WE? $89.1. DEFFERENCE FLAME PEGWMETRY: A NEW APPREAC?! 550E? C’LENECAL ANAL‘E’SES Thesis {'5}? me: {Eegsee of M. S. Eflfiéviifim 3%"? We! £56371" BE??? L-‘JU SCEéQEPKE 3.975 wgsxs ‘ 880V ENDERY 'NC. W Hep xRY BWDERS amounts," T“ ‘mm; a saw“; at siiiiiiFafillfll“! ABSTRACT CHEMICAL DIFFERENCE FLAME PHOTOMETRY: A NEW APPROACH FOR CLINICAL ANALYSIS BY Betty Lou Schoepke Although there have been great improvements in emission flame photometers over the past years, there still remains a need for increased accuracy and sensitivity in certain areas of clinical analysis. A unique concept for a flame photometer, based upon a system of continuous standardization, has been developed. This system may alleviate some of the inherent problems of emission flame analysis while providing improved accuracy, precision and sensitivity. The feasibility of the Chemical Difference Flame Pho- tometric measurement has been demonstrated and comparisons have been made with other systems which are commercially available. CHEMICAL DIFFERENCE FLAME PHOTOMETRY: A NEW APPROACH FOR CLINICAL ANALYSIS BY Betty Lou Schoepke A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Pathology 1975 DED ICATION To John Morris Schoepke, my companion, my friend, and my lover ii ACKNOWLEDGEMENTS With gratitude and appreciation, the author wishes to thank the multitude of people who have made generous contributions towards the completion of this investigation, as follows: Dr. John F. Holland, Associate Professor, Biochemistry, my research advisor, who so willingly contributed his time, enthusiasm and expertise to direct this investigation. Without his input for a practical solution to a real clinical problem, my graduate career might not have started, let alone have been completed. Dr. C. Cleon Morrill, past Chairman of the Department of Pathology, who provided me with an NIH Traineeship and funds to attend the 1972 Pittsburgh Conference on Analytical Spectroscopy, the highlight of my graduate education. The Education and Research Fund, Inc., of the American Society for Medical Technologists, who awarded me the Clay Adams Research Grant to fund this investigation. Mr. Chester McGlynchey, representative for Scientific Products, who so generously provided me with an IL-343 Flame Photometer for the duration of this study. Mr. Robert Brooks, past Director of the School of Medical Technology, who made funds available for me to attend the 1972 ASMT Convention and two preconvention workshops. iii Mr. James W. Maine, Supervisor of the technical services for the Department of Biochemistry, who along with the men of the electronic and machine shops provided a most congenial environment to learn to use a soldering iron and operate a drill press. Dr. Robert Foy, Technical Director of Laboratories at Sparrow Hospital, who donated a retired Coleman 21 Flame Photometer which somehow ended up in Malawi, South Africa. The Analytical Chemistry Groups of Dr. Stanley R. Crouch, Associate Professor, and Dr. Christie G. Enke, Professor, who pro- vided an organized and enthusiastic learning environment academically and "Cleveburghly." My other committee members, Mrs. Martha T. Thomas, Assistant Professor, for her persistent support and calm guidance; and Dr. John F. Dunkel, Associate Professor, for his willingness to join the committee "at-the-last-minute“ and to limit the number of questions asked during the oral thesis defense. Miss Joann M. Finn and Miss Mary Ann Dietrich for their friend— ship. Also Miss Janice M. Fuller for her expedient and efficient typing ability. Last but most, my husband John, for his continued encouragement towards my professional and personal growth. And my children Shannon and Kyle, whose expressions of pleasure and excitement that mom works with such "neat" things in the lab and finally finished her degree have been most encouraging. iv TABLE OF CONTENTS Page INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . H Flame Emission Theory. . . . . . . . . . . . . . . . . . . Historical Account of Flame Emission Spectrometry. . . . . Typical Instrument System. . . . . . . . . . . . . . . . . The Flame . . . . . . . . . . . . . . . . . . . . . The Photometer. . . . . . . . . . . . . . . . . . . 1 Flame Emission Accuracy, Precision and Sensitivity . . . . 28 common-4 DEFINITION OF STUDY . . . . . . . . . . . . . . . . . . . . . . 32 MATERIALS AND METHODS . . . . . . . . . . . . . . . . . . . . . 36 Experimental I . . . . . . . . . . . . . . . . . . . . . . 36 Solution Preparation. . . . . . . . . . . . . . . . 36 Instrument Operation. . . . . . . . . . . . . . . . 37 Experimental II. . . . . . . . . . . . . . . . . . . . . . 38 Solution Preparation. . . . . . . . . . . . . . . . 38 Experimental III . . . . . . . . . . . . . . . . . . . . . 38 Instrument System . . . . . . . . . . . . . . . . . 38 Sampling System . . . . . . . . . . . . . . . . . . 38 RESULTS AND DISCUSSION. . . . . . . . . . . . . . . . . . . . . 42 Experimental I, II and III . . . . . . . . . . . . . . . . 42 SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 APPENDIX. . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 REFERENCES. . . . . . . . . . . . . . . . . . . . . . . . . . . 53 VITA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 Table 1 2 3 LIST OF TABLES Comparison of properties for two burner types. . Temperatures for various fuel and oxidant mixtures . Wavelength, ionization potential and ionization percent for Na, K and Li in two different flame temperatures . Detection limits for Na, K and Li by emission flame pho tome try 0 O O O O I O O O O O C O O O O O O 0 Recording of data for Experimental I and II. . . Comparison of average error for the SD and CDFP measurements . . . . . . . . . . . . . . . . . . working standard preparation for Experimental I. WOrking standard preparation for Experimental II vi Page 12 l3 16 29 43 45 51 52 Figure 10 11 12 13 LIST OF FIGURES Excitation and emission energy levels for sodium . . . Block diagram for an emission flame photometer . . . . Total consumption burner and sample cup. . . . . . . . Pre-mixing chamber for a Méker burner. . . . . . . . . Flame processes. . . . . . . . . . . . . . . . . . . . Ebert mount monochromator. . . . . . . . . . . . . . . Typical vacuum phototube . . . . . . . . . . . . . . . Typical circuits for photocurrent measurement. . . . . Simulated recorder response for a chemical difference flame photometer measurement . . . . . . . . . . . . . Components