How to understand medical research.
This article will first help you better understand how tests are interpreted if a screening test, such as for thyroid, anemia, lead, development or autism, comes out "abnormal" on your child. You can ask your doctor how "normal" is defined for the test. You can also ask about the test's accuracy--how "sensitive" is it at detecting real cases and how "specific" is it at ruling out the condition in a normal person? Your doctor probably won't have test accuracy data readily available, but he or she can give you a rough idea of whether "Abnormal" means "Panic," "We need more tests to know for sure," or "It's probably a false alarm."
Even more importantly, if your child develops a serious illness like cancer, cystic fibrosis, or diabetes, you may want to read and understand information from original research sources. This article goes on to explain some basic statistics and quality issues that doctors consider when they read research in medical journals.
Screening and Health Tests
An ideal screening test is safe, finds trouble in a person who has no symptoms, and finds trouble in time to treat and prevent suffering or death.
No test is 100% accurate. Potential downfalls of medical tests include:
* False positives (diagnosis of a disease that you don't have)
* False negatives (failure to detect a disease that you really have)
* Complications of the test (e.g. colon perforation in colonoscopy)
* Diagnosis too late to make a difference
Normal ranges for most blood tests are the central 95% of a bell-shaped curve of a set of people who took the test. An "abnormal" value may just lie outside this arbitrary range (top 2.5% or bottom 2.5%), but doesn't necessarily mean you have a disease.
Other "normal" ranges are set as desirable values for good health, such as blood sugar (under 100 is "normal") or cholesterol.
Doctors interpret a test based on the probability that the patient has a condition, plus the reliability of the test. For example, a positive Lyme antibody test is likely to be a true positive in a Boy Scout who had a tick bite at camp in New England and now has a rash. It may be falsely positive in a long-term nursing home resident in a state with low incidence of Lyme disease.
Also keep in mind that tests to find rare conditions in young children, such as newborn hearing and blood tests, are often very sensitive, but not very specific. (They have a high false positive rate).
The detection "bar" (threshold) is purposely set very low to minimize a "missed" diagnosis. Case in point: an abnormal hearing screen or thyroid test in a newborn is more likely to be a false positive than a true positive. Abnormal followup tests, or an abnormal test in a child with known risk for a disorder, increases concern for a true positive.
Developmental screening tests are usually standardized to identify children who are reaching milestones more slowly than normal and may need closer followup or expert evaluation. An abnormal screen doesn't necessarily mean anything is wrong, but a referral to early intervention, a therapist or specialist may be indicated if your child has abnormal results, especially if they are at higher risk of a problem due to perinatal complications or family history of a suspected disorder.
Evaluating Research Reports
All people are NOT equal biologically. The most meaningful research studies have participants similar to the patient in age, sex, race, and disease severity.
"Statistically significant results" in a medical study just means the findings likely didn't happen by chance. The practical significance--whether the results are worth acting on--is quite different. A statistically significant herbal treatment for influenza might reduce symptom duration by only half a day.
Remember the difference between "absolute risk" vs. "relative risk." Change in absolute risk of a problem is the most helpful criteria for making decisions. You need to know your baseline risk to judge the importance of a relative change in risk.
For example, a drug might lower relative risk of cancer by 50%. It may not be worth lowering two-in-a-million risk to one in a million, but the drug might save your life if it reduced your risk from 20% to 10%. Another example: A vaccine increases risk of fever from 1% to 2%. You still have a 98% chance of not getting a fever. The absolute risk of fever rises 1%, but the relative risk goes up 100%!
Quality ranking of doctors or hospitals needs to be taken with a grain of salt. Great doctors at university hospitals may see the sickest patients. Rates of complication for their patients may be worse than the chance of problems in patients at your community hospital. Health plan ratings of a doctor's care that are based on billing records or patient feedback don't include all the information from the medical record. Moreover, some authorities are concerned that "quality" in a health plan's eyes may mean "This doctor doesn't spend much money on patient care, so she's good for business."
Doctors and patients need to make decisions together by balancing the "pros" against the "cons" of a given treatment. No treatment is 100% effective with no side effects ... but if you believe the benefit is worth the risk, then the side effects may seem more tolerable.
An example illustrates statistics that are often discussed in medical studies.
Statistical significance The chance that the study's results are just coincidence. If a study has P<0.05, for example, this means there's less than 5% chance that the study's results are coincidence, and more than 95% chance that results are truly related to the experimental intervention.
Absolute risk reduction or rise (ARR) The actual rate of improvement or worsening caused by an experimental intervention. Formula: Percent of patients who respond to control minus percent of patients who respond to experiment.
Number needed to treat (NNT) or harm (NNH) The number of patients that must be treated for one patient to benefit or suffer. Formula: 1 divided by ARR (written as a decimal).
Relative risk reduction (RRR) The percent change between the treated group and the control group. Formula: Absolute risk reduction (ARR) divided by control response rate, as a percent.
A hypothetical study Does an apple a day keep the doctor away?
Dr. John Appleseed randomly divides 2000, disease-free school children from Your Town, USA into two groups. One thousand eat a Red Delicious apple daily while the 1000, control participants avoid apples. Over three months, two apple eaters and ten apple-free controls see a doctor. However, ten apple eaters and five controls develop mild diarrhea, which they manage at home.
P=0.01 for apple eaters making fewer trips to the doctor. There is only a 1% chance that this is coincidence.
P<0.10 that eating apples is associated with diarrhea. There is a 10% chance that this is coincidence. This is not "statistically significant" (not less than 5% chance of coincidence), but there is some reason to suspect apples as a culprit for diarrhea, and a larger study might show this better.
