Abstract
Clinical Immunohistochemistry (IHC) laboratories face unique challenges in performing accurate and reproducible immunostains. Among these challenges is the use of homemade controls derived from pathological discard samples. Such positive controls have an unknown number of analyte molecules per cell (epitope density). It is unclear how the lack of defined analyte concentrations affects performance of the control. To address this question, we prepared positive IHC controls (IHControls) for human epidermal growth factor receptor type II (HER-2), estrogen receptor (ER), or progesterone receptor (PR) with well-defined, homogeneous, and reproducible analyte concentrations. Using the IHControls, we examined the effect of analyte concentration on IHC control sensitivity. IHControls and conventional tissue controls were evaluated in a series of simulated primary antibody reagent degradation experiments. The data demonstrate that the ability of a positive IHC control to reveal reagent degradation depends on (1) the analyte concentration in the control and (2) where that concentration falls on the immunostain’s analytic response curve. The most sensitive positive IHC controls have analyte concentrations within or close to the immunostain’s concentration-dependent response range. Strongly staining positive controls having analyte concentrations on the analytic response curve plateau are less sensitive. These findings emphasize the importance of selecting positive IHC controls that are of intermediate (rather than strong) stain intensity.
Keywords: IHControl, immunohistochemistry, peptide, positive control
Introduction
The field of diagnostic Clinical Immunohistochemistry (IHC) is evolving from an inexact, qualitative collection of stains to a quantitative set of assays whose results can independently influence patient management. Published reviews, regulatory requirements, and consensus guidelines convey a similar theme: The general concepts of assay validation and control protocols that are a standard of practice in other clinical laboratory disciplines, such as Clinical Chemistry, are generally applicable to Clinical IHC.1–4 Technical hurdles, however, confound the application of some of these concepts to practice in the Clinical IHC laboratory. Assay validation and control protocols that are straightforward when dealing with blood samples are sometimes not straightforward when applied to tissue sections. For example, assay validation protocols for a serum assay typically involve checking calibration and linearity. These parameters are not easily measured for immunohistochemical stains. This article focuses on one of those unique challenges associated with IHC—the selection of positive IHC controls.
The Clinical Laboratory Improvement Act of 1988 (CLIA 88, Section §493.1202(c)) mandates that clinical laboratories performing moderately complex testing run at least two controls at different analyte concentrations.5 For example, if a serum glucose test has an analytic measurement range (AMR) from 5 to 800 mg/dl, then the clinical laboratory runs at least two controls, typically one toward the low end of that range and another in the high end. This concept is difficult to apply to Clinical IHC, a field where terms like “analytic measurement range” does not yet have meaning.
It is generally recommended that Clinical IHC laboratories use both a “low” and “high” control.4,6–8 However, the terms low and high are ambiguous because, without calibrators, there is no correlation with actual analyte concentrations. Without an external calibration standard, clinical IHC test results are expressed in terms of color intensity. It is not presently possible to routinely quantify a low or high test result with units of measure traceable to an objective standard, such as the number of molecules per cell. Consequently, difficulties arise in standardizing test results. Most notably, a “low” in one laboratory can be “high” in another. We recently described the observation that samples expressing approximately 104 molecules of human epidermal growth factor receptor type II (HER-2) per microbead can be strongly positive by one commercial HER-2 test and completely negative by another.9 This discrepancy arises because the two Food and Drug Administration (FDA)–cleared commercial HER-2 tests have different analytic response curves. Differences in analytical sensitivity among FDA-cleared commercial IHC tests were previously suspected based on correlation studies10–16 but never before quantitatively measured. For example, two separate “high” controls can both yield a strong stain and still have very different analyte concentrations. Once the analyte concentration is on the analyte response curve plateau, then analyte concentration and stain intensity no longer correlate with one another.
In this study, we address the question of how these uncontrolled analyte concentrations affect the sensitivity of IHC controls in detecting primary antibody degradation. The purpose of an assay control is to detect aberrations in the reagents, procedure, or conditions that could affect the test result. Ideally, the control will be sensitive to even slight aberrations, allowing for early intervention before patient test results are compromised. To our knowledge, there are no published data describing the relationship between analyte concentration in an IHC positive control and that control’s sensitivity in detecting IHC staining aberrations.
Materials and Methods
IHControls
The IHControls have been previously described.9,17,18 Briefly, cell-sized microbeads serve as a solid surface on which HER-2, estrogen receptor (ER), or progesterone receptor (PR) are anchored. IHControls are comprised of two different microbeads: analyte-coated glass test microbeads (7–8 micron diameter) and color standard microbeads (4.5 micron diameter). The analyte-coated microbeads bear covalently linked peptide epitopes for HER-2, ER, and/or PR. Among all of the various IHControls products used in this study, peptide analytes for all of the major clinical HER-2, ER, and PR tests are represented. The microbeads are suspended in a proprietary clear liquid that hardens after application to the glass microscope slide, thereby retaining the microbeads on the glass slide during baking, deparaffinization, antigen retrieval, and IHC staining. Once dried, the droplet can be treated as one would treat a tissue sample. Each dried microliter droplet on the slide incorporates approximately 5000 analyte-coated (test) microbeads.
The IHControls microbead suspension also includes smaller color standard microbeads, which are permanently colored dark brown regardless of the IHC staining procedure. These have been previously described.17 The color standard microbeads serve as a color intensity reference for standardizing color intensity measurements by image analysis. The IHControls also include a second type of test microbead that is not immunoreactive with the antibody in question. The antigenically irrelevant microbeads serve as an unstained, internal negative control (shown in Fig. 2).
Figure 2.
Photomicrographs of the IHControls bearing 8187 molecules (MEF) HER-2, after immunostaining with HercepTest at various primary antibody dilutions. Figure 2A illustrates the appearance under normal conditions, without primary antibody reagent dilution. Examples of a stained microbead bearing HER-2 and a color standard microbead are highlighted. The faint microbead outlines of unstained microbeads are due to light refraction, not staining. The photomicrographs also illustrate the smaller color standard microbeads to which the stained test microbeads are compared. Figure 2B and C illustrate representative immunostaining with the indicated primary antibody dilution. An example of an unstained microbead that bears an antigenically irrelevant analyte is highlighted. A similar set of images using a slightly shorter photomicrographic exposure is shown in Fig. S1. Scale bar 10 µm. Abbreviations: MEF, molecules of equivalent fluorochrome; HER-2, human epidermal growth factor receptor type II.
