Making the call on assay interference
When he discusses immunoassay interference and how it can be identified in the laboratory, Julian H. Barth, MD, FRCP, MRCPath, is careful to refer only to the incidence of interference, not its prevalence. "We haven’t a clue as to what the prevalence is," he says flatly.
But Dr. Barth, a consultant in chemical pathology and metabolic medicine at Leeds General Infirmary in Great Britain, reminded an audience at the July 2003 meeting of the American Association for Clinical Chemistry that the price of interference can be steep.
"Seven or eight years ago, a company said they didn’t have a problem with interference with their hCGs," he said in a presentation on erroneous results. "This problem resulted in 12 women experiencing a huge range of completely unnecessary investigations that were performed on these patients, including D & Cs, laparoscopy, total hysterectomy, thoracotomy, and chemotherapy. And we know the costs both for patients and the laboratory and the company that made the assay."
Of immunoassays susceptible to interference, hCG is only one example. A study at his pathology service, one of the largest in Europe, found one in 200 cases of interference in thyroid testing. "My laboratory does somewhere around 75,000 thyroid tests a year, roughly 300 per working day. That’s one every day I’m missing because I don’t recognize them, and my guess is most of you don’t recognize them either. But it’s a lot of samples coming through our laboratories."
Interference has been well recognized for many years with competitive immunoassays, Dr. Barth said. "The way it was identified in the early days was by serial dilution." With the development of immunometric (noncompetitive) assays a decade ago, case reports of interference have appeared sporadically, he noted.
The difference between the immunometric assays, particularly the competitive, automated ones, and the noncompetitive assays is that the latter are not done at equilibrium. He pointed out that two-site assays are believed to be more prone to interference than single-site ones.
"We know from the literature through a regular stream of case reports. Every time a new assay is released, there are case reports describing interference. And then they become old hat, and nobody reports them any longer."
"So for example," he says, "when the assay for luteinizing hormone came out, we all knew there was a problem with LH and cross-reactivity with hCG. But there haven’t been any case reports about it for years now, so we assume there isn’t a problem."
Whether it is cross-reactivity or antibody interference, there are two ways for laboratories to identify interference, he said-what he termed the laboratory strategy and the clinical identification strategy.
With the laboratory strategy, serial samples that come through the laboratory are treated-usually by dilution, as in the case of digoxin in a study he described, or by use of anti-animal antibodies. "And a rough prevalence can be found there-for example, one in 20 or one in 30. If it’s one in 20, that’s five percent of the results of your workload that has an interference."
The clinical identification strategy asks whether the samples are critically abnormal-in other words, are they likely to cause a clinical problem? "I’m not talking about samples elevated by twofold, but those that remain within the reference range," Dr. Barth noted. The question is whether a diagnosis will be made or not made on the basis of an abnormal result.
Among the causes of interference are venous stasis, the presence of hormones and drugs such as anticoagulants, and to an unknown extent, other preanalytical factors, he said. For example, "Nobody knows what the problems with gel separators are and whether they interfere. When you ask the companies, they’ll compare their gel separator tubes with somebody else’s gel separator tubes. They don’t compare them with glass tubes."
"The nonspecificity of our assays can also cause interference," he added, citing a case that came through his laboratory about five years ago. The patient was a woman who had had samples sent to another hospital for perimenopause. "She had LH, FSH, and estradiol measured. This was repeated several times and considered to be grossly abnormal."
"On the basis of this grossly elevated estradiol, this lady had a hysterectomy and a bilateral oophorectomy, and after the operation her gonadotropins had elevated, and her estradiol remained elevated," Dr. Barth said.
"After the laboratory measured her estradiol by organic extraction, it realized that this lady had her complete pelvis removed on the basis of an assay interferent or an estradiol nonspecificity problem."
Under a national external quality assurance scheme, he recounted, a second trimester pregnancy sample was distributed around the British countryside. "Most people recognize that there is no LH in that sample, but there are five methods from three manufacturers that show elevated amounts of LH. So if you had a clinical detail of amenorrhea and you measured LH and FSH, you might well have a diagnosis of polycystic ovaries on that-when in fact you were just missing a pregnancy sample."
