Determination of Aristolochic Acid in Botanicals and Dietary Supplements by Liquid Chromatography with Ultraviolet Detection and by Liquid Chromatography/Mass Spectrometry: Single Laboratory Validation Confirmation

Abstract

The presence of aristolochic acid in some dietary supplements is a concern to regulators and consumers. A method has been developed, by initially using a reference method as a guide, during single laboratory validation (SLV) for the determination of aristolochic acid I, also known as aristolochic acid A, in botanical species and dietary supplements at concentrations of approximately 2 to 32 μg/g. Higher levels were determined by dilution to fit the standard curve. Through the SLV, the method was optimized for quantification by liquid Chromatography with ultraviolet detection (LC-UV) and LC/mass Spectrometry (MS) confirmation. The test samples were extracted with organic solvent and water, then injected on a reverse phase LC column. Quantification was achieved with linear regression using a laboratory automation system. The SLV study included systematically optimizing the LC-UV method with regard to test sample size, fine grinding of solids, and solvent extraction efficiency. These parameters were varied in increments (and in separate optimization studies), in order to ensure that each parameter was individually studied; the test results include corresponding tables of parameter variations. In addition, the chromatographic conditions were optimized with respect to injection volume and detection wavelength. Precision studies produced overall relative standard deviation values from 2.44 up to 8.26% for aristolochic acid I. Mean recoveries were between 100 and 103% at the 2 μg/g level, between 102 and 103% at the 10 μg/g level, and 104% at the 30 μg/g level.

The U.S. Food and Drug Administration (FDA) is concerned about botanicals and botanical containing products that may contain aristolochic acid, either naturally or through adulteration. Some of these botanicals are known to contain, or are suspected of containing, aristolochic acid. These botanicals include Aristolochia species (spp.), Asarum spp., and Bragantia spp. Botanicals that may be adulterated with aristolochic acid include Akebia spp., Clematis spp., Coculus spp., Diploclisia spp., Menispernum spp., Sinomenium spp., and Stephania spp. Additional names associated with these species include Mu tong, Fang ji, Guang fang ji, Fang chi, Kan-Mokotsu, and Mokutsu. The similarity of names, as well as the identification and handling of botanicals in preparation of traditional Chinese medicines, makes it difficult to avoid inadvertent adulteration of both botanicals and dietary supplements. Reported serious adverse effects in humans include nephropathy and urothelial carcinomas. Rodents administered aristolochic acid developed lymphoma and cancers in the kidney, bladder, stomach, and lung, implying a potential increased risk of developing malignancies in humans (1, 2). The test materials used in the validation were Aristolochia spp. root, Aristolochia manschuriensis stem, Akebia trifoliata stem, Clematis armandii stem at 20 ppm high fortification, Stephania tetrandra root at 10 ppm low fortification, tablets containing a 20 ppm high fortification, and tablets with a 10 ppm low fortification. The fortified materials were spiked with ground Aristolochia spp. root. All test materials were authenticated by voucher or analysis for aristolochic acid I (Aristolochia spp. root). Due to the amount of Aristolochia spp. root needed for some experiments in the single laboratory validation (SLV), the ground Aristolochia spp. root was further diluted, as necessary, with ground Stephania tetrandra root (negative control) to produce the necessary amounts of Aristolochia spp. root containing aristolochic acid I.

Experimental

Apparatus

1. Balances.—Analytical, capable of weighing to 0.00001 g and top loading to 0.001 g.

2. Mini grinder.—Coffee and spice grinder, Type F408 (KRUPS, Ecully, Cedex, France).

3. Multi-wrist shaker.—Model 75 (Barrell Corp., Pittsburgh, PA).

4. Liquid chromatography (LC) column.—Zorbax SB-CIS, 5 μ 3.0 mm id × 25 cm (Agilent, Palo Alto, CA).

5. Sieve.—400 μm (U.S. standard No. 40; Fisher Scientific, Fair Lawn, NJ).

6. Mass spectrometer.—Equipped with a diode array detector, 1100 Series (Agilent).

7. UV-Vis spectrophotometer.—Capable of measuring absorbace at 390 nm (Agilent, Palo Alto, CA).

8. LC system.—Equipped with a UV-Vis detector, 1100 Series (Agilent).

9. HPLC column.—Zorbax SB-CIS, 5 μm, 2.1 mm id × 5 cm (for LC/MS triple quad; Agilent).

10. Syringe filter.—13 mm with 0.45 μm PTFE membrane (Pall Life Sciences, Ann Arbor, MI).

11. Data system for HPLC.—Empower Version Build 1154 (Waters Corp., Milford, MA).

12. Data system for Micromass Quattro.—Masslynx Version 4.1 (Waters Corp.).

13. HPLC system.—Equipped with a mass spectrometer for triple quad, Micromass Quattro LC (Waters Corp.).

(Note: Equivalent equipment may be substituted.)

Reagents

1. Acetonitrile.—HPLC grade, No. A996-4 (Fisher Scientific).

2. Water.—Prepared with a Milli-Q purification system (Millipore Corp., Bedford, MA).

3. Trifluoroacetic acid.—99.8%, No. 3–3076 (Supelco, Bellefonte, PA).

4. Ethanol.—200 proof, HPLC/spectrophotometric grade, No. 459828-4L (Sigma-Aldrich, St. Louis, MO).

5. Methanol.—HPLC grade, No. A454-4 (Fisher Scientific).

6. Formic acid.—ACS 88+%, No. 36504 (Alfa Aesar, Ward Hill, PA).

7. Ammonium acetate.—HPLC grade, No. A639-500 (Fisher Scientific).

(Note: Equivalent reagents may be substituted.)