The IHControls test microbeads are manufactured at a series of different analyte (peptide) concentrations that differ by approximately one log, from 106/microbead (the highest concentration) to 102 (the lowest concentration). Detailed explanations of analyte quantification with the IHControls have been previously described.9 Briefly, each of the HER-2, ER, or PR analytes are synthesized as peptides with a single fluorescein. These peptides are chemically conjugated to the microbeads. By determining the number of fluorescein molecules per bead, we learn the analyte concentration. Analyte concentrations on the microbeads are traceable to an already-existing microbead fluorescence standard called “molecules of equivalent fluorochrome” (MEF).
Photomicroscopy
Images were acquired as previously described.9 Briefly, we used either (1) a Nikon Eclipse E400 microscope fitted with a Spot Imaging Solutions RT-cooled charge-coupled device (CCD) color camera, Model 2.3.0 (Diagnostic Instruments, Inc.; Sterling Heights, MI), or (2) a Zeiss Axioskop microscope fitted with a Spot Imaging Solutions Insight Gigabit CCD camera (Diagnostic Instruments, Inc.). For any single experiment, the same camera was used for photomicroscopy. Before photomicroscopy, the camera was white-balanced and a flat-field correction was performed. For brightfield photomicroscopy of IHControls, the microscope optics are first set for Köhler illumination. Tissue controls are photographed with Köhler illumination. For IHControls, once Köhler illumination was established, the condenser aperture is then opened wide because the microbeads have more than sufficient contrast. With this adjustment, unstained test microbeads are faintly visible alongside stained microbeads. The camera software was set at a gamma of 1.0, using manual (fixed) photographic exposure times. We found that slightly over-exposing the images improved the accuracy of microbead quantification by creating a slightly whiter background. This increased image contrast. The benefit of the slightly whiter background is that the image segmentation step in the MatLab program more rapidly locks in on the relevant microbeads for quantification. Avoiding occasional spurious image segmentation errors improved the accuracy of the stain intensity measurement. This adjustment does not affect the final readout because stain intensity is calculated relative to an internal optical standard (the color intensity microbead). Whole slide imaging was not used. Each slide’s color intensity was measured by averaging three images per spot (slide). Each data point in the “Results” section represents the mean ± standard deviation (SD) of triplicate slides.
IHControls Stain Intensity Image Quantification
To promote consistency, we kept the photomicroscopy settings constant within each experiment. This includes both the optical settings, such as condenser and illumination apertures, and camera settings, such as exposure time. For quantification of IHControls stain intensity, we developed a custom algorithm embedded in MatLab (MathWorks Corp.; Natick, MA), as previously described.17 The algorithm measures image intensity of the test microbeads’ rims relative to the color standard microbeads’ rims. Consequently, IHControls stain intensity is expressed as a ratio. A score of 1.0 means that the test microbeads, stained for HER-2, ER, or PR, are equally intense in color (expressed in mean pixel intensity) as the color standard microbeads. A score ≥1 represents a strong stain intensity.
Tissue Controls
For ER and PR, archival formalin-fixed, paraffin-embedded (FFPE) tissue blocks were obtained from the Tissue Biorepository of the Department of Pathology and Laboratory Medicine, Tufts Medical Center, under an approved institutional review board (IRB) protocol. For ER and PR tissue controls, normal uterine endometrium was used. We measured only the stromal cells, as they express lower levels of ER and PR than the glandular epithelium. HER-2 controls were comprised of two different breast carcinomas, one expressing HER-2 at a 3+ level and another expressing HER-2 at a 1+ level.
ER/PR Image Quantification of Tissue Sections
ER and PR (nuclear) staining was quantified using a custom algorithm in Image-Pro Premier 9.2 (Media Cybernetics, Rockville, MD). To quantify ER and PR stain intensity, a color deconvolution algorithm was first applied to the image, using a color vector for the brown color of diaminobenzidine (DAB). The color vector was part of a subroutine in the Image-Pro Premier software. The color intensity of the image pixels was then plotted on a histogram, on a 0 to 255 scale. On this initial scale, darker stain intensities had lower numbers. A threshold channel number was manually identified on this histogram that included immunostained nuclei but not other unstained elements in the image. In other words, we segmented stained nuclei from other image elements by virtue of the degree of brown staining. Appropriate image segmentation was verified by visual inspection. With very faint or absent immunostaining, determining the most appropriate threshold with a precise channel number was imprecise. Regardless, we found that this did not materially affect the data. All of the image segmentation threshold settings that seemed reasonable for such faintly stained images produced similarly low numbers. After segmentation of the image nuclei, we measured the mean pixel intensity of those pixels. We then subtracted the mean pixel intensity from 255 so that darker stain intensities will have higher numbers and weak stain intensities will have lower numbers. We found this scale to be more intuitive.
Parenthetically, we initially attempted to apply a more sophisticated image algorithm that took advantage of the hematoxylin counterstain (for unstained nuclei). We initially hoped that the counterstain would help segment the images, identifying unstained nuclei. However, we found substantial variability in the counterstain from different manufacturers, thus confounding the analysis. For this reason, we adopted the aforementioned algorithm that did not incorporate hematoxylin as a factor for image segmentation. The single color (brown) image segmentation algorithm was found to be reliable, at least when used with the ER/PR tissue control (uterine endometrium).
Before using this algorithm in these experiments, we validated it against the interpretive score provided by a panel of observers, using a series of different images of various intensities. Validation testing was repeated in several different experiments. The accuracy of the algorithm closely aligned with the panel. When the image quantification scores were graphed against the panel consensus scores, there was a strong positive correlation with slopes of 0.92 to 1.8. The correlation coefficients (R2) were >0.90. Precision data across multiple experiments yielded coefficients of variation (CVs) ranging from 6% to 12%. For these reasons, we believe the ER/PR nuclear stain intensity measurement algorithm to be accurate and precise.
HER-2 Image Quantification of Tissue Sections
Stain intensity for HER-2 (tissue sections) was measured using ImageJ (National Institutes of Health [NIH]; Bethesda, MD) with the ImmunoMembrane plugin.19 The program is freely available. We previously described details of its use.17 The algorithm includes color segmentation for distinguishing the brown color associated with DAB from the blue color associated with the hematoxylin counterstain. We used the program’s image intensity score, expressed on a relative 0 to 10 scale.
