`
`Reduction of Aluminum Levels in Dialysis Fluids Through the Development and
`Use of Accurate and Sensitive Analytical Methodology
`
`MICHAEL L. McHALSKY•\ BARRETT E. RABINOW, SCOTT P. ERICSON,
`JOSEPH A. WEL TZER, and SHARON W. AYO
`
`Truenol Laboratories, Inc., Morton Grore, Illinois
`
`ABSTRACT: Various analytical methods of determining the aluminum content of CAPD (continuous
`ambulatory peritoneal dialysis) solutions and associated raw materials were developed and compared. The
`methods include two graphite furnace atomic absorption procedures and an aluminum-lumogallion fluores(cid:173)
`cence technique that can quantitate aluminum in CAPD solution at, and in some cases below, the 1-µg/liter
`level with good accuracy and precision. These methods were then used to examine the possible contributions
`to aluminum content from all aspects of the production of CAPD solutions at Travenol Laboratories, Inc.
`The major source of aluminum to the low, but measurable, levels in CAPD solutions was lactic acid. With the
`use of USP grade or equivalent ingredients, especially a low-aluminum-content lactic acid, these solutions
`can be prepared with aluminum levels below 10 µg//iter.
`
`Introduction
`
`Concern about trace levels of aluminum in parenteral
`products has arisen in connection with contamination in(cid:173)
`curred from the container (1, 2) closure (3) or the drug (4,
`5) itself. Excessive intake of aluminum from dialyzing
`fluids (6-17) and aluminum containing phosphate binders
`(18-26) has been associated with dialysis dementia and
`bone disease in patients with chronic renal failure. The
`maximum aluminum levels in dialysis fluids which have
`been considered as "safe" by various researchers have
`ranged from 10 to 50 µg/liter (10, 15-17, 21). The part
`her billion accuracy required of analytical methods used
`to study these aluminum levels is available to only a few
`laboratories willing to expend the effort to achieve reliable
`results. In contrast, the prevalence of expensive atomic
`spectroscopic instrumentation would suggest trace metal
`analysis to be a routine undertaking. The numerous con(cid:173)
`tradictory reports within the literature imply that while
`such determinations may be widespread, inaccuracy is
`pervasive (27). The erroneous values adversely affect our
`understanding of the clinical problem, thereby preventing
`rational determination of what constitutes safe levels of
`aluminum.
`The present studies describe reliable analytical method(cid:173)
`ology necessary to assess the concentration of aluminum
`in dialysis fluid products. In the absence of a recognized
`reference standard for aluminum in this matrix, the reli(cid:173)
`ability of the methods was established through interlabor(cid:173)
`atory and intermethod comparisons. This required cor(cid:173)
`roboration of results by redundant methods based upon
`dissimilar physical principles.
`
`Received June 7, 1985. Accepted for publication December 23, 1986.
`t:>. Author to whom inquiries should be directed.
`
`Utilizing these validated methods, aluminum levels in
`DIANEAL®, peritoneal dialysis fluid manufactured by
`Travenol Laboratories, Inc., were found to be far lower
`than the 50- to 100-µg/liter levels implicated in patient
`cases involving excessive aluminum loading. To assure
`continued low levels, a program was undertaken to evalu(cid:173)
`ate sources contributing to the small, though measurable,
`aluminum levels in DIANEAL® solutions. This process
`was simplified by the numerous worldwide manufacturing
`facilities, that lent themselves to comparative analysis.
`During this study, several plants were found with exceed(cid:173)
`ingly low levels in their product. The focus of the research
`program was then altered to reduce aluminum levels from
`all plants to that of the facility with the lowest values.
`Many factors were evaluated as possible contributors of
`aluminum to the final product: raw materials, water puri(cid:173)
`fication system, mixing, pumping, filtering, filling, steril(cid:173)
`ization, container material, and time of storage.
