Tumor Marker Assays
The Significance of Normal Range Studies

Normal range studies serve two main purposes. First, they help to characterize an assay's performance in a fundamental way (method verification), and may prove a valuable source of clinically relevant insights. Second, they provide one basis for making sense of individual patient results (interpretation), even in certain contexts where adopting the upper limit of normal as a decision level would be inappropriate.

These two benefits are especially evident for tumor marker assays - though only if the normal range studies are carefully designed and executed, of course. (The value of such studies depends critically on the quality of the assays involved, on the character and size of the study population, and on proper data analysis.)

Accordingly, DPC has undertaken extensive multicenter, multivariate reference range studies for several IMMULITE® tumor marker assays. The main results established thus far have been summarized in a DPC technical report (catalog number: ZB148).¹ This article draws on material from the report to illustrate the nature and significance of normal range studies in the context of tumor marker assays.

Interpretation
It is widely accepted that tumor marker assays are generally inappropriate for population screening, due to inadequate clinical sensitivity and/or specificity. Many such assays have more firmly established applications in the monitoring and follow - up of various types of treatment.

Consequently, some authorities have advocated that upper reference limits be established for certain groups other than normals, e.g. for curative prostatectomies,² where an upper limit for PSA on the order of 0.10 ng/mL or less is expected. (This is well below the upper limit for normal men and underscores the need for exceptional, low - end precision - "third generation" sensitivity - if an assay for PSA is to yield precise and meaningful results when monitoring a subject's progress after radical prostatectomy and other therapies.³)

Nevertheless, normal range studies for tumor markers are also relevant to the interpretation of individual patient results. The upper limit of normal often represents a major landmark, even when not serving as a cutoff. Thus, successful treatment for a cancer is often followed by the return of relevant tumor markers to normal circulating levels, as with CEA. Likewise, for most prostate cancer treatments, a urologist would look for PSA levels to drop to well below normal.

Results reported to the physician should therefore be accompanied by a characterization of normal limits applicable to the assay in use. Laboratory report forms usually supply this as a centile - either the 95th or 97.5th for most tumor marker assays - indexing the upper limit of normal for adults, or for each of several relevant subgroups.

Method verification/validation
Precision and accuracy represent the two fundamental dimensions of assay performance which can be assessed via analytical studies ("analytical" in the sense of requiring no special attention to the clinical status of the samples employed). Reference range studies make a third basic contribution towards validating the assay, usually by exploring the assay's ability to reproduce well - established, clinically relevant, group - based distinctions or trends. Examples are given below for IMMULITE CEA, IMMULITE Third Generation PSA, and IMMULITE OM - MA.

Accuracy is typically assessed via recovery studies, based on a reference preparation, and/or by comparison studies against a reference method. For several tumor markers including CA15 - 3, CA19 - 9 and CA125, however, neither reference methods ("gold standards") nor reference preparations exist. These assays report in "arbitrary" units, making it imperative to generate assay - specific normal range limits as a guide for the physician faced with interpreting patient results.

Moreover, the significance of the usual analytical approach to assessing accuracy is substantially reduced in this context, forcing method verification to rely more heavily on a combination of precision and reference range studies.

Validation: AFP
Alpha - fetoprotein (AFP) assays provide an excellent example of how normal range studies can be essential even when analytical studies of accuracy do apply. Thus, for AFP, there exists a well - established reference preparation,⁴ with a well - defined conversion between mass and International Units; and major collaborative efforts have aimed at improving and standardizing various aspects of AFP immunoassay design.⁵

Even so, a 1996 survey of laboratory practice in the United Kingdom revealed a surprisingly broad range of variation in the upper limits of normal quoted for AFP results.⁶ Evidently, the same limits cannot be assumed to apply to all assays for this tumor marker in spite of a common approach to standardization - even if some of the highest values in the survey are dismissed as the legacy of an older technology. (We expect an evolution towards lower AFP reference limits as assays become more sensitive and specific, and less susceptible to so - called matrix effects.)

The survey also stands as a reminder of how difficult it can be to disentangle the impact of genuine demographic differences from differences attributable to variation in the design and quality of reference range studies, or to variation in data analysis and presentation; for the survey also showed considerable variation across laboratories in the limits quoted for the same assay.

