jump to Step 4: Breaking it down

In this chapter, I will first show you how to go on Med-Line and access the contents of medical journals that are of interest to you. You will need a computer that is hooked up to the Internet and moderate skills. I will show you how to break down the jargon of a clinical trial abstract into something that is understandable. If it seems like an overwhelming task, take a deep breath and go slowly. The researchers doing the work did not learn all they know in a day, they took years. Like any other skill, reading medical research is a skill that improves over time. Genuine interest - whatever the motivation for that interest - is all you need to keep you going.

Step 1: PubMed

Thanks to computers and the Internet, this job of accessing primary sources is now far more efficient. Instead of spending hours at a medical library sifting through numerous books and periodicals by hand, the National Library of Medicine has catalogued and indexed the contents of thousands of medical journals. This database is called Med-Line and it is accessible to the public through PubMed.

Med-Line is a wonderful resource for the home health researcher. It is the National Library of Medicine's premier database with listings from medical journals that cover the fields of medicine, nursing, dentistry, veterinary medicine, and the health care system. Med-Line contains citations and author abstracts from more than 4,000 biomedical journals published in the United States and 70 other countries. There are over 11 million citations in the database dating back to the mid-1960's. The coverage is worldwide, but most records are from English-language sources or have English abstracts.

The URL - or web site address - for PubMed is http://www.ncbi.nlm.nih.gov:80/entrez/query.fcgi

There will be a text box where you can enter a search, but for the sake of being efficient it is a good idea to put limits on whatever search we are going to do. Getting 12,000 results to a query (yes, this has happened to me) is just a tad overwhelming. PubMed has numerous search filters and using them saves time and frustration. Click on "limits" which is just under the text box where you enter your search subject.

First, type in your subject search in the text box, for example "homeopathy". You may also use the sample search I have included below. Searches may be done for work done by a particular researcher, in which case, put in their last name first, then first initial. Do not put in any punctuation marks. Searches may also be done by journal if you know a particular journal title, for example, the Journal of the American Medical Association.

Tag terms - this is where you decide where you are looking for the search word. Choosing 'title word', will get you fewer entries than 'text word'. You may be looking for work by a particular researcher and this would be where you choose 'author name'

Here is a sample search. I have included a number of these searches in the remainder of the book. Where I have not designated particular limit, please use the default position.

In text box: acupuncture

• Dates - published in the last 2 years
  
• Abstract - do not click in the box
• Language - English
  
• Publication Types - clinical trails
  
• Tag terms - title
  

This will give you at least 40 studies to view. There will likely be more by the time you, the reader do this search. The sample searches I've included will give you 20 to 80 studies to examine. Your entries will look like this

Gao H, Yan L, Liu B, Wang Y, Wei X, Sun L, Cui H. Related Articles
Clinical study on treatment of senile vascular dementia by acupuncture.
J Tradit Chin Med. 2001 Jun;21(2):103-9.
PMID: 11498896 [PubMed - indexed for MEDLINE]

The most common settings in limits that I use are "title word" when searching for a particular subject, "clinical trials", "English" since I am not bi-lingual (yet), "human" and viewing only studies with abstracts. Naturally it is important to have spellings correct! To refine your search you may add terms using "AND", "OR", and "NOT". These must be capitalized. You might say "protein AND hyperinsulinemia" which will cause the search function to look for both these terms in the same title. However, do not include the quotation marks! If you say "protein OR hyperinsulinemia" your search will become even wider, because the search function will then show you all studies that have either of these terms. To narrow the search, you may exclude words. You might say "protein AND diet NOT renal ". This will give you titles that have both "protein" and "diet" in their titles but exclude any studies that mention "renal".

Step 2: Some Vocabulary

Before we start looking at real studies, let's take a look at how studies are put together. There are many different types and structures for studies. They all have their strengths and weaknesses and give us different sorts of information. Some methods may be combined in a single study.

