The Cancer Immunotherapy Biomarker Testing Landscape Open Access

The immune system is divided into 2 distinct response components, innate and adaptive. The innate immune system responds rapidly but indiscriminately and includes neutrophils and macrophages, whose receptors recognize foreign antigens such as microbial products. In contrast, the adaptive immune response is highly targeted and precise, with a vast antigen response diversity achieved through the rearrangement of B-lymphocyte immunoglobulin genes and T-lymphocyte antigen receptor genes.⁸  Although slower to develop, the incredible selectivity of the adaptive immune system appears to underlie the clinical effectiveness of cancer immunotherapy.

Although associations between infections and cancer regression had been made as early as the 1700s, William Coley, MD, is widely acknowledged as the “Father of Immunotherapy” because of his systematic, lifelong study of induced bacterial infections as therapy for sarcoma patients, beginning in 1891.⁹  Although Coley firmly believed in the tie between severe infections and cancer regressions, the mechanisms underlying the responses he observed were not elucidated until the mid to late 1950s with the realization that the immune system could recognize and attack tumors based on specific antigens.^10,11 

Our current mechanistic understanding of immune-mediated cancer elimination is based on the widely accepted process referred to as the cancer immunity cycle (Figure 1).¹²  This process, which is dependent on the adaptive immune system, is initiated when necrotic and/or apoptotic tumor cells release immunogenic neoantigen proteins into the tumor microenvironment. These neoantigens are recognized as foreign by the immune system, and are engulfed and processed by dendritic cells, which migrate through lymphatics to lymph nodes to prime and activate tumor-specific cytotoxic T-cell responses. Tumor-specific CD8⁺ cytotoxic T cells then traffic back to the tumor to find and destroy their cancer cell targets, triggering release of additional tumor neoantigens and renewing the cancer immunity cycle.

Interruption of the cancer immunity cycle can clearly occur and results in immune evasion, uncontrolled tumor growth, and clinical progression. The success of cancer immunotherapy to date has come from a careful elucidation of this cycle, including the specific parameters and mechanisms underlying immune responsiveness and resistance, an approach that will continue to be necessary for further advancement of the field. The first of the cancer immune mechanisms to be understood and successfully targeted were the immune regulatory checkpoints.

A complex system of regulatory checkpoints has evolved in humans to ensure immune homeostasis and prevent autoimmunity. Central to this system of regulation are immune checkpoint proteins present on T lymphocytes and antigen-presenting cells. More than a dozen immune checkpoint proteins have been discovered, some functioning to activate the immune system and others functioning to down-regulate it.¹³  The pioneering work of Jim Allison, PhD (MD Anderson Cancer Center, Houston, Texas), led to the discovery that cytotoxic T-lymphocyte associated protein 4 (CTLA-4) acts as an inhibitory checkpoint molecule, functioning as a brake on antitumor immune responses.¹⁴  Dr Allison hypothesized that inhibiting CTLA-4 via antibody blockade would abrogate suppression of immune priming and enhance the immune response to cancer.¹⁴  Initial positive results in animals led to subsequent clinical validation of CTLA-4 inhibition in metastatic melanoma and to eventual FDA approval of ipilimumab in 2011.¹⁵  Dr Allison and Tasuku Honjo, MD, PhD (Kyoto University, Kyoto, Japan) were jointly awarded the 2018 Nobel Prize in Physiology or Medicine for their discovery of cancer therapy by inhibition of negative immune regulation.¹⁶  In 2014, the FDA approved nivolumab, which inhibits the programmed death receptor-1 (PD-1), a checkpoint molecule that directly regulates T-cell effector activity in the tumor microenvironment. There are currently several FDA-approved checkpoint inhibitor therapies targeting CTLA-4 as well as PD-1 and its ligand, PD-L1 (Table 1),^17,18  with numerous additional immune checkpoint inhibitors in clinical development.⁵ 

The PD-L1 ligand and its cognate receptor, PD-1, are components of an immune inhibitory axis that, when activated, negatively regulates T-cell signaling, effector function, and killing capacity.¹³  Tumor cells can activate this pathway through expression of PD-L1 on the cell surface via 2 distinct mechanisms, a constitutive mechanism triggered by genomic alterations (see Table 2) and an inducible mechanism termed adaptive immune resistance.^13,19  In the latter case, tumor surface expression of PD-L1 occurs in response to the release of the cytokine interferon γ (IFN-γ) from T cells upon tumor recognition. In this way, tumors avoid immune-mediated destruction by activating a naturally occurring immune regulatory pathway. Reactivation of T-cell proliferation and effector function through antibody blockade of PD-1/PD-L1 has been conclusively shown to yield durable clinical benefit to patients across multiple tumor types.²⁰ 

Five anti–PD-(L)1 therapies (atezolizumab, avelumab, durvalumab, nivolumab, and pembrolizumab) have been approved by one or more global regulatory agencies across multiple cancer indications (Table 1).⁴  A recent systematic review and meta-analysis by Khunger et al²¹  examined 41 PD-1/PD-L1 inhibitor clinical trials and found that PD-L1 IHC positivity in tumor cells and immune cells was predictive of favorable response across all tumor types. Generally, PD-L1 expression is associated with a greater likelihood of benefit from anti–PD-(L)1 therapy, but, in contrast to traditional companion diagnostic markers such as HER2 or ALK, patients with PD-L1–negative tumors also benefit from these therapies up to 20% of the time, depending on the specific study, therapy, and indication.^22–24  In response to these data, the FDA created a novel diagnostic category termed complementary diagnostic, defined (in a draft definition given at the 2016 American Society of Clinical Oncology annual meeting) as “a test that aids in the benefit-risk decision-making about the use of the therapeutic product, where the difference in benefit-risk is clinically meaningful.”²⁵  In contrast, the FDA defines a companion diagnostic as an “in vitro diagnostic device or an imaging tool that provides information that is essential for the safe and effective use of a corresponding therapeutic product.”²⁶ 