Absolute risk reduction (ARR) for doctor visits in apple eaters: 10/1000 (controls who saw a doctor/all controls) minus 2/1000 (apple eaters who saw a doctor/total number apple eaters) = 8/1000, which is the same as 0.008 or 0.8%. There was less than 1% drop in risk of seeing a doctor.
NNT (Number needed to treat) is 1/ARR=I/0.008 = 125, meaning 125 people have to eat apples for one person to avoid a doctor visit.
ARR (absolute risk rise) for diarrhea in apple eaters: 10/1000 (apple-eaters who had diarrhea) minus 5/1000 (controls who had diarrhea) = 5/1000 = 0.005 = 0.5%. Thus, compared to controls, five more apple eaters per thousand children developed diarrhea.
Number needed to harm (NNH) is 1/0.005 = 200:200 people have to eat apples for one to get apple-related diarrhea.
Conclusion: A small benefit out-weighs the smaller risk.
Note that the relative risk of seeing a doctor is reduced by 90%. RRR = ARR/control event rate = (0.009/0.01) = 0.9 = 90%. Relative risk reduction is often quoted by reporters since it makes small results seem more dramatic.
Also note that there are limits to the conclusions you can draw from this hypothetical study. It wouldn't apply to a child with severe asthma who takes three inhalers and sees her doctor every month, and may not apply to children who live in other towns or who eat Jonathan or Gala apples.
The study also could be biased if the apple eaters tended to get more fresh air and exercise and were less likely to be overweight than the control group. And in the real world, you'd end up with some kids skipping their daily apple in the apple-eater group, and some in the apple-avoider group sneaking their favorite fruit.
Questions to ask about health research:
1. Why was the study done? What did researchers already know?
2. Who were the subjects, researchers, and sponsors? (Are subjects similar to you? Do researchers have credible credentials? Who funded the study? Could conflicts of interest cause biased reporting?)
3. How was the study done? (See Levels of Evidence to judge study's quality.)
4. How many people were in the study? There's strength in numbers.
5. Where were they studied? Primary care offices, or university clinics? Patients at teaching hospitals often have more severe cases than local hospitals.
6. What was studied? POEMs (Patient-Oriented Evidence that Matters) focus on health outcomes: rates of chicken pox disease in children who received the chicken pox vaccine, or neurologic condition of children after treatment for lead poisoning. Other research looks at test results: antibody levels against chicken pox, or percent reduction in lead levels. Test results can be important, but they are weaker evidence than changes in health status.
7. When was the study done? And for how long? Was the study long enough to find a difference? Cholesterol levels can change in three months, but you may need three years to detect a change in heart attack rate.
Levels of Evidence (Best to Worst)
1. "Meta-analysis" or "systematic review" of many similar research studies: These reviews combine information from smaller studies to create a pool of hundreds or thousands of patients' data. Conclusions from a multi-study review are usually much stronger than conclusions from one small study.
2. A single randomized controlled trial: (RCT) Researchers randomly divide patients into two similar groups. The experimental group receives a new treatment. The control (comparison) group receives traditional care, or an inactive treatment that mimics the experiment. Both groups are followed over time. The larger the difference, the stronger the evidence.
3. A cohort study: A large group of people are followed for years to see how often a condition develops, and to discover the factors that affect the condition's behavior. Examples of cohorts: all children in Ohio with a birth weight under 1,500 grams; all children hospitalized with asthma in three cities in 2001.
4. Case-control studies: These retrospective (look-back) studies compare cases (people with disease) to controls (people without disease), seeking differences between the groups.
5. Cross-sectional studies (surveys): look at a population at one point in time.
6. Small case studies: describe several patients with a disease.
7. Expert opinion: This is only as good as the evidence it is based on. Often, expert advice is from a "BOGSAT"--"a Bunch Of Guys/Gals Sitting Around Talking."
8. Single cases or testimonials: Don't trust the claim "It worked for me, it'll work for you, too!"
On the Web--Good Sites for Good Health
I often refer parents to Medlineplus.gov, familydoctor.org, mayoclinic.com, intelihealth.com, kidshealth.org, kidsgrowth.org, specialty societies and non-profit organization websites. For specific needs, check the sites in the grey box, or ask your doctor or a librarian to help you find understandable, reliable information about your condition.
Finding quality information
Medical specialty societies
National Organization of Rare Disorders
Labs, radiology tests, drug information
Lab tests www.labtestsonline.org
Radiology (x-ray) tests
Drug info & interactions
National Library of Medicine Public Portal
Complementary and alternative
National Center for Complementary and Alternative Medicine
Experimental clinical trials
Pregnancy and reproductive
American College of OB/GYN
American Academy of Pediatrics
American Psychiatric Association
American Psychological Association
Dr. Pector is a family physician, freelance medical writer and founder of Spectrum Family Medicine in Naperville, IL. She is married and the mother of two living sons. Following the premature birth of her second son, a surviving twin, she became involved in advocacy and support for parents coping with prematurity, bereavement, or multiple births. She is on the Board of Directors for National Perinatal Association and serves on the medical advisory board of Mothers of SuperTwins.
|Printer friendly Cite/link Email Feedback|
|Author:||Pector, Elizabeth A.|
|Publication:||Pediatrics for Parents|
|Date:||May 1, 2009|
|Previous Article:||Vascular birthmarks: what every parent should know.|
|Next Article:||Liquid medication dental dangers.|