Epidemiological studies examine statistical patterns of disease in populations. These cannot draw conclusions about the causes of these patterns of disease.Case control studies compare people with a particular illness to similar healthy people matched by age, gender, and other factors. This helps define the risk of contracting that illness.Population studies look at people in particular geographical areas and catalogue types of disease and death rates for that group.Prospective studies look at groups of people and wait to see who gets sick or dies and from what sort of illness.Multi-Risk Factor studies look at a spectrum of behaviors that correlate with a given illness. It may also examine the effectiveness of multiple interventions used together to cure an illness.Retrospective studies often use the case control method to look at people who are already ill and compare them to a similar group by looking back in time at their lifestyles.Double blind studies are those in which neither the participants, nor the experimenter know whether the participant is taking the medication or a placebo until the conclusion of the study.Trials are any study in which there is some form of intervention. Clinical trials only use humans, as opposed to pre-clinical trails, which use animals, bacteria, cells etc.Crossover studies are those where participants spend half their time in the group receiving intervention and half their time in the group receiving the placebo.

Randomized studies are those in which the participants are randomly assigned to either the control group or the intervention group.Placebo-Controlled studies are those in which an inactive pill or technique is provided to the control group.Reviews or meta-analysis are when researchers analyze a group of studies on a particular subject for results and efficacy. A study consists of five parts: abstract, objective, design, results, and conclusion. The abstract is a brief overall description of the study including the objective, design, results, and conclusion. Abstracts are far more available then full studies. The objective is the purpose of the study, what the researchers are hoping to prove. The design is how the study is structured including the type of trial, number of participants or subjects, the criteria for subject selection, the amount of time the study ran, and the techniques used. The results are the mathematical analysis of the numbers to determine if the results are statistically significant. The conclusion is a brief description of what the researchers learned from the trial.

Whether or not the results of a study are considered to worth noticing is called statistical significance and it depends on how many people (or animals) in the study had a positive result from the chosen intervention or treatment. But just having a majority of positive results would not be enough. If 3 out of 5 people had a positive result, it would not be statistically significant. Statistical significance is a matter of degree. It indicates the odds that the result was due to chance. Significance is measured by the probability or p value. When in studies, p value would look like this: <.1 or <.05. This translates as "Less than 10 percent probability that the result was from chance" or "Less than 5 percent probability that the result was from chance". In general a p value of <.05 is considered to be significant. The greater the number of participants in any given study, the greater the likelihood that the results will not be from chance.

Step 3: Things to Watch Out For

All studies have flaws, which need to be acknowledged. I have chosen some common ones and supplied real studies to illustrate them.

Multi-risk factor trials
Multi risk factor trials examine a spectrum of behavior, or a group of interventions. I'll use diet to illustrate. If a group of people are cutting a particular item out of their diet and they also do other healthful things such as cut out sugar or quit smoking, then it is not an accurate test of the danger of that particular item if their health subsequently improves. If one lost weight or felt great on the T'n'T diet, it would be hard to tell if it was the Tea or the Tomatoes that was having the most effect. Ideally, these factors would be separated and the participants would consume either Tea OR Tomatoes. To make our study perfect, we would also have a control group - a group that would be our base line for comparison - who just ate a Standard American Diet, and a group who consumed BOTH Tea and Tomatoes. We might even find that the effects were not cumulative, which would mean that comsuming either Tea or Tomatoes had a good effect, but that eating both did not double the benefits.

In 1990, Dr Dean Ornish did a small multi risk factor test on a group of 28 patients who had high cholesterol and blockages of their arteries. The control group was given traditional medical counseling and the intervention group was given a program of low-fat vegetarian diet, stopping smoking, stress management training, and moderate exercise. They were given no cholesterol lowering medication. Eighty percent of the participants had small reductions in the size of their coronary lesions after a year. An excellent outcome, but which of these interventions did the most good? And would the results have been the same if only one intervention was used? We cannot know for sure without isolating each factor in its own test.

When these tests are cited in secondary sources as a reason to do this or that for your health, the authors are going to emphasize one point over another. In a book about vegetarianism I found this study cited as a good reason to be vegetarian, and Dr Ornish himself argues the value of this approach. In a book on mind-body interaction, I found this study cited as a good reason to do meditation for stress reduction. I even found a web site on the benefits of exercise that cited the study as well. None of this is bad, but it is important to look closely at any study being cited and evaluate it for yourself.

