We prepared all reagents using laboratory deionized water (≥18 MΩ·cm, measurement of the resistance of DI water) filtered in-house using an Aqua Solutions Water Purification system (Jasper, GA). Sample diluent, carrier, and rinse solutions consisted of 0.01% ammonium pyrrolidine dithiocarbamate (APDC, Fisher Scientific, Waltham, MA), 0.4% (v/v) tetramethylammonium hydroxide (TMAH, 25% in H₂O, Alfa Aesar, Ward Hill, MA), 1% ethanol (Pharmco Products Inc., Brookfield, CT), and 0.05% Triton X-100 (Sigma Aldrich, St. Louis, MO). We added 20 μg/L of gallium (Ga) and scandium (Sc) internal standards (Inorganic Ventures, Christiansburg, VA) to diluent and carrier solutions. We used single element stock standards of Cr and Co from various sources (Inorganic Ventures, Christiansburg, VA; National Institute for Standards and Technology (NIST), Gaithersburg, MD; SPEX CertiPrep, Metuchen, NJ; High-Purity Standards, Charleston, SC to prepare calibration standards in 2% v/v nitric acid (HNO₃, environmental grade, GFS Chemicals, Columbus, OH) and 1% v/v double distilled hydrochloric acid (HCl, 30–35%, GFS Chemicals, Columbus, OH). All calibration standards are traceable to the NIST. We purchased human whole blood with ethylenediaminetetraacetic acid (EDTA) from Tennessee Blood Services Corporation (TBSC), Memphis, TN for matrix matching of the calibration curve and preparing in-house quality control (QC) material. For matrix matching, we used non-spiked blood, referred to as “base blood” in this paper. The sample blood tubes collected from 100 donors for Cr and Co leaching from stainless needles study were purchased from TBSC.

For blood sample analysis we used a calibration curve covering analyte concentrations up to 15 μg/L consisting of five calibrators with concentrations of 0.5, 1.5, 5.0, 7.5, and 15 μg/L ⁹. To prepare a set of calibrators, we used 0.25 mL of the standard, 0.25 mL of base blood, and 4.5 mL of diluent. We prepared in-house QC samples, long-term stability samples, and the 1st and 2nd blood tube samples by mixing 0.25 mL of the blood sample, 0.25 mL of water, and 4.5 mL of diluent. For the lot screening procedures, we used different calibration levels (0.5, 1, 2.5, and 5.0 μg/L) since the required concentration range is smaller. To prepare a set of these calibrators, we used 0.25 mL of the standard and 4.5 mL of diluent. For the lot screening sample, we used 0.25 mL of aqueous sample (screening solution) plus 4.5 mL of diluent.

We prepared two sets of pooled samples using human whole blood (EDTA disodium salt used as anti-coagulant) purchased from the TBSC (Institutional Review Board [IRB] TB-12-01, expiration November 18, 2020). We distinguished these as the low blood (A) pool, spiked with low concentrations of Cr and Co analytes, and the high blood (B) pool, spiked with higher concentrations of Cr and Co. Table SI in Supplemental Information (SI) summarizes the calculated concentration results for the two pools over 50 analytical runs on two different instruments and by two analysts. Calculated Cr and Co concentrations in pool A are 2.58 μg/L and 1.66 μg/L and for pool B are 12.26 μg/L and 8.89 μg/L, respectively.

We reserved specific portions of each of the A and B pools to use as long-term stability study samples. We used additional samples from each pool with each analytical run as bracketing quality control (QC) material samples. For our quality assurance and quality control programs, patient samples (or any samples) are bracketed with this QC material. If bracketing QC concentration falls outside of its 3SD range, the analytical run is marked as “fail,” and the sample analysis is repeated. In practice, we use more sophisticated evaluation criteria based on modified Westgard rules^{32, 33}. In this paper, we refer to 3SD range of QC material as a “quality assurance limit” (QAL).