`
`Experimental
`
`Instrumentation
`
`Two graphite furnace atomic absorption spectrophoto(cid:173)
`meters (GFAAS), an inductively coupled plasma atomic
`emission spectrophotometer (ICP-AES), and two spectro(cid:173)
`fluorometers were used in the work reported here. The
`Perkin-Elmer Model 603 atomic absorption spectropho(cid:173)
`tometer with HGA 2100 graphite furnace included a
`Model AS-I autosampler. The excitation source was the
`309.3-nm line from an aluminum hollow cathode lamp
`(Perkin Elmer) set at 12-ma current and was adjusted to a
`slit width of 0. 7 nm. To correct for background interfer(cid:173)
`ences to the aluminum signal caused by the high salt and
`organic matrix of the dialysis fluids, a deuterium back-
`
`Vol. 41, No. 2 / March-April 1987
`
`67
`
`Eton Ex. 1070
`1 of 9
`
`
`
`ground correction was used. Nonpyrolytic graphite tubes
`(Perkin Elmer # 0290-1633) were found to be more sensi(cid:173)
`tive than the pyrolytically coated ones (Perkin Elmer
`# B0091-504). The temperature program used with the
`HGA 2100 graphite furnace was ramp 30 s to 120 °C,
`hold for 20 s, ramp 25 s to 1200 °C, hold for 35 s, then
`atomize at 2650 °C for 9 s. Argon flow through the tube
`was 30 cc/min except at atomization when the Argon
`interrupt mode was turned on. The peak height was used
`for the absorbance measurement with the signal moni(cid:173)
`tored for 5 s during atomization.
`Later work involved the use of a Perkin-Elmer Model
`Z5000 spectrophotometer with an HGA 500 graphite
`furnace and AS-40 autosampler. The background was
`compensated by using Zeeman effect background correc(cid:173)
`tion. With the availability of this instrument the assay was
`modified to use a pyrolytically coated graphite tube (Per(cid:173)
`kin Elmer # B0109-322) with a L'vov platform (Perkin
`Elmer# B0109-324). An aluminum hollow cathode lamp
`set at 15-ma current was used as in the 603/2100 system.
`The temperature program for the Model 5000 furnace
`was ramp 25 s to dry at 110 °C, hold 20 s, ramp 20 s to
`char at 350 °C, hold 15 s, ramp 30 s to char at 500 °C,
`hold for 10 s, ramp 20 s to char at 1500 °C, hold 25 s,
`atomize at 2700 °C, hold for 7 s, ramp 1 s to cool at 20 °C
`for 15 s. The Argon flow through the tube was 300 cc/min
`except at atomization when the Argon interrupt mode was
`used. Peak area was used for absorbance measurements
`over a 6-s integration time.
`Inductively Coupled Plasma-Atomic Emission Spec(cid:173)
`troscopy (ICP-AES) was used as an alternate method for
`measuring aluminum levels in certain raw materials. The
`specific instrument used was the Model JY38P (Instru(cid:173)
`ments S.A.), which includes a Model ICP 2500 Inductive(cid:173)
`ly Coupled Plasma source, maximum power 2.5 mW at
`27.12 MHz (Plasma Therm, Inc.), a Czerny-Turner 1-
`meter monochromator with 2400 g/mm holographic grat(cid:173)
`ing, and a spectral range of 190-750 nm using the Hama(cid:173)
`matsu R928 photomultiplier tube as a detector. Emission
`intensities for aluminum were measured at the 396.15-nm
`line. Due to the low aluminum concentrations present in
`the CAPO solutions and the low sensitivity oflCP-AES as
`compared with GFAAS, the inductively coupled plasma
`technique was used only for those raw materials which
`contained greater than 0.1 µg/ g of aluminum.
`The determination of aluminum down to the µg/liter
`level could be performed by the aluminum-lumogallion
`fluorescence method similar to one previously described
`by Hydes and Liss (28). The instrument used for these
`experiments was the Perkin-Elmer 650-105 Fluorescence
`Spectrophotometer. Excitation and emission wavelengths
`used for the aluminum-lumogallion complex were 472 nm
`and 568 nm, respectively. The method of multiple stan(cid:173)
`dard additions was used in this case since the competition
`of the lactate ion with lumogallion for complexation of
`aluminum resulted in low aluminum values when deter(cid:173)
`mined with external standards. The actual procedure used
`is summarized in the following description.
`Four 50-mL aliquots of the sample were added to poly(cid:173)
`ethylene containers along with 0.5 mL of an acetic acid
`
`(Baker, Ultrex)/sodium acetate (Fisher) buffer (4 Min
`acetate and adjusted to a pH of 5), and either 0.25 or 0.5
`mL of 0.02% lumogallion (Pfaltz and Bauer) solution.
