Factors that the researcher intends to impose or observe based on the research purpose, which can act on the research object and cause direct or indirect effects

Subjective imposition or objective existence

Ideally, only one treatment factor should be involved in a study, and other influencing factors should be controlled as confounding factors.

The processing factors should be standardized, that is, the processing factors should be consistent throughout the entire process of a study and cannot be changed at will.

Basic unit to receive processing

Except for phase I drug clinical trials, which use healthy people as the research subjects, other drug clinical trials and medical device clinical trials use patients as the research subjects.

The experimental plan must clearly define the conditions for inclusion of research subjects and include clear inclusion and exclusion criteria to ensure the homogeneity of research subjects.

All research objects that meet the inclusion criteria are the population of the study, and those selected for research are samples. It is necessary to ensure that the samples are representative.

In clinical studies that use healthy people or patients as research subjects, it is also necessary to ensure that the research subjects have good compliance.

The objective reactions and results of processing factors acting on the research object are generally expressed through some kind of observation index.

The selected observation indicators should be able to objectively, effectively and accurately reflect the effects of treatment factors. Improper indicators will make the results lack scientificity and reliability.

It is best to choose objective indicators as the main efficacy indicators in experimental studies to reduce the adverse effects of subjective impressions from both doctors and patients.

Subjective indicators can be quantified or graded

Visual analogue scale/score (VAS): Draw a 10cm horizontal line on the paper. The left side of the horizontal line is 0, indicating no pain; the right end is 10, indicating the most severe pain, and the middle part is varying degrees of pain

Sensitivity: when the treatment factor exists, the observation index can reflect its experimental effect

Specificity: The observation indicator does not show its experimental effect when the treatment factor does not exist

Purpose: as a reference level to set off the treatment factor effect of the experimental group

The control group established in the study must be balanced and consistent with the experimental group. In addition to different treatment factors, the distribution composition of other important non-treatment factors in the control group must be as consistent as possible with the experimental group, and the distribution of the control group and the experimental group must be consistent. The susceptibility and chance of occurrence of the disease under study must be equal among the research subjects. The detection, observation methods, diagnostic standards and randomization methods of the two groups must also be consistent.

Control type

Ensure the representativeness of the sample and ensure that the distribution of a large number of uncontrollable non-treatment factors between each treatment group is as balanced and comparable as possible.

Random sampling: means that every subject in the population has the same chance of being selected, so that the sample can be representative of the population

Random allocation: The fundamental characteristic of experimental research that is different from observational research is that each research subject has the same opportunity to enter each group, which is an important means to ensure balance and consistency among experimental subject groups. It can not only ensure that the research subjects in each treatment group are balanced and comparable in terms of various non-treatment factors such as demographic characteristics and disease severity, but also avoid the subjective factors of the researcher from interfering with the trial grouping.

Conduct multiple experiments in the same experimental conditions to improve the reliability and accuracy of the experiment

Reproducibility: Reliable experimental results should be reproducible under the same conditions. Unrepeatable experimental results are not scientific.

Sample size: Reliable experimental conclusions cannot be obtained only from the results of a few examples. Accidental results obtained in individual cases cannot be regarded as inevitable rules. There must be a sufficient number of observation units to make the results stable.

In order to avoid the interference of subjective factors and psychological factors of researchers, subjects, efficacy evaluators, diagnostic result evaluators, statistical analysts, etc. on the research results, in order to control the bias generated during the clinical trial process and interpretation of the results.

Double blindness: Neither the researcher nor the subjects know the trial grouping situation, which can truly avoid the subjective psychological influence of both the researcher and the subjects.

Single blindness: Only the researcher knows the trial groupings

Third blind evaluation (third blind evaluation): both researchers and subjects know the trial grouping situation, mainly limiting the efficacy evaluators or diagnostic result evaluators (reading medical impact personnel, endpoint committee, etc.)

Open trial (non-blindness test): clinical studies that do not use blinding methods, such as surgery and chemotherapy, etc., where blinding methods cannot be used.

Blind analysis: Only the group code of each subject is revealed after the trial. The statistical analysts only know whether the subject is group A or group B, but do not know who is in group A and group B. Experimental group, what is the control group, in order to control the subjective interference of analysts

Implementation means: placebo technology, capsule technology, double simulation technology

Definition: Regardless of individual differences, only one treatment factor is arranged in an experiment, so it is also called a single-factor design or a group design. Homogeneous research subjects are randomly assigned to multiple treatment groups for experimental research, or Comparative observational studies using random samples from several different populations

A completely randomized design involves only one treatment factor, which can have two or more levels, and the sample sizes of each treatment group can be equal or unequal. The design is most efficient when the sample sizes of each group are equal.