of a gravity flow sampling system . . . . . Amplifier schematic for experimental III . . . . . . . Tracing of the CDFP output . . . . . . . . . . . . . . Tracing of the CDFP output for 1 mEq/l sodium difference . . . . . . . . . . . . . . . . . . . . . . vii Page 11 15 19 20 23 34 4O 41 46 47 INTRODUCTION Flame Emission Theory Flame emission spectrometry (flame photometry) is a useful analytical method in many fields and is of particular importance in the area of applied biology and medicine. While more than 70 metallic elements can be evaluated by this method, requests for analysis of the blood electrolytes, sodium (Na+) and potassium (K+) outnumber all others in the clinical laboratory. Although the methods of flame photometry have an inherent sen- sitivity, precision and rapidity, the actual measurement depends upon complex, thermally dependent processes. The thermal energy required for the emission phenomenon to occur is supplied by the exothermic, oxidative reactions occurring within the flame. Normally, at room temperature the electrons of an atom are at their lowest energy level, the ground state. As the sodium atom, symbolized in the ground state as Nao,a is placed in an environment of sufficient thermal energy, the electrons will be raised to an allowed higher energy level, the excited state of which is symbolized * as Na . The process of raising an electron from one energy level to aThis is a simplified version of the notation used by spec- troscopists. An explanation of these symbols can be found in most quantum mechanics texts and in the manual prepared for the American Society for Medical Technology (ASMT) Instrumentation Wbrkshop Program (1) . l 2 another is called excitation. Mathematically, this process is repre- sented as * Na + AB = Na 0 where E symbolizes energy. In the excited state the Na* atom is unstable and immediately loses its excess energy by either a non- radiative or a radiative process. For example, excess energy may be transferred from one atom to another in a sequence of collisions that in essence dissipate the excess energy in the excited atom. This is an example of nonradiative deactivation. Flame spectroscopy utilizes the radiative deactivation process in which the excess energy is given off as electromagnetic radiation. The radiated light has a distinct set of frequencies (wavelengths) specific for the element from which it arises. For example, the light emitted by sodium is yellow with the most intense emission at a wavelength of 589 nm. Wavelength (A) and frequency (v) are related as given in the following equation: A0 = g—- 10 = length of wave in nanometers o c = light velocity v0 = frequency of wave in hertz (Hz) The process whereby an electron passes from a high energy level to a lower energy level by a radiative process is called emission. The mathematical representation of this process is * Na -———->-Nao + AB = NaO + hv 3 where h is a constant derived by Planck (2) in his radiation law which establishes the proportionality between energy and frequency. Relaxa- tion is another term given any process by which an atom loses energy. Figure 1 illustrates the excitation and emission energy levels of sodium. One should note that the ground state can be excited to any permitted excited state as indicated by the energy lines above it. Also, with sufficient energy, as found in high temperature flames, the sodium atom can become ionized. In other words, if the atom is excited enough to lose an electron, the sodium ion (Na+) is left behind. This is an undesirable process for flame photometric measurements as ions have a different relaxation pathways and the spectrum becomes more complex. Usually the emitted wavelengths resulting from transitions between the ground state and the excited state, resonance transitions, are used for analysis whenever possible. Up to this point only one atom involved with one transition has been discussed. Unfortunately, when large numbers of atoms are involved there are many chemically complex and thermally dependent processes taking place at one time. Each of these atoms has numerous allowable transitions back to the ground state. In order to better understand and describe these interrelated processes, various laws have been formulated. Most important of these is the Boltzmann distribution law (2). This law states the relationships of the number of atoms in the excited state [Nao] as a function of the energy difference between the two states. Normally, in the flame only a small portion of the atoms are excited. NH------------E ION NA F El Figure 1. Excitation and emission energy levels for sodium. 5 [Na*1/{Na01 % 1% Keep in mind that this 1% may represent tens of billions of atoms. Each transition which involves a radiative deactivation will emit a discrete wavelength or energy. If these energies are sorted and recorded for visual inspection, the result is the emission spectrum for a particular element and will appear as bright lines against a dark background. The emission spectrum is unique for each metal and forms the basis of qualitative analysis. Consequently, components of mixtures of metals can be identified by interpretation of their characteristic emission spectra. In addition to the allowability of a specific transition, there is a probability function (2) which describes how often that transi- tion will occur. Since allowed transitions do occur with different probabilities, the intensity of various spectral lines for one element will vary. For any specific spectral line, the intensity will vary with the concentration of that metal in the flame. Obviously, this feature forms the basis for quantitative analysis. The intensity of a specific spectral line is dependent upon several factors, the major ones being: 1. Energy of excitation available, i.e., flame temperature 2. Nature of the metal and other species excited 3. Nature of the flame gases 4. Concentration of the metal. For good quantitative analysis it is essential that the variation due to the first three factors be minimized! 