of clinical trials are often ambiguous in the current literature. The resulting confusion can lead to erroneous conclusions and misguided decisions, especially when activities that are operational or observational are evaluated in a probabilistic sense as inferential. The authors seek to define "interim analysis" and "monitoring" in a mutually exclusive fashion. These definitions will then provide the opportunity to review and categorize existing clinical trial practices and procedures. This will clarify such issues as "when to look" and "when to pay a price" (e.g., test size and power) and characterize such issues in the context of pharmaceutical industry drug development.

Use of fluoxetine, a selective serotonin-uptake inhibitor, in the treatment of obesity: a dose-response study (with a commentary by Michael Weintraub).

Pharmacologic measures which increase serotonergic activity in the brain decrease food consumption and lead to decreased weight in animals. Fluoxetine, an inhibitor of serotonin reuptake, decreases food intake in animals and is associated with weight loss in depressed and otherwise healthy obese patients. To determine the most effective daily fixed dose which causes weight loss in nondepressed obese patients, fluoxetine (10, 20, 40 or 60 mg) or placebo was administered once daily for 8 weeks to 655 patients consisting primarily of women (mean age 40 years, mean weight 95 kg). Diet and activity were not controlled. The placebo-treated patients lost 0.6 +/- 2.3 kg. With the 60-mg fluoxetine dose, patients lost an average of 4.0 +/- 3.9 kg (P less than 0.001), with intermediate responses at the lower doses. Weight loss was proportional to the initial body mass index (weight/height squared). There were no statistically significant differences between any fluoxetine treatment group and placebo for discontinuations from the study. There were statistically significant dose-dependent increases in reports of asthenia, somnolence and sweating. Thus, fluoxetine 60 mg daily appears to be potentially effective for use in weight reduction.

Effect of fluoxetine on psychomotor performance, physiologic response, and kinetics of ethanol.

The effects of fluoxetine, a specific serotonin reuptake inhibitor, on the psychomotor performance, physiologic response, and kinetic disposition of ethanol were examined. Fluoxetine (30 or 60 mg) with ethanol (45 ml absolute alcohol per 70 kg body weight) did not alter the plasma or blood concentrations of fluoxetine or ethanol, respectively, when compared with levels after either drug alone. There was no significant effect on standing or recumbent blood pressure or heart rate after single or multiple doses of fluoxetine alone or in the combination. Single or multiple doses of fluoxetine had no effect on the psychomotor activity (stability of stance, motor performance, or manual coordination) or subjective effects of alcohol. Data indicate that fluoxetine does not inhibit ethanol metabolism nor does it have any effects on its psychomotor activity.

Computed tomography in the prediction of outcome in head injury.

To determine the prognostic significance of computed tomographic (CT) findings in head injury, retrospective analysis was performed in 128 randomly selected severe head-injury patients managed with a standardized protocol. The minimal criterion for entry into this study was that the patients were unable to obey simple commands or utter formed words. Serial CT was performed on admission and 3-5 days, 2 weeks, 3 months, and 1 year after injury. A scale of severity of abnormalities was devised taking into account the size of the traumatic lesions on CT. The CT findings using the proposed scale were correlated with the clinical outcome and analyzed using linear logistic regression. Other characteristics such as midline shift, multiplicity, and corpus callosum and brainstem lesions were not included in the analysis either because they did not affect the prognosis or because too few of these lesions were present for statistical analysis. The correct prediction rate of outcome using the proposed scale for CT findings alone was found to be 69.7%. When CT findings were combined with the Glasgow Coma Scale score this rate was increased to 75.8%.

Fluoxetine: clinical pharmacology and physiologic disposition.

Fluoxetine (30 mg), administered for 7 days to normal volunteers, produced a 66% inhibition of tritiated serotonin uptake into platelets. Plasma concentrations of fluoxetine correlated positively with inhibition of serotonin uptake. Fluoxetine is well absorbed after oral administration in both the fed and fasted states and demonstrates dose proportionality. Fluoxetine disappears from plasma with a half-life of 1-3 days; its metabolite norfluoxetine has a plasma half-life of 7-15 days. After administration of 14C-fluoxetine, approximately 65% of the administered dose of radioactivity is recovered in urine and about 15% in feces. Fluoxetine, given as a single dose or in multiple doses over 8 days, did not produce significant effects on the plasma disappearance of warfarin, diazepam, tolbutamide, or chlorothiazide. Coadministration of fluoxetine and ethanol did not result in an increase from control values in the blood ethanol levels, nor did it produce significant changes in physiologic, psychometric, or psychomotor activity. Pharmacokinetics of fluoxetine in the elderly and normal volunteers appear to be similar. In addition, pharmacokinetic analyses in patients with varying degrees of renal impairment did not show significant differences from healthy subjects.

Clinical pharmacology of tomoxetine, a potential antidepressant.

