Pavlik Reviews Immunotherapy for Cutaneous Carcinomas in Case Studies

Anna C. Pavlick, DO, MS, MBA

Anna C. Pavlick, DO, MS, MBA

Anna C. Pavlick, DO, MS, MBA, spoke with a group of physicians about the systemic therapy options, including immunotherapeutics, for treating patients with cutaneous carcinomas in a recentTargeted Oncologylive case-based peer perspectives discussion. Pavlick, the assistant director of clinical research education and co-director of the Melanoma Program, NYU Langone Health, Perlmutter Cancer Center, explained treatment options based on 2 case scenarios of patients with cutaneous malignancies.

Case 1

An otherwise healthy, single, 69-year-old man, fair-skinned, retired construction worker presented to his primary care physician with what he described as a wound behind his ear that was not healing. He reported first noticing it at least 4 months earlier and complained of recent onset of numbness in the area. He had an ECOG performance status of 1.

The visible ulcerated lesion was 4.5 cm in diameter and >5 mm deep. No palpable nodes were detected. A biopsy confirmed a poorly differentiated, infiltrative, postauricular, cutaneous squamous cell carcinoma (CSCC) lesion with 7-mm invasion into subcutaneous fat.

Targeted Oncology:When taking the patient’s history based on this physical examination, how important is it to take note of neuropathic changes?

Pavlick:Squamous cell carcinoma very commonly presents with perineural invasion, so it will grow along the nerves. Many times, when it is invasive—especially in this case in which the [lesion] is very deep and very large—it is exceedingly difficult to clear those surgical margins and [remove] all the carcinoma. It is possible to remove the superficial lesion, but the patient will have squamous cell tracking all the way across that nerve.

When a patient presents with the symptoms of a stabbing, burning, or neuropathic type of pain, [the surgeon] has to real­ize that what [is present] is way bigger than what you are seeing on the surface. You should probably image it to get a better idea of what you are dealing with.

What additional testing would you order to aid treatment planning?

You would start with routine laboratory tests. An MRI is very sensitive to picking up neural invasions, making it a better test compared with a CT scan, especially in the head and neck [region]. But we are dealing with an older population, and many patients in this population have pacemakers, so an MRI may not be available. But if it’s possible, an MRI is the ideal test to tell you about the extent of the perineural spread.

What are the high-risk features of this patient?

The high-risk features include the neuropathic [involvement], the size of the lesion, the location, and ulceration. This is bad; this patient is not going to do well. The location refers to the head and neck area and the fact that he is a man. If considering surgery, the question is, “Can you get clear margins?” And if you can get clear margins, how big is your defect going to be and will that deformity be something the patient is willing to accept and live with? Just because we can cut it out, does not always mean that we should. Patients need to understand exactly what is going to be cut off.

A patient of mine was sent to me after undergoing surgery from the head-and-neck surgeon. She had a 2-cm lesion that was ulcerated, but she [also] had significant perineural pain above her eye, ear pain, and perioral numbness. They sent her for an MRI and she had a gigantic tumor that was in her maxillary sinus that was tracking every single nerve on the left side of her face, [including] the trigeminal nerve. Fortunately, she did not have any invasion into her sinus, and this looked very resectable from the surface. But upon imaging, it was revealed to take up half her face, and they did talk to her about a hemifacectomy. Fortunately, she said, “Can I talk to somebody else?” So they sent her up to see me to see if anything else could be done.

[Something] to mention about squamous cell is that it does take a village to make the right decision. I think when it comes to a lot of skin cancers, [the patient] benefits from a big multidisci­plinary team. I am fortunate enough to have my head-and-neck specialists, my dermatologists, my oncology surgeons. We all talk to each other and we all trust each other. Fortunately, when the ENT [ear, nose, and throat; otolaryngology] surgeons saw how extensive this tumor was, they thought, “Maybe we should get an opinion from the oncologists.”

Many of my surgeons think that because I use immunother­apies that I can fix everything, so there is no need for them. In fact, sometimes they send me patients and I call them and say, “You need to cut this out.” And they reply, “Can’t you fix this with­out cutting it out?”

Case 1 (continued):

The recommended definitive surgical approach involved auriculectomy, but the patient declined.

What would you administer in the case that the patient does not want to have surgery?

