`
`Aluminum in Parenteral Products: Analysis, Reduction, and Implications
`for Pediatric TPN
`
`BARRETTE. RABINOW.t., SCOTT ERICSON, and TIM SHELBORNE
`
`Baxter Healthcare Corporation, Round Lake, lllinois
`
`ABSTRACT: Performance and cost of the best available analytical methodology for the measurement of
`aluminum in parenteral products are presented. Typical levels of aluminum in representative solutions are
`summarized. A methodical approach to the minimization of aluminum contamination in the manufacturing
`process is considered in light of aqueous aluminum chemical considerations. Results of long-term clinical
`follow-up studies of infants maintained on currently manufactured TPN solutions indicate no adverse
`pathology arising from aluminum.
`
`the variation among the means was attributed to analyti(cid:173)
`cal problems.
`The analytical methodology has evolved extensively
`since that time. Because of the essential role that the
`analysis plays in implementation of an enforceable stan(cid:173)
`dard, it is well to briefly review the improvements made to
`overcome the wide variability shown by previous investi(cid:173)
`gators during the last decade. The technique that most
`commends itself for our purposes is graphite furnace
`atomic absorption spectrometry. It is specific, can be
`made sufficiently precise, and is automatable. In this
`methodology, Figure 1, a liquid drop of sample is heated
`at various stages, to successively dry the sample, char off
`the organic interferences, and char off the low boiling
`inorganic salts. Finally, the temperature is sharply raised
`to volatilize the aluminum in the sample and atomize it, so
`that it intercepts a beam oflight at the specific wavelength
`of aluminum. This results in the atomic absorption pro(cid:173)
`cess that leads to estimation of mass by reduction of light
`intensity (2).
`
`Atorn,ze
`
`Clean-out
`
`Char
`(inorganic)
`
`3000
`
`2000
`
`TEMP.
`(°C)
`
`1000
`
`Ory
`
`This paper will attempt to present an integrated under(cid:173)
`standing of a number of seemingly unrelated facets of the
`issue concerning levels of aluminum in parenteral solu(cid:173)
`tions. The strides made in the analytical methodology,
`responsible for making routine measurement possible, will
`be shown as will the cost to implement this technology in a
`quality control setting. The aqueous chemistry of the met(cid:173)
`al will be examined to understand the complexity of the
`reactions involved. This is necessary to effectively trouble(cid:173)
`shoot a manufacturing process for the purpose of mini(cid:173)
`mizing the aluminum levels in the product. A survey of
`typical values of aluminum to be expected in some repre(cid:173)
`sentative solutions will be shown. Finally, the results of a
`long-term follow-up study involving over a hundred pedi(cid:173)
`atric patients maintained on TPN over a ten-year period
`will be presented in an attempt to evaluate the likelihood
`of demonstrable pathology accruing from TPN admix(cid:173)
`tures, as currently compounded.
`
`I. Quantitative Methods
`
`Until relatively recently, the acquisition of accurate
`and precise determinations of aluminum was considered
`something of an art. Versieck and Cornelis ( l) have re(cid:173)
`viewed the literature on the estimation of aluminum in
`human plasma or serum. The 17 reviewed papers had been
`published between 1960 and 1979 and showed mean alu(cid:173)
`minum concentrations ranging from 3.72 to 1460 ppb.
`Since electrolyte levels in humans are regulated within
`narrow limits by the kidneys and other systems, most of
`
`[EDITOR'S NOTE: This is the fourth in a series of four presentations
`on "Aluminum in Parenteral Products" from the PDA Annual Meeting,
`Chicago, IL, October 1988. They are being published as a group in this
`issue.]
`Received November 23, 1988. Accepted for publication March 3,
`1989.
`"" Author to whom correspondence should be addressed: Barrett E.
`Rabinow, Ph.D., Baxter Healthcare Corporation, Baxter Technology
`Park (WG3-25), Route 120 and Wilson Road, Round Lake, IL 60073.
`
`0
`
`25
`
`50
`
`75
`100
`TIME (sec)
`Figure 1-Typical graphite furnace atomic absorption heating pro(cid:173)
`gram.
