The Evolving Role of PD-L1 Testing in Immunotherapy of Metastatic Melanoma and Non-Small Cell Lung Cancer

By Bill Bowman, J.D.

The Treatment of Cancer is Becoming More Precise

The movement toward precision medicine in oncology has accelerated over the past few years, driven by the emergence of multiple agents that target specific biomarkers across a wide array of malignancies, as well as the availability of companion diagnostics used in molecular testing. The standard of practice has evolved such that most oncologists routinely test their patients for the more established biomarkers in certain tumor types, e.g., EGFR mutations and EML4-ALK in metastatic non-small cell lung cancer (mNSCLC) and BRAF mutations in metastatic melanoma (mM).1

The Emergence of Immuno-Oncology

More recently, immunotherapy – unleashing the body’s immune system to attack malignant cells – has risen to the fore as the latest exciting development in the treatment of cancer. The introduction of the PD-1 checkpoint inhibitors Opdivo® (Bristol-Myers Squibb) and Keytruda® (Merck) over the past year has already had a major impact on the market dynamics for mM and mNSCLC. Both Opdivo and Keytruda work by blocking PD-1 on the surface of immune cells, disrupting its interaction with PD-L1 on cancer cells. This lifts a natural brake on the immune system (i.e., PD-1), allowing it to kill cancer cells.

To date, each of these brands has captured a large share of the mM market, particularly among pre-treated patients.2 Opdivo was just recently cleared for use in combination with Yervoy® for frontline treatment of mM. In the mNSCLC arena, Opdivo has leveraged its FDA approval for treatment among second-line or greater squamous cell patients to become the dominant player in this segment. Both PD-1 inhibitors gained full approval for second-line or greater NSCLC in October 2015 and early tracking data indicate that both brands (particularly Opdivo) have already made substantial inroads in the non-squamous market.3

Yet, the early commercial success of these PD-1 inhibitors has been accompanied by uncertainty and questions with respect to patient selection. 

Clinical Trial Data Based on PD-L1 Status

> PD-L1 Expression in mNSCLC

As illustrated in Table 1, data from the KEYNOTE-001 trial involving Keytruda in patients with advanced NSCLC demonstrated the highest overall response rate (ORR) of 45.2% among patients with high levels of PD-L1 (defined as >50% of tumor cells expressing PD-L1).4 Results from the CHECKMATE-057 trial of Opdivo vs. docetaxel in pre-treated patients with non-squamous cell NSCLC also demonstrated a benefit among patients expressing PD-L1, albeit at lower levels. On the other hand, in the KEYNOTE-001 study, clinical activity was observed in a smaller group of patients who were PD-L1-negative or who expressed only low levels of this protein; and in the CHECKMATE-017 and 063 trials, patients with previously treated squamous cell NSCLC derived a benefit from Opdivo regardless of PD-L1 status.5   


>PD-L1 Expression in mM

In mM, the impact of PD-L1 positivity is also not clear. In the KEYNOTE-001 mM trial, a cut-off of >1% was used to determine PD-L1 positive status and 71% of 125 evaluable patients who underwent testing met this threshold.  PD-L1 positivity was associated with increased response rates and progression free survival (PFS). Yet again, responses were also seen in patients who were PD-L1 negative.6 Due to the less stringent criteria used to assess PD-L1 positivity, it may be more difficult to assess the impact that PD-L1 expression played in treatment outcomes in this study.        

As a result of this uncertainty, some clinicians question whether PD-L1 is the most appropriate biomarker to determine eligibility for immunotherapy. The clinical guidelines for ASCO and the NCCN regarding mNSCLC and mM do not currently recommend that use of these checkpoint inhibitors be restricted to patients who are PD-L1 positive, even though the Keytruda label for use in mNSCLC requires that patients must “…express PD-L1 as determined by an FDA-approved test…”.7   Clearly, there is a lack of consensus on the role that PD-L1 status will play in the use of PD-1 inhibitors, as well as PD-L1 inhibitors in development.    

In response to this dynamic and evolving landscape, many oncologists are already testing for PD-L1 expression among their patients with mNSCLC and mM and are working closely with pathologists to decide how to test for this biomarker and how to interpret the results. Early data suggest that PD-L1 status appears to be playing a role in helping to drive treatment decisions among certain patient populations.

