Cholesterol tests: how accurate are they?
Several months ago, 58-year-old Sandra Miller (*1) had her blood cholesterol measured for the first time. The result--198 (*2)--sounded like good news. Her doctor told her this meant that she fell into the "desirable blood cholesterol" category (under 200).
If Sandra's cholesterol level had topped 200, her doctor would have urged her to go on a cholesterol-lowering diet. As it is, she'll continue to eat foods such as ice cream, bacon, and cheese as often as she wants.
What Sandra, her doctor, and millions of others don't know is that cholesterol tests are not as accurate as many believe. Assuming her doctor's lab meets the government's guidelines, Sandra's blood cholesterol might actually be anywhere from 168 to 228.
Interest in cholesterol and its role in heart disease has never been higher. Over the next five to seven years, industry analysts estimate that there will be a 20-fold increase in cholesterol testing.  But because of laboratory error, many people may get inaccurate results--so inaccurate that, like Sandra, they are given the wrong advice or treatment.
Recently, NIH set guidelines for how accurately labs should measure cholesterol. The current goals mean the results from your blood test could be as much as 15 percent higher or 15 percent lower than your actual cholesterol level. This wide margin of error raises the specter that millions of Americans now having their blood cholesterol measured could be misclassified--and mistreated--as low-risk when they should be high-risk...or vice versa.
For example, the average blood cholesterol is 211 for men and 215 for women. Some "average" men will get cholesterol test results as high as 243--which misclassifies them as "high-risk." Many others will get a test result as low as 179--which misclassifies them as "low-risk." In fact, the average man should be classified as "borderline high" risk (see graph at right).
Assessing Accuracy. Since 1963, the College of American Pathologists has conducted proficiency surveys to determine how much cholesterol test results vary from lab to lab. The labs' performance is judged in two ways. "Precision" is defined as the ability of a lab to get similar results on a single test sample each time it is run. "Accuracy" is defined as a lab's ability to match the target, or true, value of a test sample. Proficiency testing surveys measure accuracy only.
In a recent survey, CAP sent a blood sample with a cholesterol level of 265 to 5,434 labs. Even after excluding about five percent of the results which were far off the target, the remaining results ranged from a low of 229 to a high of 312--as much as 18 percent off the target value, and a total difference of more than 80. 
According to Herbert Naito, chairman of the National Institutes of Health Laboratory Standardization Panel on Blood Cholesterol Measurement, most manufacturers of the machines that analyze cholesterol have already achieved good precision. "What labs need to work on is accuracy," he says.
But the current standard for accuracy allows for a five percent testing error. And the current standard for precision allows for roughly a 10 percent error. Together, they mean that in 95 percent of cases, cholesterol test results may be as much as 15 percent higher or 15 percent lower than the actual value.  (In the remaining five percent of cases, the results may be even more off target.)
That isn't good enough, especially when other factors cause test results to vary. Among other things, cholesterol levels in a person's blood naturally fluctuate over time.
Congressional Criticism. Fortunately, Congress has recently become a severe critic of poor lab performance. Representative Ron Wyden (D-OR), who chairs the House Subcommittee on Regulation and Business Opportunities, held hearings on medical laboratories last March. Under Wyden's questioning, Guy C. Glenn of the College of American Pathologists acknowledged, "lack of accuracy in lab testing is a serious problem."
"Clearly, the lab industry's current self-policing system is not working," charges Wyden. "Right now, there are thousands of labs all over the country that are essentially unregulated."
Currently, an estimated 15,000 to 20,000 hospital and independent labs, and 100,000 to 150,000 physician office labs, perform dozens of diagnostic tests ranging from Pap smears, to white blood cell counts, to cholesterol tests.
But federal regulations cover only those hospital and independent labs that participate in the Medicare program. Under the Clinical Laboratory Improvement Act (CLIA), the government also regulates labs that do interstate business. But it has failed to set standards for all other labs--including those in doctors' offices.
Outside the System. Physician office labs, the fastest-growing segment of the lab industry, are not subject to any federal regulation at all. Over the last several years, the number of physician office labs in the U.S. has skyrocketed, in part because doctors prefer to do their own lab work. That way, they can cut out the middleman and get the full profit themselves. And even though the government doesn't regulate their labs, doctors with Medicare patients can still bill the government for full reimbursement.