We used a Thermo® iCAP™ Qa (upgraded to an iCAP Qc) ICP-MS with collision cell (QCell™, Thermo Fisher Scientific, Waltham, MA) equipped with a peristaltic pump, a PFA-ST MicroFlow Nebulizer, a peltier-cooled cyclonic spray chamber, a 2.0 mm quartz injector, platinum sampler cone, platinum skimmer cone, and skimmer cone insert 3.5 mm for all experiments. The instrument parameters can be found in our preceding Cr and Co method paper ⁹. We used an SC-4DX FAST autosampler (Elemental Scientific Inc., Omaha, NE) for the uptake of diluted blood samples for analysis and control of the FAST sample introduction timing. See our method paper for detailed information on the FAST method ⁹. Our laboratory used > 99.999% argon (Specialty Gases Southeast, Airgas, Atlanta, GA) for the plasma and nebulizer gases. We operated the collision cell in KED mode at 5 mL/min helium gas flow (99.999% grade, Airgas South, Kennesaw, GA). We prepared final dilutions (calibrators, QC, and unknown samples) for analysis using a Hamilton Microlab 625 advanced dual syringe diluter (Hamilton Company, Reno, NV), equipped with a 10 mL dispensing syringe and a 500 μL sampling syringe.

The Cr and Co lot screening procedure is to test 50 devices from a manufacturing lot to determine if the lot is suitable for use in collecting, analysis, and storage of blood samples. To prepare a device for Cr and Co lot screening, we either put screening solution in the device (e.g. vacutainers and cryovials), or we pass a screening solution through the device (e.g. needles, syringes). For Cr and Co lot screening we use water as a screening solution. Before use we test the water to make sure there are no detectable levels of Cr and Co. The volume of screening solution is determined by the amount of blood sample that would either pass through the device or be stored in a container. The contact time of the screening solution with each device depends on the type of device. For instance, for cryovials, we uncap the vials, add screening solution, and recap the vial. We invert the 25 odd-numbered cryovials and let the devices sit overnight (8 to 12 hours)¹⁷. If contamination is present, this allows us to determine whether it comes primarily from the vial (the even units that were left upright), the cap (all inverted vials), or both. A variety of types of devices are screened using this methodology. With needles, we pass the screening solution through the needle into a prescreened evacuated blood tube, thus the contact time with the screening solution is identical to the contact time of blood being passed through a needle¹⁷.

For Cr and Co lot screening we use the same instrumentation and quantification method as for analysis of blood samples. We prepared QC for the lot screening process by diluting NIST 1640a standard reference material (NIST, Gaithersburg, MD) with water using a 1 + 9 dilution (1 mL of NIST 1640a plus 9 mL of DI water), referred to hereafter as “1640a_10x.” QC passing criteria can be found in Table SII. If Cr or Co 1640a_10x concentrations are not within ± 10%, the analytical run fails and must be repeated.

For the Cr and Co lot screening procedure; we allow no more than 5% of the units to be defective in an acceptable lot with a 90% confidence interval¹⁷. We use Equation 1 to determine criteria for the maximum (max) allowable contribution to total concentration of the analyte that can be present in the device or container for it to pass lot screening. We consider a lot failing if it has more than one result for the analyte of interest above the max allowable contribution. If no devices are above the max allowable contribution, we consider the lot acceptable for use in the Cr and Co biomonitoring method. If only one device of the screened 50 fails, we use our scientific knowledge to decide whether this failure should be considered an outlier.

The “expected population mean” represents the expected mean concentration of the analyte of interest in the population to be studied. We use 0.151 μg/L and 0.56 μg/L for Co and Cr, respectively. The Co value is based on the geometric mean for blood Co from the 2015-2016 NHANES cycle³⁰. Since Cr does not have a geometric mean due to the majority of the values being below the LOD, we use 50^th percentile based on 107 samples analyzed by our laboratory in 2014. The “max percent contribution” is the percentage of the analyte allowed to be present in the device tested, in our laboratory we set that value to 10%. The “volume sample in device” stands for the amount of sample that will either pass through the device or be stored in the container tested. The “volume screening solution” is the amount of screening solution that passes through the device or is aliquoted in the storage container during screening. If the calculated “max allowable contribution” is below the limit of detection (LOD) of the method (Cr - 0.41 μg/L and Co – 0.06 μg/L), then the “max allowable contribution” is equal to 1.5 times LOD of the method (Cr – 0.62 μg/L and Co – 0.09 μg/L).

