UBC Faculty Research and Publications

A review of research on the use of desk-top analysers for cholesterol screening Archibald, Chris 1991

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-Archibald_Chris_Review_Research_Cholesterol.pdf [ 5.95MB ]
JSON: 52383-1.0045254.json
JSON-LD: 52383-1.0045254-ld.json
RDF/XML (Pretty): 52383-1.0045254-rdf.xml
RDF/JSON: 52383-1.0045254-rdf.json
Turtle: 52383-1.0045254-turtle.txt
N-Triples: 52383-1.0045254-rdf-ntriples.txt
Original Record: 52383-1.0045254-source.json
Full Text

Full Text

A REVIEW OF RESEARCH ON THE USE OFDESK-TOP ANALYSERS FOR CHOLESTEROL SCREENINGBCOHTA91:2B.C. Office of Health Technology AssessmentCentre for Health Services & Policy ResearchS 184 • Koerner Pavilion2211 Wesbrook MallVancouver, B.C., Canada V6T 285Chris ArchibaldDecember 1991This document was prepared under the direction of the British Columbia Office of HealthTechnology Assessment (BCOHTA) staff. A draft version was reviewed by two membersof the Steering Committee and an expert in laboratory medicine. The reviewers'comments were then incorporated into this revised version. The BCOHTA takes soleresponsibility for the contents of this document; comments or suggestions are welcome.IntroductionThe association between an elevated total serum cholesterol and an increasedrisk of coronary heart disease is well established and there has been muchrecent interest in cholesterol screening as a public,health,measure to identifyhigh-risk individuals and refer them for appropriate treatment. Concurrentwith this interest has been the development of portable desk-top analyzerswhich can rapidly measure total cholesterol in blood taken .by means of asimple finger prick. This new technology offers the advantages of minimalpatient discomfort, more convenient testing sites due to instrumentportability, and the availability of test results within minutes which obviatesthe need for a return visit to discuss test results. A number of authors havesuggested the use of of such analyzers to implement recommendations forgeneral cholesterol screening, although the implications of this suggestionhave not been fully discussed.l-3This paper will examine the use of desk-top cholesterol measurement withrespect to the reliability and validity of the results as well as the implicationsof this new technology for cholesterol screening in general.1. Diagnostic test characteristics of cholesterol desk-top analyzersThere are two desk-top cholesterol analyzers available on the market forwhich information on their performance characteristics could be located: theBoehringer Mannheim Reflotron and the Abbott Vision analyzer.I Most ofthe published information refers to the Reflotron analyzer, and so this modelwill be examined in further detail.Proponents of desk-top analyzers emphasize their practicality compared tolaboratory testing where it is usually necessary for the patients to make aspecial visit to have their blood drawn. It is appropriate, therefore, tocompare the total cholesterol results obtained using capillary blood in theReflotron analyzer with results obtained from venous blood drawn at thesame time and analyzed in a quality-controlled lipid laboratory. A number of2studies have carried out this comparison and found the correlationcoefficients (r) between the two data sets to range from 0.92 - 0.99,1,2,4Reliability was assessed by calculating the coefficient of variation (100 xstandard deviation/mean) for repeated analyses on the same sample; valuesfor the Reflotron were 1-2% for trained technical personnel and 2-6% foruntrained-personnel (such as medical -office clerks -and- secretaries).5 Fewpapers have examined the direction of any bias inherent in the Reflotronresults and those that have report inconsistent results: Bachorik et al.4 foundthe Reflotron values to be 1-4% lower than the laboratory values in two citiesstudied (with sample sizes of 107 and 275) , Jones et all found them to be 8%higher (but based on only nine samples), and another study by Bachorik'sgroup found them to be 6% higher (sample size of 290).6 No adequateexplanation of this variation has been offered, but it may be due to therelatively small sample sizes involved in the studies and/or the variabilitythat has been observed between reference laboratories (see later).The above estimates of reliability and validity do not address the question ofinterest here which is the extent to which the Reflotron results misclassifypeople compared to the laboratory results. This is of particular importancesince proposals for cholesterol screening call for certain cut-off values belowwhich the individual is reassured and above which the individual is referredfor further investigation and possible treatment. Only two studies permit anestimation of the sensitivity and specificity of the Reflotron measurementrelative to the laboratory analysis of venous blood. Bachorik et al.4 comparedresults from two Lipid Research Center laboratories that used Centers forDisease Control (CDC)-standardized methods with Reflotron desk-top results,and found the sensitivity of the Reflotron analyzer to range from 0.93 to 0.99and specificity to range from 0.80 to 0.93 (using a cut-off of 5.2 mmol/L or 200mg/dL). This particular cut-off was chosen since the National CholesterolEducation Program in the United States stated that desirable total cholesterollevels were those below 5.2 mmol/L, which is approximately the mean forthe adult US populationf