But Representative Wyden has proposed legislation to pull all labs, including physician office labs, under the federal regulatory umbrella. If Wyden's amendment passes Congress, then all labs will be regulated under CLIA (see coupon).
Wyden has proposed that all labs:
* participate in a proficiency testing program,
* undergo annual inspection, and
* meet specified quality control and personnel requirements.
This broad-brush legislation is badly needed. For consumers, it would mean that all hospital, independent, and physician office labs would be responsible for meeting the standards the government says are necessary for producing reliable test results.
States' Regs. States have the right to regulate labs more stringently than the federal government. But few do, according to a report issued this year by the General Accounting Office.  With the notable exceptions of Pennsylvania and Wyoming, states regulate labs unevenly. Hospital and independent labs are held to far higher standards than are labs in doctors' offices. Here's how the current patchwork system of state regulation works:
* Physician Office Labs: 36 states don't regulate labs in doctors' offices; of the 14 that do, most regulate group practices only.
* Independent Labs: 22 states regulate independent labs; 13 do not. The remaining 15 states have limited regulations. For example, some states only regulate labs that do syphilis testing.
* Hospital Labs: Hospital labs are the most highly regulated of all. Sixteen states regulate them fully. Twenty-nine have limited regulations. But five states--Washington, Iowa, Louisiana, South Carolina, and Vermont--do not regulate hospital labs or any other type of lab. In these states, a test result carries absolutely no promise of reliability.
Even when states do regulate labs, their authority is often weak. Some states, for example, have quality control and personnel requirements only for specific kinds of tests. And states rarely shut down a lab for doing bad work.
New York State. Only a few states do a good job of regulating labs. One is New York. In this state, the health department independently tests the work of labs through proficiency surveys two to five times a year, and through annual inspections. The health department has the power to order labs that fail to meet accuracy standards to stop performing tests, and, if necessary, it can close a lab altogether. Yet even in this highly regulated state, lab performance is far from perfect.
According to health commissioner David Axelrod, New York stopped labs from performing tests on 94 occasions last year when proficiency surveys found they failed to meet standards for accuracy. Each time a survey is done, 10 to 15 percent of New York labs fail one type of test.
Labs that fail go through a "remediation" process which, depending on the problem, can include retraining of lab personnel or replacement of defective equipment. According to Axelrod, the remediation process almost always results in a lab passing the next proficiency survey.
But even New York doesn't touch individual physicians' labs. At a hearing last March before the House Subcommittee on Regulation and Business Opportunities, Axelrod said the lack of regulation of labs in doctors' offices nationwide has created a situation "fraught with potential for abuse."
Physician office labs are "increasingly the basis of clinical decisions," said Axelrod. "Mandatory proficiency testing and licensure is the only way to assure accountability of labs to the public."
But in most states, no one, including the health commissioner, knows how well or how poorly labs are doing. That's because many labs choose not to participate in a proficiency testing program. Indeed, CAP's program for physician office labs, called EXCEL, has only about 2,000 participants. That amounts to less than two percent of all doctors' labs.
The Personnel Factor. Catherine Cohen, of the American Society of Clinical Pathologists, says that the expertise of lab personnel is another critical factor afecting testing performance. Yet the "personnel factor" is often ignored. "A study published last year clearly demonstrates that lab-trained medical technologists tend to produce more accurate test results," says Cohen. 
"But in many physician office labs, doctors, nurses, and other non-medical personnel with no lab training end up doing the tests," she explains. "That's a problem."
Josephine Bartola, director of lab improvement for the Pennsylvania department of health, concurs with Cohen's view that medical technologists are best qualified to perform lab testing. "There's no question that those with lab training do better," she says. "Without a lab background, operators don't fully understand the need to calibrate equipment, the concept of quality control, or what to do when something goes wrong."
Public Screenings. According to Bartola, lack of lab experience can be an acute problem for a fast-growing, popular form of cholesterol testing: public screenings.
Sponsored by organizations ranging from the American Heart Association to Lever Brothers, makers of Promise margarine, cholesterol screenings are now attracting millions of Americans. "Since screenings aren't done in a lab environment, we feel that the training of those who are doing the actual testing is very important," says Bartola.