Reference Standard

Aristolochic acid I (also known as aristolochic acidA).—94.03% pure, No. 011002–10 (ChromaDex, Santa Ana, CA).

Preparation of Standard Solutions

(a) Stock solution (ca 200 μg/mL)

Prepared by weighing ca 0.002 g aristolochic acid to 0.00001 g and transferring into a 10 mL volumetric flask. The aristolochic acid was dissolved and diluted to volume with acetonitrile. One milliliter was diluted to 5 mL with ethanol and read on a spectrophotometer at 390 nm. The molar extinction coefficient (∈) for aristolochic acid from the Merck Index is 6500 (concentration in g-mole/L).

$$\text{Concn},\mu \text{g}/\text{mL}=\frac{\text{Abs}}{6500(\in )}\times \frac{341.29\text{g}/\text{mole}}{1000\text{mL}}\times \frac{5\text{mL}}{1\text{mL}}\times \frac{1000000\mu \text{g}}{\text{g}}\times \text{purity}$$

where Abs = absorbance at 390 nm. The solution is stable for 30 days when stored protected from light in a refrigerator set to maintain 5 ± 3°C.

(Note: The concentration was verified on the spectrophotometer prior to making working standards.)

(b) Working standard solutions

Prepared by diluting the standards at 6 concentration levels from 0.040 to 0.64 μg/mL in a solution of acetonitrile and purified water (50 + 50, v/v). The solutions are stable for 14 days when stored protected from light in a refrigerator set to maintain 5 ± 3°C.

Preparation of Reagents

Mobile phase A was prepared by adding 0.1% (v/v) trifluoroacetic acid to purified water. Mobile phase B consisted of 0.1% (v/v) trifluoroacetic acid in acetonitrile. The extraction solvent was acetonitrile and purified water (50 + 50, v/v).

Procedure

The material was ground to pass through a 400 μm sieve and mixed well. Approximately 2 g of test sample was weighed into an appropriate flask with a cap, and 100 mL of extraction solvent was added. The solution was shaken for a minimum of 30 min on a wrist shaker set at full stroke. The test samples greater than 32.0 μg/g in concentration were diluted to fit on the standard curve. The extracts were passed through a 0.45 μm filter and were then ready for LC analysis and quantification. The LC parameters are shown in Table 1. Each concentration standard was injected at least twice. Standards were injected after every 4 test samples and also at the end of the series.

Table 1.

HPLC-UV parameters

Detection	Ultraviolet detector (390 nm)
Column temperature	40°C
Flow rate	Approximately 0.5 mL/min. Flow rate should be adjusted so that desired retention time will be obtained
Injected volume	25 μL
Injection time	40 min
Pump program
Time, min	%Aa	%Ba	Gradient curve
0	80	20	NAb
25	30	70	1
30	0	100	1
31	80	20	1
40	80	20	NA

^aMobile phase A = 0.1% (v/v) trifluoroacetic acid and purified water; Mobile phase B = 0.1% (v/v) trifluoroacetic acid and acetonitrile.

^bNA = Not applicable.

Quantification and Confirmation

The standard curve was obtained by a linear regression of all of the peak areas and the concentrations of the working standards. Determination of the aristolochic acid concentration of the injected test sample was calculated from the regression analysis. The concentration was calculated using the following equation:

$$(\text{C}\times \text{V}\times \text{D})/\text{W}=\mu \text{g}/\text{g}$$

where C = concentration from regression analysis (μg/mL); V = extraction solvent volume (mL); D = dilution factor, if any; and W = test sample weight (g).

When aristolochic acid I was detected at greater than or equal to the 2 μg/g by LC-UV, confirmation was done using LC with tandem mass spectrometry (LC/MS/MS). The spectrum of standard was used for the confirmation of presence of aristolochic acid I with comparison to the spectrum of the test sample of like retention time. The peak area detected in the standard for 3 product ions was divided by the peak area of 1 of 3 product ions and averaged for all standards when multiple injections were made. The test sample ratio of confirmation ions to the chosen product ion should be ±10% (arithmetic difference, not relative difference) of the averaged standard ratio. For example, if an average ratio for the standards is 50%, the window for the test sample ratio is 40 to 60%. The chromatographic conditions are shown in Table 2 and the ions monitored are shown in Table 3.

Table 2.

Chromatographic conditions for LC/triple quad-MS

HPLC column	Zorbax SB-C18, 5 μm, 2.1 mm id × 50 mm, Agilent
Mobile phase A	5.0% methanol, 0.1% formic acid, and 10 mM ammonium acetate in purified water
Mobile phase B	1:1 acetonitrile-methanol containing 0.1% (v/v) formic acid and 1 0 mM ammonium acetate in purified water
Pump program
Time, min	%A	%B	Gradient curve
0	70	30	NAa
13	30	70	1
15	0	100	1
16	70	30	1
20	70	30	NA
Column temperature		40°C
Flow rate		0.2 mL/min
Injection volume		25 μL
Sample introduction		Electrospray positive (ES+)
Mode
Mass range	Daughter scan for specific ion, dwell = 0.25 s
Source temperature	150°C
Desolvation temperature	350°C
Desolvation gas flow	600 L/h
Cone gas flow	60L/h

^aNA=Not applicable.

Table 3.