We used human whole blood based in-house QC material to assess the long-term storage stability of Cr and Co whole blood samples. We stored aliquots of the pooled materials (A and B) for one year at four different temperatures: −70°C, −20°C, 4°C, and RT. We evaluated −70°C and −20°C, since these are established long-term storage temperatures ^{35, 36}. We evaluated the recommended short-term storage temperatures of 4°C and room temperature (RT) to account for cases when samples are left at RT during sample processing or when refrigerators are not available. We chose not to evaluate samples stored at an elevated temperatures because based on a previous blood metals stability study for other elements, we know that the blood matrix becomes viscous or dry after two weeks when stored at 37°C ³⁷. Aliquots of the A and B pools were analyzed at 1 day; 1, 2, 4, and 6 weeks of storage; and at 2, 4, 6, 8, 10, and 12 months of storage (“E” stands for time event. E1=1 day; E2=1 week; E3=2 weeks; E4=4 weeks; E5=6 weeks; E6=2 months; E7=4 months; E8=6 months; E9=8 months; E10=10 months; E11=1year). At each time event, we analyzed three aliquots of A and B in duplicate per each temperature condition. After original analysis, we did not choose these aliquots for subsequent analysis. We bracketed each stability analytical run with QC material. We implemented the 3SD range quality assurance limit (QAL) (Table SI) to examine the long-term stability of A and B aliquots. If Cr and/or Co concentrations in A and B stability samples fall outside of our established QAL, Cr and/or Co whole blood samples are not considered stable. If they fall within our QAL, these samples are considered stable.

For statistical analysis, first we used a general linear regression model to fit the long-term stability data. The model included time (categorical) and temperature (categorical) as covariates and the measurement value was the outcome. Further, we evaluated the study results using the equivalence test ³⁸. We used Cr and Co concentration values from −70°C as the “reference” and compared Cr and Co concentration values from temperatures (−20°C, 4°C, RT) against the values at −70°C. The hypotheses were as follows: H0μTest/μRef<0.8orμTest/μRef>1.25 H10.8≤μTest/μRef≤1.25 with μ_Test: Cr and Co concentration values at −20°C, 4°C, and RT and μ_Ref: values at −70°C.

Next, we used Cr and Co concentration values from day 1 of storage (E1) as the “reference” and compared the Cr and Co concentration values from the other 10 events (E2=1 week; E3=2 weeks; E4=4 weeks; E5=6 weeks; E6=2 months; E7=4 months; E8=6 months; E9=8 months; E10=10 months; E11=1year.) against the values in E1. The hypotheses were as follows: H0μTest/μRef<0.8orμTest/μRef>1.25 H10.8≤μTest/μRef≤1.25 with μ_Test: values at 1, 2, 4, 6 weeks and 2, 4, 6, 8, 10, and 12 months and μ_Ref: value at 1 day.

If the ratio of μ_Test/μ_Ref was less than 1, we tested the hypothesis of μ_Test/μ_Ref < 0.8. If the ratio was greater than 1, we tested the hypothesis of μ_Test/μ_Ref > 1.25. If the p-value was less than 0.05 for hypothesis testing, we rejected the null hypothesis (H₀) and accepted the alternative hypothesis (H₁:0.8 ≤ μ_Test/μ_Ref ≤ 1.25), meaning that the test storage condition results were statistically equivalent (no difference) to the reference storage condition results. We conducted the statistical analysis using SAS software ³⁴.

Since developing the Cr and Co biomonitoring method in 2014 ⁹, we have screened 156 manufacturing lots to determine their suitability for laboratory and field use (data collected from November 2014 through January 2018), Table 1. Out of these lots, 22 failed for Cr and/or Co. Data in Table 1 reveal interesting trends and failure percentages for different types of devices. For all the screened devices our calculated “maximum allowable contribution” was below the LOD, thus we used 1.5 times LOD values of Cr - 0.09 μg/L and Co - 0.62 μg/L. We tested 58 manufactured lots of evacuated blood collection tubes and obtained a 21% failure rate (10 lots failed for Co (17%) and 4 for Cr (7%)). Evacuated tubes are widely used as blood collection devices. Since the possibility of contamination is high, prescreening them prior to use is particularly important. For both Cr and Co, we tested 6 lots of transfer pipettes with a 17% failure rate (1 lot) and 11 lots of polypropylene tubes with an 18% failure rate (2 lots). We analyzed only a few lots of these devices, but failures occurred, demonstrating the need for screening before use. We analyzed one lot of microvettes (fingerstick blood collection device) which failed for Cr. Six lots of alcohol pads were analyzed and two lots failed for Cr and Co. From our experience with other metal screening methods, alcohol pads often have unacceptable levels of metal contamination. Needles had 8% failure rate; out of 49 lots screened, four lots failed. Cryovials (16 lots), luer-adapters (7 lots), and syringes (2 lots) had no failures. The highest individual screening results from the screened lots from 2014 through 2018 for Cr were needles (4.62 μg/L), vacutainers (2.34 μg/L), and microvettes (1.54 μg/L) and for Co were vacutainers (0.92 and 0.89 μg/L) and alcohol pads (0.67 μg/L) , Table SIV. These preliminary results reveal that a significant number of lots had failures, thus supporting the necessity of screening all devices that come in contact with a patient sample, either in the laboratory or in the field, to reduce the likelihood of contaminating a patient sample