Four separate PD-L1 assays^27–30  have been developed and approved in association with the clinical development of the different FDA-approved anti–PD-(L)1 therapies and cancer indications, leading to distinct therapy-indication-test combinations (Table 3). In addition, a fifth assay based on the 73-10 PD-L1 clone is currently in development in association with the PD-L1 inhibitor avelumab. This is a rapidly evolving field, however, and changes to the current PD-L1 testing landscape are expected with approvals of additional immunotherapies, therapeutic indications, and assays. Based on each combination of assay, therapy, and indication, the PD-L1 scoring measure and threshold for positivity (ie, cutoff) can vary. For example, the companion diagnostic assay based on the 22C3 clone for the pembrolizumab non–small cell lung cancer (NSCLC) first-line indication uses a tumor proportion score with a 50% cutoff, whereas the companion diagnostic assay based on the SP142 clone for atezolizumab in the triple-negative breast cancer indication assesses tumor-infiltrating immune cells at a 1% cutoff (Table 3). The most current version of the manufacturer's package insert and/or the interpretation guide for each assay should be used for clinical testing and scoring. In addition to the FDA–approved assays, nonapproved PD-L1 antibody clones have been independently developed as laboratory-developed tests, also known as laboratory-developed assays, per College of American Pathologists and Clinical Laboratory Improvement Amendments of 1988 guidelines.^31,32  A 2017 comprehensive review of PD-L1 laboratory-developed assays demonstrated variable concordance between them, concluding that standardization is needed before PD-L1 laboratory-developed assays can be recommended for routine clinical use.³³  Because FDA labeling of PD-1/PD-L1 immunotherapies with an associated companion diagnostic requires tumors to be tested for PD-L1 as determined by an FDA-approved test, use of PD-L1 laboratory-developed tests in these indications would be considered off-label use of the respective therapy.^34,35 

Based on the associated therapeutic efficacy and safety data, the FDA has designated PD-L1 as a complementary diagnostic for some specific PD-(L)1 inhibitor indication intended uses and as a companion diagnostic for others (Table 3). In addition, the FDA has demonstrated that it will adapt drug and diagnostic labels based on emerging clinical and safety data. For example, in June 2018, the FDA restricted the use of pembrolizumab and atezolizumab in patients with locally advanced or metastatic urothelial cancer to those whose tumors express PD-L1 and who are not eligible for cisplatin-containing therapy. This restriction was based on the finding in 2 ongoing clinical trials that PD-L1–low patients treated with anti–PD-(L)1 therapy had decreased survival compared with patients receiving chemotherapy.³⁶  The labels of both drugs and the associated diagnostic assays were revised accordingly, thereby converting PD-L1 from an optional complementary diagnostic to a mandatory companion diagnostic for this specific clinical indication.

Several studies have assessed the analytic and technical concordance between the on-market PD-L1 IHC assays. The Blueprint Project was an industry-academic partnership sponsored by the International Association for the Study of Lung Cancer designed to provide information on the analytic comparability of the various PD-L1 assays. Phase 1 of the project examined 39 NSCLC cases scored by 3 pathologists across 4 PD-L1 antibody clones (22c3, 28-8, SP142, and SP263).³⁷  The results demonstrated comparable tumor cell staining among the 22c3, 28-8, and SP263 assays, with the SP142 assay exhibiting fewer stained tumor cells relative to the other 3 assays. Variability of immune cell staining was greater than for tumor cell staining across all assays, and in 37% of cases (14 of 38), a different PD-L1 classification would have been made depending on which assay and cutoff combination was used. Extending the comparison to 81 additional lung cancer cases, phase 2 of the project validated the results of phase 1, with the 22C3, 28-8, and SP263 assays again demonstrating highly comparable staining results and the SP-142 assay demonstrating a lower proportion of tumor cell staining.³⁸  Additionally, the 73-10 assay was found to have a higher PD-L1 analytic sensitivity versus the other assays. Intrareader reproducibility was high for tumor cell PD-L1 scoring (intraclass correlation coefficient, 0.86–0.93) but low (intraclass correlation coefficient, 0.18–0.19) for immune cell scoring. The study also found a high concordance between PD-L1 scoring using glass slides versus digital images (Pearson correlation >0.96) and good agreement in assessing PD-L1 in cytology cell block material (intraclass correlation coefficient, 0.78–0.85).

Rimm et al³⁹  conducted a prospective, multi-institutional, pathologist-based assessment of 28-8, 22c3, SP142, and E1L3N PD-L1 IHC assays (the SP263 PD-L1 assay was not included in this comparison). The study, which was sponsored by the National Comprehensive Cancer Network and funded by Bristol-Myers Squibb, found that the 22c3-based assay showed a slightly but statistically significantly lower staining level than the 28-8 and E1L3N assays, but the authors ultimately concluded that the 28-8, 22c3, and E1L3N assays are “essentially equivalent.” Consistent with the Blueprint project findings, the SP142 assay was associated with a significantly lower mean PD-L1 expression score in tumor cells. The study also found that interreader reproducibility across assays was excellent for tumor cell PD-L1 expression, but poor for immune cell expression.

Recently, Munari et al⁴⁰  conducted an interclone comparison of the 22c3 and SP263 PD-L1 assays on 198 NSCLC cases, concluding that the 2 assays are not interchangeable, with statistically significant differences seen at both the 1% and 50% cutoffs. At the 50% cutoff, approximately half of the cases positive for SP263 would have been interpreted as negative for 22c3. Although the authors were not able to perform a direct correlation with response to anti–PD-(L)1 therapy, they concluded that these differences could lead to “possible underestimation of patients suitable for pembrolizumab therapy.”

Although these analytic comparison studies are valuable, no study to date has examined cross-assay concordance against clinical outcome endpoint measures such as response rate, progression-free survival, or overall survival in a common PD-(L)1 therapeutic trial. Assessment and validation of multiple PD-L1 assays and cutoffs across multiple PD-L1 inhibitors would allow a determination of true clinical sensitivity and specificity and therefore represents a priority in the field of cancer immunotherapy.

PD-L1 heterogeneity, both intratumoral⁴¹  and intertumoral,^42,43  has been described and represents a challenge for reliably assessing PD-L1 in the diagnostic setting. Ilie et al⁴⁴  performed a comparative study of the PD-L1 status between surgically resected specimens and matched biopsies of NSCLC patients. They found a discordance rate of approximately 50%, with the majority of the discrepancies showing PD-L1–negative biopsies and PD-L1–positive surgical resections from the same patient (75% of the discordances were due to differences in immune cell scoring). Therefore, PD-L1–negative tumors that respond to immune checkpoint inhibitors may represent heterogeneous and/or temporal expression that was not detected, or potentially not present, at the time of initial specimen testing.⁴⁵  Radiation therapy, chemotherapy, and small-molecule kinase inhibitors have all been shown to induce PD-L1 expression^46–48  and therefore could represent biological causes of pretreatment versus posttreatment PD-L1 testing discrepancy.

