When students know that some (even most) published papers should be thrown into the dustbin, they are sure.

They are usually surprised when it cannot be used to guide clinical practice. The thesis is listed in the first block diagram of this paper.

Some common reasons for being rejected by magazines with peer review procedures.

At present, most papers published in medical journals are more or less written in the standard IMRAD format: that is, in the past,

Text (why did the author decide to do this research), method (how did the author do this research and how did he analyze it?

The results), the results (what the author found) and the discussion (what is the significance of the results). if

If you want to judge whether a paper is worth reading, you should review the scientific research design of the methodology part of this paper, and

It is not to review the importance of research hypotheses, the possible impact of research results or the depth of discussion.

Strict evaluation

Many EBM textbooks 2-6 and medical articles written by Sackett and his colleagues are published in JAMA magazine.

The user's guide 7-2 1 introduces the evaluation method (strict evaluation) of the quality of scientific research methods in detail.

If you are an experienced magazine reader, most of the structured lists provided by these authors are self-evident.

Make love. If you are not, please try to answer the following basic questions.

The main points of this article

Many papers published in medical journals have potential serious defects in methodology.

When judging whether a paper is effective or not and whether it is related to clinic, first determine what clinical problems the paper expounds.

title

Double-blind randomized controlled clinical trials should be used to explain the problems related to drug treatment or other medical interventions.

Longitudinal cohort study is needed to explain the prognosis. When expounding the problems related to etiology, it is necessary to

Cohort study or case-control study

Although the case report is not rigorous in methodology, it can be completed quickly and remind doctors to pay attention to medication.

adverse effect

Question 1: Why is this research conducted? What clinical problems did the author elaborate?

The preface of a scientific research paper should briefly explain the background of this research. For example, "Gro

Mmet insertion is a common operation in children, so some people think that not all operations are available in clinic.

It is necessary. "This explanation should be followed by a brief review of the published literature.

Why was the paper rejected for publication?

. This study did not elaborate on important scientific topics.

. This study is not original (others have done the same or similar research)

. This study does not really verify the hypothesis put forward by the author.

. Another type of research should be carried out.

. Difficulties in implementation (for example, screening research objects) made the author change the original research plan.

. The sample size is too small.

. There is no control group in the study or the control group is insufficient.

. Statistical analysis method is wrong or inappropriate.

. The conclusion drawn by the author from the research data is incorrect.

. There is an obvious conflict of interest (one of the authors or sponsors may gain economic benefits by publishing this article)

And there is not enough evidence to prove that there is no prejudice.

. The writing level of the paper is too poor to understand.

The author should clearly state the hypothesis to be tested in the preface of the article, otherwise it should be clearly stated in the methodology department.

This does mean. If it is assumed to be expressed in a negative way, such as "adding 2 to the maximum dose of iodourea treatment"

Metformin can't improve the control of type 2 diabetes ",which is the so-called zero hypothesis.

When the authors of a study begin to study, they seldom really believe their hypothesis that they are invalid. be like

Ordinary people, they usually begin to reveal the differences between the two aspects they study. But the way scientists use it

Yes, "let's assume that there is no difference first, and then try to deny this theory." If you follow Carl Pope,

This hypothesis-deduction method (establishing invalid hypothesis and then testing it) is a scientific research method.

The most basic element 22.

Question 2: What kind of research has been done?

First, determine whether this paper describes original research or second-hand (or comprehensive) research.

. The original research report first-hand studies the data, and the second-hand research summarizes the first-hand research and draws lessons from it.

Draw a conclusion. Most of the research published in medical journals are original research, which can usually be divided into the following three categories.

. Experiments: In experiments, experimental measures are carried out on animals or volunteers in a simulated and controlled environment.

ok；

. Clinical trial: In the clinical trial, first, a group of patients were intervened, such as drug treatment,

Then follow up this group of patients and observe their situation;

. Survey: In the survey, the researchers sample a group of patients, medical workers or other people.

Run some tests.

The second box in this paper shows some common terms used to describe research design.

Second-hand research includes:

. Review, can be divided into:

(non-systematic) review: summarizing the original research;

Systematic review: summarize the original research according to the predetermined and strict methods;

Summary analysis: Synthesize the data of more than one study.

. Guide: Draw conclusions about how clinicians should operate from the initial research.

. Decision analysis: the tree structure of probability is established by applying the original research results of medical workers and patients.

Choice of clinical treatment 24-26

. Economic analysis: using the results of original research to determine whether a certain treatment measure is reasonable in resource application.

.

Terms used to describe the design characteristics of clinical research

Parallel comparison between groups-each group received different treatments, and the two groups began to study at the same time; Results obtained

The two groups were compared and analyzed.

Paired comparison-Subjects receiving different treatments are paired to balance potential confounding factors, such as age.

And gender; The results were analyzed by the differences between matched subjects.

Comparison of the subjects themselves-check the subjects before and after treatment and compare the results with the research.