All organizations doing cholesterol screenings in Pennsylvania must work with a lab in the state. The lab is required to tell the department of health in advance where the screening is being held. According to Bartola, "this is so we can do spot-check inspections to assess the accuracy of the [screening's] results."
Furthermore, she says, if medical technologists are not performing the tests, then "we require detailed complete information on the training of these people. We also want to know what quality controls are in place if non-medical technologists are doing the screenings."
No Federal Guidelines. So far, the government has no regulations like Pennsylvania's. Beth Schucker is overseeing a pilot study on cholesterol screening funded by NIH. The results will help formulate future federal guidelines.
"One question the study asks," Schucker says, "is how well do the desktop analyzers used for screenings work on-site? They're very good machines, and we know they can work well in laboratories. But what happens if I drive the analyzer in a station wagon to a site, plug it in, and assign someone with no knowledge of clinical chemistry to operate it? How accurate are the results then?"
But the government's guidelines will be voluntary, not mandatory like Pennsylvania's. More important, they won't be issued until 1989. By then, how many millions of people will have been screened--perhaps inaccurately?
Standardizing Laboratories. While Congress is just waking up to the fact that no one is policing medical laboratories, NIH is making some effort to set voluntary guidelines. Herbert Naito's Laboratory Standardization Panel is writing guidelines for labs to help them improve their performance.
According to Naito, by 1990, all labs should report results that are "only" nine percent higher or nine percent lower than your actual cholesterol level. That's an improvement, but still far from reassuring. That narrows Sandra Miller's cholesterol from a 168-to-228 range to a 180-to-216 range. The new guidelines, like the old ones, will be strictly voluntary. There's nothing to say that labs have to meet these standards, and nothing will happen to them if they do not.
Still, NIH's guidelines are likely to be taken seriously by the lab industry. And the agency is insisting that all labs and manufacturers begin using certified reference materials developed at the National Bureau of Standards and other top labs in the country.
Certified reference materials are vials of frozen and freeze-dried serum (the fluid part of the blood) that have been checked for accuracy with the best testing methods available. These vials are to cholesterol testing what the atomic clocks at the National Bureau of Standards are to your watch. To be sure of the correct time, you'd have to check your watch against NBS's atomic clocks, and adjust your watch whenever necessary.
Naito and others want manufacturers and labs to use these extremely accurate serum samples to check the accuracy of their equipment. Right now, many manufacturers and labs don't use certified reference materials. Instead, they make up their own materials, probably because it's cheaper and easier to do so.
Humans Are Variable. Even without lab error, cholesterol levels vary anyway. "Biological variability is not a problem of faulty lab measurement--it's just the way someone's cholesterol level actually changes over time," explains Paul Fischer of the Medical College of Georgia.
According to Gerald Cooper of the Centers for Disease Control, the average person's cholesterol level varies within a range of about plus or minus 13 percent. Others, including Fischer, think biological variability may be only plus or minus five percent. "It's not a well-studied phenomenon," Fischer says.
Biological variability cannot be eliminated. But you can control it to some degree. According to Naito, women should not have their cholesterol measured until three to four months after giving birth, because levels are likely to be higher than normal during pregnancy. Similarly, Naito recommends waiting eight weeks after a heart attack to measure cholesterol.
"Drugs, including estrogens, high blood pressure medication, and anabolic steroids, affect blood cholesterol. So does serious illness or trauma," Naito notes.
Other sources of variability can be minimized, he says, by properly taking the blood sample. "It's important that the patient be in a sitting position for 10 to 15 minutes before having blood drawn," says Naito. "This can affect cholesterol levels by as much as 10 percent."
Naito also stresses that your finger should not be "milked" for blood, because this can dilute the sample. And samples should be analyzed as soon as possible.
HDL and LDL. NIH recommends further testing for people whose total blood cholesterol level is 240 or above (or 200 for those who have evidence of heart disease--such as a previous heart attack or angina--or two other risk factors).  In those cases, LDL ("bad") cholesterol and HDL ("good") cholesterol levels should be measured. CSPI recommends HDL and LDL tests for anyone whose total cholesterol is 200 or above (see "Measuring Cholesterol, Step by Step").
