`
`PD-1 Blockade in Tumors
`with Mismatch-Repair Deficiency
`D.T. Le, J.N. Uram, H. Wang, B.R. Bartlett, H. Kemberling, A.D. Eyring,
`A.D. Skora, B.S. Luber, N.S. Azad, D. Laheru, B. Biedrzycki, R.C. Donehower,
`A. Zaheer, G.A. Fisher, T.S. Crocenzi, J.J. Lee, S.M. Duffy, R.M. Goldberg,
`A. de la Chapelle, M. Koshiji, F. Bhaijee, T. Huebner, R.H. Hruban, L.D. Wood,
`N. Cuka, D.M. Pardoll, N. Papadopoulos, K.W. Kinzler, S. Zhou, T.C. Cornish,
`J.M. Taube, R.A. Anders, J.R. Eshleman, B. Vogelstein, and L.A. Diaz, Jr.
`
`ABSTR ACT
`
`BACKGROUND
`Somatic mutations have the potential to encode “non-self” immunogenic antigens.
`We hypothesized that tumors with a large number of somatic mutations due to
`mismatch-repair defects may be susceptible to immune checkpoint blockade.
`METHODS
`We conducted a phase 2 study to evaluate the clinical activity of pembrolizumab,
`an anti–programmed death 1 immune checkpoint inhibitor, in 41 patients with
`progressive metastatic carcinoma with or without mismatch-repair deficiency. Pem-
`brolizumab was administered intravenously at a dose of 10 mg per kilogram of body
`weight every 14 days in patients with mismatch repair–deficient colorectal cancers,
`patients with mismatch repair–proficient colorectal cancers, and patients with mis-
`match repair–deficient cancers that were not colorectal. The coprimary end points
`were the immune-related objective response rate and the 20-week immune-related
`progression-free survival rate.
`RESULTS
`The immune-related objective response rate and immune-related progression-free sur-
`vival rate were 40% (4 of 10 patients) and 78% (7 of 9 patients), respectively, for mis-
`match repair–deficient colorectal cancers and 0% (0 of 18 patients) and 11% (2 of
`18 patients) for mismatch repair–proficient colorectal cancers. The median progres-
`sion-free survival and overall survival were not reached in the cohort with mismatch
`repair–deficient colorectal cancer but were 2.2 and 5.0 months, respectively, in the
`cohort with mismatch repair–proficient colorectal cancer (hazard ratio for disease
`progression or death, 0.10 [P<0.001], and hazard ratio for death, 0.22 [P = 0.05]). Pa-
`tients with mismatch repair–deficient noncolorectal cancer had responses similar to
`those of patients with mismatch repair–deficient colorectal cancer (immune-related
`objective response rate, 71% [5 of 7 patients]; immune-related progression-free sur-
`vival rate, 67% [4 of 6 patients]). Whole-exome sequencing revealed a mean of 1782
`somatic mutations per tumor in mismatch repair–deficient tumors, as compared with
`73 in mismatch repair–proficient tumors (P = 0.007), and high somatic mutation loads
`were associated with prolonged progression-free survival (P = 0.02).
`CONCLUSIONS
`This study showed that mismatch-repair status predicted clinical benefit of immune
`checkpoint blockade with pembrolizumab. (Funded by Johns Hopkins University and
`others; ClinicalTrials.gov number, NCT01876511.)
`
`The authors’ full names, academic de-
`grees, and affiliations are listed in the Ap-
`pendix. Address reprint requests to Dr.
`Diaz at the Johns Hopkins Sidney Kim-
`mel Comprehensive Cancer Center, 1650
`Orleans St., Rm. 590, Baltimore, MD
`21287, or at ldiaz1@ jhmi . edu.
`
`This article was published on May 30, 2015,
`at NEJM.org.
`
`N Engl J Med 2015;372:2509-20.
`DOI: 10.1056/NEJMoa1500596
`Copyright © 2015 Massachusetts Medical Society.