In general, Cr and Co contamination of human blood samples collected using stainless steel needles is considered low; however, when measuring very low concentrations of these metals for establishing reference ranges, even small levels of contamination could have an effect ^{11, 15, 21, 24, 27–29}. Therefore, we conducted a study comparing Cr and Co concentration in 1st and 2nd consecutively collected blood tubes using the one needle device to detect if there is any Cr and Co leaching out of the needles during blood draw procedure.

Our laboratory analyzed 200 samples from 100 donors collected by TBSC. TBSC personnel collected two sample tubes per donor: 1st and 2nd blood tubes were collected consecutively using the same medical set up (needle/luer adaptor), Figure 1a and b show graphic illustration of the Cr and Co concentration difference between 1st and 2nd blood tubes in patient samples (actual concentrations can be found in Table SIII). Figure 1a shows slightly higher results for Cr concentrations for the 2nd blood tube over the 1st in some donors, making the mean concentration difference negative (Table 2). Figure 1b shows the difference in Co concentrations between the 1st and 2nd blood tubes, there is no visual trend.

After the data collection, we performed paired sample t-test analysis to determine any statistically significant difference in Cr and Co concentrations between the 1st and 2nd blood tubes (Table 2). The paired sample t-test produced, p-values below 0.05, indicating there is a statistically significant difference between Cr and Co concentrations in 1st and 2nd blood tubes. The mean concentration difference between the 1st and the 2nd blood tube for Cr was negative 0.191 μg/L, meaning that the average Cr concentration in the 2nd blood tube was higher than in the 1st (Table 2). If Cr was leaching out of the stainless steel needle, we would expect mean Cr concentration in the 1st tube to be higher than the concentration in the 2nd, however, we saw the opposite effect. Out of 100 samples analyzed for the 1st blood tube, Cr concentration in 90 samples (90%) was below the method’s LOD limit (0.41 μg/L). For the 2nd blood draw, out of 100 samples analyzed, Cr concentration in 78 samples (78%) was below the LOD limit. For Co, the average concentration difference was positive 0.024 μg/L, indicating the Co mean concentration for the 1st tube was higher than the Co concentration for the 2nd tube. However, since the concentration difference is very small 0.024 μg/L compared to the LOD for Co being 0.06 μg/L, we think this difference is possibly due to variations in instrumental response, sample preparation, or other experimental parameters and not leaching. Further, we consider the Cr and Co differences (0.191 μg/L and 0.024 μg/L, respectively) would not have clinical significance in patient sample analysis. The healthy population reference ranges for Cr and Co in blood are < 2.1 μg/L and <0.6 μg/L, respectively²⁹. When analyzing MoM patient samples, clinically significant concentration differences in measuring hip prosthesis performance before and after the hip implant are considered to be 1 μg/L for Cr and 0.5 μg/L for Co ²⁸. Therefore, our Cr and Co concentration differences would have minimal impact on either of these concentration ranges. The data in this study indicated that there was no scientifically and clinically significant Cr and Co contamination from the limited set of stainless steel needles used.

Since most human blood samples from donors had Cr concentration below LOD, we decided to mimic the real-world collection in the lab (in-vitro), this time using spiked Cr and Co blood, 1.5 μg/L and 0.9 μg/L, respectively that is closer to 95^th percentile concentrations in 2015-2016 NHANES³⁰. Two blood tubes were collected consequently using same needle (200 blood tubes were collected using 100 needles). The concentration difference results for Cr and Co for 1st and 2nd blood tube have no clear trend (Table 2 and Figure SII). We performed paired sample t-test analysis to determine any statistically significant difference in Cr and Co concentrations between the 1st and 2nd blood tubes (Table 2). The paired sample t-test produced p-values above 0.05, indicating there is no statistically significant difference between Cr and Co concentrations in 1st and 2nd blood tubes collected in-vitro. We conclude that no Cr and Co leaching occurred into the blood sample using this particular set of stainless steel needles.

