Effects of Wavelength Error and Spectral Band Width on Measurement

Document Sample
Effects of Wavelength Error and Spectral Band Width on Measurement Powered By Docstoc
					CLIN.CHEM. 25/3,


1 (1979)

Effects of Wavelength Error and Spectral Band Width on Measurement of Alkaline PhosphataseActivity
Richard S. Schifreen and Robert W. Burnett

We re-examined the effects of wavelength error and spectral band width on the measurement of alkaline phosphatase activity. The method we studied is relatively
insensitive to these two factors, a conclusion we base on


to spectral



of 5 nm and 20 nm. The

both experimental results and theoretical analysis. These findingsre inconflict a witha recently published report [Loft
et al.,Gun. Chem.

apparent absorbances predicted by this model were then compared with experimental data. Calculations were made with a Hewlett-Packard 9825 computer with nonlinear regression and with Simpson’s Rule numerical integration algorithms supplied by the vendor. Experimental Conditions

24, 938 (1978)],and we suggest a
stray light

possible explanation for this. Additional Keyphrases: variation, source of

of serum alkaline phosactivity was recently published as a Selected Method (1). In attempting to use this method with the GEMSAEC centrifugal analyzer, Lott et al. (2) reported large phatase

method for the determination

The conditions of the Selected Method (1) for serum alkaline phosphatase measurement were used for analysis of a pooled serum sample. In addition, we prepared a set of simulated reaction solutions, using known concentrations of 4nitrophenol in water as the “sample,” but otherwise adhering to the Selected Method. Spectrophotometric measurements were done with either

systematic biases, which were correlated with the wavelength setting of the instrument’s monochromator. Because the reported magnitude of this effect did not appear to be consistent with our previous experience or that of others (3), we decided to re-examine the effects of spectral band width and positive and negative wavelength errors on the measurement of serum alkaline phosphatase activity. We compared our experimental results with previously reported results and with predictions based on an accepted theoretical model.

a Cary 16 or Cary 219 spectrophotometer (Varian Instruments, Palo Alto, CA) equipped with water-jacketed cuvets and a temperature controller. We verified wavelength accuracy of the spectrophotometers by examining mercury emission lines, and used National Bureau of Standards SRM 930 filters to verify absorbance accuracy.

Theoretical Model
The two lines in Figure 2A show the dependence of apparent activity on wavelength at spectral band widths of 5 nm and 20 nm, as calculated from the theoretical model described above. We normalized all values, taking the calculated #{163}4 at a nominal wavelength of 402 nm and at a spectral band width of 5 nm as 100 units. The natural band width of 4-nitrophenol is approximately 66 nm. This is sufficiently broad so that even at a spectral band width of 20 nm, the observed #{163}4402 nm is predicted at to be only 4% less than the true #{163}4 would be obtained at that a very narrow spectral band width. One might expect that the predicted A, and therefore apparent activity, would be highest at 402 nm and symmetrically lower at adjacent wavelengths. In this case, however, the theoretical results show some asymmetry, especially with a 20-nm spectral band width. The background absorbance of 4-nitrophenyl phosphate, which changes quite rapidly with wavelength, introduces asymmetry into the theoretical curve. At 408 nm the predicted apparent activity at a 20-nm spectral band width is 8% less than that at a 5-nm spectral band width, while at 396 nm the apparent activity is predicted to be exactly the same at the two band widths.

Materials and Methods
Theoretical Model
The effect of wavelength and spectral band width on absorbance measurements may be predicted if the shapes of the slit distribution function and of the absorbance band of interest are known. A conventional approximation assumes a symmetrical triangular shape for the slit distribution function and a gaussian shape for the absorbance band (4). In the case of the alkaline phosphatase reaction mixture, however, the very steep background absorbance curve due to the substrate, 4-nitrophenyl phosphate, must also be taken into account. Figure 1 shows absorbance curves for 4-nitrophenol and 4nitrophenyl phosphate corresponding roughly to concentrations that are encountered in alkaline phosphatase assays. We fitted the observed absorbance curve of 4-nitrophenyl phosphate to a power function, and added to it varying proportions of a gaussian function centered at 402 nm, which approximated the contribution of 4-nitrophenol. Following the gen-

eral approach
merically function


by Brodersen

(4), we integrated


over the wavelength interval of interest the resulting and triangular slit distribution functions correHartford, CT

Clinical Chemistry Laboratory, Hartford Hospital, 06115. Received Nov. 3, 1978; accepted Jan. 3, 1979.