6 Historical Account of Flame Emission Spectrometry, Melville (1752) first described the yellow light of sodium in flames. Kirchhoff and Bunsen (1856) correlated the colors of the spectrum with specific emission lines. Their work laid the founda- tion for the use of flames in qualitative analysis. The Bunsen spectroscope was one of the first devices used in an attempt to quantitate the emission phenomenon. Again, visual inspection was required to identify spectral lines while peering through an eyepiece. Then in 1928—30 Lundegardh (3) developed the first photoelectric flame photometer for quantitative analysis in application to medical problems. The first commercial flame photometer was developed by Waibel—Zeiss in 1938 (4). However, it was not until 1945, motivated by the medical needs during World war II, that Barns and co-workers (5) developed the clinical methods of flame photometry as they are practiced today. The reader is referred to the biennial reviews in the Journal of Analytical Chemistry April issue. Alternate years are concerned with applications and instrumentation. Both contain excellent references for flame emission spectrometry. Typical Instrument System Although there are nearly 80 different models of flame photome- ters or flame spectrophotometers available, they all can be related to the block diagram of a typical flame instrument system as depicted in Figure 2. .uoumfiouonm mEmHm :oflmmflEm am How Bayonet xoon .N musmwm mmBmEOBomm WEB middm mmB EMBmMm mmUDDmZ43 w EDOdem mmHhHAm2¢ UZHAQEdm The Flame The sample—excitation system is comprised of an aspirator (sprayer), burner and flame. The aspirator is merely a capillary tube placed in a suitable aqueous sample. It serves as a means of solution transport from the sample cup to the flame. The burner-flame may be one of two types, a turbulent flow (total consumption) or a laminar flow (pre-mix). Figure 3 pictures a total consumption burner. The sample is aspirated (nebulized) and carried into the flame by a Venturi effect created by the oxidant (at A) as it passes over the capillary resulting in a partial vacuum. A distinction is made between aspiration and atomization in flame photometry. Aspiration applies to the mixing of a liquid and a gas or gases which results in a mist. Atomization means the breaking of bonds to form individual atoms. Fuel enters the burner at B and is mixed with the sample mist and oxidant at the orifice of the burner. The turbulence of this system is due to the mixing of the mist and gases, the velocity at which the gases are forced from the burner, and the rate of combustion. As its name implies, the total sample is consumed in the flame. The major advantages of this burner are the small area in which the flame is concentrated, the efficiency with which the spray enters the flame, the high sensitivity attainable, and ease of cleaning. The disadvantages include the production of varying sized droplets which affect vaporization and atomization efficiency, the change in turbulence when spraying the sample which disturbs flame equilibrium, and the flame background which affects the signal to noise ratio to 3 fuel A oxidant Figure 3. Total consumption burner and sample cup. 10 attain reproducible results. The flame turbulence may allow some sample droplets to pass through the flame unaffected. It also affects the mixing of flame gases with the ambient air which may alter the flame temperature, and users may complain about the audible noise created by the burner. Consequently, flame tempera- tures with a total consumption burner often may be lower than that for a pre—mix Méker type burner due to incomplete combustion, mixing of flame gases with ambient air, and spraying of sample into the flame (6) . The pre-mixed flame uses a chamber type nebulizer for mixing of the sample and combustion reactants resulting in a flame with a stable appearance and having considerably less flame background and audible noise. The diagram in Figure 4 shows the pre-mixing chamber for a Méker burner (7). While the air jet nebulizes the sample, the larger mist particles impinge on the chamber wall and are drained away. The remaining, rather uniform, droplets are mixed with the fuel and pass into the flame. The major disadvantage to the chamber type pre-mixed flame is that 90% or more of the aspirated sample goes "down the drain." Also, cleaning is more of a chore than with the total consumption burner. Table 1 from Alkemade (8) summarizes the comparison of properties of the two burner types. The fuel-oxidant requirements are dependent upon the burner type and the element(s) to be analyzed. Table 2 lists a few fuel- oxidant combinations and the temperatures they produce. The most common mixture in clinical flame photometers where sodium and potassium are the elements of interest is propane-air. 11 TO BURNER\ PROPANE AIR DRAIN @ TUBE SAMPLE Figure 4. Pre—mixing chamber for a Méker burner. 12 Table 1. Comparison of properties for two burner types Property Premixed flame Un-premixed flame Structure distinct zones blurred zones Character of gas Noise Mixing with ambient air Homogeneity Equilibration of the flame gas Suitable for gas—mixtures with high burning velocity Spraying efficiency Influence of spraying on flame prOperties Suitable for spraying inflammable solvents Desolvation of drOplets in flame laminar no moderate good fairly good no 1-15% weak no (nearly) complete turbulent yes strong bad fairly good yes 10-100% marked yes (usually) incom- plete 13 Table 2. Temperatures for various fuel and oxidant mixtures Fuel Oxidant Temperature, °C Propane air 1925 Acetylene oxygen 3050 Propane oxygen 2800 Hydrogen air 2100 14 In the burner-flame numerous processes occur as the sample is converted from a spray into an optical signal. If one were to follow a single droplet containing only water and a compound such as sodium chloride as it is carried upward through the flame, the following processes would be observed: NaCl-desolvation of solute Na gas-vaporization of solute Na -atomization Na -excitation hv 589 nm-emission The significance of these processes will be explained later. Alkemade (9) has designed a diagrammatical representation of the flame process listed above, as shown in Figure 5. In addition, many other flame processes, not necessarily advantageous to emission, may occur. Among these are ionization, formation of compounds with flame gases and self-absorption. Table 3 lists the ionization percentages at different tempera- tures for three elements. Ionization is determined by the bond energy and the ionization potential of the element. The amount of ioniza— tion increases with an increase in flame temperature but is a reversible process as given by the equation + - Na —-——+-Na + e o +————- Consequently, if an excess of electrons (e-) can be introduced into the flame the population of ground state