The DNA MMR system recognizes and corrects insertion, deletion, and base pair mismatches that occur during DNA replication.⁴⁹  Deficiencies in MMR are primarily caused by inactivation of one or more of the 4 main proteins: MLH1, MSH2, PMS2, and MSH6. Mismatch repair deficiency (dMMR) can be inherited or acquired (sporadic) and leads to accumulation of 2 main types of mutations in the DNA: missense mutations throughout the genome and changes in the length of microsatellite regions.^50,51 

Although inherited MMR defects can be caused by inactivating mutations of any of the 4 MMR genes, MLH1 and MSH2 mutations account for approximately 90% of cases, with mutations in PMS2 and MSH6 accounting for most of the remaining cases.^52,53  Sporadic MMR defects are usually caused by epigenetic silencing of MLH1 via promoter methylation and are frequently associated with BRAF V600E mutations.^54,55 

Mismatch repair deficiency was first detected in colorectal cancer (CRC) but can occur in many other tumor types^56,57  (Figure 2), with a prevalence of 4% across all adult solid malignancies.⁵⁸  Tumors with a significant frequency of dMMR include endometrial, gastric, small intestinal, colorectal, cervical, prostate, bile duct, liver, and thyroid carcinomas; neuroendocrine tumors; and uterine sarcomas.⁵⁸  In one study, endometrial, gastric, and small intestinal cancers were more likely to have MMR defects than colon cancer.⁵⁹ 

Inherited deficiencies of MMR are the cause of hereditary nonpolyposis colon cancer, also known as Lynch syndrome, a genetic syndrome with a predisposition to cancer, especially CRC and endometrial carcinoma (52%–82% and 25%–60% lifetime risk, respectively). Approximately 8%–16% of dMMR CRC cases are associated with Lynch syndrome, with the remaining cases representing sporadic CRC.^60–63  Identifying cases of Lynch syndrome is critical given the increased cancer risk not only for the patient (ie, for other tumors), but also for the patient's family members. Therefore, it is recommended that all cases of CRC and endometrial carcinoma undergo screening for MMR defects.^64–66  If dMMR is detected, further testing should be performed to determine if the defect is inherited or sporadic after the appropriate genetic counseling and consent is obtained.

Microsatellites, also known as short tandem repeats or simple sequence repeats,^67,68  are short repetitive DNA sequences that occur commonly throughout the genome, primarily within noncoding intergenic regions and introns. During DNA replication, transient dissociation and subsequent misaligned reassociation of the replicating DNA strands can cause alterations in the length of these microsatellites. These alterations are usually corrected by the MMR system, but in the setting of dMMR can remain uncorrected, resulting in MSI.⁶⁹  Therefore, MSI is a genomic consequence of dMMR, with the severity of instability stratified by the number of affected microsatellite markers into 3 groups: MSI high (MSI-H), MSI low, and microsatellite stable (see MMR and MSI Testing Methods below).⁶⁹ 

Testing for dMMR and MSI has become a mandatory component of identifying patients most likely to respond to immunotherapy targeting PD-(L)1 in certain indications. The first clinical observation suggesting that MMR status predicted clinical response to anti–PD-1 blockade came from a phase 2 study of the PD-1 checkpoint inhibitor pembrolizumab.⁷⁰  Le et al⁷⁰  hypothesized that the efficacy rate of PD-1 blockade in CRC was associated with MMR status and initiated a phase 2 clinical trial to evaluate pembrolizumab in patients whose tumors were dMMR or MMR proficient (pMMR). The results of this study showed a dramatic difference in the objective response rate (ORR) of dMMR versus pMMR patients, with a 40% ORR in the dMMR CRC patients versus 0% ORR in the pMMR CRC patients. Additionally, dMMR non-CRC patients, including those with ampullary, endometrial, small bowel, and gastric cancers, also experienced a high rate of benefit with a 71% ORR.

Currently, 2 cancer immunotherapy drugs are FDA approved for dMMR tumors based on IHC or MSI polymerase chain reaction (PCR) testing. In May 2017, the FDA granted accelerated approval to pembrolizumab for treatment of unresectable or metastatic dMMR or MSI-H solid tumors that have progressed following prior treatment, the first cancer site–agnostic approval granted from the FDA.⁷¹  The data for pembrolizumab were based on the results from 5 single-arm multicenter trials (KEYNOTE-016, -164, -012, -028, and -158) of 149 total patients (90 CRC patients, 59 with 14 other cancer types) showing an ORR of 40%, which was similar irrespective of tumor type.⁷¹  In these studies, determination of MSI-H or dMMR status for the majority of patients was prospectively performed via PCR for MSI-H status or IHC for dMMR. The second FDA approval related to MMR deficiency occurred when nivolumab was approved in July 2017 for CRC that has progressed following therapy, based on the CheckMate 142 study of 53 CRC patients showing an ORR of 32% in the overall population.^72,73 

The association between dMMR response to cancer immunotherapy appears to be driven by the probabilistic creation of nonself, cancer-specific antigens termed neoantigens (see the Cancer Neoantigens section). These neoantigens, in the presence of a cancer immunotherapy–enabled activated T-cell immune response, are sufficient to elicit antitumor immunity. Mismatch repair deficiency is simply one mechanism that can create increased tumor mutations, thereby increasing the probability of a neoantigen that can be processed and recognized by the adaptive immune system. For example, dMMR tumors harbor a 10- to 100-fold higher rate of mutations compared with pMMR tumors.^74,75  High tumor mutational burden (TMB) (see the TMB section) can be created via other defects in the DNA editing machinery, such as loss of polymerase exonuclease function via POLE and POLD1 mutations (see the POLE and POLD1 Mutations section).

