Bodily Fluid Analysis of Non-Serum Samples using Point-of-Care Testing with iSTAT and Piccolo Analyzers Versus a Fixed Hospital Chemistry Analytical Platform

Abstract

Introduction:

Forward deployed military medical units can provide sophisticated medical care with limited resources. Point-of-Care Testing (POCT) may facilitate care and expedite diagnosis. This study assessed the accuracy of results for POCT for non-serum samples (pleural, peritoneal, and cerebrospinal fluid) using iSTAT and Piccolo hand-held devices compared with results obtained using a hospital chemistry analyzer.

Methods:

Pleural, peritoneal, and cerebrospinal fluids obtained during routine care were simultaneously analyzed on a Vitros 5600 automated clinical chemistry hospital analyzer, iSTAT, and Piccolo POCT devices.

Results:

POCT results were highly correlated with the Vitros 5600 for pleural fluid LDH, glucose, and triglycerides (TG); for peritoneal fluid bilirubin, TG, glucose, albumin, and protein; and glucose for cerebrospinal fluid.

Conclusion:

POCT results for non-serum samples from pleural, peritoneal, and cerebrospinal fluid correlate with standard hospital chemistry analysis. The results of this study demonstrate potential for possible new diagnostic roles for POCT in resource-limited environments.

Introduction

The delivery of quality healthcare in developing countries has dramatically increased during the last 10–15 years.¹ This growth is related to global health initiatives and advances in medical technology.² The advent and expansion of point-of-care technology illustrates how diagnostic acumen in austere or low-income environments has changed. Point-of-care testing (POCT) offers rapid and accurate biochemical testing on serum samples. POCT has well established clinical utility for serum chemistry and blood gas analysis, as well as for the assessment of diabetes, pregnancy, HIV, and malaria.³

Most point-of-care technologies have targeted blood tests, yet comprehensive laboratory testing (to include the evaluation of non-serum samples) continues to require conventional bench top analytic platforms, whose cost and size limit their practicality in resource constrained environments. Only two studies have compared POCT to traditional biochemical analysis of non-serum samples in humans. Kohn, et al, demonstrated that pleural fluid pH can be accurately performed on an iSTAT POCT device.⁴ Wockenfus, et al, demonstrated that the iSTAT is comparable to a validated blood gas analyzer for pleural fluid pH analysis.⁵ An animal study on equine synovial fluid accurately measured lactate and glucose but was limited in the number of samples (N = 8).⁶ Overall, there is scant evidence-based literature regarding the use of POCT on non-serum samples, and those uses are not approved by the Food and Drug Administration (FDA).

The purpose of this study was to compare the accuracy of iSTAT and Piccolo devices to a standard hospital bench top analyzer using analyte measurement, in order to determine if the POCT devices have a potential role in the analysis of non- serum samples such as pleural, peritoneal, and cerebrospinal fluid.

Methods and Materials

Existing pleural, peritoneal, and cerebrospinal fluids collected for standard of care analysis were stored at −20°C for further POCT analysis. The samples were thawed and analyzed on an Ortho Clinical Diagnostics Vitros 5600 automated clinical chemistry analytical platform, considered be the analytical reference standard in this study. All samples were simultaneously analyzed using an iSTAT and/or Piccolo POCT hand-held device, employing test-specific cartridges for individual analytes. The iSTAT used EC8+ cartridge, and the Piccolo used Lipid Panel, Basic Metabolic Plus, and General Chemistry 13 cartridges.

Pleural fluid analysis measured glucose and pH on the iSTAT using an EC8+ cartridge. The Piccolo POCT device measured LDH on the Basic Metabolic Panel Plus cartridge. Amylase, protein, and albumin were measured using the General Chemsitry13 cartridge, and the Lipid Panel cartridge was used to quantify triglycerides (TG) and cholesterol levels in pleural fluid.

