施貴寶 (BMY) 2017 Q4 法說會逐字稿

Good day, and welcome to the Bristol-Myers Squibb 2017 Fourth Quarter Results Conference Call.

Today's conference is being recorded.

At this time, I would like to turn the conference over to Mr. John Elicker, Senior Vice President, Public Affairs and Investor Relations.

Please go ahead, sir.

Thank you, and good morning, everybody.

Thanks for joining early on a Monday morning.

We have a lot of information to discuss today, both from our Q4 earnings, our 2018 outlook as well as results from -227.

We are going to be using a slide deck today, which we normally don't do.

So if you are on our distribution list, we mailed it to you about 15 minutes ago.

The slides are also available on our website.

With me this morning are Giovanni Caforio, our CEO; Tom Lynch, our Chief Scientific Officer; Charlie Bancroft, our Chief Financial Officer; and Murdo Gordon, our Chief Commercial Officer.

Giovanni, Tom and Charlie will have prepared remarks and then will be available with Murdo for Q&A.

On Slide 2 is our safe harbor language.

I'll read it real quickly.

The presentation contains statements about the company's future plans and prospects that constitute forward-looking statements for purposes of the safe harbor provisions under the Private Securities Litigation Reform Act of 1995.

Actual results may differ materially from those indicated as a result of various important factors, including those discussed in the company's most recent annual report on Form 10-K and reports on 10-Q and Form 8-K.

The documents are available from the SEC, our website or from the Investor Relations group.

Any forward-looking estimates represent our estimates only as of the date and should not be relied upon as representing our estimates as of any subsequent date.

While we may elect to update forward-looking statements at some point in the future, we specifically disclaim any obligation to do so even if estimates change.

The presentation also contains certain non-GAAP financial measures adjusted to include certain cost expenses, gains, losses and other specified items.

Reconciliations of these non-GAAP financial measures to the most comparable GAAP measures are available on our website.

Giovanni?

Thank you, John.

Good morning, everyone.

We have a lot to discuss today, so let me share an overview of the call.

As you've seen, we've made an exciting announcement this morning on our lung cancer program.

We consider today's results as a breakthrough in cancer research, and we're excited for what this means for patients and for the treatment of lung cancer.

This result is a true example of the innovation that is core to our strategy.

Based on our understanding of this disease, we made bold and innovative changes to our program as the science evolved.

Today's results validate our approach.

I'm really proud of what our R&D organization has accomplished to advance the understanding of biomarkers in the treatment of lung cancer, and Tom will walk you through the details in a few minutes.

We also announced a very good quarter and strong performance overall in '17, which Charlie will share in more detail.

From my perspective, we are starting '18 with good momentum in our business, and we see a number of opportunities in oncology and outside of oncology that we'll be focusing on this year.

And as I started to discuss in San Francisco last month, looking beyond '18, there are multiple growth drivers that position us well for the longer term.

I'll talk more about this a little later in the call.

And with that, I'll hand it over to Tom to discuss today's announcements.

Tom?

Thank you, Giovanni.

If I could ask everyone to turn to Slide 6. Today, we are excited to announce that Study -227 met the co-primary endpoint of improved progression-free survival for the Opdivo-Yervoy combination versus chemotherapy for patients with high tumor mutational burden or TMB regardless of PD-L1 expression.

And remember that we're looking at both histologies in CheckMate -227.

And as we also announced, the DMC has recommended that the trial continue to completion for overall survival in a PD-L1 selected population in Part 1a.

While we plan to present the results at an upcoming medical meeting, what I can tell you now is that this PFS data is highly statistically significant and clinically meaningful.

I strongly believe that this will make a difference for patients with advanced lung cancer.

Let's just put this into context.

When I first began seeing lung cancer patients 30 years ago, the only options were chemotherapy and radiation, both marginally effective and significantly toxic.

Since then, lung cancer has become a group of related diseases defined by distinct biomarkers that drive biology and treatments, EGFR, ALK, ROS1, RET and PD-L1, just to name a few.

Today, we add TMB to this list to define a subtype of patients who clearly derive benefit from combination immunotherapy treatment regardless of their PD-L1 status.

As I've consistently said, Bristol-Myers Squibb is a science-driven company.

These results demonstrate the strength and importance of our translational capabilities and validate the innovative changes we made to the trial design in light of how quickly the science was evolving.

Perhaps most importantly, the results today further validate the role of the CTLA-4 mechanism.

Lung cancer is now the third tumor where we've shown benefits with the Opdivo-Yervoy combination in a randomized trial.

And we look forward to the overall survival analysis from -227 as the data matures in the TMB population.

Now if I could ask you to turn to Slide 7. What I'd like to do is tell you -- or answer the question, why did we choose to include TMB in our analysis plan?

First, from a mechanistic perspective, TMB measures the number of somatic mutations that are present in the DNA of a tumor compared to DNA of normal tissue.

We believe some of these mutations can drive expression of neo-antigens and that these neo-antigens could be more visible to the immune system and, thereby, drive an immune response.