Sometimes when the lesion is very large and invasive, we give neoadjuvant therapy with the intent to shrink it down to make it more resectable and less likely to cause a deformity. The patient did not want to undergo the surgery, so he was offered cemiplimab [Libtayo]. It is an anti—PD-1 monoclonal antibody. Its only indica­tion is for patients with metastatic CSCC or locally advanced CSCC who are not candidates for curative surgery or curative radiation.

It is a standard-dose, anti—PD-1 antibody. It targets PD-1 on tumors and has its own set of adverse events [AEs] that differ from those of chemotherapy. AEs are not as extensive, but patients still get a rash, diarrhea, itching, pneumonitis; you will see more pneumonitis if you give this to a smoker. Different tumor types have different predominant AEs. Collectively, immu­notherapy causes autoimmune-type AEs.

I tell my patients that if you are administered immunother­apy, you are going to get the “-itises”: dermatitis, pneumonitis, hepatitis, nephritis, colitis, meningitis. [Meningitis] is the one that is often overlooked, especially when patients come in with subtle AEs such as headaches or a mild stiff neck. When these patients go to the emergency department and they receive a spinal tap, the spinal fluid is going to be [filled] with leukocytes. When you subtype them, they will all be T cells. It is an amazing sight because you have brought all these T cells into the area. Steroids are usually prescribed and they get a whole lot better.

Case 1 (continued):

Cemiplimab 350 mg every 3 weeks was initiated.

Do you need to consider staining for anti—PD-L1 on the tumor?

No. What we know, at least for melanoma, is very different from what we know in lung cancer, and it is different from what we know about squamous cell cancer. When we consider giving anti—PD-1 agents to patients with metastatic lung cancer, something that we always do ahead of time is anti–PD-L1 staining on the tumor. If the tumor has a lot of expression of PD-L1, the patient will probably respond to immunotherapy. If they are low- or negative-expressing, you probably do not want to give them the drugs.

When it comes to melanoma, PD-L1 status does not matter. You can be a 0% expresser but still respond to immunotherapy. When it comes to squamous cell carcinoma, which has even more muta­tions than melanoma, 80% to 90% of patients with squamous cell carcinoma are going to be overexpressers of PD-L1. So we do not routinely test for it in squamous cell carcinoma. These patients behave much more like patients with melanoma than patients with lung cancers. Microsatellite instability does not correlate with response, either.

What do you expect this patient's response to be?

For this patient, we should expect a very good response. Response rates are typically beyond 50%. We would expect a good response because of the disease type and knowing that these tumors have high mutational burdens and that they are big PD-L1 overexpressers.

The one thing I have yet to find when using cemiplimab is a patient who develops an endocrinopathy. Those are the ones that are interesting because they are permanent.

Is this usually prescribed by dermatologists?

No, most dermatologists do not have infusion centers. They do not want to deal with the “-itises” that can occur. Some derma­tologists do, depending on geographic location. In Europe, some dermatologists [will prescribe]; they are called dermato-oncologists and they would be double-boarded [in both special­ties]. Most dermatologists here in the United States would refer to an oncologist.

How are these agents metabolized?

These drugs are usually pathway metabolized. Part of the issue with these agents is that because most of the metabo­lism is hepatic, when you treat patients with liver metastases, and their liver function test results are elevated because of the disease, a high risk of liver toxicity exists because the agent is hepato-metabolized.

Does this type of patient have a standard of care?

There is none for treating a patient with invasive squamous cell cancer. We extrapolate the data and we talk about giving 5-FU [fluorouracil], platinum chemotherapy, radiation, or cetuximab [Erbitux]. However, when you look at those data, the response rates are only in the 20% to 30% range, with no durability.

The nice thing about these [anti—PD-1 agents] is their dura­bility. If you hit the home run, you hit the home run. I have patients who I treated 3 or 4 years ago on these trials who have remained disease-free. And [these agents] are a lot more toler­able than platinum.

For most of the squamous cell cancer cases, after 2 cycles, you start to see a response.

If the patient does not achieve a response, what would be your next step?

You could fall back to opting for surgery, but with someone with perineural invasion, you probably do not want to do that. You could consider chemotherapy; you could consider cetuximab and radiation. You fall back to what we had, knowing that response rates have been low.

What is your approach to monitoring and managing treatment-related toxicities in patients receiving cemiplimab?