`
`125
`
`150
`
`132
`
`Journal of Parenteral Science & Technology
`
`Eton Ex. 1071
`1 of 8
`
`
`
`~NNTA~:;:Rl~~s~:.......... ............. _..l[. .................. -
`. r··
`
`ABSORBANCE
`
`"
`
`\.: -
`
`Temperature
`
`Absorbance
`
`(C)
`
`Time
`
`Figure 2-Furnace temperature (left) and absorbance (right) vs. time
`profiles, characterizing several modes of instrumental per(cid:173)
`formance: (a), slowly heated furnace tube; (b), slowly heat(cid:173)
`ed furnace tube with L'vov platform; (c), rapidly heated
`furnace tube.
`
`The problems suffered by earlier investigators occurred
`because the heating profiles shown in Figure 2 for the
`furnace temperature were too slow in attaining a plateau
`of stable instrumental performance. The sample alumi(cid:173)
`num atomic cloud, formed during the rapidly changing
`furnace temperature, resulted in large variability of the
`results. The problem was solved in two ways. The heating
`· rate of later model furnaces was increased so as to move
`the onset of stable operating temperature earlier in the run
`cycle. Additionally, the atomization of the sample was
`delayed as shown in Figure 2. The sample is deposited on
`top of a L'vov platform, Figure 3, rather than on the
`bottom of the furnace tube which is what is directly heat(cid:173)
`ed. This delays onset of volatilization until the establish(cid:173)
`ment of equilibrium instrumental conditions (3).
`But the major improvement occurred in background
`correction capability. As depicted in Figure 4, despite
`attempts to remove the matrix earlier in the heating pro-
`
`1500°C .....:.------
`
`l&J a:: ::,
`I-"' a::
`
`l&J
`0..
`::::IE
`l&J
`I-
`
`TIME-
`Figure 4-lnterfering absorbance superimposed upon heating profile.
`
`gram, nonspecific absorption will occur during atomiza(cid:173)
`tion due to residual interferences. Earlier optical designs
`permitted too broad a wavelength range around the absor(cid:173)
`bance peak to enter the monochromator. Also, they cer(cid:173)
`tainly could not differentiate between absorption due to
`aluminum and that due to extraneous background absorp(cid:173)
`tion occurring at the apex of the aluminum peak. This
`capability was accomplished with the introduction of the
`polarized Zeeman technique (2).
`As shown schematically in Figure 5, a magnetic field is
`placed around the furnace tube. This causes the absorp(cid:173)
`tion line of aluminum to split into several component lines,
`separated by wavelength. At the same time, the absorp(cid:173)
`tion lines are now polarized; they will absorb light only if
`(in addition to a wavelength match) the light is similarly
`polarized in the same direction with respect to the direc(cid:173)
`tion of the magnetic field. For example, the central line in
`Figure 5 will absorb only light which has its electric vector
`polarized parallel with respect to the magnetic field. To
`take advantage of this, the unpolarized light from the
`analyzing lamp is split into components polarized parallel
`and perpendicular to the magnetic field, by interposing a
`rotating polarizer in the optical path. Only the parallel
`component is absorbed by the sample aluminum, while the
`nonspecific background will absorb both components
`
`Figure 3-L'vov platform.
`
`Vol. 43, No. 3 / May-June 1989
`
`Figure 5-Schematic diagram of polarized inverse Zeeman graphite
`furnace atomic absorption spectrometer.
`
`133
`
`Eton Ex. 1071
`2 of 8
`
`
`
`TABLE I. GFAAS-Aluminum Assay Validation Composite
`Average for Days 1-3
`Cone. Precision % Recovery
`9.9 ppb Al
`CV(%)
`Sample ppb Al
`
`Instrument
`
`Zeeman 5000
`
`Model603/2100
`
`3.66
`4.7
`1
`6.07
`2.5
`2
`5.3*
`1
`l l.7
`39.3
`2.8*
`2
`* Spike recoveries used to correct sample concentrations.
`
`100.6
`102.7
`63.3
`62.9
`
`equally. This is the principle behind this enhanced back(cid:173)
`ground correction technique (2).
`The improved performance, attendant with polarized
`Zeeman capability, is apparent in Table I (4). (Full ex(cid:173)
`perimental details are found in the reference cited.) Levels
`of aluminum in Dianeal® CAPD solution are analyzed by
`two graphite furnace atomic absorption instruments, one
`with and one without Zeeman (but with deuterium) back(cid:173)
`ground correction. Without Zeeman, a recovery of only
`60% is found for a 10 ppb aluminum spike intentionally
`added to the sample. With Zeeman capability, the recov(cid:173)
`ery is increased to 100%. Techniques that have low recov(cid:173)
`ery usually have variable recovery as well, which leads to
`high coefficients of variation. This is the case here, as a
`C.V. of almost 40% is found for the conventional instru(cid:173)
`ment, but a respectable 6% is found for the Zeeman in(cid:173)
`strument.