It is important to note, this conclusion does not apply to all blood collections. Lot screening results have shown the presence of contamination in some manufactured lots of needles. In the future, we plan to test multiple lots of needles and repeat these experiments using a lot of needles with known levels of Cr and Co contamination.

We analyzed A and B aliquots stored at −70°C, −20°C, 4°C, and RT at 11 separate time intervals (E1=1 day; E2=1 week; E3=2 weeks; E4=4 weeks; E5=6 weeks; E6=2 months; E7=4 months; E8=6 months; E9=8 months; E10=10 months; E11=1 year) during 1 year period. Both Co and Cr analyte concentrations for all aliquots were within our established QAL (Table SV and Table SVI). At 4 and 10 months at −70°C, Cr concentration in B aliquots trended lower but was still within QAL, 10.82 and 10.69 μg/L, respectively (Table SVI). At 8 months, no A and B aliquots stored at RT could be analyzed. Fibrin micro clots had formed, making the matrix highly viscous; therefore, the samples could not be pipetted. Thus, we ended further analysis at this temperature.

To evaluate further the stability of Cr and Co human whole blood samples as a function of storage time and temperature, we conducted a statistical analysis of the data. First, we used a general linear regression model, including time and temperature as covariates to fit the data, with the measurement value as the outcome. Figure 2 presents model based Cr and Co mean estimate concentrations and the 95% confidence intervals as a function of temperature and time (concentrations can be seen in Table SVII and Table SVIII). No clear trend is present in Cr and Co concentrations in A and B pools as a function of temperature (Figure 2a and b). In Figure 2d, the Cr concentration in B pool at 1 year (E11) jumps slightly but is still within QAL.

We examined and compared Cr and Co recovery trends of bracketing QC with long-term stability A and B aliquots as a function of time over the one year to determine whether this jump is a sign of instability or the result of variations in experimental parameters (Figure SI and Figure 2). The trend in Cr and Co concentrations of bracketing QC and A and B long-term stability aliquots over the same time is similar (Figure SI and Figure 2). At E11, there were Cr concentration spikes in bracketing QC (Figure SI); therefore, we think outside factors, such as instrumental drift/fluctuations and variance in the sample preparation procedure unrelated to stability influenced the Cr spike at E11.

We evaluated the data using the equivalence test. Ratios of μ_Test/μ_Ref versus temperature and time are plotted in Figures 3 and 4, respectively. Numerical data can be found in Table SIX and Table SX. All p-values for both time and temperature were < 0.001 (Table SIX and Table SX). Since the p-values < 0.05, we rejected the null hypothesis (H₀) and accepted the alternative hypothesis (H₁: 0.8 ≤ μ_Test/μ_Ref ≤ 1.25). Figure 3 shows equivalence test ratios of μ_Test/μ_Ref for Cr and Co results in A and B pools falling well within the equivalence boundaries, meaning Cr and Co concentrations in both pools at −20°C, 4°C, and RT were statistically equivalent to the Cr and Co concentrations at reference storage temperature −70°C. (Room temperature was only evaluated for the first 6 months.) Figure 4 displays equivalence test ratios of μ_Test/μ_Ref for Cr and Co concentration in A and B pools within the equivalence boundaries, meaning Cr and Co concentrations in both pools over one year were statistically equivalent to the Cr and Co concentrations at the reference storage of one day. The variation represented by error bars (95% confidence interval) in μ_Test/μ_Ref ratios at different temperatures (Figure 3) was slightly lower than the variation in μ_Test/μ_Ref ratios different times (Figure 4). Since μ_Ref is in the same analytical run for temperature but in a different analytical run for time than μ_Test, this slight variation was possibly due to instrumental drift/fluctuations, variance in sample preparation procedure (different analyst, reagents, and calibrators), or any other experimental parameters that can change daily. Overall, we concluded Cr and Co human blood samples are stable at temperatures −70°C, −20°C, and 4°C for one year and at RT for six months.