atoms will be increased. An excess of electrons will occur upon addition of an element to the sample solution which is more easily ionized than the metal to be 15 ZOHmmHZN ZOMBflNHEOBm .mmmmoooum oEmHh ZOHB¢BZNEG¢Mh WHHI BmUHMW T uorqeroossrq NOIIVLIDXS rf" _ ZOHBHmomZOUHmflJA L m. . a mon o o u u I. s z 3 e m n o o moms, 1 u m. e u mzoem s . S .. o a To 9 .4. To me WWQDUWAOE homomma mums10‘6Q) is used to produce a voltage drop (E ) from the photo- current (i). This voltage is amplified and the:Lused to Operate a readout device. This voltage type of measurement has two distinct disadvantages. One, the anode to cathode potential (ET) changes with photocurrent or light intensity and may cause nonlinear detec- tion characteristics. Two, since the large load resistor is directly across the operational amplifier (0A) input terminals, it may act as an antenna and introduce error in the measurement by electronic noise pickup. In circuit B of Figure 8 a current to voltage operational ampli- fier replaces the load resistor. This system has two advantages: the non-linear detection is eliminated because the potential from the battery (EB) equals the potential across the tube (ET) and is independent of current in the tube; also, since there is virtually no resistance across the amplifier inputs, the scheme is less suscep- tible to noise pickup. In both circuits the readout device which concludes the measurement may be a meter, a recorder, a digital display, or a computer printout. The relationship between the readout and the concentration of an element in solution may be complex. The readout will be propor- tional to sample concentration resulting in a reproducible working 23 J_ l VOLTAGE MEASURING DEVICE hv p j. CURRENT MEASURING DEVICE Figure 8. Typical circuits for photocurrent measurement. 24 curve if and only if the following variables are held constant or are optimized. l. The transport parameters such as the flow rate of sample (influenced by oxidant pressure and solution viscosity), the flow rate of unburned gases, and the expansion of flame gases, must be controlled and the effects of varying the concentration of the element in solution must be minimized. 2. Desolvation and vaporization of the sample is determined by the aspiration and vaporization efficiency of the burner used. These two variables will influence the sensitivity of the system and obviously must remain uniform. 3. Atomization or the fraction of free atoms formed is influenced by numerous processes which will have a great effect on the final readout. These processes were described in the discussion on flame theory. Heiftje and Malmstadt (11) have developed a unique isolated drop technique for pictorial study of the vaporization process leading to the formation of free atoms. 4. Excitation occurs only if the temperature of the flame is of sufficient energy to excite the metal atoms. As described above, the Boltzmann distribution law relates the ratio between the number of free and the number of excited atoms as a function of temperature. Therefore, the thermal excitation process converts a number of free atoms into excited atoms. 5. Emission results from the radiative deactivation of the excited atom to produce a radiance specific for the particular element. The wavelength of this line is directly related to the energy of 25 excitation and usually results from resonance transitions. The radiance of a spectral line is defined as the radiant power emitted per solid angle per unit area. 6. The passage of radiation to an entrance slit of a monochroma- tor or through a filter occurs following emission. Because the flame emits in all directions, only a small portion is observed at the entrance slit. The fraction of the radiant power entering the mono- chromator is determined by the area of the propagating spherical radiation that falls on the entrance slit and the solid angle (numerical aperture) that the monochromator can accept. 7. When considering the passage of light from the entrance slit to the exit slit, one must assume that the monochromator is set for the appropriate wavelength or that a proper filter has been chosen. Since no optical surface is perfect, some radiation may be lost through the filter or, in the case of the monochromator, from the mirrors and grating or prism. The amount of radiant power at the exit slit is dependent upon the efficiency by which the monochromator optics transmit the light, the relationship of the wavelength setting and spectral band pass to the center of the spectral line and a factor determined by the entrance optics. From the exit slit of the mono- chromator the radiation is transmitted to a transducer or photodetector. 8. The sensitivity factor of the photodetector is determined by the response of the cathode to light photons which impinge on its surface. Usually the cathode surface responds to one out of five photons or has an efficiency of 20%. The photodetector is actually a light measuring device which measures different intensities of 26 emission and converts the optical signal into an electrical signal. With appropriate amplification of this electrical signal and an accurate readout device, the relationship between sample concentra— tion and readout will be reproducible. The medical technologist has a choice of commercial flame pho- tometers having a wide range of sophistication, sensitivity and price. Those designed primarily for clinical applications commonly fall into one of two categories: direct reading or internal standard instruments. With the direct reading instrument the readout responds in a manner directly proportional to the radiation intensity of the species being measured. With the internal standard instrument the readout is a ratio between the compound being measured and a com- pound which has been added to serve as a known reference. Although direct reading instruments commonly employed in the clinical laboratory offer sufficient precision for sodium and potas- sium analysis, they are less accurate than internal standard instru- ments. They are influenced more by fluctuations in critical instru- mental parameters and interferences from chemical, spectral or background luminescence and physical variations such as viscosity or burner clogging, the adverse effects of which are minimized by the ratio process employed in the internal standard instrument. In order to minimize changes of various effects in the flame, the blank and standards should be of similar composition as the sample. A perfect match is not possible. As a result, with direct reading instruments, upon aspiration of the sample the flame is greatly perturbed due to the difference in composition between it and the blank and the standards. 