`The method developed in our laboratory, Table II, fea(cid:173)
`tures acidification of the sample in nitric acid, and utiliza(cid:173)
`tion of magnesium nitrate, a matrix modifier. This makes
`for a more forgiving method in terms of tolerating widely
`varying sample types ( 5). The performance of the method
`when applied to the analysis of four types of parenteral
`
`TABLE II. Polarized Zeeman Graphite Furnace Atomic Ab(cid:173)
`sorption Spectrophotometric Procedure and Pro(cid:173)
`gram
`
`Sample preparation: Dilution with 4% HNO3/0.4%
`Mg(NO3h
`Instrument = Perkin-Elmer Zeeman 5000AAS with HGA-500
`Graphite Furnace and AS-40 Autosampler
`External Standards= 0, 5, 10, and 20 ng/mL in 4% HN03/
`0.4% Mg(NO3)2
`
`Sample Size = 20 µL
`
`Tube Type = Carbon pyro with
`L'vov Platform
`
`Wavelength= 309.3 nm
`Mode= Abs
`Int. Time = 4.0 sec
`
`Slit= 0.7 nm
`Current = 25 mA
`Signal = Peak Area
`
`Step
`
`Temperature
`(QC)
`
`Ramp Time
`(s)
`
`Hold Time
`(s)
`
`1
`90
`2
`100
`3
`500
`4
`1500
`5 (Read)
`2550
`6
`2600
`7
`20
`Ar Int. Gas Flow= 300 cc/min except for Steps I and 5.
`
`10
`10
`30
`20
`0
`
`10
`25
`10
`10
`5
`4
`5
`
`TABLE III. Comparison of Baxter and Alfrey Methods for
`Determination of Aluminum in Parenteral Solu(cid:173)
`tions
`
`Alfrey
`Baxter
`Mean C.V. Within-Unit Mean C.V.
`(µg/L)
`C.V (%)
`(µg/L)
`(%)
`(%)
`
`5.4
`14.5
`280.0 14.5
`47.8
`6.0
`
`0.4
`2.5
`1.3
`
`13.3 19.4
`317.0 12.2
`50.7
`2.8
`
`<I
`
`<2
`
`10% Travamulsion
`25%Albumin
`Heparin sodium
`(1000 U/mL)
`5% Dextrose
`
`solution is shown in Table III. Also shown for comparison
`are the results of Dr. Allen Alfrey, to whom we sent a
`number of samples from the same lot of each of the four
`solutions. The mean values found by both methods agree.
`The coefficient of variation, representing container to
`container variability is shown in the first set of parenthe(cid:173)
`ses for the Baxter data, and is compared to that for the
`Alfrey method. The second set of parentheses for the
`Baxter data indicates C.V. for repeated determinations
`from a single container, and is indicative of intrinsic vari(cid:173)
`ability in the instrumental and sample preparation.
`These data are comparable to those reported by Koo,
`Table IV (6), who found that LVP's typically contain less
`than 50 ppb and that most of the problem occurs for
`phosphate salts, both sodium and potassium, as well as for
`gluconate. It is clear that in order for any method to be
`useful in determining aluminum levels in clinically impor(cid:173)
`tant solutions, validation work should be designed with
`these matrices in mind. As part of work performed on
`behalf of a PDA task force to evaluate analytical method(cid:173)
`ology for aluminum in parenteral solutions, we compared
`the performance of the method developed in our laborato(cid:173)
`ry to that of one submitted by Dr. Ted Rains of the
`National Institute of Standards and Technology, Table V.
`Phosphate containing solution, as well as amino acids,
`with and without electrolytes, a simple electrolyte solu(cid:173)
`tion, and heparin were subjected to the inter-method
`study. As it was originally submitted, the NIST procedure
`
`TABLE IV. Sources of Aluminum Contamination in Paren(cid:173)
`teral Nutrition Solutions
`
`Product
`
`Number Tested
`Sample/Lot/Mfg.