`The bottles were spiked with 0, 5, 10, and 20 ppb of an
`aluminum standard, respectively, for a multiple standard
`additions determination. A fifth bottle containing 50 mL
`of deionized distilled water, to be used as a reagent blank,
`was similarly treated. The bottles were transferred to a
`water bath held at 80 °C for 1.5 hr. The samples were then
`allowed to cool to room temperature and analyzed with a
`spectrofluorometer within 24 hr of preparation. The
`straight line obtained by a regression analysis was extrap(cid:173)
`olated back to zero emission intensity where the Al con(cid:173)
`centration of the sample could be obtained once correcting
`for the reagent blank.
`
`Contamination Control
`
`Avoiding aluminum contamination is essential to any
`assay with measurements at the µg/liter level. A critical
`factor for successful aluminum contamination control was
`the use of properly cleaned plastic labware for sample
`handling and storage. Samples and standards were al(cid:173)
`lowed to contact only suitably cleaned plastic labware.
`The following cleaning procedure was found most eff ec(cid:173)
`tive in removing aluminum from plasticware. The labware
`was soaked in a low-metal content laboratory detergent
`(NRS-250 Norell, NJ), rinsed with distilled water, then
`soaked for at least 2 hr in a 10% nitric acid bath, thorough(cid:173)
`ly rinsed with deionized distilled water, dried face down
`on rubber matting, and stored in sealed plastic bags. The
`plasticware was rinsed again with deionized distilled wa(cid:173)
`ter just prior to use. Containers were left open for as short
`a time as possible to avoid airborne contaminants. Eppen(cid:173)
`dorf micropipetters with disposable plastic tips were used
`throughout. The polystyrene autosampler cups (Perkin
`Elmer) were also soaked in acid and rinsed with deionized
`distilled water.
`
`Standards
`
`A 1000-mg/liter aluminum stock· solution was pre(cid:173)
`pared by dissolving Al(NO3)3•9H2O (J. T. Baker, 99.1%
`by EDTA titration) in deionized distilled water. This was
`stable for at least 6 months. Intermediate and calibration
`standards were prepared in polymethylpentene (PMP)
`volumetric flasks (Nalgene). A reagent blank and 2-, 5-,
`10-, and 20-µg/liter aluminum standards were typically
`employed, although standards as high as 100 µg/liter of
`aluminum also have been used. For use with PE603/2100,
`all samples and standards were made 0.5% v /v in nitric
`acid (Baker Ultrex grade). In a later modification of the
`assay for use with the Perkin Elmer Z5000 instrument, in
`addition to the nitric acid, recrystallized Mg(NO3)i
`(Fisher Scientific ACS grade) was also added as a matrix
`modifier at a concentration of 0.05% w /v similar to the
`procedure described by Manning et al. (29, 30).
`
`Sample Preparation
`
`Preparation of CAPO solution samples for aluminum
`determination required special procedures to avoid con(cid:173)
`tamination. The solutions are distributed either in glass
`
`68
`
`Journal of Parenteral Science & Technology
`
`Eton Ex. 1070
`2 of 9
`
`
`
`instrument. A common set of DIANEAL solutions was
`analyzed each day on each instrument over a period of 3
`days. A composite of the results of this study is given in
`Table I-A. DIANEAL solutions containing approximate(cid:173)
`ly 2.5 and 5.0 µg/liter of aluminum were used along with
`these same samples spiked with 9 .9 µg/liter of added
`aluminum. Recoveries of 52-68% for the 9.9-µg/liter
`spiked samples were obtained on the 603 instrument,
`while recoveries of 98-104% were obtained on the Zee(cid:173)
`man 5000 when determined against external aluminum
`standards. Using the spike recoveries to correct the assay
`values obtained on the Model 603, however, resulted in
`aluminum concentrations that agreed quite closely with
`the values obtained directly on the Zeeman 5000 instru(cid:173)
`ment. This study showed that accurate aluminum results
`could be obtained with the 603/2100 system if a standard
`addition technique was used. The graphite furnace meth(cid:173)
`odology was also compared with other analytical proce(cid:173)
`dures used for the determination of aluminum concentra(cid:173)
`tions. The G FAAS procedure was compared with Induc(cid:173)
`tively Coupled Plasma-Atomic Emission Spectroscopy
`(ICP-AES). This comparison was performed on solutions
`containing higher aluminum levels (>50 µg/liter) since
`the detection of aluminum by ICP-AES was 30 µg/liter.
`Results of this comparison are also given in Table 1-B.