A doctor should never prescribe a drug to treat high blood cholesterol after measuring total blood cholesterol alone. Only when LDL cholesterol is 160 or above (or 130 if the person has evidence of heart disease and two other risk factors)--and only when diet has failed to lower cholesterol--should a doctor consider prescribing a drug.
"I recently spoke with a friend of mine," Paul Fischer recounts, "a 76-year-old man. His doctor measured his blood cholesterol, found it to be 270, and immediately put him on Mevacor [a cholesterol-lowering drug]. That's exactly how cholesterol measurements should not be used."
Unfortunately, according to Naito, labs are no better at measuring HDL and LDL cholesterol than they are at measuring total blood cholesterol. A test that measures HDL and LDL cholesterol averages $50 to $60, as opposed to about $8 for a total blood cholesterol test.
What You Should Do. While cholesterol tests are less than reliable, the results are still valid. When you go to get tested, here's what you need to know.
1. A Diet for Everyone. More than half of all Americans have blood cholesterol levels above 200. A large share of the blame can be placed squarely on our national diet, rich in artery-clogging foods such as meats, cheeses, fried foods, eggs, pastries, and ice cream.
The American Heart Association and other major health organizations recommend that everyone over the age of two get no more than 30 percent of their calories from fat, and no more than 10 percent of their calories from saturated fat. The inaccuracy of cholesterol tests is just one more reason why everyone--not just those with "high" cholesterol--should follow this sound advice.
2. Multiple Measurements. To get an accurate reading of your cholesterol, have it measured several times. According to CDC's Cooper, it's best to have it checked two to three times at least one month apart. Average the results.
3. What the Numbers Mean. If the lab testing your cholesterol meets NIH guidelines, you can safely assume that you are at low risk if your cholesterol is under 170. If it's 170 or above, a testing variability of plus or minus 15 percent means that your true level could be above 200. A cholesterol of 200 would put you in the "borderline high" risk category.
4. Diet Before Drugs. According to NIH, if your cholesterol is high and you switch from a typical American diet of about 40 percent calories from fat to the AHA diet, then your cholesterol level could drop 30 to 40 points within three months.  But to really reduce your risk of heart disease, you'll have to keep your cholesterol low permanently.
The treatment of first choice for high blood cholesterol is diet. If the AHA diet doesn't lower your cholesterol enough, cut your saturated fat to six percent or three percent. Your total fat will probably drop below 30 percent, but focus on minimizing saturated fat. Eat more fish, fruits, vegetables, beans, and grains.
Only if diet doesn't work should the alternative--cholesterol-lowering drugs--be tried. That's because all drugs have side effects and risks (see January/February, 1988). Moreover, because the drugs must be taken every day, they are expensive.
Every minute, another American dies of a heart attack. Cholesterol tests are not perfect, but they do provide some important clues as to whether you are at risk. If you are, do something about it--now.
(*1) Sandra Miller is a fictitious name.
(*2) All cholesterol values represent milligrams of cholesterol per deciliter of blood.
 Personal communication from Kerri Keaton, Boston Biomedical Consultants, Inc., Waltham, MA.
 Proceedings from the American Association for Clinical Chemistry conference, January 18-20, 1988.
 Clin. Chem. 34: 193, 1988.
 Report from General Accounting Office, H8-0111, February 29, 1988.
 J. Am. Med. Assoc. 258: 357, 361, 1987.
 Arch. Intern. Med. 148: 36, 1988.
Lulu Kay Arimi (Member): Accuracy of your tests. 7/22/2010 4:48 PM
I had a lipid panel test done at the local Hospital in California and it showed my Triglycerdie to be 159 and my total cholestrol to be 298, which was too high as a result of drinking a few days prior to the testing. So I took the test again after 7 days. This time I had no drinks at all, ate salad for dinner every night no meat at all for one week. The results were: My total cholestrol was 279, so a little lower than the week before due to change of diet but my Triglyceride was now 260 from 159 the week before. I just can't understand how that could have happened. Should I trust the test which was done at the hospital or the one which was done at the doctors office (this lab is located in a building with at least 50 doctors)Please let me know.