`
`Al
`(ug/L)
`
`Sterile water
`Dextrose water (5-50%)
`Cryst. amino acids (5-10%)
`Soybean oil emulsion
`S<Xiium chloride
`Sodium acetate
`Sodium phosphate
`Sodium lactate
`Potassium chloride
`Potassium phosphate
`Calcium gluconate
`Calcium chloride
`Magnesium sulfate
`
`7/7/3
`13/13/3
`17/17/3
`4/4/1
`
`3/3/1
`1/1/1
`7/5/2
`2/2/1
`9/9/5
`4/4/2
`11/11/5
`5/4/3
`5/5/3
`
`<5
`<5
`<5-47
`<5
`<5-5
`<5
`<5-2370
`184
`
`<5-17
`90-2300
`1100-5600
`5-19
`
`<5-5
`
`134
`
`Journal of Parenteral Science & Technology
`
`Eton Ex. 1071
`3 of 8
`
`
`
`TABLE V. Comparison of Baxter and NIST Methods for Determination of Aluminum in Parenteral Solutions
`NIST
`Method of Addition
`(µg/L)
`
`Solution
`
`Mean
`
`Baxter
`(µg/L)
`
`16.0
`
`8.5% Travasol® inj.
`with electrolytes
`8.5% Travasol® inj.
`without electrolytes
`Plasma-Lyte® solution
`1.5
`Heparin lock
`<0.4
`Sodium phosphate
`250.0
`* Dilution factor increased to improve reproducibility.
`
`10.5
`
`SD
`
`1.5
`
`1.7
`
`1.4
`
`1.7
`
`Mean
`
`21.0
`
`9.7
`
`2.4
`<3.6
`302.0
`
`SD
`
`0.2
`
`4.5
`
`4.2
`
`60*
`
`NIST
`External Standards
`(µg/L)
`
`Mean
`
`18.6
`
`9.5
`
`2.3
`<3.6
`193.0
`
`SD
`
`1.2
`
`4.2
`
`4.0
`
`14*
`
`featured a method of addition technique. This is fine for
`imparting a robustness to the method in terms of matrix
`compensation for a wide variety of different solution
`types, but it makes for lower sample throughput. One
`must assay not only each sample, but each sample inten(cid:173)
`tionally spiked with a known addition of aluminum at
`several different levels. To evaluate the capability of the
`NIST method to be run faster, we compared its perfor(cid:173)
`mance utilizing external standards. Comparable results
`were found for the mean values of all three methods for
`the first four solutions, although the NIST methods have
`a somewhat high CV at 4% for repeated injections.
`Irreproducible results were found with the NIST meth(cid:173)
`ods when applied to the phosphate solution. We suspected
`that the acceptable performance from the Baxter method
`arose from the incorporation of the matrix modifier
`MgNO3 in the sample preparation. This reduces sensitiv(cid:173)
`ity of the method to phosphate in the sample matrix (5).
`As stated by Slavin et al., the mechanism of reduction of
`interference effects is attributable to imbedding of the
`aluminum in a matrix of magnesium oxide. This delays
`vaporization of the analyte until the magnesium oxide is
`vaporized. In any event, lacking this modifier in the sam(cid:173)
`ple dilution step, the NIST procedures incurred wide vari(cid:173)
`ability. We attempted to mitigate this problem by increas(cid:173)
`ing the sample dilution factor, in an attempt to dilute out
`the phosphate interference. This was somewhate success(cid:173)
`ful, in that the means now found by the NIST methods did
`not differ statistically from that of the Baxter method,
`although a standard deviation of 60 ppb resulted. The
`lesson here is that even though a method appears to work
`well for the determination of aluminum in four sample
`matrices, there is no guarantee that it will work on the
`fifth.
`Assay variability is particularly a problem with meth(cid:173)
`ods operating at the trace level. Sample contamination,
`adsorbtion to the walls of the sample container, and limi(cid:173)
`tations in sensitivity of the analytical method become
`more onerous as the analyte concentration is decreased.
`To underscore this are the results of an analysis of fifty
`interlaboratory round robin studies conducted by the As(cid:173)
`sociation of Official Analytical Chemists (7), shown in
`Figure 6. Each study involved at least twenty laboratories.
`All different kinds of analytical methods as applied to
`many types of analytes were studied. The major factor
`leading to assay variability was identified as the concen-
`
`tration level of the analyte sought. Precision for the major
`ingredient of a dosage form poses no problem. But as the
`concentration of the substance sought decreases, the inter(cid:173)
`laboratory coefficient of variation increases, following a
`power law. At the ppb level, a C.V. of 40% was found. It is
`clear that especially at trace levels, a careful validation of
`the analytical method is essential.