There are 2 main methods of screening for MMR functional defects: IHC for the 4 MMR proteins MLH1, MSH2, PMS2, and MSH6,^76,77  and molecular PCR testing to detect MSI.^78–80  Each method has unique limitations and diagnostic pitfalls, with IHC being dependent on fixation conditions and accurate interpretation and MSI requiring normal tissue comparison and sometimes microdissection.⁸¹  Analytically, these methods show comparable performance with an approximately 5% to 10% false-negative rate each.⁸²  Mismatch repair IHC results can be falsely negative in the setting of mutations (eg, in MLH1) that result in an antigenically intact but functionally inactive protein, and MSI can be falsely negative in the setting of intratumoral heterogeneity and inadequate microdissection.⁸³  The choice of which method to use at a given institution usually depends on factors such as cost, availability, and turnaround time. Although using both tests will detect slightly more cases than either test alone, the benefit is small. In the setting of Lynch syndrome screening, a dMMR or MSI result is typically followed by sequencing of specific MMR genes, except in the case of MLH1 loss, when BRAF V600E and/or promoter methylation testing can be used to confirm somatic MMR deficiency.⁸⁴ 

Although IHC-based assessment of MMR status has been available for many years in the context of Lynch syndrome screening, it is unclear what, if any, changes to the analytic and interpretative aspects of testing will be required for the application to cancer immunotherapy. Multiple IHC antibody clones for each of the 4 MMR proteins are available, primarily as class 1 in vitro diagnostics, with 1 commercial MMR panel having been FDA cleared for CRC as an aid in the identification of probable Lynch syndrome.^85,86  There are currently no FDA-approved MMR/dMMR or MSI assays as companion diagnostics for cancer immunotherapy, despite the fact that the FDA has approved 2 cancer immunotherapies for use in dMMR/MSI-H patient subsets.

The IHC determination of dMMR is based on identifying loss of one or more MMR proteins, indicated by absent nuclear staining in viable tumor nuclei in the presence of unequivocal internal control cell staining. Although MMR IHC staining patterns are usually consistent with the biological function of the 4 MMR proteins, pathologists should be aware of unusual staining patterns and diagnostic pitfalls.^85,87  Based on data in CRC, MMR IHC assessment in biopsy samples may be preferable to surgical resection specimens given that MMR protein staining loss has been observed posttreatment.^88–90  Although detection of BRAF V600E via IHC can play a role in differentiating sporadic CRC from Lynch syndrome,⁹¹  it does not have a role in the identification of dMMR cancers potentially responsive to cancer immunotherapy.

Microsatellite instability is traditionally assessed via PCR analysis of specific microsatellite loci to determine if a change in length due to insertion or deletion of DNA repeats has occurred in the tumor as compared with normal tissue. If the same number of DNA repeats is present in the tumor and normal tissue, the result is considered microsatellite stability. In contrast, if the number of DNA repeats differs between tumor and normal, the result is considered MSI. In 1998 the National Cancer Institute established international criteria for MSI determination in CRC based on 5 microsatellite markers, 2 mononucleotide markers (BAT25 and BAT26), and 3 dinucleotide markers (D2S123, D5S346, and D17S250).⁹²  Subsequently, an alternative panel containing 5 mononucleotide quasi-monomorphic markers was proposed that demonstrated improved sensitivity (95.8% versus 76.5%) and positive predictive value (88.5% versus 65.0%) compared with the original National Cancer Institute panel.⁹³  Using either panel, a tumor is classified as MSI-H if 30% or more of the repeats are unstable, MSI low if less than 30% of repeats are unstable, and microsatellite stable if no repeats are unstable.⁹⁴ 

Microsatellite instability can also be assessed via next-generation sequencing (NGS), and a limited number of studies have shown good sensitivity and specificity using NGS to assess MSI compared with traditional PCR methods.^95–97  The NGS assay must be designed to include microsatellite regions, use bioinformatics algorithms that can size these regions, and take into account the size distribution for microsatellite regions in normal samples due to polymerase slippage in vitro. Sensitivity of 96.4% to 100% and specificity of 97.2% to 100% have been described for both capture-based and PCR-based library preparation.^98,99  Future cancer immunotherapy trials examining response and survival rates in these discrepant cases will determine if MSI-NGS provides any advantages over the traditional 5-marker PCR panel. A clear advantage associated with MSI-NGS is the ability to couple the analysis with broader genomic analyses without adding an additional test or consuming additional tumor material.

Tumor mutational burden is a measure of somatic cancer mutation prevalence typically represented as the number of mutations per megabase. Tumor mutational burden varies significantly across and within human cancer types, with melanoma, lung cancer, and bladder cancer having the highest mutation prevalence (Figure 3).¹⁰⁰  Although TMB is related to dMMR and MSI, there is not complete overlap between them; most dMMR/MSI-H tumors have high TMB, but not all high-TMB tumors are dMMR/MSI-H.⁹⁹  For these reasons, TMB should be considered an independent parameter and should not be used interchangeably with dMMR or MSI.

The early realization that cancer immunotherapy response rates were highest among tumor types with the highest tumor mutation density led to several clinical outcome studies to further investigate this correlation. The seminal observation in 2014 by Snyder et al¹⁰¹  was that high mutational load in melanoma correlated with sustained clinical benefit from ipilimumab or tremelimumab-based CTLA-4 inhibition. Importantly, the authors noted that this correlation was imperfect, with some high-TMB patients not benefiting from therapy. Several subsequent studies, including those of Rizvi et al¹⁰²  and Hellmann et al,^103,104  demonstrated a similar association between TMB and response to immunotherapy in lung cancer. The specific immunotherapies, disease indications, and conclusions from these studies are listed in Table 4.

Beyond melanoma and lung cancer, mutational load has also been shown to be a predictor of response to checkpoint blockade in a broad array of tumor types, suggesting that high TMB could be used to identify optimally responsive patient subsets across all cancers. Rosenberg et al¹⁰⁵  found mutational load to be an independent predictor of response in metastatic urothelial carcinoma patients treated with atezolizumab. In a retrospective analysis, Goodman et al¹⁰⁶  also found that TMB was an independent predictor of outcome in a study of 151 patients, multiple immunotherapies, and more than 20 diverse tumor types including breast, cervical, ovarian, prostate, and renal cancers.

Evidence to date indicates that PD-L1 status and TMB are independent biomarkers with largely additive ability to predict benefit from immune checkpoint inhibition. Carbone et al¹⁰⁷  did not observe an association between PD-L1 expression and TMB in an exploratory analysis within the CheckMate 026 phase 3 study of nivolumab as first-line treatment of advanced NSCLC patients, finding a higher response rate in those patients with a high level of both TMB and PD-L1 expression (75%) compared with patients with a high level of either biomarker alone (32% for TMB, 34% for PD-L1) or patients with low levels of both biomarkers (16%). Hellmann et al^103,104  similarly found TMB and PD-L1 expression to be independent biomarkers in the CheckMate 227 trial¹⁰³  and the CheckMate 012 trial,¹⁰⁴  where NSCLC patients with positive PD-L1 expression (≥1%) and high TMB (>median) had significantly improved response and progression-free survival rates with combination nivolumab plus ipilimumab therapy compared with patients whose tumors had only one or neither variable.