In a later study using the same cut-off level and asimilar group of patients (those volunteering for cholesterol measurementsas a result of community publicity), these same authors estimated thesensitivity to be 0.95 and the specificity to be 0.73.6 The authors were not able3to explain this difference in specificity, but it is likely related to the reasonsmentioned above for the difference in the direction of bias.If one assumes these test characteristics are constant for different cut-off levels(different prevalences of "high" total cholesterol), then the predictive valueof this -desk-top -analyzer may be assessed -relative to -the laboratory. Forexample, the Canadian Consensus Conference on Cholesterol? recommendedintervention when cholesterol values exceed 6.2 mmol/L, which isapproximately the 75th percentile for the adult Canadian population.8,l OUsing a sensitivity of 0.95 and a specificity of 0.8 (the approximate means of. the estimates in Bachorik's two papers4,6), one may calculate a positivepredictive value of 61.3% and a negative predictive value of 98% (since theprevalence of "high" cholesterol is about 25%). That is, 61% of the resultsjudged to be high by desk-top analysis would be judged high on laboratoryanalysis, and conversely 98% of the low desk-top results would also be low onlaboratory analysis. Furthermore, 39% of the adult population would beidentified as having "high" total cholesterol rather then the laboratory­determined prevalence of 25%.2. Cholesterol screening programs in generalThere are a variety of current recommendations for the implementation ofcholesterol screening. For example, the National Cholesterol EducationProgram (NCEP) in the USA recommends screening for all adults over theage of 20 and drug therapy for those with LDL cholesterol above 4.9 mmol/L(approximately above a total cholesterol of 6.2 mmol/L).8 The CanadianConsensus Conference on Cholesterol (CCCC) recommends screening for alladults over the age of 18 as resources permit, and recommends drug therapyfor those with cholesterol levels persisting above 6.2 mmol/L after sixmonths of intensive dietary therapyfHowever, before one can consider the impact of desk-top analyzers on theserecommendations for cholesterol screening 'programs, it is necessary to firstaddress the issues involved in cholesterol. screening when laboratorymeasurements are used to determine total serum cholesterol. A recent4review of asymptomatic hypercholesterolemia by the Toronto WorkingGroup on Cholesterol PolicylO examined this question in detail and the mainissues may be summarized as follows:a. There are -recognized difficulties -ln standardization -and quality controlbetween laboratories regarding the measurement of total serumcholesterol. For example, a 1988 survey of laboratories in Ontario found arange of measurements of 4.64-7.72 mmol/L for a reference value of 6.67mmol/L.10 It was further estimated that for a patient with a total serumcholesterol of 6.5 mmol/L, the range required to encompass 95% of theresults from the different laboratories would be 6.0-7.0 mmol/L.Such sources of inaccuracy may be due to imprecision or bias. Theprecision in Ontario laboratories appears fairly good; a 1984 survey found amean coefficient of variation of 4% (the suggested "acceptable" level isS5% and "ideal" is S3%),lO Bias, on the other hand, is more variable anddepends not only on the type of analytic machine used, but also on theparticular laboratory using the machine. For example, American labsusing the Dupont-ACA machine reported values ranging from 15% belowto 3% above the value obtained by a Lipid Research Centre referencelaboratory, and Technicon-SMAC results varied from 5% below to 12%above the reference value.10. As a result, the Toronto group recommendedthe establishment of a number of lipid reference laboratories to improvestandards and quality control to reduce the problems associated withmisc1assification. In addition, it is important to note that for any givendegree of imprecision and bias, the lower the threshold value fordiagnosing "high" cholesterol, the greater the absolute number of personswho will be misc1assified since a larger proportion of the population willhave cholesterol values close to the the cut-off level.b. Even assuming no laboratory error, there is still a great deal of variation inthe relationship between total serum cholesterol and subsequent coronaryheart disease (CHD). Data from the MRFIT trial indicate that middle-agedmen in the top 20% of the serum cholesterol range (approximately greater '5than 6.7 mmol/L) have a 0.7-2.2% chance of dying from a CHD event overthe next six years, which is approximately double the chance for those inthe lower 80% of the range.10 If one then uses 6.7 mmol/L to divide meninto high- and low-risk categories, the sensitivity would be 0.35-0.40 andthe specificity 0.72-0.83 for predicting CHD mortality over the next sixyears.10 Assuming from the-above that the cumulative incidence ofcoronary mortality over a six-year period for a group of middle-aged menis 1%, then an individual with a positive test (serum cholesterol greaterthan 6.7 mmol/L) would have about a 1.7% chance of experiencing suchmortality (1.7% positive predictive value). A person with a negative testwould have approximately a 0.8% chance (99.2% negative predictivevalue).A similar picture results if one examines prediction of CHD morbidity.Using a cut-off of 6.7 mmol/L to predict morbidity