We are fortunate to have excellent nurse practitioners. And we do try to educate the families to call the clinic because if something happens, it is easier to address the problem before it becomes [an acute emergency].

I have learned to try to use reverse psychology on most patients. As you know, most patients want to get their treat­ments on time. They know they are coming back in 3 weeks, and they want their treatment, and if they tell you they had an AE, they think you might not give them a treatment.

I just turn it around and tell them that if they come in with AEs and they have not told me [about them earlier], then they are not going to receive their treatment.

Addressing the AEs as soon as they happen means the patient is going to get their treatment in 3 weeks, [because] I have 3 weeks to make them go away. That [strategy] seems to work in getting them to call [right away] when they experience AEs.

Please discuss the data from the phase I and pivotal phase II trials¹involving cemiplimab in squamous cell carcinoma.

Invasive, locally advanced, metastatic squamous cell [cancer] does not affect a large population. We started out with a phase I [trial], but when we found patients had great responses, we expanded that into the metastatic area. Then we went into the phase II [study] with patients with metastatic disease.

The majority of patients had head or neck involvement, with 18 out of 16 patients in the expansion cohort of the phase I study and 38 out of 59 patients in the metastatic-disease cohort of the phase II study. These are older patients who had [involvement on] the scalp behind the ear or on top of the ear. About 60% had some type of prior therapy. A lot of people [77%] had some radi­ation. And in the phase I study, 8 patients had distant metastasis versus 45 in the phase II study. Eight patients had only regional metastasis in the phase I study; 14 did in the phase II study. Ten patients had only locally advanced progression in the phase I study; 0 patients did in the phase II study.

A dermatologist referred a 95-year old man with locally advanced disease because he could not do anything more for him. But he is in very good shape for 95 years of age. That is the nice thing about these drugs—I do not think these drugs are affected by age limitations and performance status.

This gentleman comes into my office with a gauze turban on his head because the bleeding was so extensive. Removing the gauze revealed necrotic [tissue] and it smelled very bad. The dermatol­ogist told the man that there was not anything they could do for him, but I said that I may be able to. He had squamous cell cancer eating down to his scalp. You could see periosteum.

He has received 7 cycles of cemiplimab. He no longer wears the gauze turban. He presented with crusty scabs that were start­ing to lift, because skin heals from the bottom up. And I had a light day in the office so I said, “Can I pull those off?” and he said, “Really?” Underneath was beautiful pink skin. It looked fantastic.

Do you reach a point when you stop treating, after you have seen progress?

Nobody knows. Why do we treat for 2 years, when anyone after 3 cycles could die? There is no rhyme or reason. If I have an [older] patient who comes in and I render them free of disease based on clinical examination, biopsy that proves that they have no squamous cell cancer, or they have scans that are clear—then we apply Anna’s rule. My rule is that after the patient is clear, I will treat for 3 cycles, and then I stop. I do not believe treating people to toxicity.

If this is about quality of life and this gentleman is 95 years old, I am not looking to buy him 20 more years. I am looking to buy him quality for the time that he has got. He is feeling good right now, so let’s keep it that way.

Looking at the findings of Migden and colleagues, the phase I response data were 50% and 47% in the phase II [trial]. The most important thing is this tail at the end of the progression-free survival [curve]. Once you get people up to and beyond the first year of therapy, they are probably going to be fine. It is probably not going to come back. We only have about 3 years’ follow-up with this drug, but the line after a year remains flat.

Please describe the AEs observed in patients in the phase II portion of the trial.

The most common AE was any-grade diarrhea, reported in 16 patients (27%). Only 1 patient reported a grade ≥3 case of diarrhea. Remember that most of these patients stayed on therapy for 2 years. Most of these toxicities occurred toward the end of the 2 years. These are similar to the complaints we have heard with other PD-1 agents.

What is the duration of response that physicians can expect with cemiplimab therapy?

[We have seen a duration of response] in 50% of responders because the study has only been closed for 2 years. So far, responders are still responding.

Case 2

A 73-year-old Caucasian man presented to his physician with a right periauricular pigmented lesion with a 21-mm diameter. He had a medical history remarkable for arterial hypertension, dyslipidemia, hyperuricemia, and hypothyroidism, for which he has been medically treated. He had a right nephrectomy in 2002 for retroperitoneal hematoma that resulted in consequent chronic renal insufficiency. He otherwise maintains full autonomy in his daily activities and personal care. He is a nondrinker and former smoker.