Immunohistochemistry Staining
IHC staining was performed using three different automated immunostainers, in three separate sites. For immunostains using HER-2, ER, or PR antibodies supplied by Dako Corp./Agilent, a Dako Autostainer (Dako Corporation; Carpinteria, CA) was used. These immunostains included HercepTest, PR 636, PR 1294, and ER 1D5/2-123. The HER-2 and ER/PR PharmDx kits (Dako/Agilent Technologies; Carpinteria, CA) are sold with prediluted solutions and reagents, and were stained according to the manufacturer’s instructions. Slides were initially baked at approximately 57C to 60C for 40 min, de-paraffinized in xylene, and then hydrated in decreasing grades of ethanol. Antigen retrieval was performed using Dako’s antigen retrieval solutions provided with the HER-2 or ER/PR kits. For HER-2, antigen retrieval was performed in a 97C to 98C water bath for 40 min, as per the manufacturer’s instructions. For ER/PR antigen retrieval, the slides were processed for 25 min in a Biocare Medical Decloaking Chamber (Biocare Medical; Pacheco, CA) pressure cooker. For all subsequent steps, the manufacturer’s reagents, buffers, and instructions were followed. PR antibody 636 was purchased separately from Dako and coupled with the ER/PR PharmDx kit detection system.
For immunostains using antibodies supplied by Leica Corp. (Leica Biosystems; Buffalo Grove, IL), we performed the testing on a Bond III instrument. Slides were baked at 60C to 62C for 20 to 30 min. Deparaffinization and antigen retrieval were performed on the instrument using the manufacturer’s reagents and protocols. We used the Leica ER 6F11, PR 16, and HER-2 CB11 antibodies with the kit detection reagents.
For immunostains using antibodies supplied by Ventana Medical/Roche Corp. (Ventana Medical Systems, Inc.; Tucson, AZ), we performed the testing on a Benchmark XT. These samples were not baked, as per the usual protocol for that clinical IHC laboratory. Deparaffinization and antigen retrieval were performed on the instrument, using the manufacturer’s solutions and protocols. We used Ventana ER SP1, PR 1E2, and HER-2 4B5 antibodies coupled with the kit detection reagents.
At the end of each immunostaining protocol, the slides were removed from the instruments, dehydrated through increasing grades of ethanol, immersed in xylene, and coverslipped using Permount (Thermo Fisher Corp.; Waltham, MA).
When diluting primary antibodies, so as to simulate degradation of the reagent, antibodies were diluted in 10 mM Tris-buffered saline (TBS)/0.05% Tween-20/0.34% Brij-35/0.05% ProClin, at pH 7.6. The undiluted group represents the manufacturers’ primary antibody that is provided at the recommended concentration.
Pathologists’ Assessment of Controls
The slides associated with the HercepTest and PR 1294 immunostains were also evaluated by four pathologists (S.N., M.P., J.D.G., and K.G.C.). The purpose of the pathologists’ assessment was to compare (objective) image analysis with (subjective) manual scoring. The slides were covered with a 3 × 1 inch adhesive label. A small hole was created in the label, so that only a single control was visible on each slide. The slides were coded with a random number and organized in a random order. The pathologists were also provided with three slides showing the normal (“baseline”) level of stain intensity for the particular control being evaluated. These baseline slides demonstrated the expected level of stain intensity if the stain is working properly. Each pathologist was asked to first review the baseline slides, followed by the unknown slides. We asked the pathologists to either pass or fail each of the numbered slides. If the stain intensity of the control (on the numbered slide) is indistinguishable from that of the baseline control, then it should pass. If, on the contrary, the control has a visibly lower stain intensity relative to the baseline control, then it should fail. As each experimental group is comprised of triplicate slides, each experimental group (i.e., each dilution) in the pathologist assessment was also evaluated in triplicate.
After each of the four pathologists reviewed the first set of slides, relating to one type of control (e.g., an IHControl), all the labels were changed. New labels were applied to the slide, this time revealing a different control (e.g., a tissue control) on the same slide. Another round of review was conducted, in which a tissue control was evaluated. The pathologists evaluated only one IHC control during each round so that the appearance of one does not influence their interpretation of another on the same slide. This process was repeated until all of the evaluations were complete.
Statistical Analysis
Each data point represents the mean ± SD from triplicate slides. Each slide bears an IHControl spot containing approximately 5000 analyte-coated microbeads. To quantify a single IHControl spot, we sampled three different microscopic areas. This is analogous to sampling three fields of a patient’s breast carcinoma for assessment of HER-2 or ER/PR. From these three fields, we calculated the mean stain intensity per spot (slide). Each data point in Figs. 1B, 5B, and 5E represents the mean ± SD of three separate IHControl spots, each on a separate slide. The data for Figs. 1C, 5C, and 5F represent the mean of two separate microscopic fields. Regression curves in Figs. 1 and 5 were calculated using the linear regression tool of Excel (least mean squares method).
Figure 1.
(A) HercepTest response curve with IHControls, correlating immunohistochemical stain intensity (y-axis) as a function of HER-2 concentration (x-axis). HER-2 concentration (x-axis) is expressed in units of molecules per microbead, traceable to units of MEF. Each data point is the mean ± SD of triplicate slides, that is, three independent staining replicates. (B) Graphical representation of IHControls stain intensity after diluting the HercepTest primary antibody by the factor indicated on the x-axis. The rate of decline in stain intensity is apparent as the slope of the regression lines. (C) Graphical representation of tissue section control stain intensity after the same primary antibody dilutions. Both tissue controls and IHControls are mounted on the same slide. Abbreviations: HER-2, human epidermal growth factor receptor type II; MEF, molecules of equivalent fluorochrome.
Figure 5.
Sensitivity analysis of ER and PR controls. Panels A to C represent the ER immunostain using mAbs 1D5 and 2-123. Panels D to F represent the PR immunostain using mAb 1294. Panels A and D: Analytic response curves, correlating immunohistochemical stain intensity (y-axis) as a function of ER (Panel A) or PR (Panel D) concentration (x-axis). Panels B and E: Graphical representation of stain intensity after diluting the ER (Panel B) or PR 1294 (Panel E) primary antibodies. The dilution is indicated along the x-axis. Panels C and F: Graphical representation of a tissue controls’ stain intensity after the same primary antibody dilutions. For all panels, each data point is the mean ± SD of triplicate slides. Abbreviation: MEF, molecules of equivalent fluorochrome.
Sensitivity of the controls is defined as their ability to identify aberrations in immunostaining. In this study, we simulated one of these aberrations. Primary antibody degradation is simulated by dilution. Regardless of the specific immunostaining aberration (degraded reagents, incorrect times or temperatures, improper reagent dispense, etc.), the result is the same—a decrease in stain intensity. Therefore, we express the sensitivity of immunostaining controls as the group-specific slope of the regression line for stain intensity versus dilution. To estimate these parameters, the regression model includes covariates for group, dilution, and an interaction between group and dilution. The interaction term allows us to test the statistical significance of the differences between the slopes for the three groups.