We prepared a series of simulated reaction solutions,using 4-nitrophenol concentrations of from 0 to 200 tmol/L, 15 CLINICALCHEMISTRY,Vol. 25, No. 3, 1979 429



4NPP 1.2


80 1.3

60 0.8










0.2 4NP










Fig.1.Absorbance spectra of4-nitrophenol (4NP) and 4-nitrophenyl phosphate (4NPP) at concentrations corresponding roughlyto thoseused inthe determination alkalinehosof p phatase activity mmol of 4-nitrophenyl phosphate per liter, and other com-


396 399 402 405 408

WAVEENGTH L (NM) ponents as specified in the Selected Method. The apparent “activity” obtained with these solutions, calculated as the change in absorbance divided by the change in concentration of 4-nitrophenol, is analogous to data generated in a kinetic determination of alkaline phosphatase activity. These data are presented as solid circles (5-nm spectral band width) and open circles (20-nm spectral band width) in Figure 2A, with vertical lines denoting the experimental uncertainty in the data. The experimental results shown in Figure 2A and B have also been normalized to facilitate comparisons, with the values at 402 nm and 5-nm spectral band width being taken as 100.

Fig. 2.Relative activity ofalkaline phosphataseas a function of thewavelengthusedtofollow the reaction
A. The solid lines show the theoretical wavelength dependence at spectralband widths of 5 nm and 20 nm. Circles denote experimental results at 5 nm (#{149}) and 20 nm (0). 8. The line denotestheoretical behavior at spectral band width 5 nm. Also shown are resultsof serumalkaline phosphataseassays from this study (U) and a previous study (2) (X)

concentration of 4-nitrophenyl phosphate, which results a high absorbance despite its low molar absorptivity.


In all cases the experimental

data fit the theoretical


as well as can be expected, given the experimental imprecision. This imprecision results primarily from having to reset the wavelength and the slit width between measurements of the various solutions. Because the absorbance curve for 4-nitrophenyl phosphate is extremely steep in this region, wavelength changes of the order of ±0.02 nm significantly affect the measured absorbances. The results of alkaline phosphatase assays on the pooled serum sample are shown as squares in Figure 2B. The line shown is the theoretical line for a spectral band width of 5 nm

and is included for reference. The experimentally

determined activities are clearly in excellent agreement with the predicted behavior at the different wavelengths. The points marked with X’s in Figure 2B are taken from Table 1 of reference 2. Here the agreement with predicted behavior is not as good,with values at 396 and 399 nm being markedly discrepant. Discussion
Under the alkaline phosphatase assay conditions, the background (or blank) absorbance increases extremely rapidly as the wavelength is decreased from about 410 nm down to 395 nm and beyond. This has been pointed out by Lott et al. (2), among others, and is clearly illustrated in Figure 1 of their paper and Figure 1 of this paper. The effect is due to the high

As mentioned above, thisrapidly increasingbackground absorbance affectsthe theoreticalpredictions of apparent activity a functionof wavelength and spectralband width. as However, at a band width of 5 nm, which is the nominal spectral band width of the monochromator in the GEMSAEC centrifugal analyzer used in the studiesof Lott et al., he apt parent alkaline phosphatase activity is predicted to be only 1% lower at 396 nm and 2% lower at 408 nm than at the optimum wavelength of 402 nm. In fact, even at a spectral band width of 20 nm, the predictedbias isstillnly -1% at 396 nm, o but -9% at 408 nm. As shown in Figure 2, our experimental measurements tend to agree with these predictions. hey are T also consistent with results obtained by Penton et al.(3), who measured alkalinephosphatase activities 404,414, and 424 at nm on severalsamples. Clearly,the resultsobtained by Lott et al.cannot be explainedon the basisof a wavelength errorper se.As shown in Figure 2B, theirapparent activities ere much lower than w expected at wavelengths below 402 nm, and the biaswas much greaterat 396 than at 399 nm. Moreover, theirresultsshow much larger errors at wavelengths below 402 nm than at wavelengths an equal distanceabove 402 nm. In our opinion,the most likely explanation of thiseffectis that a significant mount of straylightwas present in the ina struments used to obtain these results. Stray light is always

a potentially serious problem with a tungsten filament light source below 400 nm, because source output falls off sharply
in this region. The problem is compounded if a detector with



low sensitivity
with photodiodes.