Several challenges need to be overcome before TMB can be readily adopted into the routine clinical environment, the most critical being testing access, technical/analytic burden, and lack of standardized methods for TMB calculation and reporting.

Most published studies demonstrating a correlation between high TMB and clinical response to cancer immunotherapies used whole-exome sequencing (WES), a methodology not widely available or practical because of its high cost and analytic burden. Targeted NGS is becoming increasingly accessible across laboratory testing environments, and, when coupled with optimized bioinformatics pipelines, can provide an accurate and practical method for detecting the sequence changes used to calculate TMB. For example, NGS-based comprehensive gene panels (CGPs) sequence specific sets of genes (typically 300–600) related to a focused disease or phenotype. These panels represent a more routinely feasible option for assessing TMB in the routine laboratory environment because of their lower cost, faster turnaround time, and lower analytic burden compared with WES. Several studies have found NGS-based TMB assessment to be an accurate surrogate of TMB by WES. Rizvi et al¹⁰⁸  studied TMB assessment by targeted NGS versus WES, finding a strong correlation between the 2 methods. Chalmers et al¹⁰⁹  found that TMB from a 1.1-megabase targeted CGP was strongly correlated (R² = 0.74) with mutation burden determined by WES. Examining the effect of various genomic parameters on quality, they found that filtering out germline alterations and rare variants was important in determining accurate TMB measurements.

The ability of NGS-based methods to accurately assess TMB is dependent on the size of the target region sequenced. Several groups have attempted to define the minimum number of genes that can be used as the basis for TMB, while remaining predictive of response to immunotherapy. Campesato et al¹¹⁰  assessed the ability of their institutional CGP (641 genes) and the Foundation Medicine CGP (315 genes) to estimate TMB, comparing the clinical outcome prediction performance with WES-based TMB from the melanoma and NSCLC checkpoint blockade trials from Snyder et al¹⁰¹  and Rizvi et al,¹⁰²  respectively. Tumor mutational burden estimates from either of the CGPs were significantly associated with durable clinical benefit in NSCLC patients treated with PD-1 blockade. Predictive accuracy of CGP-TMB for durable clinical benefit was not statistically different from that of TMB from WES, with similar area under the curve, sensitivity, and specificity between the 2 CGPs and WES. Importantly, TMB predictive accuracy was lost when smaller CGPs (<150 cancer genes) were examined, causing the authors to recommend CGPs with more than 300 cancer genes for TMB estimation. The inability of smaller gene panels to accurately estimate TMB is likely due to the sequenced region, typically in the range of 25 kilobases, representing too small a portion of the genome to accurately represent overall mutational burden.

Another challenge preventing widespread adoption of TMB as a biomarker is the lack of standardization of TMB calculation and reporting. For example, there is no standard definition of high TMB, with some studies using various absolute mutation thresholds¹⁰²  and others using ratios of mutations per megabase of DNA sequenced.¹¹¹  The Friends of Cancer Research, partnering with the National Cancer Institute, the FDA, Johns Hopkins University, Memorial Sloan Kettering Cancer Center, and multiple pharmaceutical and diagnostic companies, hosted a TMB harmonization working group meeting in May 2018 to begin the process of addressing these challenges.¹¹²  Several opportunities for TMB technical harmonization were provided, such as agreement on analytic parameters for TMB calculation, generation of a universal reference standard as an alignment tool, and agreement on statistical approaches that will lead to consistent TMB calculation and clinical interpretation.

Tumor mutational burden testing currently requires access to tumor tissue, but noninvasive blood-based means of estimating TMB are being explored. Gandara et al¹¹³  developed a novel blood-based assay to measure TMB based on single-nucleotide variants from 394 genes from cell-free DNA in patient plasma. Using blood samples from the atezolizumab POPLAR (211 patients) and OAK (583 patients) trials in second-line NSCLC, the blood-based TMB assay was predictive of clinical benefit independent of PD-L1 expression. Prospective validation of this blood-based assay as an alternative to tissue-based TMB would improve the routine implementation of mutational burden assessment in clinical practice.

Polymerase δ (POLD1 gene) and polymerase ɛ (POLE gene) are DNA polymerases involved in DNA replication during the S phase of the cell cycle.¹¹⁴  These enzymes additionally have DNA proofreading and repair functionality via an exonuclease domain that allows excision and replacement of incorrect bases and continuation of DNA replication. Mutations of the exonuclease domain disrupt the proofreading function and lead to an accumulation of mutations elsewhere in the genome, often referred to the ultramutator phenotype.¹¹⁴ 

POLE mutations associated with the ultramutator phenotype have been described in many different cancer types, including about 3% of CRCs and 7% to 10% of endometrial cancers.^115–118  These tumors have thousands of mutations across the genome, an inflamed tumor microenvironment, and upregulated PD-L1, and are associated with a better prognosis, similar to dMMR tumors. POLE mutations are predominantly found in microsatellite-stable/pMMR tumors, but one study identified them in MSI cases and found that they account for some unexplained Lynch syndrome phenotypes.¹¹⁹  POLD1 mutations appear to be less prevalent than POLE mutations and occur more commonly in dMMR tumors, but can also lead to the ultramutator phenotype.^120,121 

Given the high mutation rate, inflamed tumor microenvironment, and upregulation of immune checkpoints associated with POLE-mutated cancers, there is strong scientific rationale to assess the efficacy of cancer immunotherapy in this setting. If effective, it would bring clinical benefit to 1% to 3% of colon cancer patients not currently addressed by cancer immunotherapy. Targeted NGS or allele-specific PCR testing of POLE mutations could focus on the 3 recurrent mutations (P286R/H/S, V411L, and S459F) that represent 90% of the identified exonuclease domain mutations.¹²²  In addition to POLD1/POLE and MMR, there is a strong rationale to evaluate deficiencies in other DNA repair enzymes and mechanisms, such as O⁶-methylguanine-DNA methyltransferase (MGMT), homologous recombination, nucleotide excision repair, and base excision repair, as potential biomarkers of immunotherapy response. Several translational and clinical studies exploring these DNA damage repair pathways in this context are currently underway.¹²³ 

