Biological variation of ischaemia-modified albumin in healthysubjects

Summary

Each year, several million patients present to the emergency department with chestpain. According to figures from the USA, half of this group will be admitted, butonly approximately 20% will actually be diagnosed with acute coronary syndrome(ACS). On the other hand, 2% of patients with acute coronary syndrome will bemistakenly discharged.1-3 As patients with ACS have a relatively higher risk of majorcardiovascular events in the short term, there is considerable clinical interest andclinical research effort underway to identify biomarkers of myocardialischaemia.1,3

A blood test that could exclude the presence of myocardial ischaemia woulddramatically improve the triage process of patients with acute coronary symptoms,decrease the number of hospital admissions and reduce the overall cost ofhealthcare. Clinical studies have shown that ischaemia-modified albumin (IMA), asmeasured by the albumin-cobalt binding (ACB) test®, has been cleared by the USFood and Drug Administration as a possible early indicator of myocardialischaemia.4-7 This test is a biochemical assay based on the observationthat human albumin has the capacity to bind transition metals. In the presence ofischaemia (myocardial and elsewhere), the amino or N-terminal of albumin is modifiedand subsequently affects transition metal binding.8 This modified albumin, with a lower transition metalbinding capacity,is known as IMA.

IMA rises within minutes of the onset of myocardial ischaemia and returns to baselinewithin six hours of restoring perfusion.9Previously published studies describe IMA as a risk-stratification tool forsuspected ACS. In selected low-risk emergency department patients defined by anelectrocardiogram (which is non-diagnostic for ischaemia), a negative troponin and anormal IMA test, discharge can be considered. Hence IMA is used as a ‘rule-out’ testin selected patients with ACS.1,9

Although there are a number of studies detailing the clinical utility of IMA, itsbiological variation is still lacking. Biological variation studies are essentialprerequisites to the introduction of any new biomarker. This data should begenerated early in the course of evaluation of new tests as quantitative dataobtained can be used to set desirable analytical quality specifications, assess theutility of conventional reference ranges, and define the significance of changes inserial results.10,11

In this study we performed IMA testing on a cohort of healthy subjects over afive-week period to determine the analytical and the biological variance componentsof ischaemia-modified albumin.

Materials and methods

The subject population consisted of 17 apparently healthy volunteers; seven men, twoCaucasian, five black (age range 43−61 years, median age 50 years) and 10 women; sixCaucasian, four black (age range 26−61 years, median age 41 years). All participantswere required to complete a modified RAND 36-item Health Survey 1.0questionnaire12,13 to determine their health status and were observed to behealthy. Informed consent prior to enrolment in the study was given and the HumanEthics committee of the University of Pretoria approved the study. No exclusioncriteria were applied.

Blood samples were collected once a week for five consecutive weeks. To minimisepre-analytical variation, the specimens were collected at a constant time and by thesame phlebotomist, and the collection technique was standardised. The samples wereseparated and then frozen at −70°C within one hour of collection and were analysedaccording to the manufacturer’s instruction.

Ischaemia-modified albumin was determined using a colorimetric method, the ACBtest® on a Roche modular analyser. This test measures the cobalt-bindingcapacity of albumin in a serum sample. A cobalt solution is added to the serum.Cobalt not bound to the N-terminal of albumin is detected using dithiothreitol as acolorimetric indicator. In individuals with ischaemia, cobalt does not bind to themodified N-terminal of IMA, leaving more free cobalt to react withdithiothreitol.

The instrument was set up according to the manufacturer’s instructions and wascalibrated before the analyses were performed. Duplicate analyses were performed onall samples in a single batch. The analytical coefficient of variation for the lowcontrol was 4.22% (range 55−71 U/ml), for the medium control, 2.01% (range 69−91U/ml) and for the high control, 1.24% (range 102−130 U/ml).

The hierarchical design of analysis of variance was determined using Statistix 8.0software, with the variance components then being used to calculate the analyticalvariation (CV_A), within-subject variation (CV_I) and thebetween-subject variation (CV_G) according to Fraser and Harris. Theanalytical goals for imprecision (0.5 3 CV_I) and bias [0.25 ×(CV_I + CV_G)^1/2] were also determined.10,11The index of individuality was calculated by

(CVA2+CVI2)1/2CVGasCVA>CVI

Analyses of the measurements were performed, with gender and race (black andCaucasian) as between-subject factors and weeks (1−5) as the within-subject factor,using an appropriate ANOVA for repeated measures.

