Approaches for Performance Verification Toward Standardization of Peripheral Blood Regulatory T-Cell Detection by Flow Cytometry

We analyzed peripheral blood samples from healthy subjects, patients in intensive care units (ICUs), and rheumatology and immunology (RI) patients (Supplemental Table 1, see supplemental digital content containing 9 figures and 3 tables at https://meridian.allenpress.com/aplm in the November 2024 table of contents) for the postcollection sample stability, as well as other routine samples available in the First Hospital of China Medical University (Shenyang, China). Hemolyzed, coagulated, and chylous blood samples were excluded. The blood samples were collected in K₂-EDTA tubes and were processed within 8 hours except in the case of stability studies on the postcollection samples. The present study was approved by the ethics committee of the First Hospital of China Medical University (2022-346-2).

Tregs analysis was performed by using reagent sets from BD Biosciences (San Jose, California, subsequently referred to as BD), QuantoBio Biotechnology (Beijing, China, subsequently referred to as QuantoBio), and Cellgene Biotechnology (Hangzhou, China, subsequently referred to as Cellgene).

Allophyco (APC)–conjugated anti–CD25-Ab (clone: 2A3, 662525), fluorescein isothiocyanate (FITC)–conjugated anti–CD4-Ab (clone: SK3, 340133), peridinin-chlorophyll-protein complex (PerCP)–conjugated anti–CD3-Ab (clone: SK7, 652831), and phycoerythrin (PE)–conjugated anti–CD127-Ab (clone: HIL-7R-M21, 664400), and PE-conjugated anti–FoxP3-Ab (clone: 259D/C7, 560046) were used.

APC-conjugated anti–CD25-Ab (clone: HI25a, Z6410045), FITC-conjugated anti–CD4-Ab (clone: RPA-T4, Z6410005), PerCP-cyanin 5.5 (PerCP-Cy5.5)–conjugated anti–CD3-Ab (clone: UCHT1, Z6410026), and PE-conjugated anti–CD127-Ab (clone: A019D5, Z6410046) were used.

APC-conjugated anti–CD4-Ab (clone: RPA-T4, P010014-3-20III), FITC-conjugated anti–CD3-Ab (clone: UCHT1, P010013-1-20III), PE-cyanin 7 (PE-Cy7)–conjugated anti–CD25-Ab (clone: M-A251, P010024-6-20III), and PE-conjugated anti–CD127-Ab (clone: HIL-7R-M2, P010034-2-20III) were used.

Because CD45 is optional for Treg detection, only anti–CD45-Ab (clone: 2D1, 663488) conjugated with APC-Cy7 from BD were used in this study. Samples were analyzed on a BD FACSCanto II Flow Cytometer by acquiring 10 000 events within the CD4⁺ lymphocyte gate. Acquisition and analysis were performed by using the BD FACSDiva software version 6.

The daily setup for the flow cytometer and routine QC was established according to the manufacturers’ recommendations before carrying out performance validation for the Treg measurements. The optics, fluorescence resolution, and noise were verified by using the cytometer settings and tracking beads (BD) for daily instrument setup. The BD FACS 7-color setup beads were used to calibrate the detector voltage, measure the detection sensitivity, and establish the fluorescence compensation.

For CD4⁺CD25⁺CD127^low/− Treg detection in whole blood, the appropriate volumes of monoclonal antibody (mAb) reagents and 100 μL of EDTA-anticoagulated whole blood were added to each tube. Antibodies were added one by one and no cocktails were used. The mixtures were vortexed and incubated for 15 minutes at room temperature (RT) in the dark before adding 2 mL of 1× lysing solution (FACS solution; BD). The mixture was incubated for an additional 10 minutes at RT in the dark. After centrifugation and discarding the supernatant, 2 mL of phosphate-buffered saline (PBS) was added to the tube twice to ensure complete rinsing of the mixture. The cells were finally resuspended with 300 μL of PBS. The tubes were protected from light and stored at 4°C until measurement by the flow cytometer within 2 hours.

For CD4⁺CD25⁺Foxp3⁺ Treg detection, peripheral blood mononuclear cells were isolated with Ficoll-Paque PLUS (GE Healthcare Life Sciences). After surface staining with CD45, CD3, CD4, and CD25 mAbs, the cells were fixed with Fixation/Permeabilization Buffer Set (Thermo Fisher Scientific) for 1 hour according to the manufacturer’s protocol. Cells were then washed with permeabilization buffer and stained for 1 hour at 4°C with PE-conjugated anti-Foxp3 mAb (560046, BD) in permeabilization buffer.