Results
Evaluation of Immunostaining Control Sensitivity
In this investigation, we measure and compare the sensitivity of various immunostaining controls, using both IHControls and tissue controls. The IHControls are available across a broad range of HER-2, ER, or PR concentrations, from approximately 100 to 1,000,000 molecules per microbead. The concentrations of HER-2, ER, or PR in tissue controls are not known. For all of these experiments, we placed one or two tissue controls on the same slide as IHControls (see “Materials and Methods” section). Being on the same slide, the tissue control(s) and IHControls were exposed to the same staining treatment. Afterward, stain intensity was quantified by image analysis (see “Materials and Methods” section). Quantitative analysis was performed using all major commercial HER-2, ER, and PR antibodies.
We define “sensitivity” of a positive control in terms of its ability to reveal an abnormality in the reagents, protocol, or conditions that produce atypically (low) stain intensity. To measure sensitivity, we simulated primary antibody degradation by diluting it (see “Materials and Methods” section). For example, a 1:2 dilution simulates a 50% degradation of the antibody’s efficacy. A control that shows measurably weaker staining at a 1:2 dilution is more sensitive than a control that requires a 1:4 dilution to show a similarly weak stain intensity. Representative data (for HercepTest) are shown in Fig. 1.
HER-2 (HercepTest) Sensitivity Analysis With IHControls
Our data demonstrate that the sensitivity of an IHC positive control depends on the analyte concentration and where that concentration falls on the immunostain’s analytic response curve. Therefore, we illustrate each immunostain’s analytic response curve in conjunction with the sensitivity data. Figure 1A depicts the analytic response curve for the HercepTest stain over a three-log range of HER-2 concentrations using IHControls, as previously described.9 HER-2 concentration is graphed on the x-axis, expressed in molecules per microbead. HER-2 concentration is traceable to standardized units of MEF, as explained in “Materials and Methods” section and previously published.9 The y-axis of Fig. 1A depicts stain intensity, as measured by image analysis (see “Materials and Methods” section). IHControl stain intensity (on the y-axis) is expressed as a ratio (see “Materials and Methods” section).
Figure 1A illustrates that the two highest IHControls, with HER-2 concentrations of 1,086,658 and 77,913 molecules (MEF) per microbead, are on the analytic response plateau. Even though their concentrations are more than a log different, their stain intensities are similar. IHControls or a patient sample anywhere in this plateau region will stain at the same maximal level. Figure 1A also illustrates that samples with HER-2 concentrations between approximately 1000 and 10,000 molecules (MEF) per microbead produce a variable stain intensity, depending on the HER-2 concentration. Stain intensity in this portion of the curve is concentration dependent. We then sought to identify the optimal HER-2 IHControl concentration that would most sensitively detect dilution of the primary antibody (HER-2) reagent.
The HercepTest primary antibody dilutions are graphed on the x-axis of Fig. 1B. The various HER-2 IHControls (each having different concentrations) are graphed as separate curves in Fig. 1B. Figure 1B illustrates the effect of primary antibody dilution (x-axis) on stain intensity (y-axis). As expected, diluting the HercepTest primary antibody results in fainter stain intensity. However, there are significant differences among the various IHControls, depending on their HER-2 concentrations. The IHControl with the highest HER-2 concentration, on the plateau portion of the analytic response curve (Fig. 1B, red line), is the least sensitive in detecting reagent degradation. It shows a much slower rate of decline in stain intensity (as the primary antibody is diluted) compared with the 8187 HER-2 molecules per microbead IHControl (Fig. 1B, purple line). The 8187 HER-2 molecules per microbead IHControl are the closest to the concentration-dependent portion of the analytic response curve. It produces high stain intensity under normal staining conditions (i.e., dilution of 0), and the fastest rate of decline in stain intensity as the primary antibody is diluted. Therefore, it is the most sensitive in detecting subtle degrees of primary antibody reagent failure.
In summary, IHControls bearing antigen concentrations on the analytic response curve plateaus are less sensitive. As the analyte concentration increases along the plateau region, the controls are progressively less sensitive; greater primary antibody reagent dilutions are required before the stain intensity declines. Controls having analyte concentrations in or near the concentration-dependent region are the most sensitive.
We quantitatively compared the sensitivities of the various IHControls by testing whether the slopes of the regression lines differ significantly for the three IHControls groups. The estimated slope for stain intensity (for each doubling of the dilution) for the 8187 IHControl group (slope: −0.24, 95% confidence interval [CI]: −0.28 to −0.20) is statistically significantly less than the 77,913 IHControl group (slope: −0.15, 95% CI: −0.19 to −0.11, p=0.005), which in turn is less than the 1,086,658 IHControl group (slope: −0.08, 95% CI: −0.12 to −0.03, p=0.015). These data demonstrate a progressive increase in sensitivity as HER-2 concentration dropped closer to the concentration-dependent region of the analytic response curve.
HER-2 (HercepTest) Sensitivity Analysis Using Tissue Controls
Figure 1C illustrates the same stain intensity curve as in Fig. 1B but for tissue section controls expressing low (1+) and high (3+) levels of HER-2. Unlike the IHControls, the HER-2 concentrations in these tumor tissue sections are unknown. Tissue stain intensity is measured with the ImmunoMembrane plugin using ImageJ and expressed on a 0 to 10 relative numeric scale (see “Materials and Methods” section). Figure 1C demonstrates that for the 3+ control, stain intensity progressively declines with increasing dilutions until there is no detectable staining at a 1:32 primary antibody reagent dilution. This dilution curve profile most closely compares with the 8187 molecules IHControl group (Fig. 1B, purple line).
The 1+ HER-2 tissue control dilution curve is difficult to quantitatively evaluate because the stain intensity numbers are so low. At such low intensity scores, measurement error is more significant (as a percentage). The main reason for the error is that the ImmunoMembrane image analysis program cannot express tissue intensity scores in fractional numbers less than 1. Consequently, although the stain intensity at the 1:2 dilution is subjectively weaker than the undiluted primary antibody group, the scores still register as 1. For example, the 1:4 dilution group is calculated to have a stain intensity of 0.5 because half of the image fields had no detectable staining whereas the other half registered as a score of 1.
Representative Photomicroscopic Images: IHControls and Tissue Controls
Figures 2 and 3 depict representative images from the data set associated with Fig. 1B and C, respectively. Figure 2 illustrates the appearance of the 8187 IHControl at three different primary antibody dilutions. The HER-2-coated microbeads stain strongly in the undiluted antibody group (Fig. 2A). Successive serial dilutions result in progressively weaker staining (Fig. 2B) until there is no detectable staining (Fig. 2C). The 1:4 primary antibody dilution’s stain intensity (Fig. 2B) is easily distinguished from the undiluted group (Fig. 2A) when placed side by side. This subjective assessment is in agreement with objective stain intensity quantification data (Fig. 1B). Consequently, if this lower level of immunostaining occurred while the IHControl was serving as an on-slide control, the pathologist or histotechnologist should hopefully be able to appreciate the existence of a quality control (QC) problem. We will test this expectation in Fig. 4. As described in the Materials and Methods, Fig. 2 is slightly over-exposed to enhance contrast, for more accurate image quantification. A shorter photomicrographic exposure of the same slides is shown in Fig. S1.