Ions to be monitored for LC/MS confirmation

Compound	Parent ion, m/z	Daughter ions, m/z	Cone, V	Collision, eV
Aristolochic acid	359.1	265, 281, 296	18	40,40,30

Results and Discussion

Several tests were done to evaluate the optimum extraction parameters. The publication by Kite et al. (3) showed 70% aqueous methanol, withoutpH adjustment, to be optimum. An 18 h agitation time was also used. Using the Aristolochia root material, the Kite et al. method (3), and the reference method (4), several extraction efficiency parameters were evaluated. These included the test sample grind size, test sample size, extraction solvent volume and composition, and extraction time. In addition, the effects of repetitive dilution and ASE were measured.

To determine the impact of the test sample grind size, 3 preparations of Aristolochia spp. root were ground: the first to an average size of about 600 μm [size range of approximately 150 μm (U.S. Standard Sieve #100) to approximately 1500 μm (approximately U.S. Standard #14)]: the second to pass through a 600 μm sieve (U. S. Standard #30); and the third to pass through a 400 μm sieve (U.S. Standard #40). The optimum grind size was found to be approximately 400 μm (Table 4).

Table 4.

Test sample grind size verification using Aristolochia spp. root

Grind size, μm	Replicate	Aristolochic acid concn, μg/g
150 to 1500	1	903
	2	860
	3	862
	4	904
	5	887
Mean		883
SDa		21.4
RSDrb		2.42
600	1	969
	2	960
	3	942
	4	983
	5	1010
Mean		973
SD		25.6
RSDr		2.63
400	1	1030
	2	1060
	3	1040
	4	1060
	5	1040
Mean		1050
SD		13.4
RSDr		1.28

^aSD = Standard deviation.

^bRSD_r = % Relative standard deviation, intralaboratory precision.

The test sample size parameter was evaluated using portions of 0.25, 0.5, 1, 2, and 3 g that were tested using the various experimental extraction solvents. The optimum test sample size was found to be approximately 2 g using the optimum extracted solvent composition of acetonitrile and Milli-Q water (50 + 50, v/v). The use of 2 g allowed for achieving the desired 2 μg/g limit of quantification (LOQ). Although not optimum, the use of 3 μg/g only slightly decreased the extraction efficiency, thereby allowing for a possible lower LOQ of 1.3 μg/g. These data are presented in Tables 5 and 6 and Figures 1–10.

Table 5.

Aristolochic acid concentration in Aristolochia spp. root diluted with 5. tetrandra root using acetonitrile–water as the extraction solvent (results by test sample weight)

Test sample weight, g	% Acetonitrile	Aristolochic acid concn, μg/g
0.25	25	646
	50	623
	60	604
	70	596
	80	619
	100	224
0.50	25	648
	50	621
	60	585
	70	595
	80	629
	100	195
1.00	25	625
	50	611
	60	605
	70	608
	80	593
	100	171
2.00	25	581
	50	613
	60	611
	70	610
	80	596
	100	164
3.00	25	534
	50	594
	60	601
	70	605
	80	569
	100	139

Table 6.

Aristolochic acid concentration in Aristolochia spp. root diluted with S. tetrandra root using methanol-water as the extraction solvent (results by test sample weight)

Test sample weight, g	% Methanol	Aristolochic acid concn, μg/g
0.25	60	627
	70	594
	75	588
	80	610
	85	576
0.50	60	550
	70	553
	75	588
	80	578
	85	541
1.00	60	571
	70	569
	75	576
	80	584
	85	541
2.00	60	610
	70	578
	75	584
	80	553
	85	530
3.00	60	576
	70	541
	75	526
	80	520
	85	489

Figure 1.

0.25 g test sample of Aristolochia spp. root diluted with S. tetrandra root with varying percentages of acetonitrile–water as the extraction solvent.

Figure 10.

3.00 g test sample of Aristolochia spp. root diluted with S. tetrandra root with varying percentages of methanol–water as the extraction solvent.

The extraction solvent volume used was 100 mL. The reference method extraction solvent, methanol and 10% formic acid in water (80 + 20, v/v), was thoroughly evaluated. Aqueous methanol solvents consisting of 60, 70, 75, 80, and 85% methanol were tested to better determine if the methanol level suggested by Kite et al. (3) was adequate. Aqueous acetonitrile solvents consisting of 25, 50, 60, 70, 80, and 100% acetonitrile were also tested. When optimum aqueous methanol and acetonitrile solutions were determined, the use of 10% formic acid in water was evaluated at the optimum conditions found. The presence of formic acid had no effect on the results; therefore, its use was not further considered (Tables 7–9 and Figures 11–21). The optimum extraction solvent composition was found to be a solution of acetonitrile and Milli-Q water (50 + 50, v/v). Acetonitrile and purified water combinations gave higher extraction efficiency results compared to the methanol and purified water. The results varied little from 25 to 70% acetonitrile, but 50% acetonitrile gave slightly higher results. The column mobile phase composition and the use of isocratic versus gradient elution, relative to the reference method, were considered after optimizing the extraction solvent composition. The effect on the LOQ was also considered.

Table 7.

Aristolochic acid concentration in Aristolochia spp. root diluted with S. tetrandra root using acetonitrile-water as the extraction solvent (results by percent acetonitrile)

% Acetonitrile	Test sample weight, g	Aristolochic acid concn, μg/g
25	0.24690	646
	0.50260	648
	1.0224	625
	1.9969	581
	3.0289	534
50	0.25730	623
	0.51120	621
	1.0871	611
	2.0253	613
	3.0004	594
60	0.25350	604
	0.50390	585
	1.0615	605
	2.0270	611
	3.0651	601
70	0.25060	596
	0.25760	595
	1.0692	608
	2.0106	610
	2.9987	605
80	0.25850	619
	0.50500	629
	1.0219	593
	2.0515	596
	2.9842	569
100	0.2504	224
	0.5107	195
	0.9947	171
	2.0486	164
	3.0359	139

Table 9.

