How Does Statistics Help Create Medicines - Assignment Example

ORD243943 How Does Statistics help create Medicines? Statistics, which has remained well known as a science to study the performance of a government,is a complex discipline aiding much of modern science today. It is used for making decisions of quality and planning strategy. It provides solutions to the uncertain outcomes after identifying these. The drug development is a very lengthy and costly process and statistician is important team member along with scientists right from the beginning. The clinical trials are designed and executed with the help of a statistician. He is responsible for providing feasible research designs viz. Cross over, parallel, probabilities of getting and analyzing non-parametric, ANOVA, ANCOVA etc. of a trial ( Donohue & Ruberg, 1997). Non- parametric data do not show a normal or Gaussian distribution. For example, a bronchodilator is tested on asthma patients. The baseline and drug treated forced expiratory volume in one second (FEV1) is determined. The values are different for different subjects i.e. there is no clear trend in these data. The statistician in consultation with physician finds out that the peak-expiratory flow (PEF) is a more reliable data in such case (Senn, 2008). The other important aspects that a statistician must be good at are randomization of schemes to determine safety and efficacy and other variables of the drug under development and sample size determination from the target population (Donohue & Ruberg, 1997; Esser, 2001). Randomization is necessary to generate comparable groups of patients from a population. The new drug must establish a difference in efficacy from a control (or placebo) with statistical assurance. The parallel or Latin square designs are used for randomization in traditional statistics. A blinding is also the part of design where the patient and investigator both remain unaware of the treatment. Blinding prevent biases in data collection and analysis (Chow & Shao, 2002). They further suggest that the statistician should determine a sample size in consultation with the physician. An appropriate test and significance level is selected with a null and an alternative hypothesis. A large sample size is preferred particularly if data show great variation since it decreases type I and type II errors. The significance level is the percent wrong decision, which can be tolerated. Usually in drug development studies a 5% significance level and 95% confidence limit are taken. Donohue & Ruberg (1997) find that the statistician is not only required at the drug development planning, he actually is involved in analysis and interpretation of every individual experiment of the process. The statistician also plays a fundamental role in deciding whether the trial must continue or not. Screening and Preclinical trials The chemicals with potential medicinal activities are collected from the world over and screened for the active ingredient or chemical. It is known as screening for new drug. The statistician is involved in this phase by helping design experiments so that effective use of time and materials is made. Besides, he carries out the data analysis for identification of chemical with actual drug activity leading to the isolation of it by the scientists (Bolognese, 2008). In an example Dewitt et al (2004) report testing statistical hypothesis on parameters identified in a Low Throughput Screening using small number of potential chemicals and used the results as reliable base to find out hits from large number of compounds selected after High Throughput Screening. Drugs are also synthesized by combinatorial synthesis. Where the two chemicals are combined to give a medicine of superior efficacy with safety. The statistician works with probability theory so that all possible combination with drug activity are formed (Bolognese, 2008). The statistician works with basic scientists to test the active compound for safety and efficacy on minimum possible number of animals in the preclinical phase of study. It is necessary to remove toxic component of drug before it is tested on humans. The statistician also help in arriving to best production method, potency and storage of drug prior to the clinical trials (Bolognese, 2008). To show usefulness of statistical analysis Williams et al (1987) report a preclinical trial where interpretation was made using statistical techniques. The nephrotoxicity of amikacin [3H] was found correlated to its binding to the renal