Figure 3.
Photomicrographs of tissue control immunostaining with HercepTest after the indicated primary antibody dilutions. These images are representative examples from the data set associated with Fig. 1C. Panels A to C are of the human epidermal growth factor receptor type II (HER-2) high (3+) tissue control. Panels D to F are of the HER-2 low (1+) tissue control. The images chosen for illustration were selected as representing the dilutions showing both extremes of staining and an intermediate dilution. These dilutions differ for the 3+ and 1+ tissue controls. Scale bar 10 µm.
Figure 4.
Pathologists’ interpretation of the HER-2 IHControls and tissue controls for detecting immunostaining problems. Panel A: Pass rate of the IHControls (y-axis) after increasing primary antibody dilutions (x-axis). “Pass” means that the stain intensity appears normal, without any visually appreciable decrement. The pathologists fail the 8187 molecules IHControl group (purple line) at lower dilutions than the 77,913 molecules IHControl group (green line). Panel B: Pass rate of the high (3+) and low (1+) tissue controls after increasing primary antibody dilutions. The low (1+) control fails at a lower primary antibody dilution. Each data point in both panels is the mean ± SD of triplicate slides. Abbreviation: HER-2, human epidermal growth factor receptor type II.
Figure 3 comprises representative images of the high (3+) and low (1+) tissue controls associated with Fig. 1C. Figure 3A to C show positive tissue control immunostaining using an HER-2 3+ breast carcinoma. Figure 3D to F show positive tissue control immunostaining with an HER-2 1+ breast carcinoma. The dilutions shown for the upper (Fig. 3A–C) and lower (Fig. 3D–F) panels are not the same. For both upper and lower rows, the far left panel (Fig. 3A and D) is the undiluted antibody group. The panel at the far right (Fig. 3C and E) is the dilution group that yields no staining. The panel in the middle (Fig. 3B and D) is an intermediate between these two extremes. These images are consistent with the image quantification data of Fig. 1C.
HER-2 Controls Comparison by Subjective Assessment
As a practical matter, it is important to know how well a pathologist can appreciate stain intensity decrements in the IHControls or tissue controls. This is important because their appearances are intended to potentially illustrate an immunostaining problem. IHC controls are currently interpreted by visual (subjective) inspection, not image analysis (such as is used for Fig. 1B and C). Image quantification of IHC controls is both precise and objective but it is not a standard of practice in clinical IHC.18 Therefore, we performed additional experiments evaluating whether visual evaluation would differ from image analysis quantification. To investigate this question, four surgical pathologists at three different institutions examined and evaluated the slides.
Four pathologists reviewed the slides immunostained with HercepTest and PR 1294. These are the same slides that were used in Fig. 1B and C (HercepTest) and Fig. 5E and F (PR 1294). The slides were coded and presented in a random order, as described in the “Materials and Methods” section. Moreover, the pathologists only reviewed one control at a time. The other controls (on the same slide) were covered with a label (see “Materials and Methods” section). At the outset, the pathologists were presented with slides representing the normal, expected level of stain intensity for these particular immunostains. Each pathologist separately scored the coded slides as to whether QC passed or failed. A passing score means that the QC stain intensity of a coded slide was indistinguishable from the normal, expected stain intensity. A failing QC score means that the QC stain intensity is clearly lower than the normal, expected stain intensity.
Figure 4A shows the pathologists’ assessment data using the 77,913 and 8187 IHControls. These IHControls are the same ones characterized in Fig. 1B, after immunostaining with HercepTest. The graph represents the pooled interpretations of all four pathologists. The data show that all of the slides that were stained normally (dilution of “0”) received a score of “pass.” This is expected because these slides have the same stain intensity as the baseline slides. Figure 4A demonstrates that the pathologists were able to more readily observe a drop in stain intensity with the 8187 IHControl as compared with the 77,913 IHControl. Namely, the purple line (8187) declines before the green one (77,913) as the dilutions increase. These pathologist interpretations mirror the image quantification data of Fig. 1B, showing that the 8187 IHControl has a steeper rate of stain intensity decline.
Figure 4B describes the results of pathologists’ interpretations of the low (1+) and high (3+) tissue controls with simulated immunostain failure. The data reveal that the low-positive tissue control (dotted line) is more sensitive. Almost all of the pathologist readings detected a decrement in stain intensity at the very first (1:2) primary antibody dilution. By contrast, another doubling dilution (to 1:4) was required to detect simulated failure using the high (3+) tissue control.
Although the low (1+) tissue control is more sensitive, there is a potential drawback to using such a faintly staining positive control. With an expected stain intensity so close to zero, we do not know if the normal degree of day-to-day immunostaining variability will result in false alarms. In other words, we do not know if the decrement detected by the pathologists with the 1:2 dilution slides using a low-positive control (Fig. 4B) will periodically occur even without diluting the primary antibody. Immunostain variability is not the only possible cause of an inappropriate QC failure. We previously demonstrated that variability in the tumor itself, from sections deeper in the paraffin block, can lead to a QC failure.18
Sensitivity Analysis Using ER and PR IHControls
Summarizing the simulated HercepTest failure studies of Figs. 1 to 4, the data show that the sensitivity of a positive control is dependent on the analyte (i.e., HER-2) concentration and where that concentration falls on the analytic response curve. We tested this observation on 10 other clinical HER-2, ER, and PR immunostains. Representative data are shown in Fig. 5, after immunostaining with the ER/PR PharmDx kit. For ER, the kit uses a cocktail of two monoclonal antibodies (mAbs), 1D5 and 2-123 (Fig. 5A–C). For PR, the kit uses mAb 1294 (Fig. 5D–F). Figure 5A depicts the ER analytic response curve. Figure 5D depicts the PR analytic response curve.
Fig. 5B and E illustrate that the least sensitive IHControls, which exhibit the slowest decline upon primary antibody dilution, are those furthest from the concentration-dependent portion of the analytic response curve. These IHControls have 819,056 (ER) and 1,379,113 (PR) molecules per microbead. Both of these IHControls concentrations are represented with red lines in Fig. 5B and E. By contrast, the most sensitive control in Fig. 5B (8187 molecules, in purple) has an analyte concentration that places it at the top of a concentration-dependent portion of the analytic response curve (Fig. 5A).