over the next six years,total serum cholesterol has a sensitivity of approximately 0.35 andspecificity of 0.85 10 Assuming the cumulative incidence of coronarymorbidity over a six-year period for middle-aged men is about 5% (valueestimated from MRFIT data and data from UK Heart Disease PreventionProject presented in Naylor et al10 ), then an individual with a positive testwould have an 8.4% chance of experiencing such morbidity. A negativetest would confer a risk of about 4% (96% negative predictive value). -It is clear from these calculations that using a cut-off value of 6.7 mmol/Lfor serum cholesterol does not differentiate well between those who willand those who will not experience CHD morbidity and/or mortality. Thisproblem of poor predictive value is not solved by changing the cut-offlevel: a higher cut-off would improve specificity but reduce sensitivity(more false negatives) and a lower cut-off would do the reverse (morefalse positives). Furthermore, the main problem with a high cut-off is thatit would have little impact on the population burden of coronary heartdisease, while a low cut-off would involve further diagnostic work-up andpossibly treatment of many people never destined to get coronary heartdisease.6It is evident, therefore, that even with a perfect laboratory test for totalserum cholesterol, there are many problems associated with using suchmeasurements as guides to further work-up and treatment in an attemptto reduce CHD mortality and morbidity. Extensive research efforts arecurrently underway on other serum lipids (most notably high densitylipoprotein or HDL) to -find measurements that are more predictive ofCHD events.c. Even if coronary morbidity could be accurately predicted using totalcholesterol measurements, there is still the question of whether treatmentis effective. A number of studies have convincingly shown thatcholesterol-lowering drug therapy (and to a lesser extent dietary therapy)in middle-aged men with high cholesterol levels (e.g. above 6.85 mmol/L)can lower subsequent CHD morbidity and mortality. However, there areseveral important caveats here. Firstly, because of the above-mentionedmisclassification and prediction problems, many men will have to betreated for many years in order to prevent one CHD event. Men treatedunnecessarily will likely experience inconvenience, treatment side-effectsand "labelling" which means they may feel and/or act unwell simplybecause they have been told they have high cholesterol. Secondly, thesame studies which demonstrated a reduction in CHD mortality did notshow a reduction in all-cause mortality. While the explanation for this isstill not agreed upon, the experimental data available at present indicatesthat cholesterol-lowering therapy does not reduce overall mortality, atleast in the short-term (five years). Thirdly, the effect of cholesterol­lowering therapy on overall morbidity cannot be assessed since thesestudies did not ascertain non-CHD morbidity. And finally, no studieshave examined the effectiveness of treatment in women, younger men orolder men.In summary, a minority of the middle-aged treated forhypercholesterolemia will benefit in terms of reduced CHD morbidityand/or mortality, but even these individuals may not experience reducedoverall morbidity or mortality.7d, Even if one could solve the above problems, there still remains cost andfeasibility problems associated with any proposed program. Cost­effectiveness analyses of using drugs to treat all Canadian men in the 90th .percentile for total serum cholesterol (>7.75 mmol/L) indicate that thecosts would· be $100,000-200,000 per life-year saved (assuming overallmortality reduction is similar to CHD mortality reduction).tO This isextremely high compared to other CHD interventions: $20,000 per life-yearfor treatment of moderate and severe hypertension and $14,000 forcoronary bypass graft surgery for 3-vessel disease. to Furthermore,implementing the screening programs suggested by NCEP or CCCC wouldoverwhelm practitioners and laboratory facilities with the demand forcounselling and blood analyses.In view of all of the above, the basic recommendations of the TorontoWorking Group were twofold. Firstly, instead of general screening, a case­finding approach in physicians' offices should be adopted where cholesteroltesting is done only on those with CHD risk factors (such as smoking,elevated blood pressure, diabetes, etc). Those remaining in the 90th percentileafter intensive dietary therapy should be considered for drug therapy onlyafter they are fully informed regarding the benefits and risks of such therapy.And secondly, to reduce the population burden of CHD a population strategyis recommended whereby public health campaigns reinforce appropriate dietand lifestyle changes.3. The effect of desk-top analyzers on recommendations for cholesterolscreeningThe main advantage of desk-top cholesterol analyzers is the provision ofquick, convenient results in the physicians office. This advantage may helpaddress some of the concerns regarding the overloading of laboratory facilitiesthat were raised with respect to general screening. However, this advantagecomes at the cost of increased misclassification. On top of the poor predictivevalue of a truly high cholesterol value, there is the problem of the accurate8measurement and classification of this value. Even in large laboratoriesstaffed by professionals there are difficulties in obtaining