What is the Current State of PD-L1 Testing?

We evaluated recent adoption levels of PD-L1 testing and its impact on treatment decisions among patients with mNSCLC and mM. Our evaluation leveraged data from two sources: the BrandImpact longitudinal, mobile research panel of over 400 oncologists, which provides chart audit-like detail on treatment decisions; and, the BrandImpactDx panel of oncologists and pathologists, providing insight into molecular diagnostics, including testing rates, the types of tests used and interpretation of results. 

Based on responses from 170 matched oncologists who participated in both panels from May through October 2015, we were able to comprehensively analyze PD-L1 testing behavior among oncologists and link this information to their treatment decisions. We defined a “PD-L1 Tester” as an oncologist who has tested at least some patients with mNSCLC or mM, whereas a “Non-Tester” has never tested for PD-L1 expression.

Uptake of PD-L1 testing in mNSCLC and mM has been modest to date, especially when compared to the robust testing rates observed with more established biomarkers in these two malignancies. As depicted in Figure 1, most patients with mNSCLC and a majority of mM patients are not yet being tested for PD-L1. Among mNSCLC patients, less than one-quarter have been tested for PD-L1, as compared to nearly three quarters who have been evaluated for EGFR mutations. The proportion of patients with mM who have been tested for PD-L1 is higher than the rate seen for mNSCLC. Yet, virtually all mM patients are being assessed for BRAF mutations.


A minority of oncologists has adopted PD-L1 testing so far. Yet, their behavior varies widely by tumor type.  Nearly half of oncologists have tested for PD-L1 in at least some mNSCLC patients, which is nearly double the proportion of physicians who have tested for this biomarker in mM patients. However, among these PD-L1 testers, there is a disparity in the proportion of their patients who have been tested. Figure 2 demonstrates that when patient-level testing rates are evaluated among PD-L1 testing physicians only, 63% of mM patients are being tested vs. 33% of patients with mNSCLC. 


Various reasons have been offered to explain the current rate of PD-L1 testing and why a lower proportion of mNSCLC patients have been tested. The most frequently cited obstacle for both tumor types is the fact that PD-L1 testing is not currently recommended within key clinical guidelines. This has been reported more often by oncologists regarding patients with mNSCLC, as compared to those with mM.8   Some oncologists have mentioned the difficulty in obtaining an adequate biopsy in lung cancer patients, especially, if those patients have already undergone biopsies to test for more established biomarkers such as EGFR mutations and EML4-ALK. Others have pointed out that both Keytruda and Opdivo were initially approved for mM and were rapidly incorporated into treatment strategies for this indication. Perhaps some of these oncologists became early adopters of PD-L1 testing for mM, which may account for the higher patient-level testing rate.

Which Types of PD-L1 Tests are Preferred?

Oncologists and pathologists involved in PD-L1 testing have various methods from which to choose. Until recently, no FDA-approved companion diagnostics were commercially available, although several investigational tests have been in use. Figure 3 shows results from BrandImpactDx for PD-L1 testing in mNSCLC. Nearly one-in-four oncologists and one-in-five pathologists have been using companion diagnostics (including investigational test kits) to date, including those from Dako and Ventana. Pathologists most often referred PD-L1 testing to outside laboratories, followed closely by their use of laboratory developed tests (LDTs). Perhaps not surprisingly, oncologists often defer to the pathologist to decide on the methodology, which occurred in one-third of cases. The early October approval of two Dako companion diagnostics for PD-L1 may well change the market landscape for testing in the coming months.


How are PD-L1 Test Results Being Interpreted?

Clinicians do not yet agree on how to interpret PD-L1 test results. The threshold for determining “PD-L1 positivity” varies among different clinical trials and protocols. Some trials have used a threshold as low as >1% (tumor cells in the epithelium that express PD-L1), while others have used 50% as the cut-off. Other protocols use categories such as “low (1%-5%)”, “moderate (5%-10%)” or “high (>10%)” and results may be expressed as a percentage or within a range, instead of as simply “positive” or “negative”.9 Oncologists and pathologists may have different attitudes on how to evaluate test results.