Second, as you know, Bristol-Myers and others have generated data showing improved outcomes for patients with high TMB.

We've shown this across multiple endpoints, and those include response rate, PFS and overall survival; across multiple tumor types, including non-small cell lung cancer, bladder cancer and small cell.

We've shown it for both Opdivo monotherapy and the combination of Opdivo and Yervoy.

This supported our hypothesis around TMB being an important potential biomarker.

Now if I could ask you to turn to Slide 8. Now let me take a step back and explain how we adapted -227.

On this slide, you see the overall study design.

As you know, -227 was originally designed as 2 companion studies, Part 1a, looking at PD-L1 expressers; and Part 1b looking at PD-L1 non-expressers.

As more data emerged, we made 3 important changes to -227 to reflect the evolving science.

All of these were key.

First, we increased the size of Part 1a.

This gave us more statistical optionality and lowered the potential for imbalances in the PD-L1-positive patients.

Second, we added a second part, Part 2 to -227.

This is a study of about 750 patients who are enrolled regardless of histology and across the PD-L1 spectrum, comparing Opdivo plus chemotherapy versus chemotherapy alone.

And the third change we made was to our analysis plan for Part 1, which allowed us to include TMB, and we're disclosing details of that today.

Now if I could ask you to turn to Slide 9. As I mentioned, Part 1a and Part 1b are technically 2 companion trials running at the same time.

However, in discussions with the FDA, we executed an integrated analysis plan that brought together the Opdivo-Yervoy arms and the chemo arms across all of Part 1 in a predefined population of TMB patients.

Turn to Slide 10, please.

Within that integrated analysis plan, we defined 2 co-primary endpoints across Part 1, and that is what we're reporting today.

The first is comparing the combination of Opdivo, Yervoy versus chemotherapy for PFS in a TMB selected population.

And the second is comparing the combination versus chemotherapy for overall survival in a PD-L1 selected population.

If I could ask you now to turn to Slide 11.

For the PFS endpoint, we were able to assess the TMB status of the majority of patients across Part 1, which I'll remind you is more than 1,200 patients -- actually, it's more than 1,700 patients.

And here, you can see that of all those assessed, roughly 45% were above a cutoff of greater than 10 mutations per megabase, which is what we defined as high TMB.

In our study, TMB was evaluated using Foundation Medicine's analytically validated assay, FoundationOne.

Now I ask if you could turn to Slide 12.

We're very encouraged by today's results, which establish TMB as an important new independent biomarker for selecting patients that respond to immunotherapy.

Clearly, we'll be sharing these data with health authorities, and we look forward to presenting the complete results at a future congress.

We believe TMB could be important more broadly.

We've started to incorporate it across our I-O trials, not just for Opdivo and Yervoy but also for our next-generation oncology assets.

In fact, we're studying TMB in over 100 studies in multiple tumor types and programs.

If we could now turn to Slide 13.

As a reminder, what we're reporting today is just the first of many significant data opportunities from our broad lung program.

For CheckMate -227, the analysis for patients from Part 1b is now formally complete.

Part 1a continues for overall survival in PD-L1 patients.

Remember that the overall survival is event-driven.

We know that the interim occurred later than we expected, and we anticipate that the final results will be later this year or early next year.

And Part 2 of the trial is ongoing, with results expected in 2019.

I also remind you that the 9LA trial looking at Opdivo + Yervoy concurrent with 2 cycles of chemotherapy is also ongoing, with results expected in the second half of 2019.

And of course, we have the initiation of our IDO-Opdivo combination trials, with both Incyte's IDO and our own compound, which we're eagerly awaiting to see.

If I could now ask you to turn to Slide 14.

Beyond lung cancer, we have significant data milestones with Opdivo and the combination over the next 24 months, with trials reading out in hepatoma, gastric cancer, small cell lung cancer and head and neck cancer.

As I've said, we are a science-driven biopharma company.

I am particularly excited about our opportunities across the areas of immunoscience, fibrosis and cardiovascular disease, and we continue to accelerate the development of our most promising assets in those areas as well as immuno-oncology.

With that, I'll turn it back to Giovanni.

Thank you, Tom.

I want to reiterate that I'm really proud of what our organization has accomplished.

This is an important result for lung cancer patients, and we are very excited by the opportunity ahead of us.

Turning to Slide 16, please.

2017 was a very strong year, thanks to strong -- superior commercial execution and good progress in our clinical programs.

We're now entering '18 with good trends and momentum in our business.

Moving to Slide 17.

We are starting the year in a very strong position, with good top and bottom line growth in our business.

There are a number of pillars in our business that can grow in '18 and beyond, and I'm really excited about the prospects.

First, Eliquis is a very important and growing franchise for us.

As we exit '17, Eliquis is the leader NOAC in its approved indications and has tremendous opportunity to continue to grow.

With Opdivo, the existing business going into '18 has strong momentum.

We see continued strength across the current indications for Opdivo.