I like my changes to be analyzed.

Monoblindness-subjects don't know what kind of treatment they have received.

Double-blind-subjects don't know what kind of treatment they have received, and neither do researchers.

Cross-over-each subject received intervention and controlled treatment (in random order), during which they often used.

Untreated blank period separation

Placebo control-control subjects receive a placebo (ineffective pill), which should be in appearance and taste.

This is the same as an effective pill. Comfortable (sham) surgery can also be used in surgical clinical trials.

Factor design-this kind of research can affect more than one independent variable according to predetermined results (

Whether alone or in combination). For example, a 2×2 factor design can test a placebo alone.

The role of aspirin, streptokinase alone or aspirin plus streptokinase in acute heart attack 23

Question 3: Is scientific research design suitable for this research?

The best way to explain this problem is to consider which general scientific research field this research involves. vast

Most of the research is about one or more general areas in the box below.

The general field of scientific research

. Treatment: Test the effect of drugs, surgery, other medical services or other intervention measures.

The preferred research design is randomized controlled clinical trials.

. Diagnosis: confirm whether a new diagnostic experiment is effective (can we believe it) and reliable (we are

No, you can get the same result every time. The preferred research design is cross-sectional survey. In the cross-sectional survey,

The research object should adopt the new test method and the gold standard method for inspection.

. Screening: An examination method that can be used for large-scale crowd examination to find diseases before symptoms appear.

Value. The preferred research design is cross-sectional survey.

. Prognosis: determine what may happen to patients who find a disease early. Preferred research design

Research on vertical queuing

. Etiology: to determine whether some supposed harmful substances, such as environmental pollution, are related to the occurrence of diseases. Become the first choice

The study design is a cohort study or a case-control study, depending on the rarity of the disease, but the disease

Reports can also provide key information.

Randomized controlled clinical trial

In randomized controlled clinical trials, participants are randomly divided according to a procedure (similar to coin toss).

Assigned to an intervention group (such as a drug) or another group (such as a placebo treatment or a different drug). Both groups are

Follow-up in a specific period, and according to the preliminary findings (death, heart attack, serum cholesterol

Sterol levels, etc. Generally speaking, except for the treatment measures, the two groups are the same. Therefore, from

Theoretically, any difference in the research results is attributed to the treatment measures.

Some clinical trials comparing the treatment group and the control group are not randomized trials. Randomly assigned in these trials

This may be impossible, unrealistic or immoral-for example, comparing babies born at home with those born in hospitals.

The result of life. More commonly, inexperienced researchers compare one group (such as patients in ward A) with another.

Using this design, it is impossible to compare the two groups at the statistical level.

Compare them reasonably.

Randomized controlled clinical trials should be used to answer the following questions:

. Is this drug superior to placebo or another drug for a specific disease?

. For a specific disease, can the leaflet help patients make a better treatment plan than oral advice?

Make a wise choice

However, it should be kept in mind that randomized trials have some disadvantages (see block diagram). It should also be remembered that the results of randomized trials

Results Due to the exclusion criteria (determining which patients should not enter the study), the applicability was limited.

Then) bias; Inclusion criteria bias (selecting subjects from a specific population that does not represent the disease)

; Refusal of informed consent to certain patient groups included in the study28; Only the predetermined ones are analyzed

The "objective" final result may exclude important aspects of the quality of intervention measures; And publication deviation (

Selectively publish research papers with positive results) 29.

At present, the recommended format for reporting randomized controlled clinical trials in medical journals is 30, if that's what you wrote.

Newspapers should try their best to comply with its requirements.

Randomized controlled clinical trial design

superiority

. Allow a single change in a precisely selected patient group (such as menopausal women aged 50-60).

Quantity (such as drug treatment and placebo effect comparison) for strict evaluation.

. Prospective design (data from cases that occurred after the start of the study)

. Through hypothesis-deductive reasoning (seeking to deny, not to confirm one's hypothesis)

. Eliminate potential deviations by comparing two groups with the same baseline indicators (see below)

. Allow summary analysis (joint analysis of digital results of several similar clinical trials in the later stage)

disadvantaged

Expensive and time-consuming, so in practice:

. Many randomized controlled clinical trials have never been conducted, or too few patients have been studied, or have been studied.

The investigation time is too short.

. Most randomized controlled clinical trials are funded by large research institutions (universities or governments) or pharmaceutical companies.

Finally, these units will decide the research schedule.

. Frequent use of surrogate endpoint indicators instead of clinical measurement results will lead to "potential bias", especially

. Randomization is not ideal (see above)

. Not all qualified patients are randomly assigned (clinicians only let them know in clinical trials)

For patients who may respond well to the intervention)

. The data evaluator did not blindly analyze the random state of patients.

Group study

Exposure to specific substances (such as vaccines, drugs or the environment) in cohort studies.