We quantitatively compared the sensitivities of the various IHControls by testing whether the slopes of the regression lines differ significantly for the three different groups (Fig. 5B and E). For ER (Fig. 5B), the estimated slope for stain intensity (for each doubling of the dilution) for the 8187 IHControl group (slope: −0.40, 95% CI: −0.48 to −0.32) was not statistically significantly less than the 77,913 IHControl (slope: −0.36, 95% CI: −0.43 to −0.28, p=0.45), which in turn was significantly less than the 819,056 IHControl group (slope: −0.22, 95% CI: −0.31 to −0.15, p=0.026). These data demonstrate a progressive increase in sensitivity as ER concentration dropped closer to the concentration-dependent region of the analytic response curve.
For PR (Fig. 5E), the estimated slope for stain intensity (for each doubling of the dilution) for the 8187 IHControl group (slope: −0.34, 95% CI: −0.41 to −0.27) was not statistically significantly greater than the 77,913 IHControl (slope: −0.42, 95% CI: −0.49 to −0.35, p=0.128), which in turn was significantly less than the 1,379,113 IHControl group (slope: −0.10, 95% CI: −0.17 to −0.03, p<0.001). These data demonstrate a substantial increase in sensitivity as PR concentration drops closer to the concentration-dependent region of the analytic response curve. However, two points within the same general concentration-dependent region of the curve have sensitivities that are indistinguishable from one another.
Sensitivity Analysis Using ER and PR Tissue Controls
In addition to IHControls, we also mounted a positive tissue control comprised of endometrium on the same slides. We only quantified the stain intensity of the stromal cells, which express lower concentrations of ER and PR as compared with the uterine glandular epithelium. Figure 5C (ER) and 5F (PR) illustrate the immunostain profiles for the tissue positive controls after serial primary antibody dilutions. For ER immunostaining (Fig. 5C), the data indicate comparable sensitivity for the 8187 IHControl and tissue control. Both show almost no staining if the primary antibody is diluted 1:4. For PR (Fig. 5F), the immunostain intensity exhibits a relatively slower decline, indicating a poorer sensitivity in identifying immunostaining problems. Although the same tissue control (endometrium) is used for both ER and PR, there are two differences that likely account for this sensitivity difference: (1) The ER and PR concentrations are likely different from each other, and (2) the analytic response curves of the ER and PR immunostains are different.
PR Controls Comparison by Subjective Assessment
As ER and PR are both nuclear stains, distinct from HER-2, we again investigated how easily pathologists can recognize lower than normal stain intensity. In this second evaluation, the pathologists are evaluating nuclear (rather than membrane) staining. Both the IHControls and the tissue controls were scored in a blinded fashion, as previously described. Figure 6 illustrates the pooled pathologist scores for the PR 1294 immunostain. The scores mirror the image analysis data shown in Fig. 5E and F. Figure 6A shows that at a 1:4 dilution of primary antibody, all four pathologists, in all of the replicates, recognized the decreased IHControls stain intensity; the pass rate percentage using the IHControls at a 1:4 dilution is zero. Figure 6B shows that the tissue control at the same primary antibody dilution does not produce as definitive a result. More than half the pathologists’ scores are still “pass.” Most of the pathologists could not appreciate lower stain intensity in the tissue control at this 1:4 dilution. These data suggest that the 77,913 IHControl is more sensitive than the tissue control. We interpret this finding to imply that the 77,913 IHControl has a more optimal PR concentration for the PR 1294 immunostain than the tissue control. If it were possible to measure the PR concentration in the tissue control, we would expect it to be higher, further from the concentration-dependent region of the analytic response curve.
Figure 6.
Pathologists’ interpretation of the PR IHControls and tissue controls for detecting immunostaining problems. Panel A: Pass rate of the 77,913 molecules IHControl group (y-axis) after increasing primary antibody dilutions (x-axis). Panel B: Pass rate of the tissue control after increasing primary antibody dilutions. Each data point is the mean ± SD of triplicate slides.
Discussion
This article highlights the importance of selecting positive IHC controls having an optimal concentration of analyte. To our knowledge, this is the first published experimental test on this question. Our data indicate that the most sensitive positive IHC controls have analyte concentrations close to the concentration-dependent region of the analytic response curve. This is the region where the stain intensity exhibits the greatest change in response to increases or decreases in analyte concentration. Controls with analyte concentrations on the plateau of the analytic response curve are less sensitive. Greater amounts of reagent degradation or other perturbations in the immunostaining conditions are required before being detectable.
The data suggest that our findings apply to both the IHControls and tissue controls. The important parameter affecting the sensitivity of a control appears to be the amount of immunoreactive analyte, regardless of the solid phase support to which it is attached (i.e., tissue or microbeads). Without calibrators, we do not know the analyte concentrations in the tissue controls. Nonetheless, we were able to compare the sensitivities of IHControls and tissue controls by comparing them side by side. IHControls and tissue controls were mounted on the same slides during the experiments. They were exposed to the same immunostaining conditions. We were then able to compare the changes in stain intensity for the IHControls and tissue controls. We analyzed stain intensity both by image analysis and by pathologists’ visual interpretation. Both modes of evaluation demonstrate that, allowing for likely differences in analyte concentrations, the solid phase matrix on which the analyte is mounted is largely inconsequential to these sensitivity analyses. The sensitivity of tissue controls appears to fall within the bounds of sensitivity set by the most and least sensitive IHControls.
In comparing tissue controls versus IHControls, there are advantages and drawbacks for each. For example, tissue controls suffer from unpredictable and unknown analyte concentrations. Moreover, tissue controls are labor-intensive to procure, validate, and prepare. In contrast, they also offer certain advantages relative to IHControls. Tissue controls most closely resemble the matrix of patient samples. IHControls have an artificial matrix that differs from tissue. If there were problems of background staining, or problems with the counterstain, the tissue control would more likely detect these.
Neither tissue controls nor IHControls can address preanalytical variables. This limitation is inherent in all external controls, be they tissue sections, cell lines, or IHControls. Consequently, external controls do not provide quality assurance on cold ischemic time, formalin fixation, dehydration, paraffin embedding, and microtomy. External controls such as tissue controls and IHControls are limited to analytic and postanalytic variables. Analytic variables include degradation of a reagent, instrument error, or improper antigen retrieval. Postanalytic variables relate to microscopy and pathologist interpretation.