No doubt this can be explained largely by decisions to establish, retain or adapt in - house reference limits, instead of relying on "expected values" claims from the manufacturer. Some variation in outcome from one reference range study to another is inevitable; but too small a sample size and/or too casual a selection of subjects can aggravate the problem. Hence, rather than attempting to establish their own normal values, laboratories with limited resources may be better off adopting the limits from a large, well - designed reference range study, after verifying their applicability through suitable review and experimentation.⁷

Competing conventions
Where tumor markers are concerned, we are faced with two competing conventions for characterizing the distribution of results for normal subjects.⁸ Many laboratories prefer to quote the observed 95th centile, partly because the lower limit of normal for most tumor markers is either below the detection limit of current assays, or of no apparent clinical interest, or both; and partly to buffer against "contamination" (the presence of abnormal subjects) in the reference group. Others prefer to quote the 97.5th centile, as is customary for many analytes commonly measured by immunoassay.

In effect, the first approach characterizes the distribution of values in terms of the lower 95%, the second in terms of the central 95%; so both aim at the same degree of coverage. Figures 1 and 2 show, however, that for distributions skewed towards higher values, there can be a considerable difference between the concentration levels associated with these two centiles. (Given the competing conventions in this field, DPC has opted to tabulate both centiles.)

Validation: CEA
Carcinoembryonic antigen (CEA) levels for normal adults cannot be adequately characterized by a single set of reference limits. Men tend to have higher levels than women; smokers tend to have higher levels than nonsmokers. The differences due to sex and smoking status observed in a multicenter study based on the IMMULITE CEA assay are summarized in Table 1, in terms of selected centiles.

Table 1.

Because these differences tally with well - established expectations in the literature, the statistics provide clinical evidence that IMMULITE CEA has the accuracy and precision needed to yield appropriate results under conditions known to affect baseline CEA levels.

Indeed, it would be a source of concern if a study were unable to reproduce the expected male - female, smoker - nonsmoker differences in CEA levels. Assuming that demographic factors could be ruled out, such a failure might be due to the assay (inadequate specificity or precision); to the study design (too small a data set, or too weak a criterion for distinguishing smoker from nonsmoker); or to the data analysis (transcription errors, genuine outliers, or inappropriate statistics).

The population - based information encapsulated in Table 1 is clearly also relevant to the interpretation of individual results. Thus, for example, given these range limits, a physician would be likely to evaluate a CEA result of 6.0 µg/L for an otherwise apparently normal male subject in the light of his smoking status.

Visual representations of the study results and statistical analysis can serve both as a valuable safeguard against faulty or overly simplistic data reduction and as a source of insights into the distribution of real - world data that cannot be adequately captured in a simple tabulation.

Thus, in Figure 1, the "ladders" show precisely how the centile estimates relate to the mass of data points in the study. (The horizontal lines correspond to the centiles in Table 1.) Moreover, the figure suggests one additional factor of possible relevance: there appears to be an upward, age - related trend for CEA in both male and female smokers. If genuine, this trend might be due to the duration of smoking or the amount of tobacco consumed; but this aspect of the data has not been pursued.

Figure 1

IMMULITE CEA results for normal adults, broken down by sex and smoking status. The horizontal lines correspond to the centiles in Table 1.

Validation: PSA
For prostate - specific antigen (PSA), as for CEA, there are relevant categorical variables. Sex, obviously, is one such factor, and race is emerging as another;⁹ but age, a continuous variable, is an even more fundamental determinant of PSA reference limits in normal men.

The traditional decision limit of 4.0 ng/mL had its origin as the upper limit of normal determined in a reference range study supporting the introduction of Hybritech's immunoradiometric assay for PSA. The age distribution of the subjects included in this study has been questioned;¹⁰ but the deeper issue concerns the analysis of results.

In this matter, the article published in JAMA 1993 by Oesterling et al. was a major landmark.¹¹ The authors succeeded in characterizing the age - related distribution of PSA values in normal men - both graphically, in a "nomogram" conveying the apparently continuous variation in PSA levels as a function of age, and in a table, based on a manageable number of age brackets and suitable for laboratory report forms.