membranes. Since there was significantly lower binding to renal membranes of immature rats which showed mush less toxicity to this antibiotic. The statistical analysis indicated that nephrotoxicity potential is dependent on inherent toxicity and the binding affinity of the antibiotic. Clinical Trials: Once the drug is found non-toxic on animal models and standardized for reliable production, it is tested on humans in three trials. Phase I trial is on healthy volunteers (not having the target disease) to ascertain safety in human and its pharmacological properties. The statistician decides appropriate statistical tests, hypothesis and level of significance prior to testing of a new chemical entity on human volunteers. As an example, a phase I trial of drug whose effect is to be find on two genders as well, may be done by a two factor sequential ANOVA with four tests. These are- An interaction between the linear trend of treatment and gender (test A), The overall linear treatment for combined genders (test B), The linear treatment for each gender (test C), Test C when test A is significant else test B. For a dose- response effect F-test, t-test are conducted at a significance level that avoid or minimize type I error (Dmitreinko et. al., 2007). The phase II trial are on patients suffering from target disease to find out the appropriate doses. The phase III trials are on very large number of patients with disease to finally confirm its safety and efficacy (Bolognese, 2008). The statistician relies on appropriate endpoints of these clinical trials to proceed to next phase. PhRMA (2005) found that during assessment of the QT/QTc prolongation by the selected drugs , the end points differ in different stages of trials. QT prolongation causes malignant ventricular arrhythmia. A QT interval is measured from the beginning of the Q wave to the end of T wave in Electrocardiogram (ECG). The early phase trial data focus on safety and efficacy aspect while later trials have different end points. Depending on the requirements and objectives the sample size and observations would differ in different stages of drug development. They observed that the extent of evaluation of QT/QTc effects in Phase 2/3 should depend on the outcome from Phase 1, It is generally assumed that if a true QT/QTc prolongation signal exists, it will be dose-related. To ascertain this, a sequential testing procedure was used (Bauer et al, 1998 as cited in PhRMA, 2005:248) to identify the maximum safe dose of test product relative to placebo (Hauschke & Hothorn, 1998 as cited in PhRMA, 2005: 248) The following one-sided test will first be done for a maximum safe dose level: H01: µH µP δ versus H11: µH µP δ where µH is the mean QTc endpoint observed for the maximum safe dose of test product while µP is the mean following placebo administration, and δ is a upper limit set by regulatory guidance. The one-sided test shows that only QTc elongation is important clinically. In case H01 is not rejected in favor of H11, the following would be tested to see if the therapeutic dose of test product could be demonstrated to be safe: H02: µT µP δ versus H12: µT µP δ. Here µT is the mean QTc observed following the therapeutic dose of test product. If H01 is rejected, H02 is not of direct interest to decision making (PhRMA, 2005). They sought statistical guidance to select the optimal number and timing of ECGs for the objectives of its study based on the patient population, endpoint, statistical model, sample size, and cost effectiveness. The screening rate is highest at the phase II or the selection trials where a drug may fail to do what is expected from it i.e. cure the disease, provide relief from symptoms or kill the microbes of the disease. The sooner the ineffective drug is rejected the less is wastage of time and money (Donahue & Ruberg, 1997). . Statisticians are important members of these clinical trial teams. They help design the trials, select population, determine sample size, collect data of clinical trials , analyse these and report interpretations. They write the report along with other scientists which are submitted to regulatory authorities for approval of the drug (Bolognese, 2008). The statisticians need to satisfy their counterparts in regulatory agencies about the new drug application. They may require to further design the phase IV and V trials as advised by the regularity agencies. All these well-planned activities are necessary as today the pharmaceutical companies want to enter all major markets of