Advantages: The method is simple, the application is flexible, the operability is strong, and the number of processing groups and the sample size of each group are not limited. If the subject falls off, it will not affect other subjects, and the loss to the experimental results will be smaller than other design methods.

Disadvantages: Only one treatment factor can be designed and analyzed, and the distribution of various non-treatment factors among the groups can only be balanced by simply relying on random grouping of research subjects. Therefore, the accuracy of the experiment is low, and sometimes the balance between groups will be compromised. Relatively poor

Definition: The subjects are first divided into blocks according to matching conditions, and then the subjects in each block are randomly assigned to each treatment group.

Use the main non-experimental factors as the matching conditions instead of experimental factors as the matching conditions

Each block can have two or more subjects. When there are only two subjects in each block, it is also called a paired design.

Advantages: The subjects are divided into blocks according to the pairing conditions, thereby reducing the influence of many non-treatment factors on the experiment, increasing the balance between the groups, and reducing the experimental error; and increasing the block information, reducing the research time Individual differences between subjects can reduce sample size and improve statistical efficiency

Disadvantages: Limited by matching or compatibility conditions, it is sometimes difficult to match subjects into blocks, thus losing some of the subject's information. Moreover, when a subject falls out of a block, the information loss is relatively large.

Three-factor experimental design arranges one factor according to the rows, columns and letters of the Latin square. Generally, different levels of the treatment factors are represented by different letters. Row factors and column factors are usually used to consider block control in two directions, so Latin square design is a two-way compartmentalized design

The Latin square is a k*k square matrix arranged by k Latin letters. Each letter in each row or column appears only once. Such a square matrix is ​​called a k-order Latin square table.

Advantages: Three factors can be analyzed at the same time, two important non-experimental factors are controlled, the experimental error is further reduced, the sample size is saved, and the experimental efficiency is higher

Disadvantages: The level of each factor must be equal, and there must be no interaction between the three factors. When missing data occurs, the loss of experimental information will be greater.

One treatment factor was considered, and the impact of two non-treatment factors (experimental stage and subjects) that had no interaction with the treatment factor on the experimental results was considered.

Steps: Assume that the treatment factor has two levels A and B. First, the subjects are randomly divided into two groups. The subjects in the first group receive treatment A in stage I, and the subjects in stage II receive treatment B. The subjects in the second group Subject receives treatment B in phase I and treatment A in phase II

Washout period: The test requires that a certain interval (washout period) must be arranged between the two stages in order to eliminate the residual effects of the previous treatment and ensure that the starting conditions of the two stages are consistent. Generally requires at least greater than 7 drug half-lives

It is suitable for diseases with relatively stable conditions and disease courses that can be staged. It is not suitable for diseases with a tendency to self-heal or with a short course.

Advantages: Reduces the impact of individual differences on treatment factors, saves sample size, can control the impact of time factors (trial phase) on treatment methods, each subject receives two intervention measures, and is more in line with ethical requirements

Disadvantages: It is only suitable for diseases with relatively stable conditions and disease courses that can be staged. Since treatment must be stopped during the washout period, cases may easily drop out or drop out. Once dropped out, there will be a large loss of data.

A multi-factor experimental design that completely cross-groups multiple experimental factors at different levels can not only compare the levels of each factor, but also analyze the interaction between factors. If there is interaction between factors, it means that each factor is not independent, and changes in the level of one factor will affect the experimental effects of other factors; if there is no interaction between factors, it means that each factor is independent, and the level of a certain factor changes. , will not affect the experimental effects of other factors

The difference between factorial design and completely random design: the treatment group of factorial design is a comprehensive combination of different levels of two or more treatment factors, that is, the number of treatment groups is equal to the product of the number of levels of the factors to be studied

The 2*2 factorial design means that there are two factors, each factor has two levels, and there are 4 combinations. The two-factor factorial experimental design can be used to study whether there are differences between different levels within the two factors, especially in research Is there any interaction between the two factors?

Advantages: High efficiency. It can not only analyze whether there are differences between different levels within each factor, but also has the function of analyzing the interaction of various combinations.

Disadvantages: As the number of factors and levels increases, the number of subjects and analysis workload required increases sharply, and the interpretation of high-order interaction analysis results is also more complicated.