Packet inserts for hCG assays confirm the problem, he said. "You’ll find that all the companies say this is for use only as a pregnancy test, not as a tumor marker, and that’s why the laboratories were successfully sued in the U.S. a couple of years ago."
Digoxin is one of the most important drugs laboratories measure therapeutically, Dr. Barth continued, because the side effects of overtreatment are similar to the effects of undertreatment.
"And so we’re very concerned that we keep the right plasma concentration. But about 10 or 15 years ago there was a huge amount of interest in endogenous digoxin-like interference, particularly in neonates in pregnancy, liver disease, and kidney disease."
More recently, he said, ginseng, spironolactone, and canrenoates have been studied as causes of interference. "Now digoxin is done by an immunoassay, so this is a competitive assay and the use of dilution studies is the exactly appropriate way to investigate interference." And he described a study that identified large numbers of patients who had interference, and the majority of them were in a hospital in intensive care units being treated with spironolactone and/or canrenoates.
"The important thing is that two assays were used, and several patients were toxic with one assay and not toxic with the others. So when you’ve got two results from two different methods, how do you decide which one is the correct one?" Some spiking experiments showed, in fact, that one method for testing for digoxin was negatively inhibited by spironolactone and canrenoates.
Insulin antibodies were the culprit in a case of interference that came through Dr. Barth’s laboratory recently. "An 88-year-old female came just after Christmas to our hospital with a chest infection. Over the preceding two years she had had previous infections, rheumatic heart disease, osteoporosis, a cluster of different drugs, and when she arrived she was dehydrated with signs of pneumonia, and she was treated."
"Investigations at that time showed nothing terribly exciting; she was slightly anemic, high normal white count, urea elevated a little consistent with her dehydration, and glucose of 5.7."
The following day, after nurses found her unconscious, "someone very wisely checked her glucose on a fingerprick test and found it to be hypoglycemic. They took a blood sample prior to starting treatment with glucose and dexamethasone."
"She had the lowest plasma glucose I’ve ever seen in my life," Dr. Barth recalled. "It was .7, and she had the most strongly positive insulin antibodies that had ever been done by the reference lab we’ve used. And just to show the hypoglycemia was not due to failure to counter-regulate her system, we measured her cortisol on the final few drops of her plasma. We showed that she had plenty of cortisol.
"So this was significant antibody formation against an analyte that we’re looking at. In fact it caused the death of our patient due to hypoglycemia. There are a lot of reasons why it might happen. The one I find most attractive is that the insulin antibodies soak up insulin as it’s secreted by pancreas."
Since insulin is not biologically active, the pancreas continues to secrete it until the insulin antibodies are soaked up and completely saturated, he explained.
"You then start to lose insulin by the law of mass action. Switch off the pancreatic secretion and then the insulin is used up and taken up by cell receptors, and continues to be released because there’s no regulation of this insulin secretion for release from antibodies, and the patients become hypoglycemic."
"If patients are alert and able, they can eat to combat the effects. But our lady presumably couldn’t, and that’s why she died of hypoglycemia."
"Rheumatoid factor is something that we think of almost in the same breath as interference antibodies in immunoassay," Dr. Barth continued. "So it seems strange that it was only five years ago that the mechanism was actually sorted out. But if you take a sample with positive rheumatoid factor and run it on different assays, you get different answers, not related to the concentration of rheumatoid factor."
For example, a recent study found the only sample that wasn’t changed by the addition of polyethylene glycol was the one with a very low concentration of rheumatoid factor. "In his study, Prof. Vincent Marks, MD, took 10 donors and distributed these samples around the world and asked laboratories to measure whatever assays they could on it.
"He also gave them a large supply of antibody cleanup tubes so they could see what the effect was with and without those tubes. But despite these samples having a high titer of rheumatoid factor, only half of them were resolved by antibody tubes."
"Six percent of them were clinically significant in terms of making a change in diagnosis and 2.5 percent of these were changes within a reference range, so they probably wouldn’t have made any difference in terms of clinical practice."
Dr. Barth emphasized that interfering antibodies may be human anti-animal antibodies, but he said, "I don’t think necessarily that anti-mouse or anti-animal antibodies are the key to interference."