Figure 2 displays a comparable nomogram based on IMMULITE Third Generation PSA results from a study conducted by Dr. A. Semjonow (Münster, Germany) in 1997. The 563 subjects included (a) men with PSA values less than 4.0 ng/mL and no evidence of disease by digital rectal examination, and (b) men who failed to qualify under the first criterion but showed no evidence of a tumor on transrectal ultrasound - guided sextant biopsy. (These PSA results were subsequently combined with data from other sources for DPC's technical report.¹)

Figure 2.

IMMULITE Third Generation PSA results as a function of age.

In the analysis summarized in Figure 2, contour lines representing the 50th, 95th and 97.5th centiles were fitted globally to the PSA results as a function of age, that is, using methods which avoid the initial partitioning of results into bins by age.^12,13

Several similar studies have made it clear that age is a critical factor in PSA normal range studies for adult males. Indeed, for many years, there has been no justification for believing in a single concentration level representing the (age - independent) upper limit of normal.

It does not follow, however, that an age - related analysis should be utilized for interpretation of results. Indeed, it is widely understood that the upper limit of normal should not automatically serve as a cutoff in screening programs, though the distribution of normal values is highly relevant. A PSA value of 4.0 ng/mL continues to have a great deal of support when considered as a decision limit applicable in certain contexts to adult men irrespective of age.¹⁴

Sampling conditions: OM - MA (CA125)
Some studies of the ovarian marker CA125 have yielded upper limits of normal for postmenopausal women which are lower than the limits established for adult women generally; moreover, it has been suggested that this postmenopausal reference limit might serve as a better decision level even for younger women in certain contexts, e.g. after ovariectomy.¹⁵

Setting aside issues relating to decision limits, one assumption here is that either age or reproductive status is a factor which must be taken into account in determining CA125 reference ranges for women. However, results generated with the IMMULITE OM - MA assay for CA125 underscore the need for further examination of this matter.

A multicenter study of several IMMULITE assays followed normally cycling women on a daily basis throughout one complete menstrual cycle and showed that many such women have distinctly higher CA125 levels during menstruation and the days immediately thereafter than in the remainder of the cycle.^16,17

These results confirm and extend other studies in the literature,¹⁸ providing a somewhat more precise delineation of the magnitude and frequency of menstruation - induced CA125 elevations than was previously available.

Like age in studies of PSA in adult men or gestational age in studies of HCG in pregnant women, ovulatory cycle position could be treated as a continuous covariate, and is naturally treated as such in studies of several reproductive hormones. For CA125, however, it is more natural to regard the early days of the cycle as a transient phase to be avoided when collecting samples for determinations of this ovarian tumor marker.

From the clinician's point of view, an appreciation for the sharp, transient increases in CA125 levels which can be induced by menstruation has an obvious bearing on individual patient results, no matter whether these are interpreted longitudinally, in terms of within - subject trajectories, or against population - based limits.

Moreover, an analysis of IMMULITE OM - MA results for samples collected during the latter half of the follicular phase or during the luteal phase in the multicenter ovulatory cycle study—i.e. excluding just the early days of the follicular phase—yielded an upper reference limit comparable to the upper limit observed for the older women in a large cross - sectional study based on the same assay.

This raises the possibility that the quality of reference limits determined in population studies of CA125 may depend more on certain aspects of study design than on the choice of subgroups during the data analysis stage. In particular, it argues for closer attention to the conditions for proper sample collection in women of reproductive age.

Technical resources
DPC's multicenter, multivariate reference range studies for IMMULITE tumor marker assays are summarized in technical report ZB148,¹ which currently provides a detailed analysis of results for the following analytes: PSA, PAP, BR - MA (CA15 - 3)*, GI - MA (CA19 - 9)*, OM - MA (CA125), CEA and b2 - microglobulin.

The material is being updated to reflect additional results, notably for IMMULITE AFP, which received FDA clearance for use as an aid in the management of nonseminomatous testicular cancer. The document is being expanded and reorganized in several respects to make it a still more convenient and comprehensive electronic or printed reference on normal ranges and related information for DPC's automated tumor marker assays.

The IMMULITE reference range information presented in this technical report is also relevant to the corresponding IMMULITE® 2000 assays, due to the close relationship, in both design and performance, between these two automated systems.

Conclusion
Substantial, carefully designed, appropriately analyzed normal range studies for an assay can greatly enhance the value of individual results reported by the laboratory, as well as provide evidence for the validity and clinical usefulness of the assay.