The correlation of dMMR, MSI-H, TMB, and clinical benefit from immune checkpoint blockade is thought to represent a probabilistic relationship in which higher mutational burden within a tumor increases the likelihood of a neoantigen to which an effective T-cell immune response can be generated.¹²⁴  In this way, dMMR, MSI-H, and TMB can all be thought of as surrogate measures of immunogenic neoantigens, which may be the actual targets of cytotoxic T cells in clinical cases of immunotherapy response. In both melanoma and NSCLC patients, higher neoantigen burden is associated with improved clinical responses to immune checkpoint blockade therapy.¹²⁵ 

Based on these hypotheses and early observations, there is great interest in using tumor sequencing and bioinformatic prediction algorithms to directly identify neoantigens that can be used therapeutically in cancer vaccines¹²⁶  and diagnostically as more direct predictive biomarkers of cancer immunotherapy efficacy. Experimental data exploring the frequency, identity, and uniqueness of cancer neoantigens have revealed new information that will guide the clinical translation of this potentially important immunotherapy biomarker.

Although human T cells have been shown to react to both major histocompatibility complex class I–restricted and major histocompatibility complex class II–restricted neoantigens in a large fraction of malignancies,^127–130  only a small percentage of tumor mutations actually lead to the formation of neoantigens that are recognized by host T cells.^{126,131–133}  Additionally, it appears that immunoreactive cancer neoantigens can arise from mutations in any gene across the genome, and are not more common in driver oncogenes related to tumorigenesis.^126,129,134  Steven Rosenberg, MD (National Cancer Institute, Bethesda, Maryland), pioneered a tandem minigene approach to identify the precise neoantigens recognized by T cells from patients treated with adoptive cell transfer.^135,136  In melanoma, neoantigens were identified in 29 of 31 patients (94%), with each of the identified neoantigens being unique to the autologous patient.¹³⁵  In common gastrointestinal cancers, neoantigen-reactive T cells were identified in 62 of 75 patients (83%), with 99% of the neoantigenic determinants unique to each autologous patient.¹³⁶  Immunogenic tumor antigens associated with adoptive cell transfer complete regressions have been identified in metastatic breast cancer¹³⁷  and human papillomavirus–associated metastatic cervical cancer.¹³⁸ 

Although this area is promising, several challenges would need to be overcome before neoantigen prediction or neoantigen load could become part of routine cancer immunotherapy testing. These include lack of a complete understanding of the underlying immunobiology, the low frequency and apparent uniqueness of cancer neoantigens, and a lack of suitable bioinformatics and laboratory methods that can both provide accurate prediction and be implemented into routine practice.

Pathologists have long recognized the presence of TILs in diagnostic specimens, and many studies have explored their role as biomarkers. In 2011, Mahmoud et al¹³⁹  was the first to identify TILs as a prognostic factor in breast cancer, showing that tumor-infiltrating CD8⁺ lymphocytic density is significantly associated with improved clinical outcome independent of standard factors such as tumor grade, lymph node stage, and HER2 status. In a study of stage I to IV CRCs, Mlecnik et al¹⁴⁰  found that an assessment of CD8⁺ T cells in the tumor center and invasive margin was a novel indicator of recurrence beyond TNM staging. Rakaee et al¹⁴¹  developed and validated a hematoxylin-eosin–based TIL scoring model, demonstrating that high stromal TIL level is an independent measure of prognosis in stage I to III NSCLC patients. Smyrk et al¹⁴²  proposed TILs as a screening criterion for selecting which CRC samples require MSI testing, demonstrating a 93% sensitivity and 62% specificity for MSI-H status with a TIL cutoff of 5. Assuming a 12% MSI-H rate, they estimated that TIL assessment could reduce the number of colon cancer samples referred to MSI testing by more than 50% while still identifying 93% of MSI-H cases.

Tumor-infiltrating lymphocytes indicate an inflamed tumor microenvironment, and in the era of cancer immunotherapy have been assessed as a predictive biomarker of response. Tumeh et al¹⁹  found a significantly higher CD8⁺ T-cell density in the baseline biopsies of melanoma patients responding to pembrolizumab compared with the progression group, both at the invasive margin and at the tumor center. However, there was no CD8-positive cutoff value clearly separating responders from nonresponders. Chen et al¹⁴³  also found an increase in the density of CD⁸⁺, CD3⁺, and CD45RO⁺ T cells in pretreatment melanoma samples of CTLA-4 blockade responders compared with nonresponders, but, similar to Tumeh et al,¹⁹  could not establish a clear separation between the 2 groups. However, their study also included a separate analysis of anti–PD-1 treated patients, in which on-treatment biopsies showed highly statistically significant, largely nonoverlapping expression differences between responders and nonresponders for CD8, CD4, and CD3 T-cell subsets in addition to the immune checkpoint markers PD-L1, PD-1, and LAG3.¹⁴³  These changes were seen as early as after 2 or 3 doses in responding patients. Also observed in the early on-treatment biopsies was an enrichment of CD8⁺ T cells in the tumor center versus the tumor margin in responders compared with nonresponders, suggesting therapy-induced tumor infiltration.

Beyond the assessment of TILs by hematoxylin-eosin morphology and single-marker IHC, the development of advanced in situ multiplexing methods has enabled more comprehensive immunophenotyping of tumors and the host microenvironment, including the ability to visualize the colocalization of multiple immune-related markers.¹⁴⁴  A detailed overview of the currently available IHC multiplexing methods and technologies is beyond the scope of this article, but can be found in recently published reviews.^145,146  The historic challenge of creating a multiplexed indirect IHC assay that avoids multiple species of primary antibodies and cross-reactivity between same-species primary antibodies has largely been solved by elegant chemical and heat deactivation reactions.¹⁴⁵  Chromogenic multiplexing allows light microscopy visualization, with novel methods such as multiplexed immunohistochemical consecutive staining on a single slide and reagents such as narrow-band tyramine-conjugated dyes demonstrating the feasibility of medium- to high-dimensional analysis.^147,148  Fluorescence technologies enable even higher-order multiplexing, with up to 61 individual analytes demonstrated with an iterative staining, imaging, and chemical inactivation cycle approach.¹⁴⁹  The demonstration of fully automated fluorescence multiplexing using tyramide-based covalent detection methods offers the promise of these methods entering routine anatomic pathology practice.^150,151 

Although initial studies using these novel techniques focused on the characterization of specific intratumoral immune subsets,^{147,152–154}  some studies have begun to correlate the phenotype and cellular interactions of immune cells in the tumor microenvironment with immunotherapy outcome data as a potentially more powerful predictor of response. Giraldo et al¹⁵⁵  found in the setting of Merkel cell carcinoma that the number of pretreatment PD-1⁺ immune cells in proximity to (within 15 μm of) a PD-L1⁺ immune or tumor cell was associated with pembrolizumab response, which was not the case with CD8/PD-L1 proximity. Similarly, Johnson et al¹⁵⁶  found that a PD-1/PD-L1 interaction score, but not PD-L1 alone, was associated with anti–PD-1 response in metastatic melanoma.