He described a broader interfering phenomenon called the idiotype cascade-a series of checks and balances within the immune system to make sure nothing gets out of control.
"I think what’s happening is the immune system has got a whole system of protections to prevent certain reactions, in the same way if you think of moving your arm up and down, you have a coordinated action by opposing muscles to provide a smooth movement. So the immune system has a series of checks and balances to maintain a smooth control of antibody action."
"My feeling is that it’s problems with this whole system that interfere with our immunoassays rather than anti-mouse antibodies," Dr. Barth contended. "The manufacturers now are filling their reagents with anti-mouse antibodies anyway. They’re filling them with something to block rheumatoid factor, but we’re still seeing immunoassay interference, and I think it’s because of endogenous antibodies within this antibody cascade. And that’s where the problem’s arising-because we can’t block our own human antibodies."
He cited another case in which a man was admitted with prostatic obstruction and developed an E. coli septicemia with pain in the jaw while he was in the hospital. Thinking that the patient must be having cardiac pain, his physicians measured his troponin and found it elevated.
"Most of you would send this patient into a coronary care unit, and that’s exactly what happened to him," Dr. Barth reported. "Except his ECG was normal, his echo was normal, his angiogram was normal, and his troponin continued to rise despite him being very well and requesting discharge."
An interference was suspected after many more troponin measurements and the patient’s CK-MB and myoglobin were found to be normal. "So they tested the serum on every other immunoassay they had in the laboratory-and half were elevated and half were normal."
"The message you can take away from this case is, first of all, that interference affects lots of assays-but not all of them," he said. It led to unnecessary angiography in this man, a procedure that carries a small, but not trivial, mortality.
How long does antibody interference last in someone who has it? Dr. Barth cited a recent study that found just under half of patients, once they had antibodies, had them for a long time, while other studies have reported three weeks’ or eight weeks’ duration. A study of CA 125 for ovarian cancer demonstrated that antibody interference seemed to increase with the number of monoclonal antibody infusions that were given.
"But the important thing is that some patients with immunoassay interference have it for a long time and some people have it for a short time, and there doesn’t seem to be any way at the moment of distinguishing them."
Tumor markers are a significant source of concern regarding interference, he added, because "for patients with cancer your blood test results are the only evidence on which patients have a decision made about whether to have chemotherapy or wait for radiotherapy."
"For patients with carcinoma of the ovary, it’s absolutely critical because CA 125 is the only marker for disease assessment. You cannot do a CT or MRI scan to determine tumor volume, and yet we now know that the patients whose ovarian cancer is treated with CA 125 monoclonal antibodies have a high prevalence of interference, and we also know patients with ascites have high levels of CA 125-both of which are clinical aspects of cancer of ovary."
Dr. Barth outlined another case that illustrated the difficulties of managing interference. "This lady came to us with all the symptoms of early hyperthyroidism, with an fT4 of 13.7 and a TSH of 22. And she was started on a low dose of thyroxin. She had no TSH suppression and her T4 dose was gradually increased and she did feel better, and the dose gradually increased, and the free T4s went up a little bit whereas the TSH was completely inconsistent."
"Now most of us would say this lady doesn’t take the tablets very regularly. But after we double-checked the endocrinology with different methods, we found that there was not much difference between methods for either free T4 or TSH."
What was the problem? "The TSH values don’t look terribly different until you look at the relationship between two different methods. And unless you know there’s a negative bias between methods you are using for comparing one assay with another, you don’t necessarily notice how abnormally different results are.
"She didn’t have any parallelism when we tried diluting her TSH and she had interference with two assays," he noted. After realizing it had not recognized the method bias, his laboratory learned an important lesson: "You have to study the comparison between two methods if you’re going to use two methods to back each other up."
Assay interference is not analyte-specific. He described one patient at his institution who went to her physician reporting fatigue and mental slowing. "I spoke to the family doctor to tell him that his patient was hypothyroid with a TSH greater than 100mIU/L, and he thought they were rather odd. She wasn’t clinically hypothyroid, so we did the tests on another analyzer and the TSH was only 8.2.