the world. They also plan a global development program that consists of clinical studies conducted in a multinational setting almost all over the world to achieve this goal (Esser, 2001). The statistician also helps compare the cost of drug with alternative therapies available and its impact on the company. The statistical analysis makes the scientific data clear (Bolognese, 2008). The testing of the hypothesis that oxytocin antagonist atosiban decreases preterm uterine contraction in human established it as a useful medicine for subjects suffering frequent pains due to these contractions. It also confirmed that there are no significantly adverse side effects which later proved clinically true as well (Goodwin et al., 1994). Shao and Chow (1994) show that the shelf life of a drug is also determined by statistical methods. If a drug does not show batch to batch variation a standard shelf life ( on the label ) is sufficient. It should be less than or equal to the true shelf life. However, If there are batch to batch variation, the shelf life (label) is to be calculated individually for all batches. Adaptive/Flexible Statistical designs: Posch et al (2000) observed that adaptive experimental designs integrate the phases I and II into one trial thus expedite the drug development. The adaptive or flexible designs add or drop a treatment in the ongoing experiment and test the intersection hypothesis by a combination test. For example a trial begins with two treatments and a control. In the Interim analysis, however, only Treatment 1 is continued along with the control group. The p and q are stage wise p-values. Then the intersection hypothesis is rejected if C (p12, q12) = C(p12, q1) = 1. H1 is rejected at the multiple level  if the intersection hypothesis is rejected and C(p1, q1) = 1. H2 is rejected, if the intersection hypotheses is rejected and C (p2, q2) = 1. However, since Treatment 2 has been dropped in the interim analysis, q2 = 1. Therefore, H2 is rejected only if p2  a. Here a is critical parameter. The adaptive testing procedures have the advantage that when dropping treatments the critical boundaries for the remaining treatments can be narrowed which does not happen in traditional or frequentist statistics. Additionally, the adaptive approach allows for other design modifications such as sample size reassessment and adding new treatments at an interim analysis. The treatment with highest interim effect is selected for next stage. Bayesian designs, very well known and used adaptive designs are useful for sample size determination and Phase I/II trials. This approach uses the prior data available in literature, in calculation and analysis of ongoing research. Bekele and Shen (2005) conducted oncology trial for a new gene therapy for bladder cancer patients using Bayesian approach. They jointly modeled a binary toxicity outcome and a continuous biomarker expression outcome. The latter expression indicates biological activity of the new therapy. They created a flexible dose-activity model using data from previous experiments to correlate dose with activity as a Gaussian variable for their experiment. Statistics and the Choice of drug: The user trusts an approved drug. Since it signifies that the drug is tested for safety and effectiveness before marketing. The FDA approval assures that adequate evidences are produced for drug’s safety and efficacy (Pharma mark., 2006). A crossover study of metastatic colorectal treatment indicated that patient prefer the oral chemotherapy compared to intravenous. The former is convenient. Besides the intravenous activity peaks sooner and higher but it is accompanied by an increase in toxicity also. The FDA approval requires statistician’s role in hypothesis, experimentation (observation), interpretation, and conclusion or modified hypothesis for next experiment. Food and Drug Administration (FDA) satisfies itself by (1) evaluating the adequacy of the evidence, and (2) approving drugs with adequate evidence for marketing (Chow & Shao, 2002). A physician can trust the drugs for prescription because current drug approval regulations require extensive, extremely accurate testing to demonstrate the safety and effectiveness of drugs before marketing. Statisticians enable scientists and reviewers to establish the accuracy and validity of these tests (Bolognese, 2008). The statistical assurance on such aspects provides greater acceptance to a new drug for prescription. Lichtenberg (2001) reports that analysis of data favors new drugs for