`The large interlaboratory variability found by the
`AOAC at the ppb level would seem to be at odds with the
`rather close agreement among aluminum levels reported
`in Tables III and V of this work. The minimal interlabora(cid:173)
`tory variability found here is attributable to the select
`nature of the three laboratories whose work is presented.
`Far from representing a random sample of average ana(cid:173)
`lytical outfits, these three laboratories have spent many
`years dedicated especially to the particular analytical con(cid:173)
`cerns associated with the determination of trace levels of
`aluminum. Our laboratory, for example, is continually
`involved with resolution of erroneous results for alumi(cid:173)
`num reported by contract laboratories who analyze paren(cid:173)
`teral solutions infrequently.
`An estimate of the cost to validate a method is summa-
`
`+80
`
`i +30
`C +20
`0 .:
`~ +10 Pharma-
`- 0
`0 ..
`~
`ceutical1
`• u
`C -10
`i -20
`0
`CJ -30
`
`10-'
`Majo,
`Nu1rientl
`
`-60
`
`Concentration
`Figure 6-lnterlaboratory coefficient of variation as a function of
`concentration. [Reprinted with permission from J. Assoc.
`Off. Anal. Chem., Vol. 63, No. 6, pp. 1344-1354 (1980).]
`
`Vol. 43, No. 3 / May-June 1989
`
`135
`
`Eton Ex. 1071
`4 of 8
`
`
`
`TABLE VI. Aluminum Determination by Atomic Absorption
`Spectrophotometry
`Cost Analysis
`Method Development and Validation for 20 Solutions
`Manpower: 20 man-months
`Cost:
`$160,000
`
`Operating Costs per Manufacturing Site
`Instrumentation
`2 Polarized Zeeman graphite furnace
`atomic absorption spectrometers
`with autosamplers
`Preventative maintenance contracts
`Personnel
`2 Chemists
`Supplies
`Graphite tubes and platforms
`Graphite cones, reagents, labware
`Argon
`
`$136,000
`
`6,500/yr
`
`80,000/yr
`
`16,800/yr
`1,350/yr
`1,200/yr
`$105,850/yr
`
`Total
`
`rized in Table VI. Approximately one man-month per
`solution type is estimated. Using typical overhead expense
`figures, twenty solutions would cost $160,000 for valida(cid:173)
`tion. Capital expenditure would cost almost $140,000,
`assuming one purchased a polarized Zeeman graphite
`furnace atomic absorption spectrometer, as well as a
`backup, for those occassions when the first failed to oper(cid:173)
`ate. This would become necessary if mandatory testing
`were imposed, to prevent shutting down production simply
`because the primary analytical instrument failed. Typical
`operating costs, including maintenance contracts, sup(cid:173)
`plies, and the chemists, skilled in trace metal analysis,
`would run in excess of $100,000 per year.
`
`II. Minimizing Aluminum Levels in the Product
`
`Assuming that one were required to implement testing,
`what are the sample matrices for which validation should
`be performed? To answer this, we analyzed a typical TPN
`admixture to assess the contribution to the total alumi(cid:173)
`num level from each of the components shown in Table
`VII. It was found that the amino acids contributed about
`3% of the total aluminum. While some of the trace metal
`additives had quite high aluminum concentrations, their
`overall contribution was quite low, a few percent, by virtue
`of the small volumes employed in the admixture. Calcium
`gluconate, on the other hand, contributed about 89% of
`the total aluminum found. These figures are in substantial
`agreement with the data reported by Koo (6), who also
`found that less than 3% of the aluminum arose from the
`amino acids, and about 89% resulted from the calcium
`gluconate. Clearly, the largest reduction of aluminum
`levels in TPN admixtures would accrue from addressing
`the most heavily contaminated components.
`We have seen that phosphate salts, gluconate salts, and
`citrate salts as well, carry the highest aluminum burden.
`To understand why this is the case we must understand
`the structure of aluminum in water (8), Figure 7. In
`highly acidic aqueous solutions, aluminum exists as a
`triply positively charged ion, coordinated to six neutral
`water molecules. As such, it has a very high charge densi-
`
`TABLE VII. Component Contributions to Aluminum Level in
`TPN Mixture
`Component Contribution
`toTPN
`Al Cone.