"We thought that the method on the ACS:180 was the correct result because it fitted the clinical picture. And six months later, the results were very similar. Then we found that one result diluted down absolutely perfectly, while another result doubled as we started diluting. That confused us no end, because we thought if a result fitted the clinical picture, it was therefore right, but our assay interference test didn’t match that." These cases showed that interference can occur even when it doesn’t look as if there ought to be interference because the results fit the clinical picture, and that the interference may affect one assay but not the other. The consequences can be serious, Dr. Barth added. "There was a huge cost to the patient both in terms of the number of consultations and in the number of investigations she had."
So while identifying abnormal results is important, results are not always abnormal, he stressed. "The one thing you can do in the laboratory to sort out immunoassays is to teach your clinicians and talk to them, because unless they come to you and say ’there’s something funny with this result, it doesn’t match,’ you’re not going to look for it because they’re so sporadic. Even if it’s one in 200, you can’t look and do extra tests on that number of samples."
Other strategies for identifying interference include serial dilution, PEG precipitation, and antibody cleanup tubes. "But the way things are moving at the moment, I think alternative technology such as tandem mass spectrometry may well be the way, certainly for small molecules such as steroids."
The recent case of a 51-year-old patient illustrates another set of complications in identifying interference. "This lady had six years of amenorrhea, and was treated with three drugs all capable of interference with prolactin: pimozide, procyclidine, and progesterone-only contraceptive. So she had high LH and FSH, consistent with her being menopausal, and she had a high prolactin."
"Since the prolactin method [Abbott] AxSym cannot be double-checked by using PEG, we did it on a different analyzer, [Beckman Coulter] Access," he said. "By now, we’d learned we had to compare methods to look at the difference between them. We had a high prolactin with AxSym, not a very high result with Access."
"So we had two samples on this patient showing differences between them, and they were consistent with her having a macroprolactin. That was nothing terribly exciting-except that we then tried doing a PEG precipitation and measuring the prolactin on AutoDelfia, but there wasn’t a macroprolactin there at all. So which of those three results was correct?"
The answer was discovered after the patient stopped all her therapy and her prolactin dropped. "So she in fact had drug-induced hyperprolactinemia all along. And none of us would have guessed that with those three sets of laboratory results."
Dr. Barth and a colleague went through 5,000 patient samples at his institution and pulled out 59 as being abnormal on the basis of clinical details and previous results (if they were available). "And we treated them either by dilution, with cleanup tubes, or by an alternative method."
"Just by looking at the results and clinical details, we identified 28 of them-one in 200-as having a problem. We didn’t do all three methods on all samples, because once we demonstrated there was interference, there was no need to go ahead and do any more."
"But five were only abnormal on dilution studies, 11 were only abnormal on cleanup tubes, and nine were abnormal on both. So just because someone’s got an interference doesn’t mean to say it will come up positive with every form of investigation you try on it."
"The other message," he emphasized, "is you can’t exclude an interference. Just because somebody has normal results doesn’t mean they haven’t got interference. Interference can be detected but it cannot be excluded, because if you use a test that doesn’t pick up somebody’s antibody interference, you may just have used the wrong test."
When interference is suspected, he recommends looking for previous data if they exist, looking at clinical details, and considering whether the results are implausible.
"And my personal approach is always to get fresh sample, because often the first sample that comes through with a funny result is a funny sample."
"Above all, there’s a need to rely on clinical judgment. And it may be that the best diagnostic test is a therapeutic test."
Immunoassay interference is variable with respect to time, he said. "Sometimes it’s there, sometimes not, and it’s variable with respect to analyte. So if you take the same sample and analyze it on different methods in different laboratories, you may get different answers. Moreover, if someone has interference with one assay, it may or may not imply that other assays will be normal or abnormal."
All of these factors underscore the need for caution in interpreting laboratory results, Dr. Barth warned. The case notes in his hospital have a huge "D" written on them if the patient is a diabetic, and signs are posted on the outside of the case notes if the person is allergic to drugs. "And do you think we should also have a little sticker saying this patient has immunoassay interference as well? I don’t know the answer to that-but perhaps it’s one worth contemplating."
Anne Paxton is a writer in Seattle. For more on immunoassay interference, you can read the comments of Dr. Barth’s co-presenter, Larry Kricka, PhD, in the October 2003 issue of CAPTODAY, page 76.