27 With the internal standard instrument system the lithium, sodium and potassium signals are measured simultaneously. Electronically the sodium and potassium signals are ratioed to the lithium signal. If a change occurs in an instrumental parameter such as the aspira- tion rate, this ratio process will provide an automatic correction. For example, if the capillary becomes partially clogged, the flow rate will be decreased. This is reflected in a proportional decrease in the lithium, sodium and potassium signals, but the ratio of the signals will remain unchanged. The internal standard method of flame analysis thus has an advantage over direct reading instruments of increased long term readout stability. This enhances both the precision and accuracy of succeeding analysis. While the use of the internal standard method is widely accepted and practiced in the clinical laboratory, one should be aware that it has certain limitations that affect sodium and potassium analysis. First, the spectral lines selected for the internal standard and the element under analysis should be close together due to emission characteristics which require similar flame temperatures and detector response characteristics which vary greatly with energy. Secondly, both the internal standard and the element of interest should have nearly identical ionization potentials so that the energy fluctuations in the flame will have a similar effect on the percent ionized of each species. Also, an error will be introduced if the sample con- tains the same element used for the internal standard. For example, a request for a sodium/potassium determination on a psychiatric patient undergoing lithium therapy would be meaningless unless this medication were made known to the laboratory staff. 28 Lithium has been the element of choice for the internal standard in clinical flame photometry. While it is a satisfactory choice for an internal standard, it has some limitations. Refer to Table 3 for a list of wavelengths and ionization potentials for sodium, potassium, and lithium. Foster and Hume (12) found that a high concentration of sodium enhances the emission intensity of potassium more than that of lithium. Also, they determined that lithium in small amounts is a poor choice as an internal standard for potassium in the presence of elements having low ionization potentials. This is due to the dif- ference in ion equilibrium which is even greater than the difference between the sodium and potassium ion concentrations. Flame Emission Accuracy, Precision and Sensitivity Lowered detection limits (increased sensitivity) and increased precision and accuracy reflect progress in flame emission spec- trometry. These improvements are directly related to instrumental developments and better control of flame reactions which have resulted in a better understanding of the process involved in flame analysis. Generally, detection limits range from 10-5 to 102 ppm for flame emission methods. Table 4 (13) gives a partial list of elements analyzed by flame emission and the best sensitivity obtainable to date. These limits may vary with the type of instrument used, the burner design, the fuel requirements and other factors. Precision for flame emission methods corresponding to a rela- tive standard deviation of 1-2% is usually claimed (14). Greater precision is possible but it usually requires more elaborate tech- niques and instrumentation than used in a clinical laboratory. 29 Table 4. Detection limits for Na, K and Li by emission flame photometry Detection limit Element A, nm (Hg/m1) K 766 0.0002 Li 671 0.000003 Na 589 0. 0001 30 Accuracy of 5% or better can be expected with flame methods. Accuracy is best improved by optimization of operating conditions, reproducibility of the source radiation and stable electronics. In addition, control of systematic and random errors will contribute to improved accuracy and precision. Parsons and Winefordner (15) have written an excellent paper which discusses optimization of critical instrumental parameters. They discuss not only the parameters which F“ can be systematically varied but also the importance of those which contribute to obtaining the lowest detectable concentrations and the |" most precise analysis. Instrument manufacturers have attempted to better meet the needs of the clinician and the medical technologist by providing attrac~ tively packaged instruments, more easily used because of their simplicity and "push button" operation and more convenient readout such as digital displays. In addition, many of these instruments are adaptable to an on line automated or semiautomated system. Although these newer instruments reflect greater efficiency in data acquisition and handling than older models, they may not be playing as significant a role in diagnosis as possible. Subtle changes in the normal versus diseased state of the individual have been noted. For instance, narrowing of the normal value range for a given analysis increases the need for better sensitivity, accuracy and precision. Also, a physician may find diagnosis more easily made by comparing the laboratory findings of repeated samples from a patient rather than by comparing them to normal values set by the population in general. In other words, laboratory results may be more significant 31 if they compare an individual's present condition with his past profile rather than by comparison to the "normal" population. The detection of an incipient disease may be possible based upon small changes that result in a value that usually would not be observed using the con— ventional normal range concept. These measurements will require greater accuracy than is available with today's instrumental methods. The Standards Committee of the College of American Pathologists (CAP) compared the accuracy obtainable with the accuracy necessary for a clinical test value to be medically significant (16). Among those tests not providing a medically significant level of accuracy was that of sodium. Winstead (17) calculated the present performance as a relative error of :_4% for a sodium value of 150 mEq/l when it was shown by the CAP to require a relative error of less than