And beyond our excitement about the first-line lung results announced today, we have a number of opportunities this year that we are focused on, including adjuvant melanoma and the upcoming PDUFA for first-line renal with study -214.

With today's results, I'm even more confident that we will play a meaningful role in first-line lung cancer.

And I'm looking forward to continued data readouts later this year and into next year that will build on today's announcement.

And thinking beyond '18, we see a sustained period of growth, driven by important opportunities for Eliquis and Opdivo.

So let me remind you of what those are.

I previously mentioned renal as part of our growth in '18, and it also plays a role beyond this year in what we see as a significant opportunity for Opdivo.

It's an area where we've started to make a difference in second line and see significant growth potential in first line now based on Study -214.

With the first-line indication, Opdivo will have a presence across the continuum of care in renal cancer.

Two other areas with $1 billion-plus potential where we are focused are gastric cancer and HCC, diseases with high unmet need and poor outcomes.

Based on what we've seen from Phase I and II data on Opdivo on these tumors and anticipating data readouts from our Phase III program this year and next year, we believe these are important areas driving growth in the near and in the medium term.

In addition to this, we have important opportunities in small cell lung cancer later this year and head and neck in '19.

Finally, our emerging pipeline across all therapeutic areas is one of the most promising in our history, with several opportunities contributing to sustained growth.

We are excited with several programs moving forward, such as IDO and LAG-3 in oncology and from our known oncology pipeline, where we'll see FGF21 and TYK2 moving to the next stage of development this year.

All of what I've just described gives me confidence in our future and in our ability to continue to deliver on our strategy.

The significant opportunities ahead of us requires to ensure the right level of investment behind R&D and certain commercial capabilities, and we are doing just that.

And turning to Slide 18, please.

A year ago, I spoke about the work we were doing to evolve our operating model.

And today, I'm pleased to say we are delivering across the company with a disciplined approach to resource allocation.

We are creating a better company which moves fast and is more competitive.

We are strengthening our capabilities, particularly in translational medicine.

And we are investing in our pipeline and commercial capabilities to support future growth.

This will continue to be a critical focus going forward.

With that, I'll hand it over to Charlie, who will discuss some specifics on the quarter and how we are thinking about our financials.

Thanks, Giovanni and Tom, for your very exciting comments.

All caps off for my Philadelphia Eagles winning the Super Bowl.

Let me start by saying we delivered a very good 2017 with robust revenue and earnings growth, driven by strong execution across the company.

Moving to Slide 20.

While total revenues grew 7% for the year, our prioritized brands were up 27%, as Giovanni just highlighted.

The exceptional product performance on key brands drove the EPS growth.

With that, I'll provide some color on the strong trends we are seeing for Opdivo.

It was another solid quarter for our Opdivo franchise.

We delivered $1.4 billion of worldwide revenue and almost $5 billion in sales on a full year basis.

Similar to previous quarters, we were successful in maintaining our leading share in second-line lung and saw strong performance in other tumors such as renal cell.

While the trends are early, Opdivo has rapidly penetrated the second-line HCC market, and we are seeing good uptake of Opdivo in the adjuvant melanoma setting.

Outside the U.S., Opdivo continues to lead in key markets such as Germany, France and Japan.

Excluding the sales deferral impact in Q4 of 2016, international sales of Opdivo were up over 64% in the quarter.

With these strong Opdivo trends and today's announcement, we see meaningful opportunities for growth in both the near and medium term.

Turning to Yervoy.

We are seeing pressure on U.S. sales due to the adoption of Opdivo in the adjuvant melanoma setting.

Going forward, we expect these trends to stabilize, and our outlook remains positive with potential launches in first-line renal cell and first-line lung.

From a franchise perspective, we expect the growth for Opdivo in adjuvant melanoma will more than offset the erosion of Yervoy.

Our strong commercial execution across the portfolio also resulted in substantial growth for Eliquis in the quarter.

In the U.S., Eliquis extended its leadership position with almost 50% TRx share of the NOAC market.

Internationally, we are seeing similar trends and strong brand momentum in leading new-to-brand share in top European markets.

With this performance in mind, we believe Eliquis is well positioned for a period of sustained growth going forward.

Now turning to Slide 21 and our non-GAAP P&L.

I'll start with our gross margin, which continues to be strongly influenced by product mix.

The strong growth of Eliquis and decline in our virology franchise drove most of the pressure on our margin in the quarter.

Moving to OpEx and building on Giovanni's comments about resource allocation.

We increased investment in R&D during 2017 while prioritizing spend in MS&A.

This was enabled by our ongoing operating model evolution that we will continue to execute going forward.

With respect to our tax rate.

The favorability in the quarter was primarily driven by earnings mix.

And thinking longer term and taking into account tax reform, we expect our tax rate to be in the high teens within the coming years.

With tax reform in mind, turning to Slide 22, we see no change to our balanced approach to capital allocation.

Business development remains a top priority for us, and 2017 was a very active year in which we executed over 50 transactions.