`
`n engl j med 372;26 nejm.org
`
`june 25, 2015
`
`2509
`
`T h e ne w e ngl a nd jou r na l o f m e dicine
`
`
`
`Genome Ex. 1031
`Page 1 of 12
`
`
`
`T h e ne w e ngl a nd jou r na l o f m e dicine
`
`The programmed death 1 (PD-1) path-
`
`way is a negative feedback system that
`represses Th1 cytotoxic immune respons-
`es and that, if unregulated, can damage the
`host.1-3 It is up-regulated in many tumors and in
`their surrounding microenvironment. Blockade
`of this pathway with antibodies to PD-1 or its
`ligands has led to remarkable clinical responses
`in patients with many different types of cancer,
`including melanomas, non–small-cell lung can-
`cer, renal-cell carcinoma, bladder cancer, and
`Hodgkin’s lymphoma.4-10 The expression of PD-1
`ligands (PD-L1 or PD-L2) on the surface of tu-
`mor cells or immune cells is an important — but
`not a definitive — predictive biomarker of re-
`sponse to PD-1 blockade.4,6-8,11
`In reports of the effects of PD-1 blockade in
`human tumors, only 1 of 33 patients with
`colorectal cancer had a response to this treat-
`ment, in contrast to substantial fractions of pa-
`tients with melanomas, renal-cell cancers, and
`lung tumors who have a response.10,12 What was
`different about this single patient? We hypothe-
`sized that this patient had mismatch-repair defi-
`ciency, because mismatch-repair deficiency oc-
`curs in a small fraction of advanced colorectal
`cancers,13,14 somatic mutations found in tumors
`can be recognized by the patient’s own immune
`system,15 and mismatch repair–deficient colorec-
`tal cancers have 10 to 100 times as many so-
`matic mutations as mismatch repair–proficient
`colorectal cancers.16-18 Moreover, mismatch re-
`pair–deficient cancers contain prominent lym-
`phocyte infiltrates, a finding consistent with an
`immune response.19-22 In addition, two of the
`tumor types that were most responsive to PD-1
`blockade in a study by Topalian et al.10 had high
`numbers of somatic mutations as a result of
`exposure to cigarette smoke (lung cancers) or
`ultraviolet radiation (melanomas).23,24 Our hy-
`pothesis was correct: the tumor of the single
`patient with colorectal cancer who had a response
`to PD-1 blockade was mismatch repair–defi-
`cient.25 Therefore, we hypothesized that mis-
`match repair–deficient tumors are more re-
`sponsive to PD-1 blockade than are mismatch
`repair–proficient tumors.
`To test this hypothesis, we initiated a phase 2
`clinical trial to evaluate immune checkpoint
`blockade in patients whose tumors had or did
`not have mismatch-repair deficiency. Because
`mismatch-repair deficiency in tumors arises
`
`through two routes,26-28 we recruited patients
`with hereditary nonpolyposis colorectal cancer
`(also known as the Lynch syndrome), which re-
`sults from an inherited germline defect in one of
`four mismatch-repair genes followed by a sec-
`ond inactivating somatic change in the remain-
`ing wild-type allele. We also recruited patients
`with sporadic mismatch repair–deficient tu-
`mors, in which both alleles of a mismatch-repair
`gene are inactivated by somatic mutations or by
`epigenetic silencing.29 In either case, the neo-
`plasms that arise harbor hundreds or thousands
`of mutations.16,18
`
`Methods
`
`Patients
`Patients with treatment-refractory progressive
`metastatic cancer were recruited from three cen-
`ters for this phase 2 study (Table 1). Three co-
`horts were evaluated: cohort A included patients
`with mismatch repair–deficient colorectal ade-
`nocarcinomas, cohort B included patients with
`mismatch repair–proficient colorectal adenocar-
`cinomas, and cohort C included patients with
`mismatch repair–deficient cancers of types oth-
`er than colorectal.
`
`Study Oversight
`The protocol, available with the full text of this
`article at NEJM.org, was approved by the institu-
`tional review board at each site, and the study
`was conducted in accordance with the provisions
`of the Declaration of Helsinki and the Interna-
`tional Conference on Harmonisation Good Clin-
`ical Practice guidelines. All the patients provided
`written informed consent before study entry.
`The first author (the principal investigator) and
`the last author (the Investigational New Drug
`sponsor) were responsible for oversight of the
`study. Merck donated the study drug and re-
`viewed the final drafts of the protocol and of
`this manuscript before submission; they did not
`participate in the analysis of the data.
`
`Study Design
`This phase 2 trial was conducted with the use of
`a Green–Dahlberg two-stage design and includ-
`ed the three parallel cohorts described above.