i_2.6%. Performance of tests for potassium was found to meet the needs of the clinician for medically significant results under the existing practice of normal comparison. Winstead also discusses the rela- tionship between fluctuating flame conditions and variations of sample values. Because flame conditions are not constant, the wide variations in results of flame analysis cannot be attributed to systematic error differences in methodology, which is the case in most other clinical laboratory procedures. Therefore, the need for constant monitoring of flame conditions to produce a more accurate measurement is emphasized. DEFINITION OF STUDY The purpose of this study is to demonstrate the feasibility of a unique measurement system for a clinical flame photometer. This study may provide information to alleviate some of the inherent problems of flame emission analysis while providing improved sensi- tivity, precision and accuracy. The significant features for this measurement system are: l. The concentration of the reference (blank) solution closely approximates the normal electrolytic values of blood serum. 2. A steady state of flame operation. 3. A difference measurement between the sample and the reference solution. 4. The state of the art electronics. 5. A visual or signed presentation of data for com— parison to normal values. The reference solution approaches the "ideal“ blank because it is more similar in composition to the diluted sample than is a water-Sterox SE or a water-lithium blank. Also, upon aspiration of the sample the flame will be less perturbed due to the similarity between the sample solution and the reference solution. In this case the instrument blank is not zero but is the normal value expected for each electrolyte undergoing analysis. 32 33 A steady state of flame operation is attained by continual aspi- ration of the reference solution at a constant flow rate. The con- stant flow of operation is interrupted only by aspiration of the sample at regular intervals and then returned to the reference state. In this manner there is an immediate check of flame irregularities. Actually the system is under conditions of constant standardization rather than occasional standardization as is the case with other measurement systems. By constantly monitoring the reference solu— tion, one should realize improved accuracy and precision for flame emission analysis. A chemical difference measurement for this system implies that the measurement is either a positive or a negative deviation from the normal electrolyte concentration represented by the reference solution. While in Operation, the instrument base line or zero is set with the reference solution. Upon aspiration of a sample, the automatic zeroing is suspended, a signal greater than, the same as or less than the reference is recorded. At all times except when aspirating a sample, the measurement system automatically adjusts to the reference or zero point for each element under consideration. Figure 9 pictures the measurement as it would appear on a chart recording. The "a" represents the recording of the reference solu- tion (zero) while "b" and "c" identify the upper and lower limits, respectively, of the normal range for each element. The differen- tial type of measurement generally produces a better signal to noise ratio (S/N) than do conventional methods of amplification and offsetting. 34 Na Ca na’m Figure 9. Simulated recorder response for a chemical difference flame photometer measurement. 35 The maintenance of a high S/N ratio is possible becuase of the "state of the art" electronic components used. Amplification of the small differential signal imposes no particular difficulties. Full scale expansion for the normal range is possible providing increased sensitivity over conventional systems. Visual data presentation on chart paper provides comparison to values within and outside the "normal" range. Also, the values established as a decision level for a particular electrolyte can be printed or color coded on the chart paper. This system of data presentation provides both the technologist and the physician with an immediate evaluation of a patient's electrolyte status. In addition, this measurement technique is adaptable to an on-line automated mode for either analog or digital instrument systems. What is gained through the use of the proposed measurement system for flame photometric analysis? 1. An apparent stable operating condition for the flame for increased accuracy. 2. The sensitivity of the system is greatly improved by amplification of the readout. 3. A chemical differentiation system which is much more sensitive to minor differences in the electrolyte concentration than are the usual measurement techniques. It is the combination of l, 2 and 3 above which contributes to improved accuracy, sensitivity and precision for clinical flame analysis when using the proposed measurement technique. MATERIALS AND METHODS Two types of experiments were completed to compare the tradi- tional measurement of sodium, hereafter referred to as the single deflection (SD) method with the chemical difference flame photometry (CDFP) measurement of sodium. In addition, a third type of experi- ment was made to eStablish the feasibility of a CDFP measurement. Experimental I Solution Preparation In the first group of experiments stock sodium standards were prepared according to the Model 21 Coleman flame photometer manual,a pages 20 through 24, and did include the appropriate concentrations of potassium and calcium. The working standard concentrations of 136, 140 and 144 mEq/l sodium were prepared according to the volumes given in the Appendix. The diluent used was 15 mEq/l lithium prepared according to the manual for the Instrumentation Laboratory (IL) model 343 flame photometerb on pages 14 and 15. The lithium was added in order to make the standards compatible with both instruments. aColeman Instruments, Inc., 42 Madison St., Maywood, IL. bInstrumentation Laboratory, Inc., 113 Hartwell Ave., Lexington, MA 02173. 