He underwent surgical resection of the periauricular lesion. A lymph node biopsy was positive. The patient opted out of complete lymph node dissection.

A metastatic workup was negative. Pathologic assessment revealed an ulcerated melanoma, which had a Breslow index of 2.1 mm, Clark level of IV, and staging of pT3bN1a. 

1. his surgery, he remains active and continues golfing 3 times per week.

Please distinguish between the 2 types of cutaneous melanoma.

Melanoma comes in 2 types,BRAFmutated andBRAFwild-type. We determine the type through genotyping and immuno­histochemistry (IHC). If IHC is negative and you genotype the tissue, you still have a 50% chance of finding it is positive, so you want to make sure it is genotyped. The more difficult patients to manage are those who areBRAFwild-type because the treat­ment options are limited.

When should the clinician choose or not choose a complete lymph node dissection?

The case involves a 73-year-old gentleman with a periauricular pigmented lesion. What is important [to determine] with mela­noma is diameter and depth of the lesion. After undergoing a biopsy, it was confirmed to be melanoma. Although a sentinel lymph node biopsy was positive, he opted out of a complete lymph node dissection.

I think based on the new MSLT-II data,²which looked at patients with microscopically positive sentinel lymph nodes (>2 mm melanoma cells in the sentinel node), that if the senti­nel node had this microscopic focus of melanoma you could then be randomized to either observation or complete lymphadenec­tomy. When it came to risk of recurrence or overall survival, the findings showed no difference whether you resected the lymph nodes or not, so most surgeons are not [resecting]. Overall, we are still sampling the sentinel lymph nodes because they are a true prognostic indicator of recurrence. It helps us with staging. But if it is a tiny microscopic focus, most surgeons are not recom­mending a lymphadenectomy.

Because the patient had a positive sentinel lymph node and was not going to have a further dissection, the patient under­went imaging scans.

What is the recommended follow-up for a patient with high-risk resected melanoma?

PET scan, if insurance will pay for it, and MRI [scans] of the brain. If the insurance will not pay for a PET scan, that leaves CT for chest, abdomen, and pelvis. But absolutely, positively order an MRI of the brain. Keep in mind that although you order a whole-body PET/CT [scan], the brain is using lots of sugar and the PET/CT does a poor job of picking up brain metastases, unless [the metastasis] is quite large, at which point it will be picked up by PET/CT. You want to confirm that the patient has no brain metastases.

What do you do with a patient with stage IIIa disease? Does that change your decision?

This patient is a [stage] IIIA. What are his options? His options are either do nothing or begin adjuvant nivolumab [Opdivo]. If the patient has uncontrolled Crohn’s disease with diarrhea every day, am I going to talk to this patient about adjuvant PD-1 therapy with adjuvant nivolumab or pembrolizumab [Keytruda]? I might wait to see about his metastatic disease. If the patient is relatively well with minimal gastrointestinal issues and no overwhelming auto­immune disease, that might be an option.

I often encounter patients with rheumatoid arthritis [RA]. If you give these patients an immunotherapy agent, you could flare their RA like nobody’s business. So with a patient with stage IIIA disease and an autoimmune condition, you may want to reconsider choosing adjuvant immunotherapy because they are disease-free. If they have metastatic disease, that is a different story. If you make them miserable with the chance of rendering them disease-free or at least having their cancer under control, you are not so hesitant to give immunotherapy. But for someone with bad RA, I do not know. If they have bad RA or Crohn disease, I might sit, wait, and watch.

But our case involves a healthy gentleman, so his options are [to do] nothing, adjuvant nivolumab, or adjuvant pembrolizumab.

Would you order PD-L1 testing for this patient?

No, because in melanoma [PD-L1 status] does not matter, although if I had a dime for every patient who walked in saying, “I have my PD-L1 test here,” [I’d be rich]. I would order other molecular tests, however, such asBRAF. Let us consider where his melanoma was located. The melanoma was on his face.

Most people [develop] head and neck melanoma in sun-damaged areas of the body. If you look at melanomas in areas that have solar elastosis [an accumulation of abnormal elastin] or a lot of sun damage, they usually havec-KITmutations. If you have the ability to send genotyping, you might considerc-KIT, BRAF, andNRAS. ForNRAS, we do not have a targeted therapy yet, but we have something forc-KITandBRAF.