`The study agent, pembrolizumab, was adminis-
`tered intravenously at a dose of 10 mg per kilo-
`gram of body weight every 14 days (Fig. S1 in
`
`2510
`
`n engl j med 372;26 nejm.org
`
`june 25, 2015
`
`
`
`Genome Ex. 1031
`Page 2 of 12
`
`
`
`PD-1 Blockade in Mismatch-Repair Deficiency
`
`Table 1. Demographic and Baseline Characteristics of the Patients.*
`
`Characteristic
`Median age (range) — yr
`Sex — no. (%)
`Female
`Male
`Race — no. (%)‡
`White
`Black
`Other
`ECOG performance status — no. (%)§
`0
`1
`Cancer type — no. (%)
`Colon
`Rectal
`Ampullary or cholangiocarcinoma
`Endometrial
`Small bowel
`Gastric
`Histologic grade — no. (%)
`Well or moderately differentiated
`Poorly differentiated
`Other
`Stage IV cancer — no. (%)
`Liver metastases — no. (%)
`Median time since initial diagnosis (range) — mo
`Previous therapies — no. (%)
`1
`2
`3
`>4
`Detected germline mutation or known Lynch
`syndrome — no. (%)
`
`Yes
`No
`Unknown
`BRAF wild type — no. (%)
`Yes
`No
`Unknown
`KRAS wild type — no. (%)
`Yes
`No
`Unknown
`
`Mismatch
`Repair–Deficient
`Colorectal Cancer
`(N = 11)
`46 (24–65)
`
`Mismatch
`Repair–Proficient
`Colorectal Cancer
`(N = 21)
`61 (32–79)
`
`Mismatch
`Repair–Deficient
`Noncolorectal Cancer
`(N = 9)
`57 (34–92)
`
`5 (45)
`6 (55)
`
`8 (73)
`1 (9)
`2 (18)
`
`0
`11 (100)
`
`9 (82)
`2 (18)
`0
`0
`0
`0
`
`7 (64)
`4 (36)
`0
`11(100)
`6 (55)
`31 (6–95)
`
`0
`3 (27)
`3 (27)
`5 (45)
`
`9 (82)
`2 (18)
`0
`
`8 (73)
`0
`3 (27)
`
`6 (55)
`5 (45)
`0
`
`8 (38)
`13 (62)
`
`17 (81)
`3 (14)
`1 (5)
`
`6 (29)
`15 (71)
`
`18 (86)
`3 (14)
`NA
`NA
`NA
`NA
`
`4 (44)
`5 (56)
`
`8 (89)
`0
`1 (11)
`
`2 (22)
`7 (78)
`
`0
`0
`4 (44)
`2 (22)
`2 (22)
`1 (11)
`
`18 (86)
`3 (14)
`0
`21 (100)
`11 (52)
`58 (27–192)
`
`4 (44)
`3 (33)
`2 (22)
`9 (100)
`6 (67)
`23 (2–105)
`
`0
`4 (19)
`5 (24)
`12 (57)
`
`0
`21 (100)
`0
`
`11 (52)
`1 (5)
`9 (43)
`
`13 (62)
`8 (38)
`0
`
`1 (11)
`5 (56)
`1 (11)
`2 (22)
`
`4 (44)
`4 (44)
`1 (11)
`
`4 (44)
`0
`5 (56)
`
`4 (44)
`1 (11)
`4 (44)
`
`P Value†
`0.02
`0.72
`
`0.66
`
`0.07
`
`>0.99
`
`0.20
`
`>0.99
`>0.99
`0.07
`0.89
`
`<0.001
`
`0.64
`
`0.72
`
`* NA denotes not applicable.
`† P values are for the comparison between the cohort with mismatch repair–deficient colorectal cancer and the cohort with mismatch repair–
`proficient colorectal cancer.
`‡ Race was self-reported.
`§ Eastern Cooperative Oncology Group (ECOG) performance status is a measure of a patient’s ability to perform activities of daily living; val-
`ues range from 0 to 5, with higher scores indicating greater impairment.
`
`n engl j med 372;26 nejm.org
`
`june 25, 2015
`
`2511
`
`
`
`Genome Ex. 1031
`Page 3 of 12
`
`
`
`T h e ne w e ngl a nd jou r na l o f m e dicine
`
`Supplementary Appendix 1, available at NEJM.org).
`Pembrolizumab is a humanized monoclonal
`anti–PD-1 antibody of the IgG4 kappa isotype
`that blocks the interaction between PD-1 and its
`ligands, PD-L1 and PD-L2 (Fig. S1 in Supplemen-
`tary Appendix 1).
`Safety assessments were performed before
`each treatment. At the start of each treatment
`cycle, the total tumor burden was assessed by
`means of measurement of serum biomarkers.
`Radiographic assessments were performed at 12
`weeks and every 8 weeks thereafter. Further de-
`tails concerning the clinical protocol are avail-
`able at NEJM.org.
`
`Analysis of Mismatch-Repair Status
`Tumors with genetic defects in mismatch-repair
`pathways are known to harbor hundreds to
`thousands of somatic mutations, especially in
`regions of repetitive DNA known as microsatel-
`lites. The accumulation of mutations in these
`regions of the genome is termed microsatellite
`instability.26-28 Mismatch-repair status was as-
`sessed in tumors with the use of the MSI Analy-
`sis System (Promega), through the evaluation of
`selected microsatellite sequences that are par-
`ticularly prone to copying errors when mismatch
`repair is compromised.26-28 Additional details are
`provided in Supplementary Appendix 1.