Peritoneal fluid analytes were only measured with the Piccolo POCT device. The Piccolo measured LDH with the Basic Metabolic Panel Plus cartridge and amylase, protein, albumin, bilirubin; however, the glucose was measured with the General Chemistry13 cartridge while the Lipid Panel cartridge was used to measure triglycerides (TG).

The EC8+ cartridge was used to measure cerebrospinal fluid glucose using the iSTAT's platform.

The null hypothesis was, “The differences of the means between POCT and bench top analysis is 0.” Bland-Altman charts were constructed to assess accuracy and precision of the POCT analyzer compared to the Vitros 5600. The Bland-Altman charts plot the difference of the paired measurements for each sample against the average, and also display 95% confidence intervals.⁷ A two-sided paired t-test assessed differences between the two measurements. A mean difference significantly different from zero indicated the average bias for the POCT analyzer. For analytes with a large number of measurements below the limit of detection, results were dichotomized as detected or non-detected, and a kappa statistic assessed concordance. A significance of P ≤ .05 was used for all statistical tests. All analyses were performed using Statistical Analysis Software (SAS) v 9.2 software, SAS Institute, Cary, NC.

Correlation was determined using a correlation coefficient (r), reflecting a linear relationship of POCT (y-axis) versus control (x-axis) values with increasing correlation as 1.00 is approached. Because the sensitivity of the control and POCT analyzers varied, a concordance (k) value was calculated that represented the reliability of POCT to categorize samples as detectable or non-detectable when compared to the control, with increasing k as 1.00 is approached.

Results

Pleural Fluid

Table 1 summarizes the statistical results for pleural fluid. The number of samples for the analytes was 6–23 (n) with range of data values referenced in table 1. Statistical difference (P ≤ .05) between testing modalities was demonstrated for LDH, albumin, glucose, and cholesterol based on the null hypothesis; however, protein, TG, and pH did not demonstrate a statistical difference, with P values of .96, .08, and .11, respectively. The mean values represented positive biases for protein, albumin, glucose, and pH at 0.01, 0.27, 11, and 0.01. A negative bias was calculated for LDH (−343), cholesterol (−16), and TG (−2.7). Correlation (r) values were determined for glucose (0.99), pH (0.98), LDH (0.97), protein (0.90), albumin (0.81), TG (0.91), and cholesterol (0.91).

Table 1.

Pleural Fluid Results

	n	Mean	Std	Mean 95% CI	P-value	Precision 95% CI	Correlation (r)	Concordance (k)	Data Range
LDH (U/L)	12	−343	243	−498/ −188	< .01	−820/134	0.97	1.00	2018
Protein (g/dL)	17	0.01	0.52	−0.26/0.27	.96	−1.01/1.03	0.90	1.00	3.7
Albumin (g/dL)	17	0.27	0.35	0.09/0.45	< .01	−0.41/0.95	0.81	0.16	1.37
Glucose (mg/dL)	23	11	8	8/15	< .01	−5/27	0.99	1.00	200
Cholesterol (mg/dL)	6	−16	11	−28/−4.6	.02	−38/5.9	0.91	0.87	64
TG (mg/dL)	17	−2.7	5.9	−5.7/0.32	.08	−14/8.8	0.91	0.61	42
pH	20	0.01	0.04	−0.003/0.028	.11	−0.06/0.08	0.98	0.74	1.05

The Bland Altman chart for pH is an example of how each analyte was plotted to demonstrate the difference of the paired measurements for each sample versus the average. Figure 1 is a representation of pH. Correlation is reflected in Figure 2 for pH as linear relationship of POCT versus Vitros 5600 values, with increasing correlation as 1.00 is approached.

Figure 1.

Bland Altman Chart for pH

Figure 2.

Correlation for pH. X axis is control measurement of pH. Y axis is POCT measurement of pH

Peritoneal Fluid

Table 2 summarizes the statistical data for peritoneal fluid. Protein and TG were the only two analytes where statistical difference (P ≤ .05) between the POCT and the control was not significant with P values of .23 and .17. Glucose, bilirubin, and TG had correlation values of 1.00, 1.00, and 0.98. Albumin did not demonstrate a statistical difference, and only 3 of 16 samples were detectable on the testing modalities.