Taken together, these studies demonstrate the potential of TILs as an important cancer immunotherapy biomarker. However, adoption of TIL assessment in any of its forms into routine clinical practice will require robust methods and tools for interpretation. Addressing the need for a standardized, reproducible approach to the histopathologic assessment and quantification of TILs in routine practice, Hendry et al,^157,158  on behalf of the International Immuno-oncology Biomarkers Working Group, have proposed a hematoxylin-eosin–based methodology for assessing TILs in solid tumors for clinical validity and utility assessment. In addition, the International TILs Working Group has published methodologic recommendations on the evaluation of TILs in breast cancer, focusing on the stromal (versus intratumoral) compartment.¹⁵⁹  Beyond manual human interpretation of TILs, computational image analysis–based TIL scoring approaches have been developed, including traditional object-oriented image-segmentation methods and more advanced machine learning–based classification algorithms that rely on extensive training sets but are able to morphologically identify cellular subsets.¹⁶⁰ 

RNA-based gene expression studies have been used to identify specific transcriptional patterns and signatures in tumors and the tumor microenvironment that may help elucidate details of immune physiology underlying immunotherapy benefit. Summarized here are selected studies exploring gene expression signatures that appear to play an important role in the prediction of cancer immunotherapy response.

Taube et al^161,162  found that immune infiltrates at the interface of PD-L1–positive melanoma cells demonstrate a CD8⁺ T cell–T helper 1 cytokine mRNA signature characterized by IFN-γ expression, which was absent in PD-L1–negative melanomas. They hypothesized that secretion of the IFN-γ cytokine not only triggers antitumor effects, but simultaneously induces PD-L1 as part of an adaptive negative-feedback immune regulatory mechanism. Since this seminal discovery, the adaptive expression of PD-L1 has been observed in other tumor types, including Merkel cell carcinoma,¹⁶³  NSCLC,¹⁶⁴  and breast cancer.^165,166  Activation of the IFN-γ signaling pathway is now recognized as a necessary component of successful anticancer T-cell immunity, an association reinforced by the discovery of immunotherapy resistance mutations in Janus kinase 1 (JAK1) and Janus kinase 2 (JAK2) (see the Biomarkers of Resistance to Cancer Immunotherapy section) that prevent upregulation of IFN-γ target genes.¹⁶⁷ 

Ribas et al¹⁶⁸  developed 2 gene signatures, IFN-γ 10-gene and expanded-immune 28-gene, both of which were statistically associated with overall response rate and progression-free survival in the context of melanoma treated with pembrolizumab. In NSCLC, an 8-gene T-effector and interferon-γ gene signature was associated with improved overall survival with atezolizumab in the POPLAR phase 2 trial.¹⁶⁹  This result supports the hypothesis that preexisting immunity is a predictor of benefit from checkpoint blockade.

In addition to having a possible role as a clinical biomarker, gene expression signatures drive much of the ongoing translational research into novel mechanisms of cancer immunotherapy response and resistance that could inform rational combination strategies.

Although cancer immunotherapy delivers unprecedented levels of durable clinical benefit to select patient subsets, most patients initially treated with these therapies do not respond (primary resistance) and of those who do respond, some experience only short-lived benefit with eventual tumor relapse (acquired or secondary resistance). This section will briefly summarize some of the latest translational research into the mechanisms of immunotherapy resistance.

One type of cancer immunotherapy resistance is immune evasion resulting from tumor mutations that cause failures in immune-mediated destruction and detection of cancer cells. For example, Zaretsky et al¹⁶⁷  identified JAK1 and JAK2 mutations in 2 melanoma patients who initially responded to pembrolizumab anti–PD-1 therapy but later progressed. Mutations in JAK1/2 lead to abrogation of IFN-γ signaling and its antiproliferative effect on tumor cells. Shin et al¹⁷⁰  identified JAK 1/2 mutations as the cause of primary resistance to anti–PD-1 therapy in dMMR colon cancer and melanoma patients, demonstrating that these mutations can cause both primary and secondary forms of immunotherapy resistance.

Mutations causing impairment of the HLA class 1 antigen processing and presentation machinery appears to be another mechanism of resistance to cancer immunotherapy agents. Gettinger et al¹⁷¹  identified homozygous loss of β₂-microglobulin (B2M), critical to proper functioning of the HLA class 1 complex, in 1 of 14 lung cancer patients with acquired resistance to immune checkpoint inhibition. Zaretsky et al¹⁶⁷  also identified a single patient with B2M loss in their study of melanoma patients who progressed after initially responding to immune checkpoint therapy.

Defects in several traditional cell signaling pathways have also been correlated with immunotherapy resistance, suggesting that many of these pathways can impact immune regulation. Some pathways and genomic changes that have been associated with negative immune regulation include β-catenin activation (Spranger et al¹⁷² ), PTEN loss (Peng et al¹⁷³ ), and EGFR mutations and ALK rearrangements (Gainor et al¹⁷⁴ ). The apparent connectivity between these pathways and immune activation/suppression provides a strong rationale for combining targeted therapies and immunotherapies.