Microsoft Excel 2013 software and GraphPad Prism 9.0.0 software were used for data analysis and processing. The signal-to-noise ratio (SNR; the mean fluorescence intensity [MFI] of the positive population/the MFI of the negative population) was used to calculate the optimal concentration of Ab. The percentage difference (%Δ = | (Detection Value − Baseline) |/Baseline × 100%) was used to calculate the postcollection sample stability. The mean, the standard deviation (SD), and the coefficient of variation (CV) were used to calculate the repeatability, reproducibility, and reportable range. The mean value ± 2 × SD was used to calculate the accuracy. Consensus analysis between the results for the reagent sets from the 3 manufacturers was assessed by Bland-Altman plots and linear regression analysis. Pearson test was used for correlation analysis.

The gating strategy should contain the minimal or maximal gating boundaries to minimize data variability.^28,29  It is worth noting that the gate setting is performed consistently during the bioanalytical study and this prevents conscious or unconscious data manipulation. The gating strategy for determination of the peripheral Treg populations is illustrated (Figure 1, A through H: BD reagent set; Supplemental Figures 1 and 2: QuantoBio and Cellgene reagent sets). First, the debris and doublets were excluded (Figure 1, A); second, the total lymphocytes were gated from the side scatter (SSC) and CD45 plot (Figure 1, B) or the SSC and forward scatter (FSC) plot; third, the CD4 population was defined from the “lymphocyte” gate (Figure 1, C); fourth, the Tregs were defined as the CD25+ and CD127 low/− cells in the CD4+ T cells (Figure 1, D). Gating the CD25+ and CD127 low/− fractions of the CD4+ T-cell populations is a critical step in this assay. The gating of the CD25+ cells was done on the basis of the fluorescence minus one (FMO) control for CD25-APC as shown in Figure 1, E. The CD127 low/− population was gated with the corresponding FMO control and the fluorescence intensities (Figure 1, F). Improved resolution can be attained by subgating from the CD4⁺ T population within the CD25 versus CD127 strategy using manually placed octangular gates (Figure 1, G and H).

The optimal Ab concentration is one of the primary factors to be considered in assay optimization.⁴¹  For the Ab titrations, 2 blood samples with high (white blood cells [WBCs] = 21.39 × 10⁹ cells/μL) or low leukocyte counts (WBCs = 1.91 × 10⁹ cells/μL) were selected. The antibodies were serially diluted (1:2) in 5 to 6 dilution steps starting from the primary concentration recommended by the manufacturer. Samples of each Ab concentration were processed in duplicate, measured 3 times, and the average MFI was calculated. The SNR was also calculated. The concentration with the highest SNR was selected as the optimal concentration for each mAb reagent (Figure 2, A through J). The optimal Ab concentrations for CD45-APC-Cy7 (Figure 2, A and B), CD3-PerCP (Figure 2, C and D), CD4-FITC (Figure 2, E and F), CD25-APC (Figure 2, G and H), and CD127-PE (Figure 2, I and J), as supplied by BD, were 50, 6.25, 1.5, 6, and 12.5 μg/mL, respectively. The results for the optimal concentration of the antibodies from QuantoBio and Cellgene are shown in the supplementary material (Supplemental Figures 3 and 4).

Sample stability must be evaluated to determine how soon after collection the samples need to be processed. In this study, 5 samples from healthy donors with Treg percentages ranging from 4.29% to 6.21% (median, 5.81%), 5 ICU patients with Treg percentages ranging from 6.47% to 6.99% (median, 6.71%), and 5 RI patients with Treg percentages ranging from 3.47% to 7.09% (median, 4.54%) were selected to assess the sample stability at RT. According to the validation protocols,^42,43  the samples processed 2 hours post collection corresponded to the baseline level; the percentage difference was calculated for the individual sample results at 24, 48, 72, and 96 hours after blood collection. Postcollection sample stability was defined as the latest time point with a 20% change from the baseline.^24,43 