A later study done by Dr Gordon of the Institute for Exercise and Environmental Medicine, at the University of Texas Southwestern Medical Center, tested to see if the blood pressure lowering results of weight loss and exercise were cumulative. That is, they were looking to see if blood pressure dropped more if both interventions were used than if one were used alone. They found that the blood pressure lowering effects of exercise training and diet-induced weight loss were not additive. So one could have benefit from doing one or the other, but there was no increased benefit to doing both.

Correlation does not imply causation.
Just because one thing is associated with another, does not mean the relationship is causative. This is an issue with population and epidemiological studies. Many people with high blood pressure are over weight. This does not mean over weight people are automatically going to have high blood pressure, or that normal weight people are safe from having high blood pressure.

In a study done by Emily B. Shacter, Ph.D., for the Food and Drug Administration, she found that women who enter the hospital for heart attack die at almost twice the rate of men. What was causing this? A person with a strong feminist view point, might cry sexism. Dr. Shacter suspected the incorrect administration of drugs since there are gender differences in dosage between men and women. But when she took a closer look, she found that a disproportionate number of men die before ever reaching the hospital, which would bias the results.

Duration and number of participants
If a study has only a few participants or is of a short duration, then many studies must be done on the same theory before one can take the results to heart. In 1988, Dr Burke of the Department of Sports Nutrition, at the Australian Institute of Sport evaluated the effect of carbohydrate loading on cycling performance. The researchers found that carbohydrate loading did not improve performance of a 100-km cycling time trial. This is a rather important finding given the fact that carb-loading has been a major method in the race to improve sports performance. It is also a good example of a clinical trial contradicting popular knowledge. The only problem with the study was that it had only seven participants. Many more studies would need to be done (and have been) before this popular concept could be thrown out.

Step 4: Breaking It Down

Now we will look at a fictional abstract from a fictional study in a fictional journal, and break down the abstract into a more readable format.

J Tradit Ayur Med 2010 Jan;21(2): 56-63. Related Articles Books

A comparison of Yoga, weight training and running for prevention of osteoporosis in the cervical and thoracic spine.

Rifkin S, Posa J

Bridgeport Medical College, Bridgeport School of Holistic Medicine, Bridgeport, CT.

While lifting weights is known to increase bone density in the hip and lumbar spine, it is not known how it affects the cervical spine. Because the headstand practice in yoga provides a piezoelectric effect in the cervical vertibrae, we wished to compare the effects of both these exercise routines on women with osteopenia. Fifty-eight fully compliant (three sessions per week for 38 months) women (55.1 +/- 3.3 yr) diagnosed with osteopenia and without any medication or illness affecting bone metabolism took part in the exercise training; 25 women practiced Niravadhi style yoga (NY) including head stand. 28 women practiced a program of weight lifting (WL) All groups were individually supplemented with calcium and vitamin D. Bone mineral density (BMD) at various sites (cervical spine, hip, calcaneus) was measured by dual x-ray absorptiometry (DXA) and quantitative ultrasound (QUS). After 38 months, there were no significant differences between NY and WL for the BMD at the femoral neck (4.0% vs 3.0%.) There were however significant differences at the cervical spine (4.1% vs 1.6%, P < 0.001.) (15-43%, ). Combined BMD showed a 13.9 +/- 15.6% net increase after 3 yr. It is clear that while both running and Niravadhi yoga increase bone mineral density over a period of time, Niravadhi yoga is clearly superior for laying down bone in the cervical spine.

Reading stuff like this made my head spin the first time I looked at one. But let's break it down. Here it is again. I've put in blue the words I need to look up, the conclusion is in red, and the probability factors are in green. Researchers like to abbreviate long words whenever possible and then use only the abbreviations later. I noted these abbreviations in purple. One other thing we find in studies arelots of numbers with pluses and minuses. These are called "MeansÕ, and they are noted orange.