`
`Genomic and Bioinformatic Analyses
`Primary tumor samples and matched normal
`peripheral-blood specimens were obtained from
`a subgroup of patients with mismatch repair–
`deficient carcinomas and a subgroup with mis-
`match repair–proficient carcinomas, for whom
`sufficient tumor tissue was available for exome
`sequencing30 and HLA haplotyping. To assess
`the potential for mutant peptide binding, so-
`matic exome data combined with each individu-
`al patient’s major histocompatibility complex
`(MHC) class I HLA haplotype were applied to an
`epitope prediction algorithm.31,32 This algorithm
`provided an estimate of the total number of
`mutation-associated neoantigens in each tumor.
`Additional details are provided in Supplementa-
`ry Appendix 1.
`
`Statistical Analysis
`The primary end points for cohorts A and B were
`the immune-related objective response rate and
`the 20-week immune-related progression-free
`
`survival rate, assessed with the use of immune-
`related response criteria.33 The primary end
`point for cohort C was the immune-related pro-
`gression-free survival rate at 20 weeks (Fig. S1 in
`Supplementary Appendix 1). Immune-related cri-
`teria (i.e., one of the types of criteria used to
`evaluate immune-based therapies) are based on
`radiographic responses, and unlike Response
`Evaluation Criteria in Solid Tumors (RECIST),
`they capture newly developed lesions detected on
`radiography in the measurement of tumor bur-
`den; these criteria are defined and compared
`with RECIST, version 1.1, in Table S1 in Supple-
`mentary Appendix 1. The response rate and 20-
`week progression-free survival rate were evalu-
`ated and reported in this study with the use of
`both RECIST, version 1.1, and immune-related
`response criteria. Progression-free survival and
`overall survival were summarized by means of
`the Kaplan–Meier method. Details of the hy-
`pothesis, the decision rules for the rejection of
`the null hypotheses, decision rules for early
`discontinuation of the study in a cohort because
`of efficacy or futility, and statistical methods are
`provided in Supplementary Appendix 1.
`
`R esults
`
`Patients
`A total of 41 consecutive patients were enrolled
`in the study and treated during the period from
`September 2013 through January 2015 (Table 1).
`Recruitment included patients in pursuit of a
`clinical trial option who were known to have
`tumors with mismatch-repair defects or who
`had tumors of unknown status who were then
`tested. One patient in the cohort with mismatch
`repair–deficient colorectal cancer was enrolled
`under an institutional review board eligibility
`waiver allowing a grade 3 bilirubin level (i.e.,
`higher than the cutoff specified in the inclusion
`criteria). A total of 32 patients with colorectal
`cancer were enrolled in cohorts A and B. All
`patients with colorectal cancer had received two
`or more previous chemotherapy regimens (a me-
`dian of four regimens), except for 1 patient with
`mismatch repair–proficient cancer who had
`received one chemotherapeutic and one (non–
`PD-1–based) immunotherapeutic regimen.
`Nine patients with mismatch repair–deficient
`solid tumors other than colorectal cancer were
`enrolled in cohort C. All patients in cohort C had
`
`2512
`
`n engl j med 372;26 nejm.org
`
`june 25, 2015
`
`
`
`Genome Ex. 1031
`Page 4 of 12
`
`
`
`PD-1 Blockade in Mismatch-Repair Deficiency
`
`Table 2. Objective Responses According to RECIST Criteria.
`
`Type of Response
`
`Complete response — no. (%)
`
`Partial response — no. (%)
`
`Stable disease at week 12 — no. (%)
`
`Progressive disease — no. (%)
`
`Could not be evaluated — no. (%)‡
`
`Objective response rate (95% CI) — %
`
`Disease control rate (95% CI) — %§
`
`Median duration of response — wk
`
`Median time to response (range) — wk
`
`Mismatch
`Repair–Deficient
`Colorectal Cancer
` (N = 10)
`
`Mismatch
`Repair–Proficient
`Colorectal Cancer
`(N = 18)
`
`Mismatch
`Repair–Deficient
`Noncolorectal Cancer
`(N = 7)
`
`0
`
`4 (40)
`
`5 (50)
`
`1 (10)
`
`0
`
`40 (12–74)
`
`90 (55–100)
`
`Not reached
`
`28 (13–35)
`
`0
`
`0
`
`2 (11)
`
`11 (61)
`
`5 (28)
`
`0 (0–19)
`
`11 (1–35)
`
`NA¶
`
`NA¶
`
`1 (14)*
`
`4 (57)†
`
`0
`
`2 (29)
`
`0
`
`71 (29–96)
`
`71 (29–96)
`
`Not reached
`
`12 (10–13)
`
`* The patient had a partial response at 12 weeks, which then became a complete response at 20 weeks.