We expanded our translational capabilities.

We licensed new assets and technologies, and we entered several late-stage clinical collaborations.

The IFM acquisition, our partnership with Foundation Medicine and our new relationship with Halozyme were a few important examples of transactions we executed last year.

We also remain committed to our dividend.

2018 marks our ninth consecutive annual increase.

With respect to share repurchases, we bought back $250 million in the quarter and nearly $2.5 billion for the year.

Taken together, we returned over $5 billion to shareholders in 2017.

Turning to Slide 23 for additional color around 2018 guidance.

Our full year non-GAAP EPS guidance represents strong growth over 2017 in spite of the roughly $200 million impact from the change in accounting rules affecting how certain royalties are recorded.

On gross margin, we expect additional pressure to be driven by the continued growth of Eliquis and erosion of virology, to be somewhat offset by the growth of Opdivo.

Our approach on OpEx reflects the continued prioritization in R&D while we drive efficiencies in MS&A.

In conjunction with today's announcement, our guidance includes investments supporting the educational efforts to commercialize TMB.

A quick note on OI&E.

We have restructured a portion of our future AZ royalty rights, and therefore, we'll receive higher royalties in 2018 and 2019.

Our tax rate of 20% to 21% takes into account tax reform and, as I mentioned, the potential for further favorability over the next few years.

In conclusion, we had a very solid 2017 with strong execution across the board.

We are well positioned to grow in 2018, and today's announcement increases our conviction in the growth outlook for the company.

I'll now turn it back to John for Q&A.

Thanks, Charlie.

And I think we're ready to go to the Q&A session.

So as a reminder, we have Giovanni, Tom, Charlie as well as Murdo here for any of your questions.

Go ahead, please.

(Operator Instructions) And we'll take our first person from the queue, Seamus Fernandez from Leerink Partners -- sorry, Alex Arfaei from Bank of Montreal Capital Markets.

I'm just curious, why didn't you -- why is overall survival not being evaluated based on TMB as well?

And we know from CheckMate -026 that Opdivo monotherapy showed very encouraging results in this setting.

How come Opdivo monotherapy is also not being included in the TMB analysis?

Alex, thank you.

I'll take both of these questions.

And I think let me just take the second one -- I can do the second one probably quicker, then we'll talk a little bit about the endpoints in the first one.

So just to remind everybody, what we're reporting today are the 2 co-primary endpoints.

So we're reporting the PFS in the combination therapy, which is a co-primary endpoint in the TMB-high group.

And then we're reporting the fact that the second co-primary endpoint, which is a test for overall survival in the PD-L1 group, is continuing for maturity.

We have not seen any overall survival data.

I think that's very important to mention.

So we have not seen any overall survival data at all.

And the second thing to say is we're not going to be reporting any secondary endpoints on the call.

The data is very fresh to us.

We look forward to reporting those at an important medical meeting, and we also look forward to publishing that as soon as we possibly can.

Let me just say a few things about endpoints and trials.

So a couple of things.

I think as a physician, we think that -- I think that PFS and overall survival are both important endpoints.

Doctors use both endpoints, PFS and OS, to be able to make decisions.

I will also say, when you have PFS particularly in a very large, well-powered trial, you are able to get a very good sense of potential benefit, and that does predict good benefit with patients.

So we're very encouraged by the PFS benefit today that we see in the TMB population.

I really want to stress something else about this result.

This is highly statistically significant, and it's clinically significant as well.

And I think that's an important point to drive home with the PFS.

We look forward to seeing the OS data when it matures in the TMB group, just like I'm sure you are excited about seeing it.

And we think that, that will be something that will also be important for doctors in terms of making decisions.

I also want to stress one other point, Alex, is this was across all PD-L1-expressing groups.

And I think one of the other things we think is important is that, right now, PD-L1 is an important decision-making factor for doctors when deciding how to treat people with lung cancer, and there are some people who are not getting immunotherapy because they're PD-L1 negative.

I think what TMB gives us is the opportunity to identify patients who clearly benefit from low-dose Yervoy added to Opdivo in that setting.

We could not be more excited to tell you about that today.

Thanks, Alex.

And now we'll take our next question from Seamus Fernandez from Leerink Partners.

Well, first off, let me echo the sentiment on the Eagles from a Philadelphia fan in Boston.

It was a unique event.

So thanks for that comment there, Charlie.

Second, in terms of the difference in the study design, congratulations on the creative change to incorporate TMB.

Just hoping that you could just give us a little bit of color on whether or not you guys have a relatively complete understanding of the benefits of the combination over Opdivo monotherapy.

I know that you did comment a little bit on that, on the margin.

But the clinical significance -- is the clinical significance applied more to TMB or -- and the analysis therein or the combination of Opdivo + Yervoy?

And then secondarily, as we think about the longer-term opportunity, Giovanni, just hoping that you guys could give us a little bit of color on how you're feeling about the Opdivo + Yervoy combination as a continuous treatment given some of the safety concerns or at least the safety information that we're seeing out there.