The stability at RT was first evaluated with samples from healthy donors. The percentage differences were 16.08%, 16.91%, and 18.44% at 72 hours post collection for the reagent sets from BD (Figure 3, A), QuantoBio (Figure 3, B), and Cellgene (Figure 3, C), respectively. At 96 hours, the samples showed significant instability with the percentage difference values increasing to 18.52%, 18.77%, and 25.04%, respectively. The stability at RT was further evaluated by processing samples from ICU and RI patients. When labeling using the mAbs from BD, the percentage differences were 15.11% and 17.74% for ICU and RI samples at 72 hours and the values increased to 21.28% and 22.43% at 96 hours (Figure 3, A), respectively. Similar results were noted for the labeled antibodies from QuantoBio (Figure 3, B) and Cellgene (Figure 3, C). The average percentage difference from baseline to 72 hours post collection was less than 20% but there was significant instability at 96 hours post collection (Figure 3). In addition, sample stability at 4°C was also evaluated. The average percentage difference from baseline to 72 hours post collection was more than 20% (Supplemental Figure 5). In conclusion, RT is appropriate for storage of samples for measurement of Tregs and the proportion of Tregs in the collected samples remained stable for up to 72 hours post collection.

The challenges associated with assessing the accuracy of FCM methods are primarily due to a lack of available reference materials. Alternative approaches for assessing accuracy have been suggested including the use of proficiency samples; comparison with a reference method; interlaboratory comparison; testing disease-state samples²⁴ ; and analyzing internal control samples or commercially available QC samples.²²  In this work, for the first time, 2 commercial QC samples for the detection of lymphocyte subsets, IMMUNO-TROL Cells and IMMUNO-TROL Low Cells (Beckman Coulter, Miami, Florida), were used as candidate reference materials for Treg detection. As a liquid preparation of stabilized human erythrocytes and leukocytes (lymphocytes, monocytes, and granulocytes), they have lysing, light scatter, antigen expression, and antibody-staining properties representative of those found in human normal whole blood. This allows verification of instrument and reagent performance. IMMUNO-TROL Low Cells reagents contain low-level CD4 counts (125 ± 63 cells/μL, target value of the sample used in this study) as compared with IMMUNO-TROL Cells reagents (630 ± 180 cells/μL). It was found that the Treg population could be clearly identified in the IMMUNO-TROL Cells and the IMMUNO-TROL Low Cells samples (Supplemental Figure 6). We then established the target range for Treg detection as follows: the population ranges for Tregs (mean ± 2 × SD) were established from a labeling experiment of the IMMUNO-TROL Cells and the IMMUNO-TROL Low Cells for 20 days. For the BD reagent set, the upper limits of determination mean + 2 × SD of the Treg percentages were 6.19% and 7.19%, and the lower limits of determination mean − 2 × SD of the Treg percentages were 5.53% and 6.46%, respectively.

Next, the percentage of Tregs was measured 10 times by using the same QC materials to verify accuracy. For measurement of Tregs, for labeling by the antibodies from BD and QuantoBio, the results for the 10 measurements were all within the target range for the IMMUNO-TROL Cells (compliance rate = 100%), and 9 of 10 measurements were within the target range for the IMMUNO-TROL Low Cells (compliance rate = 90%). That is, the compliance rate for the IMMUNO-TROL Cells and the IMMUNO-TROL Low Cells were 100% and 90% (Figure 4, A through D; Supplemental Table 2), respectively. As for the Tregs labeled by the antibodies from Cellgene, the percentages of Treg measurements within the target ranges for the IMMUNO-TROL Cells and the IMMUNO-TROL Low Cells were 90% and 100% (Figure 4, E and F; Supplemental Table 2), respectively. Thus, the measurements for the Tregs labeled by antibodies from BD, QuantoBio, and Cellgene yielded excellent accuracy with a compliance rate of at least 80%.⁴⁴  Moreover, the results were in agreement with the accepted values for the QC materials and provided confidence that the Treg measurements were accurate.