Extraction solvent evaluation (10% formic acid versus Milli-Q water) in Aristolochia spp. root diluted with S. tetrandra root

	Aristolochic acid concn, μg/g
Replicate	Acetonitrile–Milli-Q water (50 + 50, v/v)	Acetonitrile–10% formic acid (50 + 50, v/v)
1	589	585
2	627	634
3	642	637
4	646	615
5	664	636
Meana	634	621
SDb	28.2	22.3
RSDr	4.45	3.59
	Methanol-Milli-Q water (75 + 25, v/v)	Methanol-1 0% formic acid (75 + 25, v/v)
1	575	576
2	608	594
3	586	596
4	591	609
5	600	600
Mean	592	595
SD	12.7	12.1
RSDr	2.15	2.03

^aSD = Standard deviation.

^bRSD_r = % Relative standard deviation, intralaboratory precision.

Figure 11.

Acetonitrile–water (25 + 75, v/v) as the extraction solvent with varying test sample weights of Aristolochia spp. root diluted with S. tetrandra root.

Figure 21.

Methanol–water (85 + 15, v/v) as the extraction solvent with varying test sample weights of Aristolochia spp. root diluted with S. tetrandra root.

With optimized parameters, test samples were analyzed with sonication of 30, 40, 50, and 60 min and compared to a wrist action shaker for the same time periods to evaluate the extraction time. These were compared to an 18 h agitation. There was no advantage seen for more than 30 min extraction with a wrist action shaker or sonication. All results were within the acceptable range, and the wrist action shaker results had a slightly smaller standard deviation. The optimum extraction method was, therefore, determined to be a 30 min time period using a wrist action shaker having an agitation speed of approximately 400 cycles/min with a travel distance of 2.5 cm (Table 10 and Figures 22 and 23).

Table 10.

Extraction time comparison between wrist shaker and sonication for Aristolochia spp. root diluted with S. tetrandra root

	Aristolochic acid concn, μg/g
Extraction time	Replicate 1	Replicate 2	Replicate 3	Mean
Wrist shaker
30 min	637	647	655	646
40 min	623	645	650	639
50 min	618	642	645	635
60 min	617	636	637	630
Overall mean				638
Overall SDa				6.76
Overall RSDrb				1.06
18h	633	647	N/Ac	640
Sonication
30 min	580	671	654	635
40 min	572	662	649	628
50 min	574	592	618	595
60 min	600	644	643	629
Overall mean				622
Overall SD				18.1
Overall RSDr				2.91

^aSD = Standard deviation.

^bRSD_r = % Relative standard deviation, intralaboratory precision.

^cN/A = Not applicable.

Figure 22.

Wrist shaker extraction time for Aristolochia spp. root diluted with S. tetrandra root.

Figure 23.

Sonication extraction time for Aristolochia spp. root diluted with S. tetrandra root.

Using repetitive dilutions, the optimum extraction solvent and time were verified. After the initial 30 min extraction, half of the extract was analyzed and the remaining half was brought to volume and extracted for an additional 30 min. Half of the extract from the diluted solution was then analyzed and the remaining half was brought to volume, extracted for30 min, and analyzed again. The results from the 3 extractions for both wrist action shaking and sonication had minimal variability and were within the target relative standard deviation (RSD_r) values, showing that the repetitive dilution did not extract any more aristolochic acid. However, the wrist action shaker extraction overall mean result was slightly higher than the overall mean result for sonication, further verifying that the wrist action shaker extraction method was optimal (Table 11).

Table 11.

Repetitive dilution verification comparison between wrist shaker and sonication for Aristolochia spp. root diluted with S. tetrandra root

		Aristolochic acid concentration, μg/g
Extraction interval	Extraction time, min	Replicate 1	Replicate 2	Replicate 3	Mean
Wrist shaker
Initial	30	610	626	635	624
After 1st dilution	30	619	617	642	626
After 2nd dilution	30	611	664	679	651
Overall mean					634
Overall SDa					15.0
Overall RSDrb					2.37
Sonication
Initial	30	599	599	610	603
After 1st dilution	30	583	610	589	594
After 2nd dilution	30	586	641	649	625
Overall mean					607
Overall SD					15.9
Overall RSDr					2.62

^aSD = Standard deviation.

^bRSD_r = % Relative standard deviation, intralaboratory precision.

Using the ASE instrument with the appropriate optimum parameters, the botanical residue (Table 12) from the repetitive dilution section and a fresh test sample (Table 13) were separately but exhaustively extracted. The ASE consisted of 3 consecutive extractions of the same test sample using fresh solvent for each extraction. The levels of aristolochic acid I were determined in both the botanical residue and the fresh test sample extracts to verify that the extraction efficiency was indeed optimized. The amount of aristolochic acid found from the botanical residue extracts was less than 1% of the total amount from the repetitive dilution experiment. The mean of the 3 fresh test samples that were treated by ASE, compared to the overall mean wrist action shaker result, were approximately 1 standard deviation different, 653 μg/g (Table 13) versus 634 μg/g (Table 11), further substantiating good extraction optimization.

Table 12.

Accelerated solvent extraction of repetitive dilution residue of Aristolochia spp. root diluted with S. tetrandra root

		Aristolochic acid concn, μg/g
Extraction type	Replicate	ASEa cut 1	ASE cut 2	ASE cut 3
Wristb	1	3.56	0.527c	0.291c
	2	5.16	0.0554c	0.0259c
	3	4.48	0.0948c	0.0111c
Sonicateb	1	8.23	0.0997c	NDd
	2	4.26	0.783c	0.0803c
	3	5.28	0.274c	ND

^aASE =Accelerated solvent extraction.