Again, the clinical significance, I think, is clear, but we're just trying to get a better sense of directionally how confident you are that the Opdivo + Yervoy regimen, particularly as a continuous combination therapy at the 6-week treatment dynamic, is something that you think is going to really work its way into the lung cancer regimen over time.

So Seamus, thank you.

And I've got to say, as a Patriots fan -- diehard Patriots fan in this sea of Eagles fans here, it's been a rough morning.

If we didn't have -227 report, I don't think I'd get through it.

So let me just address a couple of your points here that I think are well crafted and are important.

The first thing I'll say is that I feel very confident that this is both a TMB story and an Opdivo-Yervoy story, okay?

And I think both are important because, first, understand the biology, selecting the patients who are going to benefit most, TMB looks like a very powerful biomarker.

As I mentioned, we're looking at it across our spectrum of I-O agents, including our next-gen agents.

The second point comes down to the contribution of components and how important Yervoy is.

And for me, I have felt for some time that Yervoy is performing extremely well in a number of different settings: remember, benefit in melanoma; benefit in renal cell cancer; and now benefit in this case in non-small cell lung cancer, it's TMB high.

And what I can tell you is that our analysis of the data thus far makes us very confident that Yervoy is a big part of what we're seeing today.

It's not just Opdivo, but that Yervoy is part of this as well.

And I want to mention, as you said, it's low-dose Yervoy that's given every 6 weeks and has that advantage in terms of being well tolerated by patients as we move forward.

So I think that it's not just TMB or the combo, it really is a story about both.

And I'll turn it over to Giovanni for his thoughts.

Seamus, this is Giovanni.

So I just had a couple of comments.

So first of all, I want to strengthen again and reinforce what Tom was saying about the importance of the combination of Opdivo and Yervoy.

This is the third tumor type in which we have seen data that makes us really excited about the potential of this combination, melanoma, renal, now non-small cell lung cancer.

The second thing that I want to say is that I'm quite proud actually of the work that we've done in optimizing the regimen.

And I think that's really important.

The dosing schedule of Yervoy as part of the combination are really important.

Based on the totality of the data that we've seen so far, our confidence in our broad lung cancer program has only increased.

And I think that's important because, in parallel, it also strengthens overall the growth outlook for the company.

So obviously, it starts with patients, but I think this is a really important time for the company overall as well.

Thanks, Seamus.

And now we'll take our next person, Jami Rubin from Goldman Sachs.

Just a couple of questions.

Murdo, maybe for you first.

If you can address the commercial implications of TMB.

Just curious what payers, KOL, doctors are saying about TMB.

This is new and exciting.

But how typical is this in their practice?

And if you could talk about the size of the front-line lung market affected by patients with high TMB.

Your press release said that 45% of patients expressed high TMB, but is that 45% of the total market?

Or is that 45% of a share of the market?

And I ask because when we go back and look at CheckMate -026, I think you evaluated 60% of the population.

So if you could put that into perspective.

And then just lastly, maybe for you, Tom.

How confident are you now in hitting overall survival in -- for Part A -- Part 1a of the study?

And did I hear you right?

I think earlier you did say that -- in an earlier question, that you would also look at OS in TMB, although I wasn't clear.

And then just lastly, is this trial a registrational trial?

Thanks, Jami.

First off, I'm also excited that we were able to announce what we've been able to do today.

Clearly, following the science and looking at tumor mutational burden as a biomarker has been the right scientific approach in first-line lung cancer, and congratulations to Tom and his team for being able to deliver on that.

We're also, in the commercial and medical area, very excited about TMB.

It appears to be a highly predictive biomarker.

As you know, it's -- the FoundationOne test was recently FDA approved.

And in parallel to that FDA approval, they're pursuing a national coverage determination for CMS Medicare coverage and reimbursement of that test upon indication of that in -- as a companion diagnostic.

So we're very, very excited about that.

In terms of size of the front-line lung cancer population, clearly, you know the design of -227.

We designed this trial to exclude ALK- and EGFR-positive patients.

So the 45% number applies to the TMB patients that were tested in the trial.

And I think you alluded to the ascertainment rate.

While it was 60% in the trial, we have a lot of confidence that, that will rise with improvements to methodologies, pathology techniques as well as surgical techniques on the biopsies.

So we're hopeful that, that will improve over time.

And I would say that we're very ready for this.

The team at Bristol-Myers Squibb has been focused on the translational science, but also on helping educate physicians.

I would say in the academic centers, there's a high degree of awareness of TMB, and there's already quite substantial testing today.

I would say it gets much lower than that when you go out into community oncology, and that's where we're focusing a lot of our educational and continuing research efforts.

We have an ongoing safety clinical trial in first-line lung cancer looking at patients with Opdivo and low-dose Yervoy being treated after having their TMB assessed, and that's being conducted through our IO-ICON network.

We have a great partnership with a large network in community oncology.