Table 2.

Peritoneal Fluid Results

	n	Mean	Std	Mean 95% CI	P-value	Precision 95% CI	Correlation (r)	Concordance (k)	Data Range
LDH (U/L)	11	−301	398	−598/− 33.5	.03	−1114/466	0.88	0.11	1829
Protein (g/dL)	16	0.09	0.30	−0.07/0.25	.23	−0.50/0.69	0.92	1.00	2.6
Albumin (g/dL)	16	0.20	0.30	0.04/0.36	.02	−0.40/0.80	0.90	0.46	1.4
Bilirubin (mg/dL)	16	0.23	0.25	0.09/0.36	< .01	−0.27/0.72	1.00	−0.12	9.0
Glucose (mg/dL)	20	19	24	8/30	< .01	−28/65	1.00	0.64	598
TG (mg/dL)	12	18.5	43	−9/46	.17	−66/103	0.98	1.00	190

Cerebrospinal Fluid

Table 3 summarizes the data for cerebrospinal fluid. Fifty-five samples were obtained in total and glucose was the single analyte measured for cerebrospinal fluid. It did demonstrate statistical difference with a P < .01 but notably had a positive bias with a mean of 6 mg/dL and a strong correlation (r = 0.97).

Table 3.

Cerebrospinal Fluid Results

	n	Mean	Std	Mean 95% CI	P-value	Precision 95% CI	Correlation (r)	Concordance (k)	Data Range
Glucose (mg/dL)	55	6	4	5/7	< .01	−1/14	0.97	1.00	53

Discussion

Pleural Fluid

Prior studies have demonstrated utility for pH measurement in pleural fluid with POCT.^{4 , 5} In this study, no significant statistical difference was observed (P = .11) with slightly positive mean bias. POCT testing could be used at bedside for diagnostic purposes of diagnosing a parapneumonic effusion which can have a pH < 7.2.⁸ All of the other samples had a pH > 7.4 with 10 yielding a pH > 8.2 on both devices. Alkaloid sampling might be from the delay in measuring pH since some of the samples were in the core laboratory freezer for a week. pH can become more alkaloid if there is a delay > 4hrs between obtaining the pleural fluid and analysis.^9,10 A future improvement on this study would be to obtain a timely bedside measurement of pH versus a bench top control and assess the accuracy and the ability of the POCT device to correctly classify the effusion as a possible complicated parapneumonic effusion based on a pH < 7.2 given the correct clinical scenario. A timely specimen collection and analysis would be classified as < 2hrs since this should not affect the pleural fluid pH.¹¹

Glucose demonstrated a statistical difference (P < .01) and a bias of 11mg/dL on the POCT when compared bench top analysis. The clinical utility of glucose is when it is ≤ 60mg/dL, usually in the setting of a parapneumonic effusion or malignant effusion.¹⁶ Glucose measurements on POCT correctly identified 6 of 7 effusions with glucose ≤ 60 mg/dL. It would be 7 of 7 but one sample on POCT would be calculated at 62mg/dL while on the control analysis it would be 60mg/dL. Glucose has demonstrated the potential to be used for clinical assessment at the bedside in a future study.