reliable, validcholesterol results. Together these two facts lead to a very low positivepredictive value (8.4%) and a relatively low negative predictive value (96%)when trying to predict future CHD events using a cut-off of 6.7 mmol/L fortotal cholesterol-measured -in quality-controlled laboratories (see section 2b).Furthermore, there is the additional problem with desk-top analyzers of theextra misclassification caused by using these devices to estimate laboratorymeasurements. At present, there is insufficient data in the literature to allowthe combination of these two sets of diagnostic characteristics to estimate theaccuracy of prediction of future CHD events based on desk-top results(especially need data on the degree of independence of the two tests). Onemay estimate, however, that using a cut-off of 6.7 mmol/L for desk-topmeasurements to predict future CHD events, the sensitivity would beapproximately 0.30-0.35 and the specificity 0.70-0.80. The resulting predictivevalue of a positive test would be 6.4% and the predictive value of a negativetest would be 95.5%. Thus, in exchange for patient convenience, the use ofdesk-top analyzers would worsen an already significant misclassificationproblem.Proponents of desk-top analyzers argue that they are useful for generalcholesterol screening if used as a preliminary screening tool. Since theirmisclassification errors appear to be primarily false positives rather than falsenegatives (low specificity, high sensitivity), few people with high cholesterolwould be missed and the false positives could be detected at follow-uplaboratory testing. Using figures presented in section 1 of this report and acut-off of 6.2 mmol/L (75th percentile), for every 100 people presenting forgeneral cholesterol screening the use of desk-top analyzers would send 39 forlaboratory testing and reassure 61 that there cholesterol was not high enoughto warrant further testing. Compared to general screening using laboratorymeasurements from the start, the desk-top approach would have certainadvantages and disadvantages. The advantage is that 61 people would bereassured with the use of a more convenient and probably less expensive test.The disadvantages are that one or two of these 61 people would be falselyreassured (they would have high cholesterol values on laboratory testing) andthat 39 people have had an unnecessary finger-prick blood test done (since9they would have gone for laboratory testing anyway in the scheme usinglaboratory testing initially). Which choice is preferable will depend on therelative cost of the two tests and the value to the patients of an office testwhich may obviate the need for a laboratory test.Note, however,- that this presupposes the bias in desk-top analyzers leads toconsistent overestimation of total cholesterol, and insufficient studies ofthese analyzers have been done to confirm this. In addition, the question ofquality control still remains: how would the analyzers be standardized andmaintained, and what level of training would the operators have?Whichever cholesterol screening program one prefers, there may be a placefor desk-top analyzers to perform a convenient, preliminary test. Additionaldata is needed, however, before such a recommendation can be made withany confidence. In particular, information is needed in the following areas:(1) the costs of desk-top versus laboratory analysis; (2) the value to patients ofhaving a convenient test that may mean they do not require a moreinconvenient, invasive test; (3) the precise diagnostic test characteristics ofdesk-top analyzers relative to laboratory testing; and (4) how quality controlissues will be addressed for desk-top analyzers. Clinicians would also have tobe careful that the ease and availability of desk-top cholesterol analysis doesnot influence them to screen people who are not eligible for the screeningprogram.1 0References1. Nanji AA, Poon R, Hinberg I, et al. Desktop analyzers: quality of resultsobtained by medical office personnel. CMAJ 1988;138:517-20.2. Nanji AA, Sincennes F, Poon R, Hinberg I. Evaluation of the BoehringerMaanheim Reflotron analyzer. Clin Chern 1987;33:1254-55.3. Bradford RH, Bachorik PS, Roberts K, et al. Blood cholesterol screening inseveral environments using a portable, dry-chemistry analyzer andfingers tick blood samples. Am J Cardio 1990;65:6-13.4. Bachorik PS, Bradford RH, Cole T, et al. Accuracy and precision of analysesfor total cholesterol as measured with the Reflotron cholesterol method. ClinChem 1989;35:1734-39.5. Nanji AA, Poon R, Hinberg I. Physician office analyzers. Arch Path LabMed 1988;112:757-59.6. Bachorik PS, Rock R, Cloey T, et al. Cholesterol screening: comparativeevaluation of on-site and laboratory-based measurements. Clin Chem 1990;36:255-60.7. Jones A, Davies DH, Dove JR, Collinson MA, Brown PMR. Identificationand treatment of risk factors for coronary heart disease in general practice: apossible screening model. BMJ 1988;296:1711-14.8. Anon. Report of the National Cholesterol Education Program, ExpertPanel on Detection, Evaluation, and Treatment of High Blood Cholesterol inAdults. Arch Intern Moo 1988;148:36-69.9. Anon. The Canadian Consensus Conference on Cholesterol: Final Report.Can Med Assoc J 1988;139:(suppl) 1-8.1 110. Naylor CD, Basinski A, Frank J, Rachlis MM. Asymptomatichypercholesterolemia: a clinical policy review. J Clin Epidemiol 1990;43:1029­1121.12


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items