So we'll help to be able to generate data and describe the safety profile of the drug in a much larger community-based population.

And then, of course, as we pursue regulatory approval, the commercial organization will be ready to focus on a number of different important audiences, inclusive of the ones you mentioned.

Lastly, I would say, from a payer perspective, this is a very positive event.

I think both clinicians and payers have been looking for a better way to segment the first-line lung cancer market.

And I think with the advent and exciting information that we have, that we hope will be presented, as Tom said, in an upcoming scientific congress and published thereafter, we hope that we'll be able to go to payers in the U.S. and around the world and say this is a patient population that discretely benefits from Yervoy -- a low-dose Yervoy plus Opdivo.

So we're very excited about that in the commercial realm.

Thanks, Jami.

So Jami, let me just echo something which Giovanni said.

I think when you have a dataset like this, which is highly statistically significant and most importantly -- probably more importantly, clinically meaningful, I think that does increase our confidence in a number of different elements of our entire lung cancer program, not just Study -227.

Remember, we've got study 9LA, which is coming after this.

We have Part 2, which is looking at the chemo combo, and then we have our IDO studies.

So I think that's an important finding today.

I think the second thing you asked is about whether we'll be looking at OS in TMB.

And as I mentioned earlier, we will be looking at overall survival in TMB.

As you know, that's event-driven.

And when the data matures, we look forward to sharing them broadly at that point.

I think the second question you asked is about your -- our confidence in hitting overall survival in Part 1. As you know, we have not seen any overall survival data.

The hurdle for stopping a study early is usually extremely high, and again, it's event-driven.

And we look forward to seeing that data when the required number of events occur at that point.

And your final question came down to the issue of do we look at this data as registrational.

I think, as we said earlier, we are eager to share this information with regulators in the United States and in the EU because -- I think Murdo put it very nicely.

We think this is going to make a big difference for doctors and for patients.

And as someone who's treated lung cancer for 30 years, to me, this kind of data is the kind of data that doctors are asking for.

They want to be able to know which patients are going to benefit, and I think it's going to have the opportunity to really change the way we think about non-small cell lung cancer.

Thanks, Jami.

Our next question comes from Andrew Baum from Citi.

Three questions, please.

Obviously, we're operating in the dark, so we can't see the -189 or -227 or the mono data from -227, but with that in mind, 3 questions.

Number one, is [it going] to be enemy of the [great] care?

And what I mean by that is, is the low friction associated with Keytruda chemo problematic in persuading clinicians to adopt a more expensive and, arguably, more toxic therapy, which requires educational support?

And then along the same lines of that, in terms of friction, you mentioned the need for continued education but also the trial date turnaround for TMB.

To what extent can that be addressed?

And how quickly can we migrate to the liquid TMB assays there?

And then the last question is in terms of cutoffs, which picks ups on one of Tom's point about the relevance of TMB to identify patient populations in other indications.

Do you have a firm sense that this cutoff is transferable across indications?

And [structurally], this is not a (inaudible) decision as anything driven by science in terms of the identifiable number.

So is there consistency here?

Or is it going to be on an indication-by-indication basis?

Murdo, why don't you start with commercial and Tom can go?

Yes.

Thanks, Andrew.

I think the way I would approach my view of the first-line lung cancer market is we've known for a while that PD-L1 expression has been an imperfect biomarker in being able to stratify and select patients.

Now with this very large dataset, we've been able to establish the value of a highly predictive biomarker in TMB, and we have a compelling profile of Opdivo plus low-dose Yervoy as a result.

And I think that, that really is an opportunity for us to redefine the way in which clinicians, payers and patients think about first-line lung cancer.

So we're excited about that.

I think there will still be a role for PD-L1 expression to play, to understand where other options and other treatment modalities should be used.

But I think the compelling nature of these data and the predictive nature of TMB will allow us to establish a very good presence in the market.

I would also say that the partnership with Foundation Medicine has been a good one for us.

We've worked very closely with them to understand the logistics of how TMB testing and patterns of testing are currently occurring.

I would say that academic centers are doing very well in turnaround time in tissue ascertainment, and I think that, that will propagate into the community.

It will take some time.

And obviously, when physicians see a very compelling clinical profile, I think that, that is likely to lead to an acceleration in testing and an improvement in turnaround times, although our hope is that we would establish testing for TMB as a panel type of test, as a test that occurs early in the diagnosis of a patient so that it is available to the primary treating physician early in that treatment decision-making process.

And that's our goal really going forward.

So the turnaround time today will hopefully not be an impediment as we go forward.

For the liquid biopsy question, I'll turn it back to Tom.

So Andrew, thank you for your question.

I think one of the things that I would start off by saying is that I'm incredibly proud of our translational capability, as Giovanni mentioned in his remarks, and how we approach this.

Let me give you a little sense of how we got to a cutoff of 10 for TMB.

What we did was we looked at a number of studies that we have done in lung cancer.

We looked at Study -012, Study -568, we looked at Study -026.