in this region is used, as is true, for In addition, the absorbance error

tometric and fluorometric measurements. J. Res. Nat!. Bur. Stand.
76A, 491 (1972).

to unabsorbed stray light is much greater at high absorbances, and as noted previously the blank absorbance due to 4-nitrophenyl phosphate increases sharply at wavelengths below 402 nm. The observed bias at 399 nm and 396 nm in the data of Lott et al. would correspond to the presence of about 1% unabsorbed stray light. Some details of the calculation are given in the Appendix. Although this is much more stray light than is present in well-designed spectrophotometers in good operating condition, recent evidence gathered from a survey of clinical laboratories (5) suggests that
stray light is indeed expected. Although a more common problem than might be


4. Brodersen, S., Slit width effects. J. Opt. Soc. Am. 44,22 (1954). 5. Beeler, M. F., and Lancaster, R. C., CAP survey to assess the extent of stray light problems in precision spectrophotometry. Am. J. Clin. Pat ho!. 63,953 (1975).

Unabsorbed stray light produces errors in absorbance readings that are greater at higher levels of absorbance. The magnitude of the error can be calculated from the equation:







1+ r




stray light could account for the data obtained between 396 nm and 402 nm, the slight decrease in apparent activity at 405 nm and 408 nm cannot be attributed to this factor and requires a different explanation. In summary, the natural band width of 4-nitrophenol is 66 wide enough to ensure that absorbance measurements made within ±3 nm of the peak with a spectral band widthof 5 nm or less will result absorbance(and activity) errors of in less than 1% from the combined effect ofwavelengthshift per
se and finite spectral band light is a potentially serious


where r is the fraction of stray light present. A similar equation can be derived for calculating an error in #{163}4, is appropriate if one wishes to know the magniwhich tude of error attributable to stray light in an alkaline phosphatase assay, where the initial and final absorbances are both greater than zero.
= true


Il + r . 10.421





On the other



problem in alkaline phosphathse measurements because the wavelength used usually corresponds to relatively low source outputand low detector sensitivity, and becausethe background absorbance from 4-nitrophenyl phosphate may be considerable.
We thank Dr. Nelson L. Alpert for helpful discussions to this study. pertinent

where A1 is the true initial absorbance, is the true final A2 absorbance, and #{163}4defined as A2 - A1. is In reference 2, column 4 of Table 1 lists observed values of initial bsorbance. a Because A1 and A2 in equation 2 are true absorbances, we must guess at r and calculate true initial absorbance (A1) from the observed value. This is facilitated by rearranging equation 1 as:
Atrue = Aobserved

log (1 + r




#{163}4true was

For the purpose of the calculations

in this study

1. Bowers, G. N., Jr., and McComb, R. B., Measurement of total alkaline phosphatase activity in human serum. C!in. Chem. 21, 1988 (1975). Selected Method. 2. Lott, J. A., Turner, K., and Scott,., J Factors affecting measurement of total alkaline phosphatase activity in human serum, especially wavelength accuracy. Clin. Chem. 24, 938 (1978). 3. Penton, J. R., Widdowson, G. M., and Williams, Z., Problems C. associated with the need for standardization in clinical spectropho-

approximated as 0.1 at all wavelengths between 396 and 408 nm. Using values of from reference 2 and making an initial guess at r, we calculate A true from equation 3, and

in equation

2 to obtain


for /A0ha5d.

a process of trial and error we found that r = 0.011 gives a ratio of #{163}4observed at 399 nm to that at 402 nm of 82% and a
ratio of Aob85.j at 396 nm to that at 402 nm of 45%. These ratios are comparable to those actually observed, which are plotted as X’s in Figure 2B.

CLINICALCHEMISTRY,Vol. 25. No. 3, 1979


Shared By:
Description: Effects of Wavelength Error and Spectral Band Width on Measurement