`(µg/L)
`(µg/L)
`
`Percent of
`Contribution
`
`Component
`
`17.6
`10% Amino acids
`1.5
`50% Dextrose
`0
`Sterile water
`223
`Calcium gluconate
`0
`Magnesium sulfate
`0
`Potassium chloride
`0.7
`Sodium acetate
`0
`Sodium iodide
`3.5
`Selenium
`0.5
`Chromium chloride
`1.2
`Copper sulfate
`3.3
`Manganese sulfate
`0.2
`Zinc sulfate
`251.5
`Total
`* <DL means less than a detection limit of 0.4 µg Al/L.
`
`47
`3
`<DL*
`5200
`4
`<DL*
`50
`<DL*
`1170
`960
`462
`471
`60
`
`7.0
`0.6
`0
`88.7
`0
`0
`0.3
`0
`1.4
`0.2
`0.5
`1.3
`0.1
`100.1
`
`ty. It forms quite strong complexes with negatively
`charged oxyanions by electrostatic attraction, especially if
`they are capable of binding at multiple ligand sites about
`this octahedron. Such is the case for phosphate, citrate,
`and gluconate (9). In the process of preparation of raw
`materials made from these anions, complexation with alu(cid:173)
`minum is inevitable, and this is carried through to the
`final dosage form. Precipitation of aluminum salts in
`pharmaceutical preparations, buffered with phosphate,
`have been discussed previously (10, 11, 12).
`As the pH is raised, protons are hydrolyzed off the
`coordinating water molecules, leaving negatively charged
`hydroxide ions in their place. This reduces the overall
`charge on the complexed ion, and hence the coulombic
`repulsion between such species. Mutual approach is facili(cid:173)
`tated, and in fact dimers are formed (8), as two octahedra
`share an edge. As the pH rises still further the positive
`charge decreases, permitting ring structures to form (I 3)
`as each octahedron shares two edges with neighboring
`complexes. With pH in the neutral region, the charge
`decreases to zero. There is no longer repulsion among the
`species, and extended networks develop (13), attaining
`colloidal dimensions, eventually precipitating out of solu(cid:173)
`tion as aluminum oxide, Figure 8. With further rise in the
`
`Figure 7-Schematic representation of aquo-aluminum ion
`Al(H20)e3+.
`
`136
`
`Journal of Parenteral Science & Technology
`
`Eton Ex. 1071
`5 of 8
`
`
`
`Q Sampling Point
`Aluminum (µg/L)
`
`Figure 10-Analysis of water purification system for aluminum.
`
`clear that one must consider the role of pH at different
`points in the process, possible complexing agents, and
`kinetic aging effects. The first step in the process, howev(cid:173)
`er, would involve assessing aluminum levels in current
`production of the dosage form. For example, we reported
`levels of aluminum in ten lots of Dianeal® CAPD dialysis
`solution examining three units from each lot. The mean
`value was 3.0 ppb. Unit to unit RSD within a batch was
`9.2%, and interbatch RSD was 65%. This seemingly large
`variability implies a standard deviation of only 2 ppb
`between lots. Furthermore, the mean aluminum levels and
`components of variance in the product packaged in poly(cid:173)
`vin ylchloride plastic containers were similar to those de(cid:173)
`termined for the glass-packaged product (4). Up to this
`point in the analysis of the production process, one re(cid:173)
`quires only an analytical method validated for the product
`matrix.
`Examination of the role of the individual intermediate
`steps in manufacturing requires, in addition, a validated
`collection procedure for the acquisition of samples. Nei(cid:173)
`ther contamination from the container, leading to errone(cid:173)
`ously high aluminum levels, nor wall adsorption, resulting
`in false low levels, can be tolerated. One would then be in a
`position to examine the contributions from each of the
`components of the production process for the water for
`injection, Figure l 0, for example. Interestingly, even after
`the ion exchangers, one finds levels of 17 ppb aluminum.
`The glass still for the distillation efficiently removes all
`but 2 ppb, which then appears in the mix tank, the rest
`being flushed to waste from the still. The contribution
`from the water is approximately matched by that from the
`raw materials, as shown in Figure 11. Neither the pump
`nor the filter used contribute further. Upon sterilization,
`
`Figure 8-Schematic representation of network of aluminum hydrox(cid:173)
`ide octahedra (gibbsite structural pattern).