And then we looked at different cut points of tumor mutational burden that were able to best segregate outcomes and best predict for who is going to have better response and better benefit than others.

And we arrived at a cut point of 10.

Now I have to say, let's stress something else.

This is very early in the understanding of this biomarker.

This is the first important report of it applied to a randomized trial.

And remember, this was a prospective analysis that we did using archival tissue, but it was -- the analysis itself was prospective.

And I think that's important to keep in mind.

I think you have to think about other biomarkers.

Look at how we understand how to use Herceptin in patients with breast cancer.

Was it 2 plus?

Was it 3 plus?

And then it became understanding that amplification was the key issue.

And so it really is early in the process.

And while I think data I've seen at 10 is going to be the number and I hope 10 is going to be the number forever, can I promise you that someday we're not going to find out the number is really 8 or 12?

And as other companies and other investigators look at this in their datasets, might we learn more?

Absolutely.

And I remain open to collaborating across investigators in that respect.

I think the second thing about the test that's important is this is a highly validated test.

While the number might be something with time that might get adjusted for indication by indication, the test itself is highly validated, and on November 30 of 2017, it was actually approved by the FDA.

So while we need to work with Foundation on the companion diagnostic indication, and we'll obviously work with them and the FDA at that point, the test itself is solid, and we feel very good about that.

And that brings us to the last question you asked, which is when do we think blood-based assays will come in.

And I think there's great promise for blood-based assays.

I think that will make this process much easier for patients.

On the other hand, we don't quite have them yet.

We're close.

And again, we look forward to working with a number of diagnostic companies that have insights into how we can create better ways of looking at TMB.

What I can tell you today is I think that our findings in Study -227 may increase the imperative and the urgency of coming up with better ways of assessing TMB, and I think blood-based biomarkers will be one of them.

Thanks, Andrew.

Our next question comes from Chris Schott from JPMorgan.

Congrats on the data.

I'm just -- a question here just on this TMB dataset and the competitive landscape in front line.

When you consider the other datasets, some of these all-comer chemo combo datasets out there, what's the hurdle that you consider as you think about the commercial kind of opportunity here?

Basically, do we need to see improved hazard ratios in this TMB population relative to some of those overall competitor datasets, the chemo combo datasets I'm talking about here, to see physicians adopt?

Or do you think there's going to be a debate here of some physicians preferring kind of IO-IO versus IO-chemo?

So I'm trying to get a little bit of sense how you're thinking about that.

My second question was just clarifying an earlier comment on filing.

I guess, do you think there is a potential pathway to file this data earlier than that the late 2018, 2019 readout from the OS arm of the 1a study?

I'm just trying to understand a little bit again the filing dynamics there.

Chris, let me start there, and then I'll ask Murdo and Tom to jump in.

So first of all, the progression-free survival was a co-primary endpoint of the study.

And so given the strength of the data, as Tom mentioned, we will discuss that with regulatory authorities, but that was the co-primary endpoint of the trial.

And the TMB part of the study, the data is -- that data is mature and ready.

With respect to your question about competitiveness, I think what's important here, obviously, Tom has said many times, and we believe that lung cancer is a very heterogeneous disease, is a very broad set of diseases.

In fact, 30% of it is treated by physicians in the academic and hospital setting.

Up to 70% is treated in the community.

So it's obvious that, depending on competitive dataset and then our own data, there will always be physician preferences and patient profiles that are more -- indicates different treatment options.

But I think what's important here is the ability to identify a patient population that is -- that has a potential to respond to a treatment strategy with a regimen of Opdivo + Yervoy.

What's also important here is the ability to identify patients that are not likely to respond, and both are really critical.

So I think the biomarker here is very important.

It identifies a population of patients that should be treated, in our mind, based on the data with Opdivo + Yervoy.

And it also helps define which patients actually may -- which patients may benefit from a different treatment strategy because it is clear that for patients with low TMB, going for a combination of immuno-oncology agents may not be the right strategy.

So I think there is clearly room in this market for multiple treatment strategies, but the data is very compelling.

Yes.

Thanks, Giovanni.

I would also echo, the way the market is right now, there's a lot of questions on the mind of treating physicians on what to do in lower PD-L1-expressing patients in terms of treatment options available.

We know what the Keynote-021G dataset said.

We haven't, as you mentioned, seen the -189 dataset, but there's still a lot of patients in the market who have low PD-L1 expression who are not receiving an immunotherapy option.

With the TMB biomarker, we're able to look across all PD-1 expression and enrich for a population with TMB.

And we've been able to, in this clinical trial, demonstrate a benefit in PFS with Opdivo plus low-dose Yervoy.

And I think that's a very compelling treatment option for clinicians going forward.

The combination of PD-L1 and TMB may prove to also be, as Giovanni said, a very useful way to select patients out of immune checkpoint inhibition.

But given the very broad program that we have with -227 and other clinical trials that Tom has described, we will be able to answer many more of those questions going forward.

And I'm excited about that.