To determine the repeatability, samples of high, medium, and low percentages of Tregs were selected. The Tregs in each blood sample were labeled by antibodies from the 3 manufacturers and 11 consecutive measurements were performed; the CV for the last 10 measurements was calculated. A CV of 10% or less was set as the acceptable criterion according to practice and published guidelines.^24,43  For the Tregs labeled by antibodies from BD, the CVs for the samples of high, medium, and low percentages of Tregs were 5.41%, 5.36%, and 6.28% (Table 1). Similarly, the CVs for Tregs labeled by antibodies from QuantoBio were 7.64%, 6.65%, and 7.22%, and the CVs for Tregs labeled by antibodies from Cellgene were 4.02%, 3.95%, and 7.26%, respectively (Table 1). Regarding the data, the repeatability data for the Tregs labeled by antibodies from the 3 different reagent sets were acceptable with the CVs all being less than 10%.

Given that the IMMUNO-TROL Cells and the IMMUNO-TROL Low Cells can serve as QC materials, an evaluation of the reproducibility was also undertaken. Two replicate samples of the IMMUNO-TROL Cells and the IMMUNO-TROL Low Cells, and the percentages of Tregs were measured for 20 consecutive days, based on labeling by antibodies from the 3 reagent sets. The CVs for the percentages of Tregs for the 20 days were calculated. A CV of 10% or less was set as the acceptable criterion for the reproducibility.^24,43  The CVs for the IMMUNO-TROL Cells and the IMMUNO-TROL Low Cells were 2.13% and 1.70% for Tregs labeled by antibodies from BD (Table 2), 2.68% and 2.91% for QuantoBio (Table 2), and 3.58% and 3.00% for Cellgene (Table 2). These results demonstrated that the reproducibility for measurement of the Tregs labeled by antibodies from the 3 reagent sets was acceptable.

In accordance with the definition of reportable range,²⁴  the reportable range for Treg detection by FCM refers to a minimum number of events required for reproducible measurement, which can be used to validate the detection capability. Whole blood samples from 5 healthy donors for each reagent set were labeled in triplicate by using antibodies from BD, QuantoBio, and Cellgene. The numbers of cells acquired were set to 10 000, 5000, 2500, 1000, 500, 250, and 100 in the CD3⁺CD4⁺ gate for each replicate. The CV for each individual measurement and the average CV were calculated. The average CV for the percentage of Tregs under the various cell number acquisitions should be less than 20%; more than 80% of the samples had a CV of 20% or less.^29,44  The average CV was less than 20% when 10 000, 5000, 2500, 1000, and 500 events were acquired in the CD3⁺CD4⁺ gate for all the reagent sets (Figure 5, A through C). Although labeling of Tregs using the antibodies from BD (Figure 5, A) and Cellgene (Figure 5, C) gave an average CV of 20% or less, when 500 events were collected in the CD3⁺CD4⁺ gate, a great variation was observed for different samples and only 60% and 40% of samples had a CV of 20% or less, respectively. The results for the Tregs labeled by antibodies from QuantoBio showed 80% of the samples met the requirement that the CV be less than 20% when 500 events were collected in the CD3⁺CD4⁺ gate (Figure 5, B). However, insufficient CD3⁺CD4⁺ events may increase the difficulty of identifying accurately the populations of Tregs according to the established strategy (Supplemental Figure 7). Therefore, a minimum of 1000 events in the CD3⁺CD4⁺ gate should be acquired for peripheral blood. When 1000 events were collected in the CD3⁺CD4⁺ gate, the percentage of Tregs from 5 donors, labeled by using antibodies from BD, QuantoBio, and Cellgene, achieved the median percentage of 6.42% (4.31%–10.45%), 6.53% (4.97%–7.49%), and 8.26% (5.78%–9.47%), respectively.

Because Treg detection belongs to a quasi-quantitative method, samples with different levels of Treg populations were needed to evaluate the linearity for Treg detection. According to the approaches provided in Clinical Laboratory Standards Institute (CLSI) guideline H62 for generating samples, we partially stained a sample by omitting CD25 and CD127 antibodies so that the Treg population could not be detected. Then, we attempted to mix a fully stained sample with the partially stained sample in 8 different ratios (10:0, 9:1, 8:2, 6:4, 4:6, 2:8, 1:9, 0:10). The R² values were calculated after linear regression. Treg detection by the 3 reagent sets showed excellent linearity with R² ≥ 0.99 (Supplemental Figure 8).