36 37 Instrument Operation Both the Coleman and the IL—343 instruments were operated and standardized according to the directions given in their respective manuals. The Coleman Model 21 Flame Photometer is a well known direct reading filter instrument which may be added as an accessory to the Coleman Jr. spectrophotometer. Standards prepared from dried reagent grade chemicals and samples are diluted 1:200 with a solu— tion of 0.02% sterox SE which serves as a wetting agent to reduce surface tension. Standardization is accomplished by setting the reference state (zero) with a water-sterox blank. The upper limit is usually set on a precalibrated scale with the 150 mEq/l sodium standard. The standards and diluent for the IL-343 Flame Photometer are prepared from commercial solutions. Both standards and samples are diluted with a 15 mEq/l lithium solution. Standardization is accomplished by setting the lithium response in the reference state to zero while aspirating a water-lithium blank. The upper limit is set at 140 mEq/l sodium and 5 mEq/l potassium while aspirating the standard solution containing the above-stated concentrations of both elements. Following the initial standardization step, neither instrument was restandardized throughout the duration of data collection from the above prepared standards. In some instances the instruments were under continuous operation for two hours. 38 Experimental II Solution Preparation The second type of experiments is identical to the first type with one exception. In this group of experiments new standards were prepared in the manner presented in the IL—343 manual, page 15. The IL stock standard for 140 Na/S K mEq/l was appropriately diluted to make working standards of 136, 140 and 144 mEq/l sodium as given in the Appendix. Both instruments were operated as previously described. Data were taken from both instruments as in the first group of experiments. Experimental III Instrument System The third type of experiment was attempted to determine the feasibility of a CDFP measurement and required the fabrication of a new instrument and sampling system. The instrumental setup utilized only the burner and the detector from the Coleman flame photometer. In addition, an amplifier was attached to provide a zero offset, a gain and a sensitivity control. Sampling System A simple but effective gravity flow sampling system was fabri- cated to provide a positive pressure at the burner capillary. The components of the sampling system are identified in Figure 10. A Sargent SR chart recorder was used for the readout. The amplifier circuit is presented in Figure 11. 39 The IL stock standard for 140 Na/S K mEq/l was used to prepare working standards in concentrations of 136, 140 and 144 mEq/l sodium. The 140 mEq/l sodium standard was used as a background emission against which the output was continuously aspirated between the samples represented by the 136 and 144 mEq/l sodium standards. 40 o a 136 mEq/l I or ~—”’¢' 2-way valve Figure 10. Components of a gravity flow sampling system. .HHH HoucoEfiummxo you oflumfionom HoHMHHmE< .HH ousmflm + zoo. .a.....\ .2. >8H>H9Hmzmm h o. .. a $ I_I Zn W 41 RESULTS AND DISCUSSION Experimental I, II and III During the first two types of experiments data were recorded sequentially, at 30-second intervals during a 1- to 2-hour period as shown in Table 5. Mathematical comparisons between the SD and the CDFP measurements were calculated to indicate the amount of improve- ment of a continuously updated standardization over the SD measurement as follows: Single Deflection Measurement: For this calculation refer to the first group of bracketed figures in Table 5. 1. Determine the average 2. A. 136.5 B. -136.0 +0.5 3. Calculate the average Chemical Difference Flame 1. Determine the average Determine the average B = 140.5. value of B = 140.5. Subtract the known value from each recorded value: 140.5 C. 145.0 -l40.0 -l44.0 +0.5 +1.0 error for A, B and C. Photometer Measurement: of B for the first two B's = 140.4. of B for the second and third Determine A and C when calculated as a negative or positive deviation from their respective averaged B. As an example: 42 43 Table 5. Recording of data for Experimental I and II Sample Reading (mEq/l Na) **B 140.5 *A 136.5 B 140.3 ***c 145.0 Flu—— L__£: 140.7 A 136.5 8 141.4 c 145.4 B 140.6 * Represents the 136 mEq/l sodium standard. ** Represents the 140 mEq/l sodium standard. *** Represents the 144 mEq/l sodium standard. 44 For A B = 140.4 A = 136.5 A should = (E'- 4.0) = 136.4 But A = -136.5 the error for A = -0.1 For c 5': 140.5 c = 145.0 c should = (E'+ 4.0) = 144.5 But C = -l45.0 the error for C = -0.5 4. Calculate the average error for A and C. A summary of the calculated average errors from both instruments for the two sets of standards and the two types of measurements is presented in Table 8. These figures indicate the improved precision that may be obtainable with the CDFP measurement. An output of a recording from the third type of experiment is shown in Figure 12. Figure 13 pictures the responses between a 140 and a 141 mEq/l sodium solution using the CDFP instrumental setup. In all cases the CDFP average error was approximately an order of magnitude better than that for the SD method of measurement. These results represent the worse case situation in operation of an emission flame photometer. However, for acceptable accuracy with a SD measurement, intermittent standardization is required along with the assumption that there is a negligible change in flame character- istics and flow rate. During the long running time, instrumental or base line drift contributed to a larger average error for the SD 45 Table 6. Comparison of average error for the SD and CDFP measurements Average Error Experi- SD CDFP mental Instru- 136 mEq/l 144 mEq/l 136 mEq/l 144 mEq/l group ment sodium sodium sodium sodium I Coleman 21 8.74 8.89 0.74 0.63 I IL-343 3.45 3.88 0.47 0.74 II Coleman 21 3.87 4.50 0.73 0.65 II IL—343 5.75 6.38 0.51 0.54 46 ova ©MH .uomuoo ammo mop mo moflomua ova. ova .NH wusmflm mz H\wme Goa 47 141 140 1 mEq/l Na Figure 13. Tracing of the CDFP output for l mEq/l sodium difference. 