`
`pH, the overall charge becomes increasingly negative, and
`the complex solubilizes. This behavior is summarized in
`Figure 9 (14), which displays the log of the aluminum
`concentration as a function of pH. In either the acidic or
`basic regions, solubility is apparent, but is evidently limit(cid:173)
`ed in the neutral zone. If one attempted to prepare an
`aluminum solution at a concentration of 10-4 Mat a pH
`of 8, for example, the intersection of the solubility curve at
`that pH indicates that most of the aluminum would pre(cid:173)
`cipitate out, leaving less than 10-5 M aluminum solubi(cid:173)
`lized. It has also been determined that a wide range of
`pharmaceutically relevant complexing agents bind to alu(cid:173)
`minum more strongly at pH values close to their pKa
`values (9).
`. The equilibria indicated on this graph are not necessar(cid:173)
`ily attained rapidly. In fact the initially formed precipitate
`is amorphous. Upon aging, the crystallinity increases,
`thus decreasing the concentration of soluble aluminum
`with which it is in equilibrium (14). Such aging effects
`have also been observed in connection with changes in the
`gel filtration pattern of complexes of aluminum with citric
`acid (15).
`To understand a particular manufacturing process,
`from the standpoint of ah1minum contamination, it is
`
`0
`
`2
`
`4
`
`6
`
`pH
`
`8
`
`10
`
`12
`
`14
`
`16
`
`0
`
`·2
`
`~ .4
`J-
`~ -8
`CJ
`0
`...I
`
`-10
`
`-12
`
`-14
`
`[Al (OH)3 (eq)] ?
`
`Al2(0H)+:
`
`Al(OH)~
`
`WATER
`
`RAW
`
`t® •MATERIALS a. ?:::\~ ~ I ~~ -
`
`Figure 9-pH vs. concentration for various monomeric and polymeric
`Al species assumed in equilibrium with freshly precipitated
`Al(OH)3. (Reprinted with permission from Advances in
`Chemistry Series, 67, pp. 1-29. Copyright 1967 American
`Chemical Society.)
`
`0 Samphng Point
`Aluminum (µg/L)
`
`Figure 11-Analysis of process line for aluminum.
`
`Vol. 43, No. 3 I May-June 1989
`
`137
`
`. STERILIZATION
`
`Eton Ex. 1071
`6 of 8
`
`
`
`however, an apparent decrease in aluminum level is found,
`presumably due to adsorption by the container ( 4). Analy(cid:173)
`sis of product, packaged in either plastic or glass, after
`expiry indicated no significant increase in aluminum lev(cid:173)
`els (16).
`
`Ill. Clinical Implications for Pediatric Total Parenteral
`Nutrition
`
`The manufacturing analysis just presented was under(cid:173)
`taken several years ago, when concern over aluminum
`levels in the water used to dilute hemodialysis concentrate
`was first raised. This was a real, demonstrable medical
`concern, and industry responded proactively. The clinical
`necessity for controlling aluminum levels in currently pro(cid:173)
`duced TPN fluids is far less convincing. In a study of the
`heterogeneity of bone histology in parenteral nutrition
`patients, it was found that "abnormalities of bone histolo(cid:173)
`gy were seen in the parenterally fed patient, irrespective of
`whether or not Al deposition was observed .... Further(cid:173)
`more, . . . extensive Al deposition resulted specifically
`from the use of casein hydrolysate" ( 17).
`Contrary to the situation for casein hydrolysate, it has
`been determined that "Aluminum does not accumulate in
`teenagers and adults on prolonged parenteral nutrition
`containing free amino acids" (18). One might wonder
`what happens in the case of a patient on TPN who has
`switched from casein hydrolysate to free amino acids; do
`the bone aluminum levels decrease? The results of such a
`study are shown in Table VIII. Stainable aluminum was
`found upon bone biopsy of a patient maintained on casein
`hydrolysate for 24 months. Thereafter, the patient was
`switched to free amino acids, and underwent subsequent
`bone biopsy after 66 and then at 78 months. Both the
`quantifiable bone aluminum as well as the aluminum
`. stainable tissue surface decrease to zero by the last sam(cid:173)
`pling interval. Concomitant with this is an observably
`increased bone formation rate (19).