We also performed the same analysis using nine other breast cancer immunostains: ER 6F11, ER SP1, PR 16, PR 636, PR 16, PR 1E2, HER-2 SP3, HER-2 4B5, and HER-2 CB11. We previously described these other immunostains’ analytic response curves.9 The data from these other immunostains lead to the same conclusion (data not shown for space constraints), with one caveat. Some immunostains, such as 4B5 (HER-2), SP3 (HER-2), CB11 (HER-2), and 1E2 (PR), did not demonstrate an optimal concentration. Multiple IHControls concentrations were approximately equally sensitive. Nonetheless, those findings do not contradict the conclusion of this article. These particular immunostains are different because, as previously published, they have a broad concentration-dependent range.9 These immunostains do not have an analytic response plateau within the concentration range that we tested (up to approximately 106 molecules per microbead). Consequently, multiple IHControls in the concentration-dependent region of the analytic response curve are approximately equally affected by primary antibody dilution.
These findings lead to an additional surprising conclusion. The same control can be optimally sensitive in one immunostain and suboptimal in another. The same analyte concentration may be in the concentration-dependent region of the analytic response for one immunostain while being on the plateau of another. There are likely numerous reasons for the differences in analytic response curves among various commercial immunostains to the same analyte. These reasons include analytic variables such as primary antibody affinity, the composition of the detection system reagents, and reaction conditions such as the time and temperature of incubation. The fact that there are differences in analytic response curves among commercial vendors, as previously described,9 is not surprising. These commercial immunostains were developed and validated without the means to measure analytic response curves. As a result, the response profiles of different manufacturers’ immunostains to the same analyte may often not align.
Our data and conclusions generally align with, and expand upon, published clinical IHC laboratory practice guidelines. Clinical IHC laboratory guidance documents recommend using both a high and low control.4,6–8 As already described (in the “Introduction” section), the terms “high” and “low” are ambiguous. Nonetheless, tissues or cell lines staining with a low or moderate intensity will likely have analyte concentrations closer to the concentration-dependent region of the analytic response curve as compared with “high” controls. Therefore, a tissue or cell line control whose stain intensity is moderate will likely be more sensitive than a strongly staining control. By contrast, a strongly staining positive control is esthetically reassuring but generates a false confidence in the immunostain’s performance. Strongly positive IHC controls will continue to stain intensely even when the immunostain is impaired. Therefore, the utility of a “high” control (as per published guidelines) is unclear.
This study also serves as a cautionary warning for validating new IHC immunostains. Our data highlight the fact that primary antibody concentration affects immunostain sensitivity. Lower primary antibody concentrations (i.e., with increasing dilutions) were associated with lower sensitivity (requiring higher analyte concentrations). In this regard, our findings are in agreement with the data of McCabe et al.20 This means that two different clinical IHC laboratories, using the same primary antibody and identical protocols and instruments, may see different patient test results if they use a different primary antibody dilution. Patient test samples with a high analyte concentration will probably be positive regardless of the primary antibody concentration. However, patient test samples with lower analyte concentrations might yield disparate test results depending on where the patient’s analyte concentration falls on the analytic response curve.
Looking forward, the IHControls may be a useful adjunct in the Clinical IHC Laboratory for promoting standardization and reproducibility. As most of the breast cancer tests evaluated in this study are FDA-cleared in conjunction with kit controls, we do not anticipate that IHControls will be used independently, without a cell line or tissue control. Instead, the IHControls may be supplemental as on-slide controls in addition to a tissue or kit cell line batch control. This role would facilitate compliance with published practice guidelines recommending on-slide controls.2,21–24 The fact that the IHControls have well-defined analyte concentrations makes it easier to ensure that the analyte concentration is close to the concentration-dependent region of the analytic response curve.
Supplementary Material
Acknowledgments
We are grateful to Dr. Ron Zeheb for his ongoing advice during the course of this project and review of this manuscript, Ms. Drorit Bogen for administrative support, Mr. Jason Badrinarain and Ms. Annette Barry for their technical assistance in operating the Bond III and Benchmark XT immunostainers, Mr. Matthew Bogen for advice on quantitative analysis and graphical presentation, and to the National Cancer Institute, National Institutes of Health (NIH), for funding this work.
Footnotes
Competing Interests: The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Three of the authors (SRS, KV, and SAB) have a patent (ownership) interest in the IHControls technology. The other authors (AKS, AB, JDG, SN, MP, KGC, and FN) declare they have no competing interests.
Author Contributions: SRS contributed to the conception and manufacture of the IHControls and execution of experiments. KV contributed to the manufacture of the IHControls and execution of the experiments and image quantification. AKS performed the photomicroscopy and image quantification. AB contributed to the experimental design and development of quality systems for manufacture. SAB contributed to the conception of the technology, experimental design, data review, and drafting of the manuscript. JDG, SN, MP, and KGC evaluated the immunostained slides and evaluated each of the controls. FN performed the statistical analysis. All authors have read and approved the manuscript.
Funding: The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Institutes of Health (NIH), National Cancer Institute (grant number R44CA183203 to SAB) and the National Center for Advancing Translational Sciences (also NIH; grant number UL1TR001064).