Not that you would necessarily use it now, but you can, because data exist that look at adjuvant BRAF medications as well, but those are for patients in stage IIIB and IIIC, so [those are] not pertinent for this patient.

But, again, if you have a patient with severe autoimmune disease and you want to give them something to reduce their risk and they have aBRAFmutation, then you have an opportunity to give them adjuvant dabrafenib [Tafinlar]/trametinib [Mekinist].

What were the phase III studies of adjuvant therapy for melanoma?

The 3 major trials were COMBI-AD [NCT01682083], CheckMate 238 [NCT02388906], and KEYNOTE-054 [NCT02362594]. About 4 years ago, the FDA approved the adjuvant therapy ipilimumab (Yervoy). It had a lot of toxicity and many patients did not like it. But it was better than nothing and then that became the FDA standard. We very quickly tested nivolumab against ipilumumab, and nivolumab won hands down.

Most recently, pembrolizumab just received an FDA approval [based on data from the KEYNOTE-054 trial]. Interestingly, when we considered the trial designs, we did not compare the experi­mental drugs with interferon. That is because of interferon’s high toxicity, and patients do not like it. [Another issue] was that we did not know whether or not to use the high-dose interferon, in which you give them the interferon for a month followed by a year of maintenance, or just give [interferon for] the month. Do you give the pegylated version? No consensus emerged because all the oncologists hated it. Patients would rather do nothing than receive interferon.

Is either nivolumab or pembrolizumab preferred?

My preference is interval of treatment. For pembrolizumab, patients are seen every 3 weeks versus nivolumab, for which patients need to come in every 4 weeks. For some of my older patients, I put them on pembrolizumab. If I am concerned about how they are going to do, I’ll see them more frequently. The toxicity profiles are relatively the same.

What about the combination of ipilimumab and nivolumab?

Giving the patient ipilimumab [3 mg/kg] and nivolumab [1 mg/ kg] in the metastatic setting results in a 50% toxicity rate. So, you are making people pretty sick. To give that combination to a well patient gives oncologists a lot of trepidation.

We actually completed a study that compared ipilimumab [3 mg/kg]/nivolumab [1 mg/kg] and did a flip dose and gave ipilimumab [1 mg/kg]/nivolumab [3 mg/kg]; we saw that this combination resulted in similar toxicity as giving nivolumab alone [CheckMate 511; NCT02714218]. So now you have a doublet that has a toxicity profile of 10% to 15%. This study just finished last year, in which patients in 1 arm received nivolumab and the other arm received ipilimumab plus nivolumab with the flipped dose, followed by maintenance nivolumab. Whether or not that change in dosage with the 2-drug combination is going to make a difference is debatable. Toxicity was not enhanced by the small dose of ipilimumab. I am expecting that we will have those data for the American Society of Clinical Oncology Annual Meeting next year.

What about the COMBI-AD trial?

This is only pertinent if your patient has aBRAFV600 muta­tion. Only then is the patient a candidate for adjuvant dabrafenib/trametinib. Not a lot of evidence exists for using other BRAF inhibitor combinations, and the only adjuvant data are with dabrafenib/trametinib. I am not a big fan of vemu­rafenib [Zelboraf]/cobimetinib [Cotellic] because patients get confused with the dosage for each. Patients have to contin­uously take the vermurafenib, but you only take cobimetinib for 3 weeks. And patients have to take that week off. If you have a patient who is very astute, then it is fine. But if your patient just wants to take the same pills every day and get into a schedule, [it is problematic].

What are the differences amongBRAFV600E, V600K, and V600D mutations?

The E version is most commonly expressed. D and K are still expressed, but their durability to BRAF inhibitors is a little less. When you see aBRAF-mutated melanoma, it usually suggests a more aggressive tumor and the patient is more prone to developing brain metastases, and these develop more often in younger patients.

What is the danger of treating for an unconfirmedBRAF-mutated melanoma?

When we first were testing BRAF inhibitors, we gave them to everyone because we thought every melanoma used the BRAF pathway. Well, we were wickedly wrong.

If you take aBRAFwild-type patient and give them BRAF inhibitors, you will accelerate the growth of their tumors. So if you don’t know [theirBRAFstatus], you are better off not treating. Make sure that you know that they have aBRAFmutation before prescribing a BRAF inhibitor.