`† One patient had a partial response at 12 weeks.
`‡ Patients could not be evaluated if they did not undergo a scan at 12 weeks because of clinical progression.
`§ The rate of disease control was defined as the percentage of patients who had a complete response, partial response,
`or stable disease for 12 weeks or more.
`¶ The median time to response was not applicable (NA) because no responses were observed among patients with mis-
`match repair–proficient colorectal cancer.
`
`received one or more previous therapeutic regi-
`mens (a median of two regimens).
`
`Primary End Point
`The immune-related objective response rate in
`cohort A was 40% (4 of 10 patients; 95% confi-
`dence interval [CI], 12 to 74), and the immune-
`related progression-free survival rate at 20 weeks
`was 78% (7 of 9 patients; 95% CI, 40 to 97)
`(Table S2 in Supplementary Appendix 1); the
`corresponding rates in cohort C were 71% (5 of
`7 patients; 95% CI, 29 to 96) and 67% (4 of 6
`patients; 95% CI, 22 to 96). In cohort B, which
`included patients with mismatch repair–profi-
`cient colorectal cancers, the immune-related
`objective response rate was 0% (95% CI, 0 to 20),
`and the immune-related progression-free sur-
`vival rate at 20 weeks was 11% (2 of 18 patients;
`95% CI, 1 to 35). Both cohorts with mismatch
`repair–deficient cancers (cohorts A and C)
`reached the prespecified point at which the pro-
`tocol indicated that the study reached its pri-
`mary efficacy end point when 4 patients were
`free from disease progression at 20 weeks and
`objective responses on the basis of immune-re-
`lated response criteria were observed in 4 pa-
`tients (Table S2 and the Methods section in
`Supplementary Appendix 1).
`The median follow-up was 36 weeks (range, 5
`
`to 55) for patients with mismatch repair–deficient
`colorectal cancer (cohort A), 20 weeks (range, 4 to
`52) for patients with mismatch repair–proficient
`colorectal cancer (cohort B), and 21 weeks (range,
`0.1 to 49) for patients with mismatch repair–
`deficient noncolorectal cancer (cohort C). All pa-
`tients for whom the 20-week immune-related
`progression-free survival rate could be evaluated
`were followed for at least 20 weeks.
`
`Radiographic Evaluation
`Of the 10 patients with mismatch repair–defi-
`cient colorectal cancer (cohort A) who could be
`evaluated for RECIST, 4 (40%; 95% CI, 12 to 74)
`had objective responses according to these crite-
`ria (Table 2 and Fig. 1, and Fig. S2 in Supple-
`mentary Appendix 1). Patients were considered
`not to have been evaluated unless they under-
`went a radiographic scan at 12 weeks. The rate
`of disease control, which was defined as the
`percentage of patients who had an objective re-
`sponse or whose disease was stable, was 90% in
`cohort A (9 of 10 patients; 95% CI, 55 to 100).
`Of the 7 patients in cohort C who could be
`evaluated, 5 (71%; 95% CI, 29 to 96) had objec-
`tive responses as defined by RECIST (Table 2
`and Fig. 1, and Fig. S2 in Supplementary Appen-
`dix 1), and the rate of disease control was 71%
`(5 of 7 patients; 95% CI, 29 to 96).
`
`n engl j med 372;26 nejm.org
`
`june 25, 2015
`
`2513
`
`
`
`Genome Ex. 1031
`Page 5 of 12
`
`
`
`T h e ne w e ngl a nd jou r na l o f m e dicine
`
`3 of 11 patients (27%) with tumors associated
`with the Lynch syndrome had a response (Table
`S3 in Supplementary Appendix 2) (P = 0.009). No
`other baseline characteristics had a significant
`association with objective responses.
`Among the 18 patients with mismatch re-
`pair–proficient colorectal cancers in cohort B,
`no objective responses as defined by RECIST
`were observed (Table 2 and Fig. 1, and Fig. S2 in
`Supplementary Appendix 1). In this group, the
`rate of disease control was 11% (2 of 18 patients;
`95% CI, 1 to 35).
`All the patients who had a response as de-
`fined by RECIST (Table 2) also had a response
`according to immune-related response criteria
`(Table S2 in Supplementary Appendix 1).