`The GFAAS method was also compared to an alumi(cid:173)
`num-lumogallion fluorescence technique. Comparisons of
`the two techniques were for DIANEAL solutions with
`aluminum concentrations varying between 1 and 25 µg/
`liter. The results of the comparison are given in Table 1-C.
`The fluorescence technique had a precision of 7.4% at the
`6-µg/liter level with a detection limit of 1 µg/liter. The
`range was calculated from the standard error of the re(cid:173)
`gression analysis at 95% confidence.
`An interlaboratory comparison was also performed, in(cid:173)
`volving Travenol Laboratories Research and Develop(cid:173)
`ment facilities in the United States (Morton Grove, IL)
`
`bottles or VIAFLEX® (polyvinyl chloride) bags. For the
`VIAFLEX bag, one port was cut open with a razor blade
`or scalpel. Some of the contents were allowed to drain to
`waste to clean the port and minimize external contamina(cid:173)
`tion of the small aliquot used. The aliquot was then trans(cid:173)
`ferred to a small polyethylene bottle. Initially, only nitric
`acid, and later, nitric acid plus Mg(NO3h, were added as
`matrix modifiers as with the standards.
`For glass bottles, the cap and rubber septum were first
`removed from the bottle stopper. The bottle was then
`inverted and some solution allowed to drain to waste. To
`avoid contamination, aliquots were taken without inter(cid:173)
`rupting the sample stream. The aliquots were then treated
`similarly as VIAFLEX container samples. As can be seen
`by some of the data shown later in this article, e.g., see
`Table III, samples with very low aluminum concentra(cid:173)
`tions and present in either glass or VIAFLEX containers
`can be assayed by this technique without incurring outlier
`values because of contamination.
`A similar sample preparation technique was used for
`liquid raw materials such as sodium lactate solutions.
`These may be diluted if necessary to fall within the range
`of the working standards.
`
`Sample Transportation and Storage
`
`For those samples that were to be stored, or transported
`in other than their original containers, treatment was
`required to insure against losses of aluminum by adsorp(cid:173)
`tion onto the container surfaces. Polyethylene bottles were
`used to transport these samples. These bottles, as de(cid:173)
`scribed previously under "contamination control," were
`further soaked in a concentrated EDT A solution then
`rinsed with deionized-distilled water. Sufficient EDTA
`(EDT A, disodium salt, Mallinckrodt AR Grade) solution
`was then added to allow a final concentration of 0.01 % of
`EDT A in the filled bottle.
`An experiment was performed to determine whether
`0.01% EDTA biased the analysis. Both water and DIAN(cid:173)
`EAL® samples with and without added EDTA were
`spiked with 0, 10, and 50 µg/liter of aluminum. Results of
`the aluminum analysis gave 100 ± 5% recovery at both
`spiked levels and indicated no bias for the unspiked sam(cid:173)
`ples. The samples could be stored for at least 1 week
`without any noticeable losses of aluminum.
`Unopened samples were stored at ambient temperature
`in their original containers. Opened solutions were trans(cid:173)
`ferred from their original containers to the polyethylene
`bottles and EDT A added. Whenever possible, these solu(cid:173)
`tions were also refrigerated if storage for more than 1 day
`was required.
`
`Results
`
`Evaluation of the GFAAS Method
`
`An instrument-to-instrument comparison was one of
`the techniques used to evaluate the G FAAS methodology.
`The Perkin-Elmer 603 with HGA 2100 and Perkin-Elmer
`Zeeman 5000 with HGA 500 furnace described above
`were used for this comparison. In addition, somewhat
`different experimental procedures were used with each
`
`S
`IO
`NIVELLES (µg/ L Aluminum)
`Nlvelles(N=3)
`Morton Grove (N = 4) except where noted by®
`Figure 1-lnterlaboratory comparison of aluminum levels determined
`in several lots of peritoneal dialysis solution.