Literature Cited
- 1. Hammond M, Hayes D, Dowsett M, Allred D, Hagerty K, Badve S, Fitzgibbons PL, Francis G, Goldstein NS, Hayes M, Hicks DG, Lester S, Love R, Mangu PB, McShane L, Miller K, Osborne CK, Paik S, Perlmutter J, Rhodes A, Sasano H, Schwartz JN, Sweep FC, Taube S, Torlakovic EE, Valenstein P, Viale G, Visscher D, Wheeler T, Williams RB, Wittliff JL, Wolff AC. American Society of Clinical Oncology/College of American Pathologists guideline recommendations for immunohistochemical testing of estrogen and progesterone receptors in breast cancer. Arch Pathol Lab Med. 2010;134(7):907–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Hewitt S, Robinowitz M, Bogen S, Gown A, Kalra K, Otis C, Spaulding B, Taylor CR. Quality assurance for design control and implementation of immunohistochemistry assays; approved guideline–second edition. Wayne, PA: Clinical and Laboratory Standards Institute; 2011. January 31. CLSI Document I/LA28-A2. [Google Scholar]
- 3. Leong AS-Y, Leong TY-M. Standardization in immunohistology. Methods Mol Biol. 2011;724:37–68. [DOI] [PubMed] [Google Scholar]
- 4. Torlakovic E, Nielsen S, Francis G, Garratt J, Gilks B, Goldsmith J, Hornick JL, Hyjek E, Ibrahim M, Miller K, Petcu E, Swanson PE, Zhou X, Taylor CR, Vyberg M. Standardization of positive controls in diagnostic immunohistochemistry: recommendations from the International Ad Hoc Expert Committee. Appl Immunohistochem Mol Morphol. 2015;23(1):1–18. [DOI] [PubMed] [Google Scholar]
- 5. Medicare, Medicaid and CLIA programs; regulations implementing the Clinical Laboratory Improvement Amendments of 1988 (CLIA)—HCFA. Final rule with comment period. Fed Regist. 1992;57:7002–186. [PubMed] [Google Scholar]
- 6. Rasmussen O, Jorgensen R. Controls. In: Taylor CR, Rudbeck L. editors. Immunohistochemical staining . methods. 6th ed. Glostrup, Denmark: Dako Corporation; 2014. p. 165. [Google Scholar]
- 7. Hammond M, Hayes D, Allred D, Dowsett M, Hagerty K, Badve S, Fitzgibbons PL, Francis G, Goldstein NS, Hayes M, Hicks DG, Lester S, Love R, Mangu PB, McShane L, Miller K, Osborne CK, Paik S, Perlmutter J, Rhodes A, Sasano H, Schwartz JN, Sweep FC, Taube S, Torlakovic EE, Valenstein P, Viale G, Visscher D, Wheeler T, Williams RB, Wittliff JL, Wolff AC. American Society of Clinical Oncology/College of American Pathologists guideline recommendations for immunohistochemical testing of estrogen and progesterone receptors in breast cancer. J Clin Oncol. 2010;28(16):2784–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Torlakovic E, Nielsen S, Vyberg M, Taylor C. Getting controls under control: the time is now for immunohistochemistry. J Clin Pathol. 2015;68(11):879–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Vani K, Sompuram S, Schaedle A, Balasubramanian A, Bogen S. Analytic response curves of clinical breast cancer IHC tests. J Histochem Cytochem. 2017;65(5):273–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Bogina G, Zamboni G, Sapino A, Bortesi L, Marconi M, Lunardi G, Coati F, Massocco A, Molinaro L, Pegoraro C, Venturini M. Comparison of anti-estrogen receptor antibodies SP1, 6F11, and 1D5 in breast cancer: lower 1D5 sensitivity but questionable clinical implications. Am J Clin Pathol. 2012;138:697–702. [DOI] [PubMed] [Google Scholar]
- 11. Cheang M, Treaba D, Speers C, Olivotto I, Bajdik C, Chia S, Goldstein LC, Gelmon KA, Huntsman D, Gilks CB, Nielsen TO, Gown AM. Immunohistochemical detection using the new rabbit monoclonal antibody SP1 of estrogen receptor in breast cancer is superior to mouse monoclonal antibody 1D5 in predicting survival. J Clin Oncol. 2006;24(36):5637–44. [DOI] [PubMed] [Google Scholar]
- 12. Dekker TJ, Borg ST, Hooijer GK, Meijer SL, Wesseling J, Boers JE, Schuuring E, Bart J, van Gorp J, Mesker WE, Kroep JR, Smit VT, van de Vijver MJ. Determining sensitivity and specificity of HER2 testing in breast cancer using a tissue micro-array approach. Breast Cancer Res. 2012;14:R93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Gouvea A, Milanezi F, Olson S, Leitao D, Schmitt F, Gobbi H. Selecting antibodies to detect HER2 overexpression by immunohistochemistry in invasive mammary carcinomas. Appl Immunohistochem Mol Morphol. 2006;14(1):103–8. [DOI] [PubMed] [Google Scholar]
- 14. Rhodes A, Sarson J, Assam EE, Dean SJ, Cribb EC, Parker A. The reliability of rabbit monoclonal antibodies in the immunohistochemical assessment of estrogen receptors, progesterone receptors, and HER2 in human breast carcinomas. Am J Clin Pathol. 2010;134:621–32. [DOI] [PubMed] [Google Scholar]
- 15. Ricardo S, Milanezi F, Carvalho S, Leitao D, Schmitt F. HER2 evaluation using the novel rabbit monoclonal antibody SP3 and CISH in tissue microarrays of invasive breast carcinomas. J Clin Pathol. 2007;60:1001–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Rossi S, Laurino L, Furlanetto A, Chinellato S, Orvieto E, Canal F, Facchetti F, Dei Tos AP. Rabbit monoclonal antibodies: a comparative study between a novel category of immunoreagents and the corresponding mouse monocloncal antibodies. Am J Clin Pathol. 2005;124:295–302. [DOI] [PubMed] [Google Scholar]
- 17. Sompuram S, Vani K, Tracey B, Kamstock D, Bogen S. Standardizing immunohistochemistry: a new reference control for detecting staining problems. J Histochem Cytochem. 2015;63(9):681–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Vani K, Sompuram S, Naber S, Goldsmith J, Fulton R, Bogen S. Levey-Jennings analysis uncovers unsuspected causes of immunohistochemistry stain variability. Appl Immunohistochem Mol Morphol. 2016;24(10):688–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Tuominen V, Tolonen T, Isola J. ImmunoMembrane: a publicly available web application for digital image analysis of HER2 immunohistochemistry. Histopathology. 2012;60(5):758–67. [DOI] [PubMed] [Google Scholar]
- 20. McCabe A, Dolled-Filhart M, Camp R, Rimm D. Automated quantitative analysis (AQUA) of in situ protein expression, antibody concentration, and prognosis. J Natl Cancer Inst. 2005;97(24):1808–15. [DOI] [PubMed] [Google Scholar]
- 21. Rakha E, Pinder S, Bartlett J, Ibrahim M, Starczynski J, Carder P, Provenzano E, Hanby A, Hales S, Lee AH, Ellis IO; National Coordinating Committee for Breast Pathology. Updated UK recommendations for HER2 assessment in breast cancer. J Clin Pathol. 2015;68(2):93–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Wolff A, Hammond M, Hicks D, Dowsett M, McShane L, Allison K, Allred DC, Bartlett JM, Bilous M, Fitzgibbons P, Hanna W, Jenkins RB, Mangu PB, Paik S, Perez EA, Press MF, Spears PA, Vance GH, Viale G, Hayes DF; American Society of Clinical Oncology; College of American Pathologists. Recommendations for human epidermal growth factor receptor 2 testing in breast cancer: American Society of Clinical Oncology/College of American Pathologists clinical practice guideline update. J Clin Oncol. 2013;31(31):3997–4013. [DOI] [PubMed] [Google Scholar]
- 23. National Standards for High Complexity Laboratory Testing. CAP-ACP Clinical Immnohistochemistry Checklists: Part I and Part II. Ottawa, Canada: Canadian Association of Pathologists/Association Canadienne Des Pathologistes. [Google Scholar]
- 24. Lambein K, Guiot Y, Galant C, Salgado R, Colpaert C. Update of the Belgian guidelines for HER2 testing in breast cancer. Belg J Med Oncol. 2014;8(4):109–15. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.