While lifting weights is known to increase bone density in the hip and lumbar spine, it is not known how it affects the cervical spine. Because the headstand practice in yoga provides a piezoelectric effect in the cervical vertibrae, we wished to compare the effects of both these exercise routines on women with osteopenia. Fifty-eight fully compliant (three sessions per week for 38 months) women (55.1 +/- 3.3 yr) diagnosed with osteopenia and without any medication or illness affecting bone metabolism took part in the exercise training; 25 women practiced Niravadhi style yoga (NY) including head stand. 28 women practiced a program of weight lifting (WL) All groups were individually supplemented with calcium and vitamin D. Bone mineral density (BMD) at various sites (cervical spine, hip, calcaneus) was measured by dual x-ray absorptiometry (DXA) and quantitative ultrasound (QUS). After 38 months, there were no significant differences between NY and WL for the BMD at the femoral neck (4.0% vs 3.0%.) There were however significant differences at the cervical spine (4.1% vs 1.6%, P < 0.001.) Combined BMD showed a 15.6 +/- .7 % net increase after 3 yr. It is clear that while both running and Niravadhi yoga increase bone mineral density over a period of time, Niravadhi yoga is clearly superior for laying down bone in the cervical spine.

When breaking down a study look first at the conclusion. This will be at the end of the abstract and will not always be a complete sentence. Second give the abstract a quick read. If there are more than a couple of vocabulary words that are confusing, take a minute to look them up

The terminology in scientific studies is challenging to understand. A good medical dictionary is a must, but PubMed comes with its own dictionary called the MeSH (Medical Subject Headings) browser. Use the MeSH browser to look up descriptions of investigative techniques and unfamiliar terms. The link is on the left side of the page in the blue part. Here you can look up anatomy, physiology, body chemistry, and the names of techniques being used by researchers. Open a new window to PubMed so you can look up words while keeping the study you are viewing available.

Sometimes when I get definitions, I have to look up words in the definitions as well. When I looked up dual x-ray absorptiometry it only gives a definition for x-rays. I then looked up absorptiometry and found it was a noninvasive method for quantitating bone mineral content. So this must be a method for using x-rays to quantify bone mineral content.

Once you have a moderate grasp of the vocabulary, look for the probability factors. Generally a study is not going to be published if the statistical test failed, so not all abstracts have the probability factors listed. Fourth, look for abbreviations near the beginning of the abstract, and find where these have been used elsewhere in the abstract.

Last, look for the means. In our fictional study our means are 15.6 +/- .7. When it says "The mean change was 5.6 +/- .7 %" it is saying that generally the people in the study showed a 15.6 percent increase in bone mass . But some showed a 16.3 percent gain (15.6 + 0.7) and some showed a 14.9 percent gain (15.6 - 0.7.)

Lets summarize our breakdown steps:

1. Find the conclusion
2. Look up your vocabulary
3. Find the probability factors if listed
4. Locate all abbreviations
5. Evaluate your means

Once you have done all these things, you are well on your way to reading studies with confidence.

Step 5: Searching Further

Not all studies on PubMed have abstracts available. So if you see a title that looks intriguing, you will need to write it down and go to the closest university or medical library. Be sure to write the entire citation. You will probably also have to do this to view an entire study. I found a few studies that had complete text available, but these are not in the majority.

Before you trek to your nearest medical library (and for some of you that is quite a trek) give them a call and see if their catalogue is on-line. If the journal you need is not in their collection, you can save yourself a trip. The easiest method for accessing complete articles is PubMed's Loansome Docª service. This is a collaboration between PubMed and various medical libraries around the country. The researcher sets up a relationship with a library, under which the researcher can order documents that appear on PubMed, and the library will locate, copy and send them for a fee.

Each library sets its own policies, so some phones calls will be required to gather information. You will need to ask about fees and what delivery methods are available. Ask if your request will automatically be referred to another library if they do not have the requested material in their collection. Find out if there is a limit on the number of articles that can be ordered in a specified time period and on the average, how long it takes for articles to arrive.

Now you have all the basic tools you need to start using Med-Line. If this is new to you, then take small bites, and remember to breathe. I learned to use Med-Line while I was writing this book, and the breakdown method is the one I created for myself. In a few months, this database has become an easy to use resource and the first place I look for information.

In this chapter we have covered types of research, the parts of a study and some pitfalls to avoid when examining a study. Most importantly we have gone to PubMed and learned how to find studies and how to read them. After each description of a particular therapeutic approach, I have included a sample search to be applied on PubMed. Use these searches to further your understanding of how the Med-Line database works. In just a little time you will be feeling competent and confident.