### Abstract

Background : Calculating appropriate sample size is indispensable process in planning clinical trials. If the result of a comparative clinical trial cannot reject null hypothesis, four possibilities might be considered. Three of those can be the truly no difference of therapeutic effects between the two drugs, poor compliance of the patients to the prescribed drugs, and the effect of bias during the study. The last possibility can be the falsely accepting the null hypothesis due to the lack of statistical power. This paper reported power analysis with non-continuous outcome variables in comparative controlled clinical trials to point out latter problem. Methods : A clinical trial of new anti-cancer agent using two groups was simulated with binary outcome variable. Another clinical trial was simulated to evaluate two treatment for lung cancer patients. After double-blind randomization to radiotherapy alone group and chemotherapy parallel group, follow-up observations were conducted. Outcome variable is survival time. By using PASS program, the statistical power and least significant number of sample size were calculated in the two different situation. Results : The statistical power of anti-cancer agent data was 29%, which meant that sample size need to be increased. The least significant sample size for detecting the therapeutic difference between groups as statistically significant requires 200 persons. The statistical power of lung-cancer data was 39%, which also meant that sample size need to be increased. The least significant sample size for detecting the therapeutic difference between groups as statistically significant requires 84 persons. Conclusion : Inadequate sample size in clinical trials can cause wrong decision due to decrease in statistical power, so proper power analysis is needed. The program introduced in this paper can be used in design, interim analysis, and final analysis of clinical trials to calculate statistical power very usefully.

Original language | English |
---|---|

Pages (from-to) | 168-177 |

Number of pages | 10 |

Journal | Journal of Korean Society for Clinical Pharmacology and Therapeutics |

Volume | 8 |

Issue number | 2 |

State | Published - 1 Dec 2000 |

### Fingerprint

### Keywords

- Clinical trial
- Non-continuous outcome variable
- Sample size
- Statistical power

### Cite this

*Journal of Korean Society for Clinical Pharmacology and Therapeutics*,

*8*(2), 168-177.

}

*Journal of Korean Society for Clinical Pharmacology and Therapeutics*, vol. 8, no. 2, pp. 168-177.

**Power analysis for comparative controlled clinical trial with non-continuous cutcome variables.** / Kwak, M.; Lee, Y. J.; Park, Byung Joo.

Research output: Contribution to journal › Article › Research › peer-review

TY - JOUR

T1 - Power analysis for comparative controlled clinical trial with non-continuous cutcome variables

AU - Kwak, M.

AU - Lee, Y. J.

AU - Park, Byung Joo

PY - 2000/12/1

Y1 - 2000/12/1

N2 - Background : Calculating appropriate sample size is indispensable process in planning clinical trials. If the result of a comparative clinical trial cannot reject null hypothesis, four possibilities might be considered. Three of those can be the truly no difference of therapeutic effects between the two drugs, poor compliance of the patients to the prescribed drugs, and the effect of bias during the study. The last possibility can be the falsely accepting the null hypothesis due to the lack of statistical power. This paper reported power analysis with non-continuous outcome variables in comparative controlled clinical trials to point out latter problem. Methods : A clinical trial of new anti-cancer agent using two groups was simulated with binary outcome variable. Another clinical trial was simulated to evaluate two treatment for lung cancer patients. After double-blind randomization to radiotherapy alone group and chemotherapy parallel group, follow-up observations were conducted. Outcome variable is survival time. By using PASS program, the statistical power and least significant number of sample size were calculated in the two different situation. Results : The statistical power of anti-cancer agent data was 29%, which meant that sample size need to be increased. The least significant sample size for detecting the therapeutic difference between groups as statistically significant requires 200 persons. The statistical power of lung-cancer data was 39%, which also meant that sample size need to be increased. The least significant sample size for detecting the therapeutic difference between groups as statistically significant requires 84 persons. Conclusion : Inadequate sample size in clinical trials can cause wrong decision due to decrease in statistical power, so proper power analysis is needed. The program introduced in this paper can be used in design, interim analysis, and final analysis of clinical trials to calculate statistical power very usefully.

AB - Background : Calculating appropriate sample size is indispensable process in planning clinical trials. If the result of a comparative clinical trial cannot reject null hypothesis, four possibilities might be considered. Three of those can be the truly no difference of therapeutic effects between the two drugs, poor compliance of the patients to the prescribed drugs, and the effect of bias during the study. The last possibility can be the falsely accepting the null hypothesis due to the lack of statistical power. This paper reported power analysis with non-continuous outcome variables in comparative controlled clinical trials to point out latter problem. Methods : A clinical trial of new anti-cancer agent using two groups was simulated with binary outcome variable. Another clinical trial was simulated to evaluate two treatment for lung cancer patients. After double-blind randomization to radiotherapy alone group and chemotherapy parallel group, follow-up observations were conducted. Outcome variable is survival time. By using PASS program, the statistical power and least significant number of sample size were calculated in the two different situation. Results : The statistical power of anti-cancer agent data was 29%, which meant that sample size need to be increased. The least significant sample size for detecting the therapeutic difference between groups as statistically significant requires 200 persons. The statistical power of lung-cancer data was 39%, which also meant that sample size need to be increased. The least significant sample size for detecting the therapeutic difference between groups as statistically significant requires 84 persons. Conclusion : Inadequate sample size in clinical trials can cause wrong decision due to decrease in statistical power, so proper power analysis is needed. The program introduced in this paper can be used in design, interim analysis, and final analysis of clinical trials to calculate statistical power very usefully.

KW - Clinical trial

KW - Non-continuous outcome variable

KW - Sample size

KW - Statistical power

UR - http://www.scopus.com/inward/record.url?scp=0034462395&partnerID=8YFLogxK

M3 - Article

VL - 8

SP - 168

EP - 177

JO - Journal of Korean Society for Clinical Pharmacology and Therapeutics

JF - Journal of Korean Society for Clinical Pharmacology and Therapeutics

SN - 1225-5467

IS - 2

ER -