To detect the assay carryover of Tregs, one blood sample with a high Treg percentage (15.10%) and the other with a low Treg percentage (5.64%) were used for the carryover analysis to observe any impact on the low-level sample result. The sample with high Treg expression was tested 3 consecutive times; the results were recorded as H1, H2, and H3, then followed by samples with low Treg percentage, and the results were recorded as L1, L2, and L3; the carryover rate was calculated as (|L3−L1|/H3 − L3)% and the carryover rate of less than 1% was considered acceptable.^24,45  A carryover rate of 0.32%, 0.41%, and 0.96% was achieved for peripheral Treg detection labeled by reagent sets from BD, QuantoBio, and Cellgene, respectively. The carryover rates were all less than 1% (Supplemental Table 3), which indicated that there was no carryover in Treg detection by the 3 reagent sets.

The results for Treg detection in 23 samples were used for agreement analysis studies involving the reagent sets of the 3 manufacturers (BD, QuantoBio, and Cellgene). The Bland-Altman plots of the differences between the 3 reagent sets were prepared (Figure 6, A through F). It was noted that 95.65% of the differences in measurements were within the 95% limits for the CI of the differences in the means, confirming good agreement between the reagent sets for Treg detection (Figure 6, A, C, and E). The percentage of Tregs, based on using the mAbs from BD, was closely correlated with the equivalent results for the antibodies from QuantoBio (R² = 0.9680) (Figure 6, B) and Cellgene (R² = 0.9735) (Figure 6, D). Similarly, the measurement results for Tregs, using the antibodies from QuantoBio and Cellgene, showed a strong correlation, with R² > 0.95 (R² = 0.9679) (Pearson correlation coefficient, P < .001) (Figure 6, F). In conclusion, the analytical values for Tregs labeled by antibodies from the 3 commercially available reagent sets were in excellent agreement.

Foxp3 is also a well-known marker for Tregs. However, Foxp3 cannot be performed in the clinic owing to the lack of fluorescent conjugated Foxp3 antibodies approved by the local National Medical Products Administration and lack of homebrew/LDT reagent kit available. Because Foxp3 has been widely used in research to identify Tregs, we compared the clinical panels used in our study to the research panel including Foxp3. CD4⁺CD25⁺Foxp3⁺ Tregs were detected and the relationship between the frequency of CD4⁺CD25⁺CD127^low/− Tregs and CD4⁺CD25⁺Foxp3⁺ Tregs was evaluated by Pearson correlation analysis. To minimize the variation caused by the staining strategies, the same fluorescent conjugated mAbs of CD45, CD3, CD4, and CD25 were used, and Foxp3 mAbs were conjugated with the same fluorochrome of PE as CD127 (Supplemental Figure 9, A). We found that the percentage of CD4⁺CD25⁺CD127^low/− Tregs was closely correlated with that of CD4⁺CD25⁺Foxp3⁺ Tregs (r = 0.9198, P < .001) (Supplemental Figure 9, B), which is consistent with previous studies.^34,35 

Treg detection is used to study and monitor immune-related disorders.^46–48  However, protocols are lacking on how to validate the performance of the FCM-based measurements, which is an obstacle for standardization of the assay. In this study, we have undertaken a comprehensive performance validation study of the Tregs assay. More generally, this work may be viewed as a first attempt to study the performance validation of immune-monitoring assays by FCM.

The evaluation of accuracy represents the biggest challenge in performance validation given that cellular certified reference materials are not available. Documentation on best practice in clinical laboratories has been issued to facilitate consensus regarding the requirements for method development and validation of accuracy of flow cytometric assays in various contexts.^24,49,50  O’Hara et al⁵⁰  have suggested the use of QC materials for assessment of accuracy based on the availability of target values. Quadrini et al²⁹  established the target values in a QC material by using the CD-Chex Plus Normal control material and verified the accuracy of mHLA-DR measurement. In the present study, the accuracy of the Tregs assay was accessed by analyzing commercial QC materials, namely IMMUNO-TROL Cells and IMMUNO-TROL Low Cells for the first time. The target range was established by measuring Tregs in the QC materials for a period of 20 days. The IMMUNO-TROL materials can serve as QC materials and be used for assessing the accuracy for measurement of Tregs and for evaluating the consistency of results from different laboratories in multicenter studies. However, this is an attempt to standardize the Tregs assay on the basis of existing conditions. The use of certified reference materials still represents the best means for assessment of accuracy, and production of such materials is urgently needed.