`
`Vol. 41, No. 2/March-April 1987
`
`69
`
`Eton Ex. 1070
`3 of 9
`
`
`
`TABLE I. Assay Validation-Comparisons Between Methods
`IA: GFAAS (Zeeman 5000) versus GFAAS (603 W /D2)
`
`Instrument
`
`Sample
`
`Al
`(µg/liter)
`
`Precision
`CV(%)
`
`% Recovery•
`(9.9 µg/liter Al)
`
`Precision (CV%)
`Of Standard
`Addition
`
`Detection
`Limit
`(µg/liter Al)
`
`Zeeman 5000
`
`Model 603/2100
`
`1
`2
`1
`2
`
`4.7
`2.5
`5.3b
`2.8b
`
`3.66
`6.07
`11.7
`39.3
`
`100.6
`102.7
`63.3
`62.9
`
`1.7
`1.9
`8.3
`4.5
`
`0.35
`0.35
`1.2
`2.2
`
`1-8: GFAAS (Zeeman 5000) versus ICP-AES-12 Lots Lactic Acid Raw Material
`
`Added (µg/liter Al)
`
`Found (µg/liter Al)
`
`% Recovery
`
`Precision (CV%)
`
`GFAAS
`
`ICP-AES
`
`0.5
`1.0
`1.5
`0.5
`1.0
`1.5
`
`0.463
`1.009
`1.586
`0.536
`1.059
`1.596
`
`92.6
`100.9
`105.7
`107.2
`105.9
`106.4
`
`5.5
`6.1
`0.9
`1.0
`1.0
`1.6
`
`1-C: GFAAS (Zeeman 5000) versus Fluorescence (AI-Lumogallion Complex)
`
`LotofCAPD
`
`1
`
`(µg/liter Aluminum)
`4
`2
`5
`3
`
`6
`
`Lot ofCAPD
`
`1
`
`(µg/liter Aluminum)
`4
`5
`2
`3
`
`6
`
`1.44 2.62
`2.0
`0.8
`
`6.85
`4.2
`
`6.09
`d
`
`24.43
`3.5
`
`Fluorescence
`Mean rangec (±) 1.44
`2.0
`0 Versus external standards
`b Percent recoveries were used to correct sample concentrations
`c Range is 2X standard deviation of the x-intercept of the standard additions curve (blank corrected) .
`d Coefficient of variation for a sample analyzed on 5 different days was 7.4%: the ranges were from 0.8 to 4.
`e Single sample analyzed (otherwise, n = 2 for graphite furnace).
`n = 4 for fluorescence.
`
`GFFAS
`Mean range (±)
`
`1.80
`
`1.36
`0.8
`
`3.20 5.95
`3.2
`
`5.70 22.80
`0.8
`
`and Belgium (Nivelles). The Nivelles group used a Pye
`Unicam Model SP-9 graphite furnace atomic absorption
`spectrophotometer with a Model SP4-01 autosampler.
`The method was similar to the one described for the
`Perkin Elmer 603/2100 instrument, which the Morton
`Grove group used for this study, except a pyrolytically
`coated graphite tube was used and no matrix modifier was
`added. Spike recoveries ranged from 82 to 119%; there(cid:173)
`fore, no correction for recovery was required. The detec(cid:173)
`tion limit for this method was 3 µg/liter . Both laboratories
`analyzed multiple units from eight lots of CAPD solution.
`Data for each lot are presented as a rhombus (Fig. 1), the
`axes of which span the range of values found by each
`laboratory on the respective units analyzed. Optimally,
`values should lie on a straight line 45° from each axis. The
`Nivelles group analyzed three samples per lot and, except
`where indicated by circled number in Figure 1, the Mor(cid:173)
`ton Grove Group analyzed four samples per lot.
`
`Interplant Survey
`
`With the development of reliable analytical methodolo(cid:173)
`gy, a comprehensive study was undertaken to measure the
`aluminum content of a wide variety of CAPD solutions
`produced by Travenol. This included solutions produced
`at fourteen of Travenol's international manufacturing fa(cid:173)
`cilities. A summary of the results can be found in Table II.
`
`The range as well as the mean values of aluminum concen(cid:173)
`tration in CAPD solutions produced at each facility are
`given.
`
`TABLE II. Aluminum Content of CAPD Solutions-Inter-
`plant Survey
`Number of
`Lots
`(Samples
`per Lot)
`
`Al {~gLliter}
`Range
`(All Units
`Tested)
`
`Mean
`(Per Lot)
`
`Facility Method
`
`1.3- 3.0
`7 (2)
`1.98
`2
`A
`11.2-18.2
`14.0
`2 (4)
`B
`3
`2;6-20.8
`10.4
`2 (3)
`1
`C
`14.6-23.8
`19.0
`3 (3)
`3
`D
`17.2-27.9
`2 (2)
`21.5
`1,2
`E
`22.7-23.4
`23.0
`2 (1)
`2
`F
`2.6-5.9
`7 (1)
`3.97
`2
`G
`H
`1.5- 7 .2
`3.48
`10 (3)
`2
`1.3-9.2
`3.05
`10 (3)
`I
`2
`3.1-3.4
`3.2
`1 (2)
`2
`J
`1.1 - 1.5
`2 (1)
`K
`1.3
`2
`1.7-4.1
`2.7
`3 (2)
`2
`L
`M
`<0.3 - 2.6
`0.54
`10 (1)
`2
`N
`3 (1)
`1,2
`5.5-11.3
`8.57
`Method 1. P & E 603/2100 with D2 background correction.