LDH had statistically different values (P-value < .01) when comparing testing modalities with large negative mean bias (−343U/L); however, it had a strong r at 0.97. LDH had statistical differences in measurements on the POCT compared to the control due to measurement of different reaction products on each separate device. LDH is reported in international units on both devices but has different references ranges with a 99–192 U/L on the POCT and 313–618 U/L on the control. The control uses LDH as a catalyst for the oxidation of NADH to NAD+ monitored by reflectance spectrophotometry, and the activity of LDH corresponds to the LDH concentration.¹³ The POCT test for LDH was the Piccolo Basic Cartridge which uses the same reaction as above but carries out a second reaction measuring the formation of formazan by spectrophotometry.¹⁴ LDH is utilized diagnostically to characterize pleural fluid into exudative or transudative classes based on Light's criteria.¹² If using Light's Criteria with LDH being 2/3 the upper limit of normal an exudate on the POCT device will be a value ≥ 161 U/L while on the control the value will be ≥ 516 U/L. The POCT identifies 7 exudates and the control identifies 6. It would have identified the same seven but one of the value was only 503 U/L while 165 U/I on the POCT which is borderline exudative/transudative on both devices. LDH demonstrated potential for clinical use on POCT testing but further testing should be conducted to broaden the range of values and increase the sample size.

Protein had a positive bias at 0.01g/dL with no significant statistical difference (P = .96). Five of the 17 samples were < 2.0g/dL on the POCT and control. The other 12 were detectable on both devices. Protein is another clinically important analyte in distinguishing exudate from a transudate with Light's criteria, but a serum sample is needed to calculate a ratio.^12,16 Future studies should investigate if the POCT device can categorize the pleural fluid as transudative or exudative.

TG had no significant statistical difference (P = .08) with a slightly negative bias (2.7mg/dL). The highest value measured was 60 mg/dL on the control. TG are routinely measured in pleural fluid if a chylothorax is suspected and confirmed with TG level > 110 mg/dL.¹⁵ The value range needs to be expanded to assess whether the POCT is clinically useful for diagnosing a chylothorax. The low concordance was due to the control detecting TG levels below 20, while POCT could not.

Albumin levels were analyzed because they have been used in studies to categorize transudative versus exudative effusions.¹⁶ Albumin demonstrated a statistical difference observed with a (P-value < .01) with a positive mean value at 0.27mg/dL with a low concordance value leading to the conclusion that albumin has limited if no potential future use on POCT.

Cholesterol had a negative bias at 16 mg/dL with statistically significant difference (P = .02), however, only 6 samples were measurable on the control. Cholesterol can categorize transudative and exudative pleural effusions. An exudative effusion has cholesterol level > 60 mg/dL for higher specificity or > 43 mg/dL for higher sensitivity but neither is more sensitive nor more specific than Light's criteria.¹⁶ POCT correctly identified 2 of 4 exudates with a cholesterol cutoff > 60mg/dL. If the cutoff is dropped to a cholesterol level > 43 mg/dl then 6 of 6 exudates are identified. Future testing should be conducted to assess the accuracy of cholesterol measurement on POCT due to the low number of measurable samples.

Peritoneal Fluid

LDH had 11 samples with a wide data range; however, it did not have a strong correlation. This test used the same cartridge as pleural fluid on the POCT device, and there was a large difference in values due to different analyte measurements on each device.^13,14 This yielded a statistical difference between the values with a weaker correlation than pleural fluid. LDH measurement would not be particularly useful for peritoneal fluid in diagnosing spontaneous bacterial peritonitis or secondary bacterial peritonitis. A future study should be repeated with more samples.¹⁷

Protein was measured on 16 samples but only detectable on 5 samples with a positive bias and no statistical significance. Since 11 of the samples read less than 2.0g/dL on both testing modalities and 5 detected samples > 2.0 g/dL the concordance was 1.00. Only 5 samples had elevated protein levels so further testing should be conducted to assess potential use for protein measurement. If a sample is read as less than 2.0g/dL of protein on POCT testing it is most likely correct. It has demonstrated utility for use at bedside if it is negative or < 2.0g/dl but if the level is positive there is not data to clinically utilize this testing modality.

Albumin demonstrated a statistically significant difference between the POCT device and the control with a mean bias of positive 0.2 mg/dl due to POCT reporting elevated albumin levels compared to the control. Of the 16 samples analyzed only3 had a detectable albumin levels on the control and POCT device; however, the POCT device had 3 other samples with falsely high elevations when compared to the control. Albumin is diagnostically utilized to calculate a serum-ascites albumin gradient to aid in diagnosing portal hypertension.^{20 , 21} A future study could assess a serum-ascites albumin gradient calculated from a control versus the POCT testing to account for the positive bias of the device but currently the POCT device should not be used to measure albumin in ascitic fluid.