`
`Survival
`In the cohort of patients with mismatch repair–
`deficient colorectal cancer (cohort A), the medi-
`an progression-free survival and median overall
`survival were not reached (Fig. 2). In contrast,
`among the patients with mismatch repair–profi-
`cient cancers (cohort B), the median progres-
`sion-free survival was only 2.2 months (95% CI,
`1.4 to 2.8), and the median overall survival was
`5.0 months (95% CI, 3.0 to not estimable). In
`cohort C (patients with mismatch repair–defi-
`cient noncolorectal cancer), the median progres-
`sion-free survival was 5.4 months (95% CI, 3 to
`not estimable), and the median overall survival
`was not reached. A post hoc comparison of the
`cohorts with mismatch repair–deficient and
`mismatch repair–proficient colorectal cancers
`showed hazard ratios for disease progression or
`death (0.10; 95% CI, 0.03 to 0.37; P<0.001) and
`for death (0.22; 95% CI, 0.05 to 1.00; P = 0.05)
`that favored patients with mismatch repair–defi-
`cient colorectal cancer (Fig. 2).
`To evaluate whether the difference in survival
`might be due to prognostic differences, we mea-
`sured the time since the diagnosis of metastatic
`disease and the clinical performance of the
`regimen that patients had received before enroll-
`ment. We found that there was no significant
`difference between patients with mismatch re-
`pair–deficient colorectal cancer and patients
`with mismatch repair–proficient colorectal can-
`cer with respect to the duration of metastatic
`disease (P = 0.77 by the log-rank test) or the
`median progression-free survival while receiving
`their previous regimens (P = 0.60 by the log-rank
`
`Mismatch repair–proficient colorectal cancer
`Mismatch repair–deficient colorectal cancer
`Mismatch repair–deficient noncolorectal cancer
`
`0% (no change)
`
`100
`
`200
`
`300
`
`400
`
`Days
`
`A BiochemicalResponse
`200
`
`100
`
`0
`
`(%)
`
`ChangeinTumorMarkerLevel
`
`−100
`
`0
`
`B RadiographicResponse
`100
`
`Mismatch repair–proficient colorectal cancer
`Mismatch repair–deficient colorectal cancer
`Mismatch repair–deficient noncolorectal cancer
`
`20% increase (progressive disease)
`
`30% decrease (partial response)
`
`50
`
`0
`
`−50
`
`−100
`
`ofLongestDiameters(%)
`
`ChangefromBaselineintheSum
`
`Figure 1. Clinical Responses to Pembrolizumab Treatment.
`The biochemical responses to pembrolizumab treatment are shown in Pan-
`el A. Serum levels of protein biomarkers were measured at the start of each
`treatment cycle, and the values represent percentage changes from base-
`line. Each line represents one patient; patients were included if their base-
`line tumor marker values were higher than the upper limit of normal. CA-
`125 was used as the biomarker for one patient with endometrial cancer,
`CA19-9 was used for one patient with cholangiocarcinoma and one patient
`with ampullary cancer, and carcinoembryonic antigen (CEA) was used for
`all other patients. Radiographic responses to treatment with pembrolizum-
`ab, evaluated on the basis of Response Evaluation Criteria in Solid Tumors
`(RECIST), are shown in Panel B. Tumor responses were measured at regu-
`lar intervals, and the values shown are the largest percentage change in the
`sum of longest diameters from the baseline measurements of each mea-
`surable tumor. Each bar represents one patient.
`
`Patients in cohort C had faster responses
`than did patients in cohort A (median time to
`response according to RECIST, 12 weeks vs. 28
`weeks; P = 0.03). Furthermore, all 6 patients
`(100%) with mismatch repair–deficient tumors
`that were not associated with the Lynch syn-
`drome had an objective response, whereas only
`
`2514
`
`n engl j med 372;26 nejm.org
`
`june 25, 2015
`
`
`
`Genome Ex. 1031
`Page 6 of 12
`
`
`
`PD-1 Blockade in Mismatch-Repair Deficiency
`
`A Progression-freeSurvivalinCohortswithColorectalCancer
`1.0
`
`P<0.001 by log-rank test
`
`B OverallSurvivalinCohortswithColorectalCancer
`1.0
`
`P=0.03 by log-rank test
`
`Mismatch repair–deficient
`
`Mismatch repair–proficient
`
`12
`
`15
`
`0 0
`
`1 1
`
`9
`
`5 1
`
`Months
`
`6
`
`7 5
`
`0
`
`3
`
`11
`
`21
`
`9
`
`12
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`ProbabilityofOverallSurvival
`
`No.atRisk
`Mismatch repair–
`deficient
`Mismatch repair–
`proficient
`
`Mismatch repair–deficient
`
`Mismatch repair–proficient
`
`12
`
`15
`
`0 0
`
`0 0
`
`9
`
`2 0
`
`Months
`
`6
`
`6 1
`
`3
`
`8 2
`
`0
`
`11
`
`21
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`Survival
`
`ProbabilityofProgression-free
`
`No.atRisk
`Mismatch repair–
`deficient
`Mismatch repair–
`proficient
`
`C Progression-freeSurvivalinCohortwithMismatchRepair–Deficient
`NoncolorectalCancer
`1.0
`
`D OverallSurvivalinCohortwithMismatchRepair–Deficient
`NoncolorectalCancer
`1.0
`
`0
`
`9
`
`3
`
`6
`
`6
`
`2
`
`Months
`
`9
`
`1
`
`12
`
`0
`
`15
`
`0
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`ProbabilityofOverallSurvival
`
`0
`
`9
`
`3
`
`5
`
`6
`
`1
`
`Months
`
`9
`
`0
`
`12
`
`0
`
`15
`
`0
`
`No.atRisk
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`Survival
`
`ProbabilityofProgression-free
`
`No.atRisk
`
`Figure 2. Clinical Benefit of Pembrolizumab Treatment According to Mismatch-Repair Status.