Results

The mean, within-run analytical variation, between- and within-subject biologicalvariation, goals for imprecision and bias, and the index of individuality are shownin Table 1. Analysis of the repeated measureswith gender and race as between-subject factors, and weeks as the within-subjectfactor showed that gender had no significant effect on circulating IMAconcentrations (p 5 0.3146), whereas race did have a significanteffect (p 5 0.0062). A significant (p 5 0.0185)interaction was observed between gender and race. From Table 2, the relationship between gender and race is evident andit seems reasonable to ascribe the interaction to the very different outcomes formales between the races. Furthermore, the within-factor week (Table 3) was not significant (p 5 0.1915) andthe interaction in terms of week with gender and race were omitted in this finalanalysis as they were insignificant in the initial analysis and were then pooledwith the error term.

Discussion

Using a standard protocol for sample collection, pre-analytical variation wasconsidered negligible or an intrinsic component of the within-subject biologicalvariation. As shown in previous studies, estimates of biological variation in asmall group of apparently healthy subjects may be useful for a variety of purposesand these estimates should be similar across studies, in theory at least, since theresults are quantitative components of homoeostatic mechanisms in a single animalspecies.10,11,14 It is widelyaccepted that the best strategy for determining the standards of analyticalperformance (precision and bias) in order to provide optimal patient care are bestderived from data on biological variation.10,11 This study thereforeinvestigated the potential use of IMA based on these calculations.

The IMA values obtained in our population were considerably higher (Fig. 1) than specified by the kit manufacturer,but similar to values obtained in other clinical studies.2,15 The sample size ofthis study was deemed too small to allow a valid calculation of the reference range.Laboratories are encouraged by the manufacturer to establish their own optimal IMAclinical cut-off concentrations as the values may vary, depending on geographic,patient, dietary and environmental factors.16In this study, these non-random variations were not accounted for. Furthermore,serum albumin, which could affect IMA results when using the ACB test, was notmeasured in this study population. The main reason for this omission was thesupposition that the population comprised ‘apparently healthy’ individuals withnormal serum albumin concentrations.

Fig. 1.

Within-subject IMA variation over five weeks. Minimum, mean and maximum IMAvalues over five weeks are depicted for each subject. The vertical brokenline indicates the manufacturer’s decision limit (85 U/ml).

Several recent studies15,17 have shown that storage of IMA samples at 4°C or −20°C haveresulted in higher values compared to real-time analysis. Although the effect ofstoring samples at −70°C (as in this study) was not assessed by means of a stabilitystudy, it can be inferred that the effect would be similar to that reportedpreviously. This may explain why higher values were obtained compared with thevalues reported by the manufacturer. Further studies on the effect of sample storageat −70°C on IMA concentrations are indicated.

The observed precision for IMA determination (5.04%), although acceptable, shows lessthan desirable analytical imprecision (1.45%), which indicates that the currentmethod for IMA assay on the Roche modular analyser may warrant further improvementand optimisation. It was difficult to decide whether the goals for bias were met,which if met, would allow the use of common reference intervals throughout ageographical area. Standardisation of the IMA assay will, however, ensure that thesegoals are met in the near future.

Harris has shown that when the within-subject variation is greater than thebetween-subject variation, conventional reference values will be of use, but whenthe between-subject variation is greater than the within-subject variation,reference values will be of little use for monitoring change.18,19 More formally,reference values are of marked usefulness only if the index of individuality isgreater than 1.4. In this study, the index of individuality was less than 1.4, whichmay be explained by the interaction between race and IMA concentrations.

It can therefore be deduced that IMA determination may currently not be a good testfor detecting latent or early disease, since an individual may have a value that isvery unusual for him/her but still falls well within conventional population-basedreference limits. Although in practice, a decision cut-off is used for theinterpretation of IMA results, this marked individuality would still mean that falsenegative results could occur, which in theory would reduce the sensitivity of theIMA assay as an exclusion test of myocardial ischaemia. The significant interactionbetween race and IMA values obviously prompts further investigation.

Factors other than biological variation that may need to be considered beforeroutinely employing the IMA test are briefly alluded to. The rapid turn-around time(dwell time of ± 20 min) makes IMA a suitable cardiac marker for the early diagnosisof myocardial ischaemia. The specificity of IMA may, however, not be optimal as IMAmay be elevated in patients with active cancer, bacterial or viral infections,end-stage renal disease, liver cirrhosis, brain ischaemia, peripheral arterialdisease and trauma. Since the meaning of a raised IMA value is not clearlyunderstood, it has been suggested that such results should prompt the clinician toresume an ACS evaluation as per their usual standard.9

Furthermore, sample and reagent lability needs to be kept in mind. The labile natureof IMA requires that the sample be analysed within 2.5 hours of sample collection orrefrigerated/frozen until analysis. The dithiothreitol reagent and hence the fullkit is only stable for 14 days. The current high cost of the test may also limitwidespread use.

Conclusion

The ACB test® could bring a new dimension to the care and management of patientswith acute coronary syndrome. IMA can be measured accurately, reliably and within anacceptable time period to be useful in the evaluation of the patient with ACS,however, further studies for normal population distributions by gender andethnicity, and an optimum cut-off value are still required.