Glucose demonstrated a large mean bias of 19mg/dL with statistical significance (P < .01). However, there were 2 samples with larger values recorded on POCT at > 700 mg/dL and 699 mg/dl while control recorded values of 618 mg/dl and 611 mg/dL. These two results may have skewed the results and elevated the positive mean bias. The overall r was 1.00 demonstrating a relationship between testing modalities. Glucose is measured in ascites typically if there is a concern for secondary bacterial peritonitis with a level of < 50 mg/dL in this setting.²² Only 3 of the samples from the study had a reading on < 50 mg/dL and another 3 samples had a value of 50–70 mg/dL with the POCT and control agreeing for each set of 3 samples. Further testing should be conducted to assess the accuracy of glucose since it may be able to be utilized for aiding in the diagnosis of secondary bacterial peritonitis.

Bilirubin demonstrated a slightly positive mean on the 16 samples but only 3 of these samples had elevated bilirubin levels at > 1 mg/dl. The r was excellent at 1.00 but concordance was poor due to the slightly positive bias on the POCT testing which lead to detectable tests when the control read a negative result on 7 samples. Bilirubin is tested in ascitic fluid to investigate a possible choleperitoneum from a gallbladder perforation which typically causes a bilirubin concentration greater than 6 mg/dL with an ascitic fluid/serum bilirubin ratio > 1.²³ Only 1 sample in the study had a bilirubin > 6 mg/dL and at 8.7 mg/dL and 9.1 mg/dL on the POCT and control, respectively. POCT did demonstrate utility and even with a slightly positive mean bias it was able to detect elevated bilirubin levels. It should be assessed in further studies with a bedside assessment versus a control because it could impact patient care immediately.

TG testing demonstrated no statistical significance in the results between the two testing modalities but there was positive mean bias which is greatly affected by one sample where the POCT recorded 373 mg/dL and the control recorded 219 mg/dL. If this measurement is excluded since all the other measurements were less than a 109 mg/dL on both devices with 3 negative samples in agreement in both testing modalities, mean bias goes from 18.5 mg/dL to 6.2 mg/dl with an STD of 7.2 mg/dL (P = .017). It seems the testing could be used for lower TG levels but it remains uncertain of its clinical utility with levels > 109 mg/dL since there was only 1 sample which was elevated.

Cerebrospinal Fluid

55 samples demonstrated a mean positive bias of 6mg/dl with statistical significance (P = <.01) but it may not be clinically relevant. The POCT device ranged higher on every sample except for one in which the POCT measured a low CSF glucose at 20 mg/dL with a control of 22 mg/dL. Both values still fall into the low range. Normal values of blood glucose typically range from 40–80 mg/dL or a CSF to serum glucose ratio of 0.6. Bacterial meningitis typically causes a ratio of CSF/serum glucose of < 0.4.^24–26 Forty-eight of the samples were within the range of 40–80 mg/dL and three samples were > 80 mg/dL. Four samples were less than 40 mg/dL. Future studies should sample glucose at bedside along with serum glucose on POCT and control devices. Further investigation needs to be conducted in values outside of 40–80mg/dL.

Conclusion

Pleural fluid protein, pH, glucose, and TG, and peritoneal fluid protein, bilirubin, glucose, and TG correlated and had similar clinical results as the control, demonstrating potential for clinical use. Pleural fluid LDH had a strong correlation between the fixed platform and the POCT device; however, differences were noted between reference ranges and methodology. Further studies will need to be conducted to confirm the utility of LDH POCT. Cholesterol POCT requires further investigation, CSF glucose POCT results were promising, and future studies should study more samples with glucose < 40 mg/dL.

Conflict of Interest

None of the authors identify a conflict of interest.