Tomoxetine (LY139603) selectively inhibits norepinephrine uptake in animals and has activity in animal models of depression. Tomoxetine was administered in single oral doses up to 90 mg to healthy normal volunteers. In addition, normal human subjects received either 20 or 40 mg of tomoxetine b.i.d. for 1 week to evaluate the safety and pharmacologic activity of the compound in humans. At these doses, no serious drug-related adverse effects were encountered. Activity of the compound at the lower dose (20 mg b.i.d.) was evaluated by examining changes in the pressor responses to infused norepinephrine and tyramine and by determining [3H]serotonin uptake in platelets harvested from subjects receiving the compound. Pressor sensitivity to norepinephrine was increased by 261 +/- 69% of control, and pressor sensitivity to tyramine was decreased by 51 +/- 6% of control during treatment. Changes in the pressor sensitivity to norepinephrine in individual subjects were positively correlated with drug levels. There were no statistically significant changes in platelet [3H]serotonin uptake. These results indicate that tomoxetine selectively inhibits norepinephrine uptake in humans at doses which are clinically well tolerated and suggest that tomoxetine has potential clinical use as an antidepressant.

Effects of therapeutic pentobarbital coma on multimodality evoked potentials recorded from severely head-injured patients.

We sought to determine whether pentobarbital (PB) coma compromises the use of evoked potentials (EPs) in the assessment of brain dysfunction and of the prognosis of severely head-injured patients. Therefore, the effects of therapeutic PB on somatosensory (SEPs, BSEPs), visual (VEPs), and auditory (BAEPs) evoked potentials recorded from 20 patients early after injury were analyzed. Seventeen head-injured patients served as controls. EP studies were obtained shortly after admission (Mean Day 2, PB present) and approximately 2 weeks after injury (Mean Day 15, PB absent). The mean serum level of PB in the treatment group was 1.9 mg/100 ml. The drug effect was assessed by comparisons between the PB and the control groups. Statistical analyses were based on differences observed between two studies in the same patient. Analyses of covariance (F tests) were performed on data from all modalities. Wave form complexity was minimally affected by the drug. Middle and long latency components of the SEP were depressed by PB, and latencies of BSEP peaks and the early components of the SEP were delayed. The amplitude of some VEP peaks was reduced by PB. The BAEP was not significantly altered. All of the observed effects of PB were determined to be due to the hypothermia exhibited by PB-treated patients (mean temperature, 36.1 degrees C), which was not seen in the control group (mean, 37.8 degrees C). It is concluded that, with appropriate interpretation, EPs can be used to monitor brain function in head-injured patients when PB therapy is used.

The dynamics of neuronal dysfunction and recovery following severe head injury assessed with serial multimodality evoked potentials.

Serial studies of visual, auditory, and somatosensory evoked potentials (EP's) obtained from 139 severely head-injured patients up to 1 year after trauma were analyzed to ascertain whether or nor EP's can be used to monitor neurological recovery or deterioration following secondary insults. The EP data were analyzed using a grading system of abnormality developed previously, and patients were grouped by the most severe EP abnormality found in any modality during an early study (mean Day 3). The findings showed differential recovery trends depending on the severity of EP abnormality obtained on the initial study and presence of secondary insult. If EP's were normal early after injury, they remained so for up to 1 year, and these patients did well clinically. The EP's that were absent did not improve, and the patients had poor outcomes. Secondary insults did not affect the EP's or the outcomes of patients in these two groups. When EP's that were initially mildly abnormal became normal or remained no worse than mildly abnormal, patients had favorable outcomes in spite of complications. In contrast, deterioration of EP's with secondary insult indicated poor patient outcome. Severe EP abnormalities which improved over time led to favorable outcomes. However, persistence or deterioration of severe abnormalities indicated a poor outcome. Changes in EP's over time were better indicators of outcome than the presence or absence of complications. The results suggest that EP's may be used to assess neural recovery and the consequences of secondary insults to the brain. Four case reports are included to exemplify results.

Improved confidence of outcome prediction in severe head injury. A comparative analysis of the clinical examination, multimodality evoked potentials, CT scanning, and intracranial pressure.

An analysis of clinical signs, singly or in combination, multimodality evoked potentials (MEP's), computerized tomography scans, and intracranial pressure (ICP) data was undertaken prospectively in 133 severely head-injured patients to ascertain the accuracy, reliability, and relative value of these indicants individually, or in various combinations, in predicting one of two categories of outcome. Erroneous predictions, either falsely optimistic (FO) or falsely pessimistic (FP), were analyzed to gain pathophysiological insights into the disease process. Falsely optimistic predictions occurred because of unpredictable complications, whereas FP predictions were due to intrinsic weakness of the indicants as prognosticators. A combination of clinical data, including age, Glasgow Coma Scale (GCS) score, pupillary response, presence of surgical mass lesions, extraocular motility, and motor posturing predicted outcome with 82% accuracy, 43% with over 90% confidence. Nine percent of predictions were FO and 9% FP. The GCS score alone was accurate in 80% of predictions, but at a lower level of confidence (25% at the over-90% level), with 7% FO and 13% FP. Computerized tomography and ICP data in isolation proved to be poor prognostic indicants. When combined individually with clinical data, however, they increased the number of predictions made with over 90% confidence to 52% and 55%, respectively. Data from MEP's represented the most accurate single prognostic indicant, with 91% correct predictions, 25% at the over-90% confidence level. There were no FP errors associated with this indicant. Supplementation of the clinical examination with MEP data yielded optimal prognostic power, an 89% accuracy rate, with 64% over the 90% confidence level and only 4% FP errors. The clinical examination remains the strongest basis for prognosticating outcome in severe head injury, but additional studies enhance the reliability of such predictions.