An analysis of BrandImpactDx data for mNSCLC (Figure 4) clearly points out that not only do oncologists differ among themselves with respect to interpreting PD-L1 test results to classify a patient as PD-L1 positive, but there is also a divergence between oncologists and pathologists on this topic. At least half of oncologists will select a specific threshold in the low/moderate/high range (>1%, >5% or >10%) in defining a patient as PD-L1 positive.  Relatively few apply the most stringent level of >50%. In contrast, less than 30% of pathologists select a threshold in the low/moderate/high range. In fact, a majority of pathologists either defer this decision to the oncologist or state that the results may vary, depending on whether specific protocol criteria are involved.


Bearing in mind the lack of clarity regarding interpretation of PD-L1 test results, a minority of tested patients with mNSCLC and mM are being classified by their oncologists as PD-L1 positive, according to BrandImpact data. As noted in Table 2, among tested patients, only 24% of mNSCLC patients and 36% of those with mM are considered PD-L1 positive. 


How is PD-L1 Testing Influencing Treatment Decisions so Far?

Ultimately, for brand teams in the Immuno-Oncology space, the greatest value to be derived from tracking PD-L1 testing is the ability to link test results to treatment decisions. Using data from BrandImpact, we evaluated the impact of PD-L1 status on brand choice among tested patients with squamous cell mNSCLC and those with mM from May through October 2015. Only Opdivo was FDA-approved for squamous cell mNSCLC during most of the study period, although both Opdivo and Keytruda gained full marketing clearance for mNSCLC in early October.

As seen in Figure 5, patients with squamous cell mNSCLC classified as PD-L1 positive were more likely to have received Opdivo as compared to those deemed PD-L1 negative or those who were not tested. Opdivo has had its highest share at 49% among PD-L1 positive patients vs. 34% of PD-L1 negative patients and 32% for those who were untested. This is particularly noteworthy, since Opdivo’s label does not call for testing and suggests that some oncologists believe PD-L1 positive patients are likely to have better outcomes when treated with a PD-1 inhibitor.


In contrast, PD-L1 positivity has so far had a more modest impact on treatment decisions for patients with mM (Figure 6). For tested patients, shares of Keytruda and Opdivo were higher among those classified as PD-L1 positive, yet the difference was less pronounced when compared to the results for squamous cell mNSCLC.  Paradoxically, Keytruda penetration (31%) was highest among untested patients, which may reflect the fact that PD-L1 testing is not required in its label for the melanoma indication, plus most oncologists are not currently testing for this biomarker.


These apparent differences in the uptake of PD-1 inhibitors among PD-L1 positive patients with squamous cell mNSCLC and mM may be due to several factors. First, physicians do not yet agree on how to define PD-L1 positivity, let alone the role this biomarker will play in the treatment decision. Applying a lower threshold (e.g., >1%) is likely to result in a higher prevalence of PD-L1 positivity, which could diminish the impact of PD-L1 as a predictive marker. Yet, as noted above in the KEYNOTE-001 mNSCLC trial results, the greatest efficacy, measured in response rates and PFS, was reported for patients expressing high levels of PD-L1. 

Second, this exercise evaluated only a segment of the overall mNSCLC market. Both Keytruda and Opdivo are now approved to treat second-line or greater mNSCLC in patients who have received a platinum-based regimen, regardless of histologic subtype. Notably, Keytruda’s label states that patients must “express PD-L1 as determined by an FDA-approved test”, although PD-L1 expression is not defined.10 In any case, it is reasonable to expect that the presence of both PD-1 inhibitors competing head-to-head in the broader mNSCLC market will affect the rate of PD-L1 testing, as well as respective brand shares within this increasingly crowded field.

Third, both PD-1 inhibitors were initially indicated for mM and perhaps oncologists have become more familiar with using these agents in this tumor type. Yet, as seen above, a smaller proportion of oncologists are testing for PD-L1 expression in their mM patients, although physicians who are testing for PD-L1 in mM are evaluating this biomarker in the majority of their patients. The fact that neither Opdivo nor Keytruda have a requirement for testing in mM within their labels may help explain why relatively fewer oncologists have adopted PD-L1 testing in this tumor type. Further, melanoma has historically been a tumor where immunotherapy has had a key role in treatment, including cytokines and, more recently Yervoy. Perhaps in the aggregate, oncologists are more willing to prescribe a PD-1 inhibitor for mM based on their greater level of comfort with this class of agents in this malignancy, combined with perceptions about the utility of immunotherapy in melanoma, regardless of PD-L1 status.  