treatment. A study on prescribed medicines from the 1996 Medical Expenditure Panel Survey (MEPS) indicated that the use of newer medicines decreases morbidity, mortality, and health spending. It was found that newer drugs had significantly increased the life of patient and there were fewer work loss days in comparison to older drugs. What was noticeable that newer drugs reduced nondrug medical expenses substantially reducing the total cost of treatment. The success of a drug depends on many aspects according to Abrantes-Metz (2003): The newly introduced cancer drugs are less successful compared to HIV-AIDS drugs The identified routes of drug administration such as alimentary increase probability of a drug’s acceptance compared to a “non-specified/other” baseline. Drugs based on biologicals such as proteins, chemicals and natural products have more probability of success than “non specified/other” label. Drugs sponsored by a large, well known company has more probability of success. The marketing strategy is also important to make physicians prescribe the new drug. The sales representative meets the physician initially to inform him about cycle of the new drug and when to prescribe it. He keeps on visiting the physician even after providing full information. The marketing of drugs goes from informative to persuasive (Pharma Mark., 2006) REFERENCES Abrantes-Metz, R. M., Adams, C. P. & Metz, A. D. (2003). Pharmaceutical Development Phases: a Duration Analysis. March 10th, 2003Retrieved October 17,2008 Website http://www.together.net/~cpadams/Pharma_AbrantesMetzAdams_v2.pdf Bekele, B. N. & Shen, Y. A (2005). Bayesian Approach to Jointly Modeling Toxicity and Biomarker Expression in a Phase I/II Dose-Finding Trial. Biometrics, 61, 2, 343-354. Bolognese, J .A. Why can you trust the drugs you take? (and how Statisticians Contribute). Retrieved October 18, 2008 Website http://www.amstat.org/meetings/jsm/2000/usei/drugs.PDF Borner, M. (2003). Patient preference and pharmacokinetics of oral modulated UFT Versus intravenous fluorouracil and leucovorin a randomised crossover trial in advanced colorectal cancer . European Journal of Cancer , 38 ,3, 349 – 358. Lichtenberg, F. R. (2001). Are The Benefits Of Newer Drugs Worth Their Cost? Evidence From the 1996 MEPS. Retrieved October 22, 2008. Website http://content.healthaffairs.org/cgi/content/abstract/20/5/241 Chow, S. C. & Shao, J. (2002). Statistics in Drug Research: Methodologies and Recent Developments. CRC Press. DeWitt, A. (2004). Data to Knowledge in Pharmaceutical Research. Retrieved October 21, 2008. Website http://www.ima.umn.edu/modeling/mm04abstracts/final-reports/2004-team2-report.pdf Donahue, R. M. & Ruberg, S. J. (1997). Standardizing Clinical Study Designs For Accelerating Drug Development. Drug Information Journal, . 31, 655–663. Dmitrienko, A., Chuang-Stein, C. & D’Agostino, R. (2007). Pharmaceutical Stability in Using SAS: A practical Guide. SAS Publishing Esser, R. (2001). Biostatistics and Data Management in Global Drug Development. Drug Information Journal, 35, 643–653. Goodwin, T.M., Paul, R., Silver, H., Spellacy, W., Parsons, M., Chez, R., Hayashi, R., Valenzuela, G., Creasy, G. W. & Merriman, R. (1994). The Effect of the Oxytocin Antagonist Atosiban on Preterm Uterine Activity in the Human. American Journal of Obstetrics & Gynecology. 170(2): 474-478. Pharmaceutical Marketing, Chicago GSB, ,Spring 2006 Reports. Retreived October 20,2008. Website: http://www.chicagogsb.edu/magazine/spr06/YJ_40-43_Front.pdf Pharmaceutical Research and Manufacturers of America (PhRMA)QT Statistics Expert Working Team. (2005). Investigating Drug-Induced QT and QTc Investigating Drug-Induced QT and QTc Prolongation in the Clinic: A Review of Statistical Design and Analysis Considerations: Report from the Pharmaceutical Research and Manufacturers of America QT Statistics Expert Team. Drug Information Journal, 39, 243–266. Posch, M., Koenig, F., Branson, M., Brannath, W., Dunger-Baldauf, C. & Bauer, P. (2000). Testing and estimation in flexible group sequential designs with adaptive treatment selection. Statist. Med., 00:1–6 Shao, J. & Chow, S. C. (1994). Statistical Inference in Stability Analysis. Biometrics, 50, 753-763. Senn, S. (2008). Statistical Issues in Drug Development, 2nd Ed. Wiley-Interscience. Williams, P. D., Bennett, D.B., Gleason, C. R. & Hottendorf, G. H. (1987). Correlation between renal membrane binding and nephrotoxicity of aminoglycosides. Antimicrob. Agents Chemother. 1987 April; 31(4): 570-574 Read More