`Method 2. P & E 5000 with Zeeman background correction.
`Method 3. Pye Unicam SP-9 (Nivelles, Belgium)
`
`70
`
`Journal of Parenteral Science & Technology
`
`Eton Ex. 1070
`4 of 9
`
`
`
`Production Process Study
`
`To determine the contribution of various phases of the
`production process to overall aluminum levels in a CAPO
`solution, a study of the entire process was undertaken.
`Facility A of Table II was chosen for this purpose because
`solutions with very low levels of aluminum were consis(cid:173)
`tently produced there.
`Figures 2 and 3 show the principal components of the
`process and aluminum levels found at each. Figure 2
`describes the most important features of the water purifi(cid:173)
`cation system used at this location, which is typical of
`most Travenol locations. Duplicate samples were taken at
`each of 10 different sampling locations along the water
`purification process. These sampling points included the
`incoming raw water, various storage and pumping sta(cid:173)
`tions, the filter and ion exchange sections,- and before and
`after distillation. In addition, this sampling procedure was
`performed at two separate times in a single day, once near
`the beginning of a work period and once near the end.
`Figure 2 summarizes the salient features of the system
`along with the corresponding aluminum levels. The values
`indicated in Figure 2 represent mean aluminum values of
`a minimum of 4 samples taken at that point in the water
`purification system.
`Figure 3 shows the CAPO solution production and fill(cid:173)
`ing line. Samples were taken of the batch water, all raw
`materials, and batch CAPO solution following mixing and
`at various points in the filling line. Samples of this solution
`were also taken following autoclaving in PVC bags. Dupli(cid:173)
`cate samples were taken at six different locations along
`the filling lines with a summation of the results shown.
`
`Container Study
`
`CAPO solutions are distributed in two types of contain(cid:173)
`er, glass bottles and VIAFLEX plastic bags. VIAFLEX
`plastic, a formulation of polyvinyl chloride, has been in
`use as a container by most of Travenol's facilities. Table
`III contains results of a comparison of the aluminum
`
`15 ~ILTER
`PUMP
`
`@
`0 Sampling Point
`Aluminum (µg/L)
`
`Figure 2-Analysis of water purification system for aluminum.
`
`WATER
`
`+@ t MATERIALS
`
`RAW
`
`ff. ~ o ""'" , ® .....
`
`-ST£-RL-,-ZA-TION-r e
`
`1
`
`0 Sampling Point
`Aluminum (µgill
`Figure 3-Analysis of process line for aluminum.
`
`Vol. 41, No. 2 / March-April 1987
`
`TABLE III. Aluminum Levels (µg/liter) for 20 Different Lots
`of CAPO Solutions Stored in Glass versus VIA(cid:173)
`FLEX® Containers
`VIAFLEX®
`Means
`
`Glass
`
`SD
`
`SD
`
`Means
`
`1.0
`8.37
`3.33
`0.51
`0.21
`2.87
`0.15
`1.83
`1.47
`0.21
`0.21
`1.57
`0.21
`2.67
`0.20
`2.40
`0.0
`2.80
`0.15
`2.93
`0 Three units of each lot were averaged.
`
`3.42
`3.97
`2.43
`1.83
`2.37
`1.73
`4.53
`2.30
`5.13
`7.10
`
`0.45
`0.47
`0.23
`0.06
`0.42
`0.25
`0.35
`0.30
`0.17
`0.17
`
`content of DIANEAL solutions stored in VIAFLEX and
`glass containers. The analysis was performed by the P & E
`Zeeman 5000 method. Three units each of 10 lots of
`CAPO solutions from each of two facilities, one facility
`using glass and the other VIAFLEX containers, were
`compared.
`
`Raw Materials
`
`The raw materials used to prepare a typical CAPO
`solution code are given in Table IV.