Where Do We Go From Here? 

As noted in Figure 1, historical data for more established biomarkers (e.g., EGFR mutations and ALK in mNSCLC and BRAF mutations for mM) may well lead us to expect that over time oncologists will test for PD-L1 expression in a growing proportion of their mNSCLC and mM patients. Factors that should contribute to an increase in PD-L1 testing will include the availability of companion diagnostics and in the case of mNSCLC, the requirement in Keytruda’s label limiting its use to patients who are PD-L1 positive.

The growing acceptance of precision medicine and the increasing prevalence of testing across a range of biomarkers will also likely spill over into PD-L1. Yet, how PD-L1 testing will influence the oncologist’s treatment choice is a question that remains to be answered. Some observers believe that the clarity introduced by the requirement for PD-L1 testing will actually prove to be an advantage and will ultimately lead to greater usage of Keytruda.

In a similar vein, numerous clinical trials with PD-L1 inhibitors in development (such as Genentech’s atezolizumab) are incorporating PD-L1 testing into their protocols and one might anticipate that as new agents come to market, they will likewise carry a testing requirement in their respective labels. Conversely, the absence of a requirement for PD-L1 testing in the Opdivo label may prove to be a benefit for this brand, as it can be marketed to a broader patient population.

Many oncologists may indeed be inclined to use PD-1 inhibitors more widely, in light of the lack of a consensus on testing, a perceived favorable benefit-to-risk ratio for these agents, and a high level of need for new treatment options in mNSCLC and mM. Only time will tell which of these treatment strategies is likely to prevail. The role of PD-L1 testing in oncology treatment is still evolving. It will be fascinating to watch the immuno-oncology market dynamics continue to unfold moving forward.



  1. BrandImpact data (AlphaImpactRx)
  2. Ibid; Barclays reports
  3. Ibid
  4. 2015 AACR General Meeting, Abstract CT104
  5. Ibid; NEJM, 373(2): 123-135 (2015); Bristol-Myers Squibb
  6. 2015 ASCO General Meeting, Abstract 9005
  7. Clinical Advances in Hematology & Oncology, 13(10): 676-683 (2015); Keytruda prescribing information
  8. BrandImpactDx (AlphaImpactRx)
  9., September 2 (2014); BrandImpactDx (AlphaImpactRx)
  10. Keytruda prescribing information


About the Contributor

Bill Bowman is Senior Manager, Product Management, at AlphaImpactRx.

AlphaImpactRx is the leading provider of primary research-based insights, analytics and solutions to biopharmaceutical, consumer health and medical device companies.  Through the combination of its consulting expertise, innovative technology, proprietary physician panels and syndicated data assets, AlphaImpactRx provides market research to help its global healthcare clients make informed and timely business decisions to develop and market their products.

BrandImpact® is fueled by data captured through a mobile, longitudinal research model and AlphaImpactRx’s proprietary research app across a range of key therapeutic areas.  BrandImpact’s longitudinal panels of high-value physicians collect detailed information at the point of engagement for both patient and sales visits.  BrandImpact captures physician treatment decisions at the point-of-care and at a chart audit-level of granularity, including patient and disease-specific characteristics such as biomarkers.  BrandImpact’s Oncology panel has 450 oncologists and hematologists and was built to align with clients’ target lists. 

AlphaImpactRx also has a Promotional Response Practice that leverages our longitudinal and normative data assets with advanced analytics, to empower clients in improving brand performance by making more informed and effective decisions with respect to marketing strategy, message execution and field force deployment.  BrandImpactDx® is fueled by a proprietary survey fielded monthly that reaches approximately 1,200 oncologists and 600 pathologists per year.  BrandImpactDx provides in-depth information on the role of molecular diagnostics in the management of cancer and covers testing rates, methodologies used and interpretation of test results.  Linking BrandImpactDx with BrandImpact provides unprecedented insights into precision medicine and the impact of biomarker testing on treatment choice among oncologists.

For more information, please visit, or contact Stacy Mecham at or Bill Bowman at

© Caribou Publishing. All rights reserved. Reproduction in whole or in part is prohibited.