What happens if a patient is improperly treated?

Well, it just happened to me. A patient came in for a second opinion. He was treated at an outside hospital and developed 4 brain metastases. The oncologist who treated him read the pathology report incorrectly. He was reading the summary at the bottom that said, “if you have aBRAFmutation” and all he read was “BRAFmutation.” The patient’s results were on top that said “noBRAFmutation.” The patient went from 4 brain metastases to 50 brain metastases in 2 weeks.

Case 2 (continued):

The patient returned to the office in June 2016 (6 years later). On routine follow-up, the patient presented with moderate asthenia that limited his daily activity, without other relevant clinical symptoms. He had an ECOG performance status of 1.

Physical examination was unremarkable. Laboratory findings were remarkable for urea of 70 mg/dL, creatinine of 1.76 mg/dL, aspartate aminotransferase of 59 UI/L, alanine aminotransferase of 52 U/L, gamma-glutamyltransferase of 363 U/L, alkaline phosphatase of 204 U/L, and lactate dehydrogenase (LDH) of 820 U/L. 

A full-body CT scan revealed the presence of pulmonary and hepatic nodules. An MRI of the brain revealed no evidence of brain metastases. He underwent core-needle biopsy of the largest hepatic lesion in segment IVb without any complications. Pathology revealed metastatic melanoma and mutation testing that it was BRAF wild-type. 

Do you test the primary tumor for aBRAFmutation?

You usually do not test the primary tumor; you test the muta­tion.You test the nodal disease or you test when they develop metastatic disease. Most primaries are so small that you do not have enough tissue. You can take a metastasis that is in paraffin and send it for next-generation sequencing and in 5 days you receive the results. This pulls the DNA out. Sometimes you get lucky and you can do an IHC forBRAF. If the IHC is positive, [then you can treat them forBRAF-mutated disease]. The false-positive rate is exceedingly low, but the false-negative rate is exceedingly high.

So even if it is negative, you still want to send the sample for sequencing because the patient might have a V600K or a V600D mutation.

What is the prognosis for this patient? Do you give him PD-1 monotherapy or ipilimumab/ nivolumab combination?

It depends. If you are talking single agent, you are seeing a 40% to 50% response rate with toxicity of about 15% to 20%. If you are talking about dual therapy, that’s a 60% to 65% response rate but a 50% toxicity. Depending where your patient is sitting inside his head…some people say, “Doctor, I want everything. I want the most aggressive treatment that you have.” Those are the people you are going to give the combination to. Then you have the other people who say, “For the extra 15% in response rate, with an increase in toxicity from 20% to 50%...no, thank you.” So there is no single correct answer.

What is the significance of the LDH in this case?

LDH is the poor man’s way of determining if melanoma is aggressive or not. You can have a patient packed with melanoma and their LDH will be normal. However, if they have an elevated LDH, it is an indicator that this is an aggressive melanoma. They could have a handful of tiny lung metastases, but if they have an LDH that is off the wall, that means you are in trouble.

How would the patient’s performance status play into your treatment decision?

Performance status is going to influence your decision. You do not want to give an immunotherapy to people who are debilitated. If an 80-year-old comes in, I am not going to go with the ipilimumab/nivolumab combination. I had the unfor­tunate experience of giving the ipilimumab/nivolumab combi­nation to 2 different 80-year-old patients who were in great shape, and [in both,] it eventually ruined their quality of life. It cured their melanoma, but it ruined their life. One of them used to run 10 miles to work at the age of 80, but now he walks with a cane. He developed such bad neuropathy. We fixed his melanoma, but I do feel guilty every time he walks into my office. I am very biased against giving ipilimumab in full doses to people who are 80 years or older. Now, using low-dose ipilimumab—that is a different combination.

References:

1. Migden MR, Rischin D, Schmults CD, et al. PD-1 blockade with cemiplimab in advanced cutaneous squamous-cell carcinoma.N Engl J Med.2018;379(4):341- 351. doi: 10.1056/NEJMoa1805131.
2. Faries MB, Thompson JF, Cochran AJ, et al. Completion dissection or observa­tion for sentinel-node metastasis in melanoma.N Engl J Med.2017;376(23):2211- 2222. doi: 10.1056/NEJMoa1613210.