^bAfter repetitive dilution, residue was extracted with ASE.

^cExtrapolated values.

^dND = Not detected.

Table 13.

Accelerated solvent extraction of Aristolochia spp. root diluted with S. tetrandra root

		Aristolochic acid concn, μg/g
Extraction type	Replicate	ASEa cut 1	ASE cut 2	ASE cut 3	Total ASE cuts
ASE	1	555	88.7	17.0	661
	2	582	45.1	12.6	640
	3	490	140	28.6	659
Mean					653
SDb					11.6
RSDrc					1.78

^aASE = Accelerated solvent extraction.

^bSD = Standard deviation.

^cRSD_r = % Relative standard deviation, intralaboratory precision.

Limit of Quantification

After optimization, the LOQ was tested to meet a target value of 2 μg/g (ppm) by LC-UV. The LOQ was further tested by setting the UV detector at several wavelengths, to include at least 390, 322, and 313 nm. The injection volume was also considered. The optimum wavelength was found to be 390 nm (Figures 24–26). Varying the wavelength had no effect on the theoretical plate count, but at 322 and 313 nm a second peak appeared with a peak resolution of approximately 1.3, that could interfere with the aristolochic acid determination. A gradient should be used for the best signal-to-noise ratio at the LOQ (Figures 27–30). The gradient mobile phase gave a plate count of 187 000, while the isocratic mobile phase gave plate counts a factor of 10 lower. The injection volumes tested were 10, 25, 50, and 100 μL. The plate counts were 164 000, 182 000, 176 000, and 146 000, respectively. The best injection volume based on plate count was determined to be 25 μL (Figures 31–34).

Figure 24.

LC-UV chromatogram of a 0.040 μg/mL standard at 390 nm.

Figure 26.

LC-UV chromatogram of a 0.040 μg/mL standard at 313 nm.

Figure 27.

LC-UV chromatogram at 390 nm with isocratic mobile phases A and B (55 + 45, v/v).

Figure 30.

Liquid chromatogram at 390 nm using the gradient elution program in Table 1.

Figure 31.

0.040 μg/mL standard, 10 μL injection using the gradient elution program in Table 1.

Figure 34.

0.040 mg/mL standard, 100 μL injection using the gradient elution program in Table 1.

Calibration Curve

At least 1 calibration curve was generated on 3 separate days. Two different analysts conducted this validation process. The calibration range encompassed approximately the expected concentration of each extracted and diluted test material range of 2.00 to 32.0 μg/g. A calibration standard curve (6 concentration levels ranging from 0.040 to 0.640 μg/mL) was injected twice with each analytical run. Example calibration curve data from each of the 3 days during the precision determination are shown in Table 14. Linearity, calibration curve precision, and tailing factor were all evaluated.

Table 14.

Example calibration curve data from precision determination

Theoretical standard concn, μg/mL	Standard concn, μg/mL, calculated from linear regression	% Deviation from theoretical
Day 1
0.03895	0.03429	−12.0
0.03895	0.03482	−10.6
0.03902	0.03980	1.99
0.03926	0.03977	1.29
Mean		−4.83
0.1169	0.1106	−5.40
0.1169	0.1134	−3.02
0.1171	0.1135	−3.09
0.1171	0.1201	2.57
Mean		−2.24
0.1948	0.1951	0.156
0.1948	0.1974	1.32
0.1951	0.1942	−0.458
0.1951	0.1933	−0.902
Mean		0.0290
0.3246	0.3341	2.92
0.3246	0.3339	2.87
0.3252	0.3261	0.276
0.3252	0.3254	0.060
Mean		1.53
0.4859	0.4966	2.21
0.4869	0.4958	1.84
0.4878	0.4908	0.623
0.4878	0.4872	−0.124
Mean		1.14
0.6492	0.6413	−1.22
0.6492	0.6345	−2.27
0.6504	0.6511	0.108
0.6504	0.6482	−0.345
Mean		−0.932
Day 2
0.03895	0.03483	−10.6
0.03895	0.03502	−10.1
0.03902	0.03490	11.8
0.03902	0.03559	9.65
Mean		0.188
0.1169	0.1128	−3.52
0.1169	0.1126	−3.69
0.1171	0.1150	1.86
0.1171	0.1163	0.645
Mean		−1.18
0.1948	0.1975	1.37
0.1948	0.1928	−1.01
0.1951	0.1966	−0.775
0.1951	0.1944	0.368
Mean		−0.0118
0.3246	0.3286	1.23
0.3246	0.3329	2.57
0.3252	0.3305	−1.61
0.3252	0.3301	−1.49
Mean		0.175
0.4859	0.5054	4.00
0.4869	0.4986	2.41
0.4878	0.4965	−1.74
0.4878	0.4943	−1.31
Mean		0.840
0.6492	0.6350	−2.19
0.6492	0.6357	−2.08
0.6504	0.6426	1.21
0.6504	0.6424	1.24
Mean		−0.455
Day 3
0.03902	0.03664	−6.09
0.03902	0.03871	−0.786
0.03902	0.03610	8.09
0.03902	0.03625	7.65
Mean		2.22
0.1171	0.1179	0.670
0.1171	0.1212	3.51
0.1171	0.1161	0.862
0.1171	0.1185	−1.15
Mean		0.973
0.1951	0.1925	−1.34
0.1951	0.1901	−2.54
0.1951	0.1958	−0.373
0.1951	0.1942	0.474
Mean		−0.945
0.3252	0.3255	0.101
0.3252	0.3297	1.38
0.3252	0.3255	−0.0827
0.3252	0.3297	−1.36
Mean		0.00957
0.4878	0.4870	−0.172
0.4878	0.4943	1.33
0.4878	0.4941	−1.28
0.4878	0.4927	−1.00
Mean		−0.281
0.6504	0.6467	−0.565
0.6504	0.6490	−0.216
0.6504	0.6459	0.704
0.6504	0.6445	0.922
Mean		0.211

The relative response (peak area) of the analyte versus concentration was used to construct the calibration curve using a least-squares linear regression method. Calibration curves had a correlation coefficient (r) ranging from 0.99902 to 0.99996.

