PD-L1 expression predicts which solid tumor patients will respond to PD-1/PD-L1 treatment

Chicago, IL, June 6, 2016—In patients with solid tumors, a positive predictive marker for patient response to treatment with checkpoint inhibitors may be PD-L1 expression on tumor cells, with a cutoff of 5% tumor expression perhaps providing for the best predictive value, according to data from a meta-analysis presented here at the American Society of Clinical Oncology 2016 Annual Meeting.

“It’s an exciting time in oncology, and an exciting time to be a lung cancer oncologist. There have been remarkable improvements and treatments with immunotherapy for various solid tumors which are otherwise very hard to treat,” began senior author Vamsidhar Velcheti, MD, staff physician, Cleveland Clinic, Cleveland, OH.

“However, the major challenge in the field is the limited response [to] these drugs. Not everyone responds to these drugs. In fact, the vast majority of patients who get these drugs don’t respond. So the challenge in the field is to develop biomarkers to use to select patients for these treatments. These treatments are extremely expensive; they are very expensive drugs. You don’t want to be giving them to everyone out of cost-consciousness; you must choose them wisely,” he said. “That’s where my work comes in in terms of actually using the PD-L1 biomarker as an effective strategy for identifying patients who would benefit most from these drugs.

Dr. Velcheti and colleagues conducted this meta-analysis to assess the predictive value of PD-L1 IHC. They searched for clinical trials that used PD1/PD-L1 inhibitors that accrued biomarker data for PD-L1 using PubMed, Medline, and Cochrane, and included data from all major conferences until January 2015. From all phase I, II, and III clinical trials that included assessments of nivolumab, pembrolizumab, atezolimumab, durvalumab, and avelumab as treatment for non-small cell lung cancer (NSCLC), melanoma, and genitourinary cancers. They used modeling with the DerSimonian-Laird random effects model to compare odds ratios from the combined trials for response in patients who were PD-L1+ to those of PD-L1- patients.

In all, they identified 18 trials in which researchers enrolled 2,731 patients; nine involved NSCLC, four melanoma, three renal cell carcinoma, and two bladder cancer (see Table). Across all tumor types, PD-L1 expression, they found, was predictive of a favorable response (OR: 2.77; 95% CI: 2.26, 3.39; P < 0.001) in all trials except those on bladder cancer, which were few, with small numbers of patients. The largest effect was seen in NSCLC (OR: 3.33; 95% CI: 2.52, 4.40; P < 0.001).

Table

Tumor	No. trials	OR (95% CI)	P
NSCLC	9	3.33 (2.52-4.40)	< 0.001
Melanoma	4	2.28 (1.63-3.19)	< 0.001
Renal Cell	3	2.35 (1.04-5.31)	0.04
Bladder	2	1.76 (0.54-5.72)	0.35
PDL-1 Cutoff
1%	3	2.10 (0.82-5.36)	0.12
5%	3	2.72 (1.27-5.81)	0.01
10%	3	2.48 (1.04-5.93)	0.04

In addition, Dr. Velcheti and colleagues found that in the trials in NSCLC in which different thresholds for defining PD–L1 positivity were studied (1%, 5%, 10%), a cutoff of 5% may allow for maximum discrimination (OR: 2.72; 95% CI: 1.27-5.81; P=0.01).

“These are very expensive drugs. These are very promising drugs for a reason: patients who have a response tend to have a long duration of response. Most of the patients who respond to these drugs are able to mount an endogenous immune response. Patients who are not able to mount an endogenous immune response and have PD-L1 expression may not benefit from these drugs. So it is really critical to have biomarker development and incorporate biomarkers into clinical practice so that we use these drugs more rationally,” concluded Dr. Velcheti.