Toxins), select two groups (or more groups) of people, and then follow up to observe how many people in each group occur.

Some kind of disease or other consequences. The follow-up time in cohort studies is usually years (sometimes 10 years).

Yes, because many diseases, especially cancer, take such a long time. It is worth noting that random

Controlled clinical trials usually start with patients who already have a disease, but most cohort studies do.

Starting with the subjects, these subjects may or may not get sick.

There is a special type of cohort study that can be used to determine the prognosis of diseases (what might happen to people with diseases?

Consequences). A group of early patients who have been diagnosed with a disease or have positive nodules in screening tests.

Results Patients were collected (initial cohort), and then repeated follow-up was conducted to observe the incidence of different results.

(new cases occur every year) and the course of disease.

The most famous cohort study in the world was conducted by Sir Austin Bradford Hill and Sir Richard Doyle.

And later Richard Peto. This research won titles for the first two authors. They followed up.

We interviewed 40,000 British doctors and divided them into four queues (non-smokers, light smokers and moderate smokers).

And heavy smokers), all-cause mortality (any death) and specific cause mortality (a disease

Death caused) as an observation result. The preliminary report of 10 published in 1964 shows that smokers ignore their lungs.

Cancer mortality and all-cause mortality have increased greatly, and there is a "dose-response" correlation (smoking

The more, the greater the chance of developing lung cancer. They walked a long way, 3 1, and finally proved that smoking and health damage.

The connection between injuries is etiological, not accidental. The results of this important study after 20 and 40 years (for those who

The follow-up rate of subjects collected without death in 195 1 year reached an astonishing 94), which not only showed that smoking

It also shows the importance of evidence obtained from well-executed cohort studies32,3.

3。

Cohort studies should be used to answer the following clinical questions:

. Will hypertension get better with time?

. What will happen to the growth and academic performance of premature babies in the future?

Case control study

In the case-control study, patients with specific diseases were identified and compared with the control group (patients with certain diseases)

Patients with other diseases, the total population, neighbors or relatives). And then collect some kind of contact in the past

Information about the possible pathogenic factors of the disease (for example, by looking up the case records of these people or reminding them)

Medical history). Like cohort studies, case-control studies usually study the causes of diseases (what causes

Disease), rather than the treatment of the disease. Case-control studies rank low in the level of evidence (see below).

But this design is often the only choice for the study of rare diseases. In the case-control study, mainly

The difficulty and potential deviation lies in accurately determining who is the "case", because as long as one research object is wrongly assigned,

This will seriously affect the results. In addition, this design can't show causality -A and A in case-control study.

You can't prove that A caused B just because B is related.

Case-control studies should be used to answer the following clinical questions:

. Does prone sleeping position increase the risk of sudden infant death (sudden infant death syndrome)?

. Can pertussis vaccine cause brain injury?

. Can overhead cables cause leukemia?

Cross-sectional survey

We may all be asked to take part in the survey, and some people even ask us which brand we like best.

Toothpaste. Epidemiologists conduct the same survey: a representative sample of the subjects (or

Patients) to visit, through examination or research to get answers to specific clinical questions. In the cross-sectional study

In the survey, data are collected at a single time, but past experiences can be traced back, such as research.

Investigate the previous case records to find out how many times the patient's blood pressure has been recorded in the past 5 years.

Cross-sectional survey should be used to answer the following clinical questions:

. What is the "normal" height of a three-year-old child?

. How confident are psychiatric nurses to use electroconvulsive therapy for patients with severe depression?

. Is it true that half of diabetic patients are not diagnosed?

Case report

The case report describes the medical history of a single patient in the form of a story: "Mrs. B is a 54-year-old secret.

Book, she began to have chest pain in June 1995 ... "Case reports are usually combined into a case series.

In the case series, the medical history of several patients with a certain disease is described to illustrate the disease.

On one hand, the treatment of diseases, or more generally at present, explains the adverse reactions of treatment. although

Traditionally, the evidence provided by this type of research is "fast and vague", but case reports can spread greatly.

Information that may be lost in clinical trials or research 34,35.

A memorable case report

A doctor observed that two newborns were missing limbs (seal limb deformity) in his hospital. These two mothers

My parents took a new drug (thalidomide) in the early pregnancy. The doctor wants to remind everyone as soon as possible.

Pay attention to the possibility of damage caused by this drug.

Level of evidence

When making decisions on clinical intervention measures, decisions should be made according to the relative weight of different types of original research.

The standard symbol ("level of evidence") can arrange these studies in the following order:

(1) system review and summary analysis

(2) Randomized controlled clinical trials with definite results (confidence interval and threshold of clinically significant effect)

Non-overlapping)

(3) Randomized controlled clinical trials without definite results (estimated to have clinically significant effects, but credible)

The sexual interval overlaps with the threshold of clinical significance)

(4) Queuing research

(5) Case-control study

(6) Cross-sectional survey

(7) Case report