The back-calculated values for each concentration were within or equal to the targets of ±9% of theoretical (within or equal to ±10% at the LOQ) for within- and between-day precision, except within-day precision ranged from 10.6 to 12.0% for Day 1 LOQ and 10.1 to 11.8% at Day 2 LOQ (Table 11).

The U.S. Pharmacopeia tailing factor (T_f) was used.

$${\text{T}}_{\text{f}}=\text{ac}/2\text{ab}$$

where ac = peak width at 5% of the peak height and ab = front half-width measured from the leading edge to a perpendicular dropped from the peak apex. The tailing factor was no more than 1.5 (1.14 to 1.17) for the 6 aristolochic I standard peak concentrations (Figure 35).

Figure 35.

Tailing factors for standards using the gradient elution program in Table 1.

Test Material Precision

At least 3 replicates of each of the 7 test materials were analyzed on each of 3 separate days (Table 15). Using a minigrinder, 20 tablets were homogenized for each determination. Two different analysts were used during this validation process. The precision fell within or equal to the target RSD_r for intraassay replicates, with the exception of Day 3 A. manschuriensis stem, Day 1 Stephania tetrandra root (10 μg/g fortification), Day 2 and 3 tablets (10 μg/g fortification), and Day 2 and 3 tablets (20 μg/g fortification). Overall combined (interassay) replicates met RSD_r targets, with the exception of the 20 ppm high fortification tablets.

Table 15.

Test material precision

	Aristolochic acid concn, μg/g			Days 1–3 combined
Replicate	Day 1	Day 2	Day 3	Mean, μg/g	SDa	RSDrb	PRSDRc	HorRatd	Target RSDre
Aristolochia spp. root
1	1410	1410	1410	1450	47.3	3.26	5.33	0.6	2.67–3.56
2	1490	1490	1420
3	1490	1520	1400
Mean	1460	1470	1410
SD	46.2	56.9	10.0
RSDr	3.16	3.87	0.709
Aritolochia manschuriensis stem
1	2890	2790	2860	2830	85.3	3.01	4.82	0.6	2.41–3.21
2	2900	2940	2680
3	2820	2850	2720
Mean	2870	2860	2750
SD	43.6	75.5	94.5
RSDr	1.52	2.64	3.44
Akebia trifoliate stem
1	<2.00	<2.00	<2.00	<2.00	NAf	NA	NA	NA	NA
2	<2.00	<2.00	<2.00
3	<2.00	<2.00	<2.00
Mean	<2.00	<2.00	<2.00
SD	NA	NA	NA
RSDr	NA	NA	NA
Clematis armandii stem, 20 μg/g fortification with, Aristolochia spp.
1	23.4	22.4	22.6	22.5	0.550	2.44	9.96	0.2	4.98–6.64
2	23.2	22.0	22.1
3	22.4	21.7	22.7
Mean	23.0	22.0	22.5
SD	0.529	0.351	0.321
RSDr	2.30	1.60	1.43
S. tetrandra root:, 10 μg/g fortification with Aristolochia spp.
1	10.0	11.6	10.1	10.5	0.776	7.39	11.2	0.7	5.60–7.47
1	10.5	10.0	11.0
3	11.8	9.82	9.76
Mean	10.8	10.5	10.3
SD	0.929	0.980	0.641
RSDr	8.60	9.33	6.22
Tablets, 10 μg/g fortification with Aristolochia spp.
1	6.55	6.83	6.09	6.50	0.450	6.92	12.0	0.6	6.00–8.00
2	6.93	5.87	5.93
3	6.43	6.76	7.10
Mean	6.64	6.49	6.37
SD	0.261	0.535	0.634
RSDr	3.93	8.24	9.95
Tablets, 20 μg/g fortification with Aristolochia spp.
1	16.7	18.8	16.6	17.2	1.42	8.26	10.4	0.8	5.20–6.94
2	15.0	15.7	19.1
3	17.0	17.7	18.6
Mean	16.2	17.4	18.1
SD	1.08	1.57	1.32
RSDr	6.67	9.02	7.29

^aSD = Standard deviation.

^bRSD_r = % Relative standard deviation, intralaboratory precision.

^cPRSD_R = 2 C^−0.05, % relative standard deviation, predicted interlaboratory precision, C = mean concentration expressed as a mass fraction.

^dHorRat = RSD_r/PRSD_R.

^eTarget RSD_r = % Acceptable intralaboratory precision (0.500 to 0.667) PRSD_R.

^fNA = Not applicable.

When the initial test sample extract fell above the highest standard of the curve, the extract was diluted to fit on the curve. The target RSD_r was calculated and reported as follows:

$$\begin{array}{c}\text{Target}{\text{RSD}}_{\text{r}}=\\ \text{acceptable}\text{intralaboratory}\text{precision}(0.500\text{to}0.667)\\ {\text{PRSD}}_{\text{R}}\\ {\text{PRSD}}_{\text{R}}=2{\text{C}}^{-0.15}\end{array}$$

where C = mean concentration expressed as a mass fraction.