`Kaplan–Meier curves are shown for progression-free survival in the cohorts with colorectal cancer (Panel A), overall survival in the co-
`horts with colorectal cancer (Panel B), progression-free survival among patients with mismatch repair–deficient noncolorectal cancers
`(Panel C), and overall survival among patients with mismatch repair–deficient noncolorectal cancers (Panel D). In both cohorts with
`mismatch repair–deficient tumors, median overall survival was not reached. Patients in the cohort with mismatch repair–proficient can-
`cers had a median progression-free survival of 2.2 months (95% CI, 1.4 to 2.8) and a median overall survival of 5.0 months (95% CI, 3.0
`to not estimable). Patients with mismatch repair–deficient noncolorectal cancers had a median progression-free survival of 5.4 months
`(95% CI, 3 to not estimable).
`
`test) (Fig. S3 in Supplementary Appendix 1). We
`also performed an additional multivariate analy-
`sis of progression-free and overall survival to
`examine the difference in outcomes between
`mismatch repair–deficient colorectal cancer and
`mismatch repair–proficient colorectal cancer,
`adjusting for elapsed time since the initial diag-
`nosis. The magnitude of the hazard ratios for
`disease progression or death (hazard ratio, 0.04;
`95% CI 0.01 to 0.21; P<0.001) and for death
`
`(hazard ratio, 0.18; 95% CI, 0.03 to 1.01; P = 0.05),
`representing the differing effects of pembroli-
`zumab between mismatch repair–deficient tu-
`mors and mismatch repair–proficient tumors,
`was maintained after adjustment for this poten-
`tial difference.
`
`Safety Assessment
`Adverse events occurring in more than 5% of
`patients are listed in Table 3. Events of clinical
`
`n engl j med 372;26 nejm.org
`
`june 25, 2015
`
`2515
`
`
`
`Genome Ex. 1031
`Page 7 of 12
`
`
`
`T h e ne w e ngl a nd jou r na l o f m e dicine
`
`interest included rash or pruritus (24%); thyroid-
`itis, hypothyroidism, or hypophysitis (10%); and
`asymptomatic pancreatitis (15%). Although the
`numbers were small, thyroid-function abnor-
`malities were limited to the cohorts with mis-
`match repair–deficient cancer (Table 3).
`
`Tumor Markers
`In the two cohorts with colorectal cancer, it was
`possible to evaluate levels of carcinoembryonic
`antigen (CEA) before enrollment; in 29 of 32
`patients, these levels were above the upper limit
`of normal (3 mg per deciliter). Substantial de-
`creases in CEA level occurred in 7 of the 10 pa-
`tients with mismatch repair–deficient colorectal
`cancer and in none of the 19 patients with mis-
`match repair–proficient colorectal cancer in whom
`CEA could be evaluated (Fig. 1, and Fig. S4 in
`Supplementary Appendix 1). Among patients
`with mismatch repair–deficient noncolorectal
`cancer, levels of tumor markers (CEA, CA19-9,
`or CA-125) were elevated above the upper limit
`of normal in 4 patients. Declines in CA19-9 or
`CA-125 of more than 70% occurred in 3 of these
`4 patients. Tumor marker kinetics in all three
`cohorts are shown in Figure 1. The degree of
`CEA decline after one dose (between day 14 and
`day 28) of pembrolizumab was predictive of
`both progression-free survival (P = 0.01) and
`overall survival (P = 0.02). The CEA response oc-
`curred well in advance of radiographic confirma-
`tion of disease control (range, 10 to 35 weeks). In
`contrast, patients who had disease progression
`had rapid biomarker elevation within 30 days
`after the initiation of therapy. Thus, changes in
`CEA levels significantly preceded and correlated
`with ultimate radiographic changes.