`This reversibility in clinical manifestation has also been
`demonstrated in infants maintained on TPN." ... Alumi(cid:173)
`num ... had been identified as an inadvertant contami(cid:173)
`nant of the casein hydrolysate solutions that were used in
`the past ... a recent study in which five of the originally
`described patients again received bone biopsy and bone
`turnover metabolic tests ... showed normal bone turn(cid:173)
`over, compared to previous ones in which aluminum con(cid:173)
`tent in blood and bone was elevated ... " The authors
`conclude, "Metabolic bone disease has not been clinically
`apparent in our patients during the past 8 years" (20).
`
`· TABLE VIII. Bone Characteristics, upon Repeated Biopsy, of
`One Patient on TPN Who Changed Nitrogen
`Source
`
`Aluminum
`Duration
`TPN mm/mm2
`(mos.)
`tissue
`
`(%)
`
`Bone Formation
`Rate
`mm/mm2 tissue/
`d X 10-3
`
`24
`66
`78
`
`2.28
`0.51
`0.00
`
`77
`17
`0
`
`0.006
`0.008
`0.097
`
`Nitrogen
`Source
`
`Casein
`Amino acids
`Amino acids
`
`138
`
`In a review of ten years experience in over a hundred
`pediatric patients, maintained on long-term home paren(cid:173)
`teral nutrition, " ... the only modifications have been a
`change in 1981 from casein hydrolysate as the protein
`source to balanced amino acids ... long-term weight gain
`and growth was best in those who received 90% or more of
`their nutrition parenterally .... Although the average
`patient received HTPN for nearly 2 years, eight have
`received HTPN for more than 5 years, and four for a
`decade. School problems have been limted, and intelli(cid:173)
`gence quotients among the HTPN patients are usually
`normal. Two of the children are mentally gifted and at(cid:173)
`tend special schools because of their advanced capability
`. . . . Of those individuals on TPN for 5 or more years,
`serious liver disease is not present in any of them" (21 ).
`In conclusion, the results of long-term follow-up studies
`of infants on TPN indicate that the aluminum loading
`that has apparently been demonstrated for the premature
`infant, is reversible, leaving no clinically residual prob(cid:173)
`lems. This may very well be a consequence of improved
`renal function as the premature infant increases weight,
`and improves, attendant with TPN.
`
`References
`I. Versieck, J., and Cornelis, R., "Normal levels of trace elements in
`human blood plasma or serum," Anal. Chim. Acta, 116,217 (1980).
`2. Welz, B., Atomic Absorption Spectrometry, 2nd ed., VCH, Wein(cid:173)
`heim, 1985.
`3. L'vov, B. V., Pelieva, L.A., and Sharnopolsky, A. I., "Decrease in
`the effect of the base during atomic absorption analysis of solutions
`in tube furnaces by evaporation of samples from a graphite substra(cid:173)
`te," Zh. Prikl. Spektrosk., 27, 395 ( 1977).
`4. McHalsky, M. L., Rabinow, B. E., Ericson, S. P., Weltzer, J. A., and
`Ayd, S. A., "Reduction of aluminum levels in dialysis fluids through
`the development and use of accurate and sensitive analytical metho(cid:173)
`dology," J. Parenter. Sci. Technol., 41, 67 (1987).
`5. Slavin, W., Carnrick, G. R., and Manning, D. C., "Magnesium
`nitrate as a matrix modifier in the stabilized temperature platform
`furnace," Anal. Chem., 54,621 (1982).
`6. Koo, W.W. K., et al., "Aluminum in parenteral nutrition solution(cid:173)
`sources and possible alternatives," J. Parenter. Enteral Nutr., 10,
`591 (1986).
`7. Horwitz, W., Kamps, L. R., and Boyer, K. R., "Quality assurance in
`the analysis of foods for trace constituents," J. Assoc. Off Anal.
`Chem., 63, 1344 ( 1980).
`8. Hem, J. D., "Aluminum species in water," Adv. Chem. Ser., 73, 98
`(I 968).
`9. Hasegawa, K., Hashi, K., and Okada, R., "Physicochemical stabil(cid:173)
`ity of pharmaceutical phosphate buffer solutions II. Complexation
`behavior of Al(III) with additives in phosphate buffer solutions," J.
`Parenter. Sci. Technol., 36, 168 (1982).
`10. Hasegawa, K., Hashi, K., and Okada, R., "Physicochemical stabil(cid:173)
`ity of pharmaceutical phosphate buffer solutions. I. Complexation
`behavior of Ca(II) with additives in phosphate buffer solutions," J.
`Parenter. Sci. Techno/., 36, 128 (1982).
`11. Hasegawa, K., Hashi, K., and Okada, R