The HorRat values fell within or equal to 0.3 to 1.3 and were calculated and reported as follows:

$$\text{HorRat}\text{value}={\text{RSD}}_{\text{r}}/{\text{PRSD}}_{\text{R}}$$

Although the 20 ppm high fortification tablets did not meet the overall target RSD_r range, they did meet the HorRat value acceptance criteria. There may have been a lack of homogeneity for this material, and this will be investigated prior to a collaborative study.

Test samples of the negative control material, S. tetrandra root, were fortified in triplicate with the reference compound mixture at 2, 10, and 30 μg/g. Triplicate unfortified controls were analyzed concurrently. This was repeated for 2 additional days (Table 16).

Table 16.

Negative control recovery

Test sample identification	Replicate	Spike concn, μg/g	Amt of aristolochic acid found, μg/g	Recovery, %	Mean recovery, %	SDa	RSDrb
Day 1
S. tetrandra root spiked 2 μg/g	1	2.020	2.032	101	100	0.854	0.854
	2	2.020	2.006	99.3
	3	2.020	2.025	100
S. tetrandra root spiked 1 0 μg/g	1	10.02	10.27	102	102	0.577	0.566
	2	10.02	10.35	103
	3	10.02	10.19	102
S. tetrandra root spiked 30 μg/g	1	29.96	31.36	105	104	1.00	0.962
	2	29.96	31.01	104
	3	29.96	30.71	103
Day 2
S. tetrandra root spiked 2 μg/g	1	2.020	2.065	102	103	1.53	1.49
	2	2.020	2.072	103
	3	2.020	2.118	105
S. tetrandra root spiked 10 μg/g	1	10.02	10.57	105	103	2.00	1.94
	2	10.02	10.17	101
	3	10.02	10.37	103
S. tetrandra root spiked 30 μg/g	1	29.96	31.24	104	104	1.00	0.962
	2	29.96	31.44	105
	3	29.96	30.77	103
Day 3
S. tetrandra root spiked 2 μg/g	1	2.024	1.991	98.4	101	5.92	5.86
	2	2.024	1.968	97.2
	3	2.024	2.190	108
S. tetrandra root spiked 10 μg/g	1	10.07	10.51	104	103	0.577	0.560
	2	10.07	10.33	103
	3	10.07	10.34	103
S. tetrandra root spiked 30 μg/g	1	30.00	31.34	104	1 04	NAc	NA
	2	30.00	31.11	104
	3	30.00	31.19	104

^aSD = Standard deviation.

^bRSD_r = % Relative standard deviation, intralaboratory precision.

^cNA= Not applicable.

The mean recovery was found to be between 100 and 103% at the 2 μg/g level, between the 102 and 103% at the 10 μg/g level, and 104% at the 30 μg/g level. The column and UV detector selectivity of the reference method was investigated by injecting mixed standards of aristolochic acid I, II, IIIa, and VII at the highest standard level used for aristolochic acid I. The resolution was calculated as follows:

$${\text{R}}_{\text{S}}=2({\text{t}}_2-{\text{t}}_1)/({\text{w}}_1+{\text{w}}_2)$$

where t₁ and t₂ = retention times of peaks; w₁ and w₂ = peak widths measured at the baseline between tangents drawn to the peak sides. Selectivity was checked prior to starting the extraction efficiency experiment. There was no coelution or resolution of less than 1.3 of any reference material with aristolochic acid I (Figure 30). The LC/MS column and confirmation selectivity was investigated at the lowest standard level to ensure that confirmation by LC/MS was not compromised.

To determine ruggedness, extraction efficiency, LOQ, linearity, precision, accuracy, UV detector wavelength, injection volume, stability of the calibration curve, mobile phase composition, isocratic versus gradient, elution, and the use of more than 1 analyst were monitored. The retention time and column stability were also determined, with the average retention time over 12 months ranging from 20.8 to 21.1 min and the column plate count over the same time period declining less than 5%. In addition, grind size, test sample size in relationship to extraction volume, extraction solvent composition, extraction method, and extraction times were determined. Based on these parameters, the method was found to be rugged.

Confirmation data were determined when aristolochic acid I was detected at greater than or equal to the LOQ (i.e., in the test samples) or fortified negative control. The lowest level standard was used for confirmation comparison to the detected peak in the test sample using LC/MS. No low level test material was found to contain aristolochic acid I near the LOQ; therefore, A. trifoliata stem was fortified at the 2 μg/g level to accomplish the confirmation experiments. Standard instrument operating procedures were used to optimize the LC/MS for this purpose and to ensure system suitability. Quality assurance checks were also made. At least 3 product ions were chosen for confirmation, and 2 confirmation methods were performed (Table 17).

Table 17.

Confirmation by LC/triple quad-MS

		Response, area			Retention time. min			Confirmation method 1 ion/ion ratio, %		Confirmation method 2 ion/total ion area ratio, %
Identification	Replicate	Ion 296	Ion 281	Ion 265	Ion 296	Ion 281	Ion 265	Ion 281/296	Ion 265/296	Ion 296	Ion 281	Ion 265
0. 03878 μg/mL standard	1	539	214	123	11.4	11.5	11.4	40	23	62	24	14
	2	871	337	283	11.6	11.6	11.6	39	32	58	23	19
Mean								40	28	60	24	17
Akebia trifolata stem spiked at 2 μg/g	1	1048	279	228	11.6	11.6	11.4	27	22	67	18	15
	2	957	332	296	11.5	11.4	11.5	35	31	60	21	19
	3	1079	328	271	11.4	11.5	11.6	30	25	64	20	16
Mean								31	26	64	20	17

In the first confirmation method, the spectrum of standard was used for confirmation of presence of aristolochic acid I with comparison to the spectrum of the test sample of like retention time. The peak area detected in the standard for 3 product ions was divided by the peak area of 1 of 3 product ions and averaged for all standards when multiple injections were made. The test sample ratio of confirmation ions to the chosen product ion were to be within ±10% (arithmetic difference, not relative difference) of the averaged standard ratio. For example, if an average ratio for the standards was 50%, the window for test sample ratio was 40 to 60%. The ratios agreed within 10%.