`
`Genomic Analysis
`The analysis of whole-exome sequences showed
`a mean of 1782 somatic mutations per tumor in
`patients with mismatch repair–deficient cancer
`(nine patients), as compared with 73 mutations
`per tumor in patients with mismatch repair–
`proficient cancer (six patients) (P = 0.007 by
`nonparametric Wilcoxon test) (Fig. S5 in Sup-
`plementary Appendix 1 and Table S3 in Supple-
`mentary Appendix 2). Most of these mutations
`(63%) are predicted to alter amino acids.
`These mutations were then assessed for their
`immunogenic potential in the context of each
`patient’s MHC haplotype. We identified a mean
`of 578 potential mutation-associated neoanti-
`
`gens from the tumors of patients with mismatch
`repair–deficient cancers; 21 such neoantigens
`were identified in tumors from patients with
`mismatch repair–proficient cancers (Table S3 in
`Supplementary Appendix 2). The percentage of
`potential mutation-associated neoantigens among
`all somatic mutations was similar in the two
`cohorts (a mean of 32% in patients with mis-
`match repair–deficient cancer and 29% in pa-
`tients with mismatch repair–proficient cancer).
`High numbers of somatic mutations and poten-
`tial mutation-associated neoantigens were asso-
`ciated with longer progression-free survival and
`with a trend toward objective response (Fig. S5
`and Table S4 in Supplementary Appendix 1).
`
`Immunohistochemical Analysis
`The expression of CD8 and PD-L1 was evaluated
`within the tumor and at the invasive fronts of
`the tumor in an immunohistochemical analysis
`in the 30 cases in which tumor tissue was avail-
`able (Fig. S6 in Supplementary Appendix 1). Tu-
`mors from patients in cohorts A and C con-
`tained a greater density of CD8-positive lymphoid
`cells than did tumors from patients in cohort B
`(P = 0.10) (Fig. S7 in Supplementary Appendix 1),
`and CD8 labeling was associated with a trend
`toward objective response and stable disease
`(Fig. S8 and Table S5 in Supplementary Appen-
`dix 1). This CD8-positive lymphoid infiltrate was
`especially prominent at the invasive fronts of the
`tumors (P = 0.04) (Fig. S7 in Supplementary Ap-
`pendix 1). Membranous PD-L1 expression occurred
`only in patients with mismatch repair–deficient
`cancer and was prominent on tumor-infiltrating
`lymphocytes and tumor-associated macrophages
`located at the invasive fronts of the tumor
`(P = 0.04) (Fig. S7 in Supplementary Appendix 1).
`The expression of CD8 and PD-L1 was not sig-
`nificantly associated with progression-free sur-
`vival or overall survival (Table S5 in Supplemen-
`tary Appendix 1).
`
`Discussion
`
`The data from this small phase 2 trial of pem-
`brolizumab for the treatment of tumors with
`and tumors without mismatch-repair deficiency
`support the hypothesis that mismatch repair–
`deficient tumors are more responsive to PD-1
`blockade than are mismatch repair–proficient
`tumors. Mismatch-repair deficiency occurs in
`many cancers, including those of the colorec-
`
`2516
`
`n engl j med 372;26 nejm.org
`
`june 25, 2015
`
`
`
`Genome Ex. 1031
`Page 8 of 12
`
`
`
`PD-1 Blockade in Mismatch-Repair Deficiency
`
`tum, uterus, stomach, biliary tract, pancreas,
`ovary, prostate, and small intestine.18,34-42 It is
`possible that patients with mismatch repair–
`deficient tumors of these types may also bene-
`fit from anti–PD-1 therapy, as may patients
`whose tumors contain other DNA repair defi-
`ciencies, such as those with mutations in POLD,
`POLE, or MYH.18,43,44
`The hypothesis that mismatch repair–defi-
`cient tumors stimulate the immune system is
`not a new idea45; it has been supported by obser-
`vations of the dense immune infiltration and
`Th1-associated cytokine-rich environment in
`mismatch repair–deficient tumors.19-22,46 A recent
`study refined these classic observations by show-
`ing that the mismatch repair–deficient tumor
`microenvironment strongly expressed several
`immune checkpoint ligands, including PD-1, PD-
`L1, CTLA-4, LAG-3, and IDO, which indicates
`that their active immune microenvironment is
`counterbalanced by immune inhibitory signals
`that resist tumor elimination.47 The most likely
`explanation for both the old and new findings
`was that the immune infiltrate associated with
`mismatch repair–deficient carcinomas was di-
`rected at neoantigens. The correlation of a
`higher mutational load and a higher rate of re-
`sponse to anti–CTLA-4 in melanoma41 and anti–
`PD-1 in lung cancer48 provides further support
`for the idea that mutation-associated neoantigen
`recognition is an important compo