Melanomas demonstrating primary resistance to checkpoint inhibitors were found to display a specific transcriptional signature referred to as innate anti–PD-1 resistance.¹⁷⁵  Hugo et al¹⁷⁵  found that this transcriptional signature is characterized by increased expression of genes involved in the regulation of extracellular matrix remodeling, angiogenesis, wound healing, cell adhesion, and mesenchymal transition. Activation of the innate anti–PD-1 resistance signature was also demonstrated in lung adenocarcinoma, colonic adenocarcinoma, renal clear cell carcinoma, and pancreatic adenocarcinoma.¹⁷⁵ 

Acknowledging that the outcome of cancer-immune interactions is based on multiple independent parameters that can each vary between patients, Blank et al¹⁷⁶  have proposed a “cancer immunogram” visualization framework as a more comprehensive assessment of immune responsiveness and mechanisms of resistance. The immunogram contains 7 initial parameter classes: tumor foreignness, general immune status, immune cell infiltration, absence of checkpoints, absence of soluble inhibitors, absence of inhibitory tumor metabolism, and tumor sensitivity to immune effectors. Resistance to immunotherapy is predicted when any of the 7 parameters is unfavorable, and only overcome with the appropriate targeting of the relevant defect. Should this immunogram or a similar tool prove to be valuable in the selection of responsive patients, its clinical adoption will require the multifactorial integration of tumor genomic, immunohistochemical, and peripheral blood data by anatomic and molecular pathologists.

The human microbiome consists of the trillions of microorganisms that colonize the skin, mucosal surfaces, and gastrointestinal tract. The role of the microbiome and its associated impact on human health and response to cancer therapy has been proposed and explored in the past, but until recently has been controversial. Recent conclusive data supporting the role of the microbiome as a biomarker of response to cancer immunotherapy have reignited scientific interest in this field and triggered significant clinical validation efforts.

Routy et al¹⁷⁷  established a correlation between gut microbiota and response to checkpoint inhibition, finding a significant shortening of OS and progression-free survival in NSCLC and urothelial carcinoma patients treated with broad-spectrum antibiotics prior to anti–PD1/PD-L1 therapy. Additionally, the authors found that gut microbiota diversity, as measured by shotgun-sequencing gene count or metagenomic species levels, correlated with clinical response. They found enrichment of specific bacterial genera in responders versus nonresponders, with Akkermansia muciniphila having the most significant association with favorable clinical outcome. They concluded that the gut microbiome markedly influences PD-L1 blockade outcome and hypothesized that certain bacteria such as A muciniphila may reinforce intestinal barrier integrity, reduce systemic inflammation, and improve immunosurveillance.

Matson et al¹⁷⁸  used 16S ribosomal RNA gene amplicon sequencing and species-specific quantitative PCR to study stool samples from 42 metastatic melanoma patients receiving immune checkpoint therapy, finding 8 bacterial species more abundant in responders and 2 species more abundant in nonresponders. Supporting the findings of Routy et al,¹⁷⁷  they identified A muciniphila in 4 patients, all of whom were responders. They propose that multiple specific bacteria may contribute to improved antitumor immunity, and that the optimal biomarker may be a ratio of beneficial bacteria that activate the immune system to nonbeneficial bacteria that negatively regulate immune responses. The authors conclude that the commensal microbiota may be a useful biomarker to predict response to checkpoint blockade therapy, with patient responder-associated bacteria having distinct local and systemic effects on innate and adaptive immunity.

Gopalakrishnan et al¹⁷⁹  prospectively collected microbiome samples from metastatic melanoma patients prior to anti–PD-1 therapy. Within-sample gut microbiome diversity was found to be significantly higher in responders versus nonresponders, with greater diversity also correlating with prolonged progression-free survival. Specific compositional differences associated with anti–PD-1 therapy were identified, with Clostridiales order, Ruminococcaceae family, and Faecalibacterium genus enriched in responders and Bacteroidales order enriched in nonresponders. Exploring potential mechanisms underlying these associations, the authors found a statistically significant correlation between CD8⁺ T-cell infiltrates in the tumor and abundance of Faecalibacterium genus, Ruminococcaceae family, and Clostridiales order in the gut, suggesting enhanced systemic and antitumor responses mediated by increased antigen presentation and improved effector T-cell function. Finally, germ-free mice transplanted with stool from responders to anti–PD-1 therapy had higher CD8⁺ T-cell density and improved response to immune checkpoint blockade compared with those transplanted with stool from nonresponders, suggesting a causal link between a favorable gut microbiome and immune checkpoint therapy.

Collectively, these studies provide compelling evidence that the diversity and composition of the gut microbiome could serve as a biomarker of cancer immunotherapy response. Although additional validation via clinical trials is necessary, the data also suggest that modulation of the gut microbiome may have therapeutic potential for cancer patients receiving immune checkpoint blockade.

The field of cancer immunotherapy is still young and in evolution, but has already transformed clinical oncology by providing durable long-term survival benefit to previously untreatable cancer patients. Because our ability to accurately identify the minority of patients who currently derive benefit from these therapies remains very limited, there is a critical need to research, develop, and translate biomarkers predictive of response.

From the landscape of cancer immunotherapy biomarkers reviewed in this article, only PD-L1, MMR, and MSI testing can be considered validated and routine because each is associated with FDA-approved therapies. Tumor mutational burden remains emerging, with issues of sensitivity/specificity and standardization yet to be resolved. Polymerase δ and ɛ mutations are rare events but may help clarify which pMMR patients derive benefit. Assessment of immunogenic cancer neoantigens appears to be critical to understanding the extent of tumor foreignness, but their detection and identification are limited by current bioinformatic and genomic technology. Tumor-infiltrating lymphocyte assessment, especially enabled by multiplexed characterization of immune cell subsets, is a promising emerging biomarker that sits squarely in the domain of the anatomic pathologist. Finally, transcriptional signatures of immune responsiveness, biomarkers of cancer immunotherapy resistance, and the microbiome are all promising early emerging biomarkers of immunotherapy response that require additional scientific, analytic, and clinical validation.

The biomarkers reviewed here represent only a small proportion of the genomic, proteomic, and immunologic parameters being explored in this dynamic and evolving field. The breadth of exploration, in both the therapeutic and diagnostic realms, is an indicator of our still nascent understanding of the mechanisms underlying immune-mediated cancer destruction. Adherence to the translational medicine philosophy, whereby laboratory research leads to novel therapeutic and diagnostic hypotheses that are then clinically tested to inform subsequent scientific investigations, will continue to drive cancer immunotherapy progress, ultimately leading to better outcomes for patients.

As the field of cancer immunotherapy continues to develop, pathologists bear an important responsibility to lead and support studies that explore and validate novel biomarkers. The current dynamism of this field represents an opportunity for pathology to establish itself as the immunotherapy biomarker center of excellence in clinical medicine, ensuring the accurate, standardized, and timely implementation of immunotherapy diagnostics through education, training, and collaboration.

The authors would like to gratefully acknowledge Molly Hansen, CT(ASCP), for her valuable assistance during the development of this manuscript.