`For this study the solid raw materials obtained from six
`facilities (A-F from Table II) were combined in the re(cid:173)
`quired proportions to form a laboratory-prepared CAPO
`solution representing each facility. This permitted analy(cid:173)
`sis by a method already validated for a CAPO matrix. The
`results for these solutions, as obtained using the Perkin
`Elmer 603/2100 instrument, ranged from 1 to 4 µg/liter
`of aluminum. The only raw materials missing from the
`solutions were the plant water and sodium lactate.
`Sodium lactate was analyzed separately from the other
`components of CAPO solution since it existed in liquid
`form and is itself manufactured from two raw materials,
`sodium hydroxide and lactic acid. Of all the raw materi(cid:173)
`als, sodium lactate has the greatest potential to cause
`aluminum contamination. Samples of sodium lactate in
`pretreated containers were obtained from facilities pro(cid:173)
`ducing CAPO solutions that consistently showed very low
`aluminum levels, as well as from facilities which produced
`solutions with somewhat higher aluminum levels. Figure 4
`contains results obtained from two such facilities. The
`results were obtained using both GFAAS and ICP proce(cid:173)
`dures prior to use. A study was performed to determine at
`what point the aluminum may enter the sodium lactate
`production. Samples of sodium lactate taken before and
`
`TABLE IV. Components of CAPO Solution with 3.86% w /v
`Dextrose
`Component
`
`Contents per Liter (g)
`
`Dextrose (anhydrous)
`Sodium chloride
`Calcium chloride
`Magnesium chloride
`
`38.6
`5.7
`0.26
`0.15
`
`71
`
`Eton Ex. 1070
`5 of 9
`
`
`
`Aluminum Level
`In Sodium Lactate/
`Sodium Lactate/
`Lactic Acid
`Lactic Acid
`(mg/kg)
`Contribution to
`CAPO Solution
`(p.g/1)
`
`8
`
`6
`
`4
`
`30
`
`20
`
`fl
`
`I Dianeal Range
`
`0
`Lactic Acid
`I::,. Sodium Lactate
`M, Max. Solid Raw
`Material
`Contribution
`
`M2 Max. NaOH
`Contribution
`
`I
`
`10
`
`2
`M11----1w------C:X) .......... .
`0 ............ .
`M
`2
`
`PLANT
`Figure 4-Contribution of the various raw materials to aluminum levels in CAPO Solution.
`
`A
`
`E
`
`after the filtration step were analyzed for their aluminum;
`content. These samples were obtain•ed from a facility with
`low aluminum levels in their CAPD solution and one with
`higher levels. The before and after filtration aluminum
`levels were 0.057 vs 0.054 mg/liter and 2.25 vs. 2.28 mg/
`liter for the two facilities, respectively. The aluminum
`content of sodium hydroxide (six lots from five manufac(cid:173)
`turers) was examined and found to contribute at most 1.2
`µg/liter of aluminum to the final product. Lactic acid was
`also examined (Fig. 4).
`
`Discussion
`
`Evaluation of the GFAAS Method
`
`The comparison of the graphite furnace methods dem(cid:173)
`onstrated the various problems which can be encountered
`when using this technique to perform an aluminum analy(cid:173)
`sis on a complex sample. The determination of aluminum
`using a relatively slow heating rate associated with the
`HGA 2100 furnace along with using a nonpyrolytic
`graphite tube and the nitric acid matrix modifier necessi(cid:173)
`tated applying a standard additions technique to correct
`for the low aluminum recoveries obtained from spiked
`samples. Deuterium background correction appears to be
`adequate for aluminum determination since the resulting
`aluminum values, corrected for spike recovery, closely
`matched those obtained with Zeeman background correc(cid:173)
`tion.
`Improvements in sensitivity were definitely obtained
`using the Zeeman 5000 system with L'vov platform and
`added Mg(NO3)z modifier as shown by the detection limit
`of this technique, 0.35 µg/liter, as opposed to about 2 µg/
`liter for the Perkin Elmer Model 603/2100. The results
`shown in Table I-A demonstrate the difference in preci(cid:173)
`sion between the two methods. Based on these results, the
`Perkin Elmer Model 603/2100. The results shown in Ta(cid:173)
`ble I-A demonstrate the difference in precision between
`the two methods. Based on these results, testing a sample
`containing 2.5 µg/liter of aluminum should involve a rela-
`
`tive standard deviation of about 6% if performed via the
`Zeeman 5000 method with platform and Mg(NO3)z mod(cid:173)
`ifier. Analysis of the same solution on the 603 /2100 sys(cid:173)
`tem, however, would have a relative standard deviation of
`near 40%. There was some improvement in sensitivity
`when the L'vov platform and Mg(NO3h modifier were
`used with the Model 603/2100 instrument, although it
`was still not as sensitive as the Zeeman 5000 instrument.