Key aspects on the performance of the Tregs assay are also clarified by assessing the results. First, determining the optimal concentration of antibodies is most important for validation and optimization of the reagents used in the FCM experiment. We found that the optimal concentrations of mAb as supplied in the reagent sets by the different manufacturers were significantly different. Given that most of the immune system assays performed by FCM are LDT methods—owing to the lack of commercially available diagnostic assay kits—it is necessary, for each reagent platform, to establish the optimal concentration of the antibodies before commencing the assay. Second, we demonstrated for the first time that there is sample stability at 72 hours post collection for the 3 reagent platforms, indicating that the proportion of Tregs in the sample may be considered as stable for 72 hours at RT. Schultze-Florey et al⁵¹  reported that blood samples were stable at 48 hours for CD3+ T, CD4+ T, and CD8+ T cells in a study aimed at establishing reference intervals for major lymphocyte subsets and memory T-cell subpopulations, based on good laboratory practice–conforming staining panels. HLA-DR on monocytes was reported to be stable for 72 hours post collection.²⁹  The results of the present study indicated that samples for the Tregs assay were stable for 72 hours, thus a time point of at least 72 hours could be set when the stability is to be tested for whole blood FCM assays based on surface staining. Third, excellent repeatability data were observed with CV values ranging from 5% to 6%; also, satisfactory reproducibility and linearity data for Treg detection across the 3 reagent platforms were noted. Moreover, high correlation and good agreement for detection of Tregs labeled by CD3/CD4/CD25/CD127 mAb reagent sets across the 3 different reagent sets were observed. Pitoiset et al³⁰  reported that the results for Tregs, using different FCM platforms (Navios cytometer and LSRII cytometer), were reproducible and highly correlated. The present study further indicated good consensus for the 3 reagent sets. We did not use CD45 from the 3 vendors because lymphocytes were well gated as based on FSC and SSC, and CD45 is optional for the identification of Tregs. Conclusions for the performance verification were not affected when Tregs were detected without CD45 (data not shown). In general, the above work highlights the repeatability and robustness of the Tregs assay, suggesting that the assay can be standardized, which is important for future establishment of Treg reference intervals and multicenter clinical trials.

At present, documents of the current best practices for validation of flow cytometric assay have been published by the International Council for Standards in Hematology, the International Clinical Cytometry Society, and CLSI (Guideline H62).^24,43,45,52  With the methods recommended in these documents, we have conducted assessments on postcollection sample stability, accuracy, repeatability, reproducibility, linearity, and assay carryover for Treg detection in this study. As a FCM-based immunoassay, the validity of Treg detection can be affected if patients are treated with the monoclonal antibody basiliximab to prevent rejection after renal transplant, because basiliximab binds to the α chain of interleukin-2 receptors (CD25 antigen), which interferes with the combination of fluorescent-labeled CD25 antibody and CD25 antigen on T cells. Limit of blank (LOB)/limit of detection (LOD) and lower limit of quantitation (LLoQ) are recommended to determine the detection capability in CLSI guideline H62.⁴⁵  Because LLoQ is required for the assay to detect rare events^24,43,52  and Tregs in peripheral blood does not belong to the rare cell population, we determined the reportable range of Tregs instead, which determines the minimum number of events in the upper gate at which the precision is acceptable for Treg detection.²⁴  In general, this study is a new attempt at performing validation for the multicolor FCM assay by referring to the documents of current best practices.

Our study has limitations. We did not assess the performance of the Tregs assay with intracellular Foxp3 staining because (1) no fluorescent conjugated Foxp3 antibodies were approved by the local National Medical Products Administration, and thus intracellular Foxp3 staining cannot be performed in the clinic; and (2) customized reagents are needed for the standardization of intracellular Foxp3 staining, such as 1-step intracellular staining³³  or precoated dried antibodies,³⁰  otherwise accurate quantification is not guaranteed. In this study, we found a close relationship between CD4⁺CD25⁺CD127^low/− cells and CD4⁺CD25⁺Foxp3⁺ cells, which is consistent with previous studies.^34,35,53  With the development of Treg reagents, further studies need to be carried out based on more methods and platforms.

In conclusion, this is the first study to evaluate systematically the performance of the Tregs assay based on the use of commercially available reagent sets supplied by different manufacturers. The approaches described herein could provide a practical solution to verifying the performance of other FCM-based immune-monitoring projects in clinical practice.