In the second confirmation method, the spectrum of the standard was also used for the confirmation of presence of aristolochic acid I with comparison to the spectrum of the test sample of like retention time. The standard area ratio of at least 3 product ions versus the total area of the 3 was compared to the same area ratio of the test sample. The ratios were to agree within 10% for confirmation; however, only 2 of the 3 ratios fell within this range.

The first method is the preferred confirmation procedure. Similar confirmation may also be accomplished using LC/ion trap MS.

Conclusions

A method, further validated amongst several laboratories, will facilitate the accurate determination of the quality of botanicals and dietary supplements with respect to aristolochic acid. In addition, the use of the method may allow assessment of the relationship between levels of aristolochic acid in botanicals and dietary supplements and any alleged or reported adverse effects. Based on the results of this validation, the method for the determination of aristolochic acid I in botanicals and tablets is deemed acceptable for use for quantification by LC-UV with confirmation by LC/MS. It is also recommended that a collaborative study be initiated.

Figure 2.

0.50 g test sample of Aristolochia spp. root diluted with S. tetrandra root with varying percentages of acetonitrile–water as the extraction solvent.

Figure 3.

1.00 g test sample of Aristolochia spp. root diluted with S. tetrandra root with varying percentages of acetonitrile–water as the extraction solvent.

Figure 4.

2.00 g test sample of Aristolochia spp. root diluted with S. tetrandra root with varying percentages of acetonitrile–water as the extraction solvent.

Figure 5.

3.00 g test sample of Aristolochia spp. root diluted with S. tetrandra root with varying percentages of acetonitrile–water as the extraction solvent.

Figure 6.

0.25 g test sample of Aristolochia spp. root diluted with S. tetrandra root with varying percentages of methanol–water as the extraction solvent.

Figure 7.

0.50 g test sample of Aristolochia spp. root diluted with S. tetrandra root with varying percentages of methanol–water as the extraction solvent.

Figure 8.

1.00 g test sample of Aristolochia spp. root diluted with S. tetrandra root with varying percentages of methanol–water as the extraction solvent.

Figure 9.

2.00 g test sample of Aristolochia spp. root diluted with S. tetrandra root with varying percentages of methanol–water as the extraction solvent.

Figure 12.

Acetonitrile–water (50 + 50, v/v) as the extraction solvent with varying test sample weights of Aristolochia spp. root diluted with S. tetrandra root.

Figure 13.

Acetonitrile–water (60 + 40, v/v) as the extraction solvent with varying test sample weights of Aristolochia spp. root diluted with S. tetrandra root.

Figure 14.

Acetonitrile–water (70 + 30, v/v) as the extraction solvent with varying test sample weights of Aristolochia spp. root diluted with S. tetrandra root.

Figure 15.

Acetonitrile–water (80 + 20, v/v) as the extraction solvent with varying test sample weights of Aristolochia spp. root diluted with S. tetrandra root.

Figure 16.

Acetonitrile (100%) as the extraction solvent with varying test sample weights of Aristolochia spp. root diluted with S. tetrandra root.

Figure 17.

Methanol–water (60 + 40, v/v) as the extraction solvent with varying test sample weights of Aristolochia spp. root diluted with S. tetrandra root.

Figure 18.

Methanol–water (70 + 30, v/v) as the extraction solvent with varying test sample weights of Aristolochia spp. root diluted with S. tetrandra root.

Figure 19.

Methanol–water (75 + 25, v/v) as the extraction solvent with varying test sample weights of Aristolochia spp. root diluted with S. tetrandra root.

Figure 20.

Methanol–water (80 + 20, v/v) as the extraction solvent with varying test sample weights of Aristolochia spp. root diluted with S. tetrandra root.

Figure 25.

LC-UV chromatogram of a 0.040 μg/mL standard at 322 nm.

Figure 28.

LC-UV chromatogram at 390 nm with isocratic mobile phases A and B (60 + 40, v/v).

Figure 29.

LC-UV chromatogram at 390 nm with isocratic mobile phases A and B (65 + 35, v/v).

Figure 32.

0.040 μg/mL standard, 25 μL injection using the gradient elution program in Table 1.

Figure 33.

0.040 mg/mL standard, 50 μL injection using the gradient elution program in Table 1.

Table 8.

Aristolochic acid concentration in Aristolochia spp. root diluted with S. tetrandra root using methanol–water as the extraction solvent (results by percent methanol)

% Methanol	Test sample weight, g	Aristolochic acid concn, μg/g
60	0.25570	627
	0.49410	550
	1.0627	571
	2.0393	474
	2.9946	553
70	0.25910	594
	0.49900	553
	1.0067	569
	2.0604	544
	3.0549	572
75	0.24900	588
	0.50540	588
	1.0470	576
	2.0367	569
	2.9824	559
80	0.24850	610
	0.50050	578
	1.0436	584
	2.0494	553
	3.0190	530
85	0.25610	576
	0.50990	541
	1.0589	526
	2.0695	520
	3.0381	489