48 measurement. However, immediately after standardization the error was greater for SD than for CDFP. The instrumental response to the small difference between the reference standard and the sample is a more accurate measurement than the comparison of the large difference between a blank and a standard or sample solution. The most exciting feature of operating a flame photometer in a differential mode is the apparent increase in sensitivity as repre— sented by the curves in Figures 12 and 13. The signal response from the Coleman 21 instrument was only a difference of approximately three chart divisions between the 140 and the 136 or 144 mEq/l sodium standards when measured and recorded in the conventional manner. The recording in Figure 12 represents a difference of 20 chart divisions for the same standard solutions. The response for a difference of only 1 mEq/l of sodium in Figure 13 is approximately 5 chart divisions. The negative excursions between successive solutions in Figures 12 and 13 (at the *) were produced at times of changing from one sample to another and resulted from the crudely constructed sample intro- duction system. It is clear that a more efficient sampling system should remove these artifacts. In turn, the resulting output would provide a fail-safe response in regard to the direction the sample concentration differs from the standard background solution. Although the results of this study have shown that CDFP measure- ment to be feasible, there remain areas in need of further develop- ment. Both a positive pressure reference solution introduction and a rapid automatic sample injection need to be developed. A constant flow rate during continual aspiration of the reference solution should 49 provide a more steady flame. There is an immediate check on flame fluctuations and an automatic adjustment to the reference point when using a rapid sample injection and quick return to the reference point. In addition, the time needed for sample response could be shortened. Also, the development of a micro-burner requiring small sample volumes would be an advantage to the clinical laboratory. SUMMARY In summary, it has been found that Chemical Difference Flame Photometry presents an alternate method to flame analysis that has the potential to provide greater accuracy and sensitivity. In addi- tion, it presents the capability to provide a more fail-safe method for determining the relationship of individual samples to the expected normal values. Although the research to date has been done only with sodium, simultaneous multiple analysis is not pre- cluded by any of the theory or hardware. The greatest benefit to the clinician and the patient is the potential to provide accuracy at the decision levels of sodium and other electrolyte concentra- tions. The CDFP approach is not limited to only emission or atomic absorption flame techniques but should apply to any continuous flow analysis. 50 APPENDIX 51 Table A-1. Working standard preparation for Experimental I WOrking Stock Stock Stock 1% standard 25 mEq/l Na, 1 mEq/l Ca, 1 mEq/l K, Sterox mEq/l Na (m1) (ml) (ml) (ml) 136 56.0 50 50 40 140 57.6 50 50 40 144 54.4 50 50 40 Each was prepared in a 2 liter volumetric flask and brought to volume with a 15 mEq/l lithium diluent. 52 Table A-2. WOrking standard preparation for Experimental II Working Stock IS standard 140 mEq/l Na, 1% Sterox mEq/l Na (ml) (ml) 136 4.86 20 140 5.00 20 144 5.14 20 Each was weighed and brought to volume in a 1 liter volumetric flask with a 15 mEq/l lithium diluent. REFERENCES 10. 11. REFERENCES Jeffers, D. M., Goode, S. R., and Timnick, A.: Atomic Spec- troscopy, Flame Emission Spectroscopy, Flame Photometry; Principles, Instrumentation, and Trouble Shooting. pp. 1-6 (1973), ASMT, 5555 W. Loop S, Bellaire, TX 77104. Daniels and Alberti. Physical Chemistry, 3rd edition, John Wiley and Sons, 1966. Lundegardh, H.: Die Quantitative Spectralanalyse der Elemente, Ed. I. and Bb. II, G. Fischer, Jena, 1929 and 1934. Waibel, F.: Uber Optische Methoden Zur Untersuchung des Ackerbodens. Z. f. techn. Physik., 19, 394 (1938). Barnes, R. 3., Richardson, D., Berry, J. W., and Hood, R. L.: Flame Photometry, a Rapid Analytical Procedure. Ind. Eng. Chem., Anal. Ed., 17, 605 (1945). Mavrodineanu, R., Ed.: Analytical Flame Spectroscopy. N. V. Philips' Gloeilampenfabrieken (USA Distributors, Springer- Verlag, Inc., New York). Chapt. 1, "From Sample to Signal in Emission Flame Photometry: An Experimental Discussion", by C. Th. J. Alkemade, p. 7 (1970). Instruction Manual fer IL 343 Flame Photometer. Instrumenta- tion Laboratory, p. 19. Mavrodineanu, R., Ed.: Analytical Flame Spectroscopy. N. V. Philips' Gloeilampenfabrieken (USA Distributors, Springer- Verlag, Inc., New York). Chapt. 1, "From Sample to Signal in Emission Flame Photometry: An Experimental Discussion”, by C. Th. J. Alkemade, p. 8 (1970). Ibid, p. 3. Ewing, G.: Instrumental Methods of Analysis. McGraw-Hill Book Co., New York, p. 32 (1969). Heiftje, G. M., and Mamlstadt, H. R.: A Unique System for Studying Flame Spectrometric Processes. Anal. Chem., 40, 1860 (1968). 53 12. 13. 14. 15. 16. 17. 54 Foster, W. H., Jr., and Hume, D. N.: Mutual Cation Interference Effects in Flame Photometry. Anal. Chem., 31, 2033 (1959). Mavrodineanu, R., Ed.: Analytical Flame Spectroscopy. N. V. Philips' Gloeilampen Fabrieken (USA Distributors, Springer- Verlag, Inc., New York). Chapt. 2, "Sensitivity, Detection Limit, Precision and Accuracy in Flame Emission and Atomic Absorption Spectrometry", by O. Menis and T. C. Rains, p. 50 (1970). Ibid, p. 52. Parsons, M. L., and Winefordner, J. D.: Optimization of Crit. Inst. Parameters. Applied Spectroscopy, 21, 368 (1967). Barnett, R. N.: Medical Significance of Laboratory Results. Amer. J. Clin. Path., 50, 671 (1968). Winstead, M.: Instrument Check Systems. Lea and Febiger, Chapt. 5, "Flame Photometers", p. 163 (1971). VITA The author was born in Benton Harbor, Michigan, on June 11, 1938. She graduated from Michigan State University with a B.S. degree in Medical Technology following completion of an internship in Medical Technology in 1963. She interned at St. Mary's Hospital in Grand Rapids, Michigan, where she remained on the staff until July 1965. She was employed as a research technologist for the Department of Animal Husbandry at Michigan State University from August 1965 to September 1970. She was admitted to the graduate program in Clinical Laboratory Science at Michigan State University in September 1970. During her graduate studies she was granted an NIH traineeship from the Department of Pathology, a research grant from Clay Adams, Inc., and American Society for Medical Technology Education and Research Fund and was a graduate assistant in the undergraduate pathology teaching laboratories. In addition she was a primary organizer and coordinator for developing a series of Instrumentation workshops for ASMT. Presently she is a technologist for the Clinical Toxicology Laboratory, Department of Pharmacology, Michigan State University. The author is happily married to John Morris Schoepke. She has a daughter, Shannon S. Bradley, and a son, Kyle N. Bradley. 55