The multicenter studies summarized here and in technical report ZB148 reflect DPC's commitment to generating and supporting extensive clinical studies for its automated tumor marker assays; and to regularly updating its library of reference materials dealing with the studies, enhancing the content, presentation and availability of these essential resources.

In this manner, DPC continues to strive for improvements in its service to laboratories using the IMMULITE and IMMULITE 2000 systems, and thus also to the physicians who rely on these laboratories for precise and accurate patient results and for the basic orientation needed in their interpretation.

References

1. Sibley PEC. IMMULITE tumor marker assays: multicenter reference range data. Los Angeles: Diagnostic Products Corporation, 1999. Technical report ZB148. Available at DPC's web site, www.dpcweb.com, under Technical Documents, Technical Reports.

2. Stamey TA. Lower limits of detection, biological detection limits, functional sensitivity, or residual cancer detection limit? Sensitivity reports on prostate - specific antigen assays mislead clinicians. Clin Chem 1996;42:849 - 52.

3. Diamandis EP, Yu H, Melegos DN. Ultrasensitive prostate - specific antigen assays and their clinical application. Clin Chem 1996;42:853 - 7.

4. Muenz LR, Sizaret P, Bernard C, et al. Results of the second international study on the W.H.O. alpha - foetoprotein standard. J Biol Stand 1978;6:187 - 99.

5. Nustad K, Paus E, Kierulf B, Bormer OP. Specificity and affinity of 30 monoclonal antibodies against alpha - fetoprotein. Tumour Biol 1998;19:293 - 300.

6. Sturgeon CM, Seth J. Why do immunoassays for tumour markers give differing results? - - a view from the UK National External Quality Assessment Schemes. Eur J Clin Chem Clin Biochem 1996;34:755 - 9.

7. National Committee for Clinical Laboratory Standards. How to define and determine reference intervals in the clinical laboratory; approved guideline. Wayne, PA: NCCLS, 1995. NCCLS Document C28 - A.

8. Stenman UH. Prostate - specific antigen, clinical use and staging: an overview. Br J Urol 1997;79 Suppl 1:53 - 60.

9. DeAntoni EP, Crawford ED, Oesterling JE, et al. Age - and race - specific reference ranges for prostate - specific antigen from a large community - based study. Urology 1996;48:234 - 9.

10. Dalkin BL, Ahmann FR, Kopp JB. Prostate specific antigen levels in men older than 50 years without clinical evidence of prostatic carcinoma. J Urol 1993;150:1837 - 9.

11. Oesterling JE, Jacobsen SJ, Chute CG, et al. Serum prostate - specific antigen in a community - based population of healthy men. Establishment of age - specific reference ranges. JAMA 1993;270:860 - 4.

12. Harris EK, Boyd JC. Statistical bases of reference values in laboratory medicine. New York: Marcel Dekker, 1995.

13. Wright EM, Royston P. Calculating reference intervals for laboratory measurements. Stat Methods Med Res 1999;8:93 - 112.

14. Dalkin BL, Ahmann FR, Kopp JB, et al. Derivation and application of upper limits for prostate specific antigen in men aged 50 - 74 years with no clinical evidence of prostatic carcinoma. Br J Urol 1995;76:346 - 50.

15. Fritsche HA, Bast RC. CA 125 in ovarian cancer: advances and controversy. Clin Chem 1998;44:1379 - 80.

16. Sibley PEC, Vankrieken L, et al. Impact of the menstrual cycle on BR - MA (CA15 - 3) and OM - MA (CA125) values, as determined by automated chemiluminescent assays on the IMMULITE Analyzer [abstract 385]. Clin Chem 1999;45(S6):A109. Full presentation available at DPC's web site, www.dpcweb.com, under Technical Documents, Scientific Posters.

17. Sibley PEC. OM - MA (CA125) and ovarian cancer. News & Views (DPC) 1999 Summer;13(3):12 - 4. Available at DPC's web site, www.dpcweb.com, under Technical Documents, News & Views, Summer 99. Also available, in printed form and at the web site, as Technical report ZB195.

18. Meden H, Fattahi - Meibodi A. CA 125 in benign gynecological conditions. Int J Biol Markers 1998;13:231 - 7.

*Available outside the US.