`Superior performance of the Model 5000 could be attrib(cid:173)
`utable to either Zeeman correction or finer control of
`furnace temperature.
`Comparisons of the graphite furnace procedures with
`other analytical methods further show the validity of the
`GFAAS techniques when applied to the determination of
`aluminum in CAPD solutions and associated raw materi(cid:173)
`als. One method, ICP-AES having a detection limit typi(cid:173)
`cally about 30 µg/liter, is not as sensitive as either graph(cid:173)
`ite furnace technique and is used strictly for some raw
`materials with relatively high aluminum levels. Typical
`results for lactic acid samples with aluminum at 3 levels
`and analyzed over a 3-day period with both techniques,
`are given in Table 1-B.
`Statistically some differences were found between the
`two techniques (p = 0.146). The ICP showed a positive
`bias of approximately 5% while no bias was found for the
`graphite furnace technique. The precision was not as good
`for GFAAS (3.8%) as for ICP-AES (1.5%). The reason
`for the positive bias in the ICP data is unknown at the
`present time, but it probably has to do with the relatively
`high concentration of organics in the samples which would
`cause differences in nebulization from the corresponding
`standards. There is a large difference in detection limits of
`the two techniques (0.35 µg/liter vs. 30 µg/liter). For
`samples in the mg/liter range, therefore, a considerable
`dilution was required to bring them down to the working
`range of 10-25 µg/liter for the G FAAS analysis, thus
`making this method more prone to error at this level.
`The other comparison of techniques, that of GFAAS
`with the aluminum-lumogallion fluorescence procedure,
`
`72
`
`Journal of Parenteral Science & Technology
`
`Eton Ex. 1070
`6 of 9
`
`
`
`was by far the most interesting. Originally performed as a
`confirmation of the G FAAS procedure, the comparison
`soon demonstrated that the fluorescence method, with a
`detection limit of about 1 µg/liter, worked quite satisfac(cid:173)
`torily in its own right for aluminum determination at µg/
`liter levels in CAPO solutions. The main difficulty with
`the fluorescence procedure is that it can be somewhat
`tedious to perform and, therefore, more prone to contami(cid:173)
`nation than the graphite furnace methods. In addition, a
`multiple standard addition technique must be used to
`circumvent interference due to the presence of sodium
`lactate.
`The results in Table 1-C indicated that no bias (p =
`0.34) was detected between the two techniques. There(cid:173)
`fore, the fluorescence method was found to be a viable
`alternative for analysis when a graphite furnace instru(cid:173)
`ment is unavailable.
`Results of the interlaboratory comparison of Travenol
`Laboratories, Morton Grove, IL, USA, and Travenol
`Laboratories, Nivelles, Belgium, are given in Figure 1.
`Different samples from the same lots were analyzed by
`both groups. The effect of laboratory on the observed
`assay value was studied by a statistical model which ac(cid:173)
`counted for the main effects of laboratory and lot and
`their interaction. The laboratory effect was not found to
`be significant (p = 0.93). Differences in precision of the
`two sets of data were attributed to differences in contami(cid:173)
`nation control in the two laboratories and the higher de(cid:173)
`tection limit of the Pye Unicam instrument.
`
`DIANEAL Solution Comprehensive Study
`
`The summary of results in Table III shows a definite
`plant-to-plant variation in aluminum levels. A statistical
`analysis of the results in Table III shows a significant
`difference (p = 0.0001) between the aluminum results
`obtained on solutions produced at facilities B through F
`and those for solutions produced at the other facilities.
`Eight facilities produced CAPO solutions with aluminum
`levels consistently less than 10 µg/liter. Two of the facili(cid:173)
`ties averaged near the 10-µg/liter aluminum level, and the
`remaining three averaged above 10 µg/liter. With only a
`few exceptions, the range of aluminum values of solutions
`produced within a facility was quite narrow, suggesting
`that the possible sources of the highe