I think that over the next 1.5 years, 2 years, we will be able to help physicians out there understand exactly when and where to use what treatment option.

And I'm also pleased that we are developing all of those treatment options within our own portfolio.

Tom?

I just think I'd echo those 2 comments.

And again, I think that I look at this as being data that we eagerly look forward to sharing with the regulatory agencies.

And I think that, that answers your question about how we're going to proceed.

Thanks, Chris.

The next question comes from Tim Anderson from Bernstein.

I just want to stay on this issue of file-ability of this data.

So Bristol itself has commonly said that PFS is an imperfect predictor of OS.

And I think we've seen that in certain datasets over time by different companies including Bristol.

And if I think about TMB in -026 at least, what was published as a later analysis on OS, there wasn't -- there was a big PFS benefit looking at TMB stratification, but there wasn't much of an OS benefit.

And if I -- Tom, you kind of said we're very early in our understanding of TMB as a biomarker.

So without having OS data, I just -- I guess I'm still struggling to see how this is data that the FDA would accept for approval essentially.

And if you could maybe just clarify your comments earlier on the biomarker saying we're very early in our understanding, but at the same time, you're saying it's very validated.

So maybe that's just a technical point about validation.

And then last question is, have you seen any Opdivo monotherapy data out of -227 such that you can put Opdivo + Yervoy into context?

So Tim, thank you.

Let me just -- I can answer the second question really quickly.

And as I mentioned today, we're commenting on the 2 co-primary endpoints.

And there are a number of secondary endpoints in the study, and we look forward to sharing those at an upcoming medical meeting and publishing those and going over those in detail.

So to the first question, which, again, it gives -- I thank you for the chance to clarify my answer to Andrew's question earlier.

I -- when I'm saying we're early in the understanding of TMB, I'm not saying that we're early in understanding how to measure it and the validity of measuring it, okay?

So the FoundationOne test has been highly validated, and it's been FDA approved as a test, meaning when you test somebody's tumor for TMB, that answer you're going to get is reproducible and is meaningful in terms of being a validated test.

This is not a test that does not have the rigor of other areas.

What I was responding to Andrew's question on the question of the cutoff, could the cutoff be different in a different cancer type or different tumor type?

Sure, it could.

Okay?

We don't have enough data yet.

As I mentioned to you, we're looking at TMB in more than 100 different studies that we're doing.

And is it possible that we'll find that the TMB for colorectal cancer might be different or pancreas cancer might be different or gastric cancer might be different?

It might be.

And that's what I meant by saying we're early in TMB, not that we're early in this scenario of -227.

So I think that's important.

The second question comes down to PFS versus OS.

Now as you know, okay, many, many drugs have been approved based on the PFS.

In fact, our own -067 experience in melanoma was approved first based on PFS before we had the OS data.

And as you also know, Tim, frequently, almost always, PFS data comes before OS data comes.

Not always, but often it comes before that in this setting.

And we look forward to seeing the OS data on this trial.

We have not -- as I mentioned to you, we have not seen any overall survival data from -227 yet, and we have as much curiosity as you do about this.

What I will say is that one of the things doctors do is when they look at a study with the size and robustness of this trial, this was -- the total number of patients on Study -227 is nearly 2,400 -- 2,500 patients.

And so it's a large trial with a result which is highly statistically significant and clinically meaningful.

And in that setting, I believe, particularly when it's meeting an unmet medical need, when there are patients out there who are not getting I-O therapy, that we have excellent data for now that shows can benefit from I-O therapy, I think that's a very compelling case for why this matters.

Thanks, Tim.

Next question comes from Jeff Holford from Jefferies.

So just 3 points here.

I don't know if you're going to be able to tell us anything now about PD-L1 status.

I assume that's going to be a subgroup analysis within this.

I remember that when we looked at CheckMate -026, the high TMB, PD-L1 greater than 50% was a stunning result there.

And maybe perhaps what we're going to be looking at here is something like breast cancer where physicians get more than 2 points on a matrix like node status and hormone receptor status.

Perhaps it could be tumor mutation burden and PD-L1 status that really drives these -- the combo.

So maybe some comments on that.

Second, do you think the rationale for tumor mutation burden also works in chemo combo and triple combo with the Opdivo, Yervoy, chemo?

There's no reason why that rationale shouldn't work just as well.

And then last -- few questions about the file-ability of this data here.

Do you not think that the Keynote-021G approval acts as a strong precedent for file-ability here?

So let's take these a little bit -- we'll take them by -- in order.

So first, I think, Jeff, your comment about PD-L1 status and TMB and the interaction between the 2, I think, is a very interesting one.

We've just got this data, so we look forward to looking at that and looking at many of the other potential understandings and relationships we can have once we have some time with the data to take a look at that.

But I do think there's a possibility that, that might help to find populations that might benefit.

I think one thing Murdo said earlier and Giovanni echoed as well, which I think is really important, which is you could imagine a scenario where a combination of biomarkers be it PD-L1 and TMB, could define patients who do not benefit from I-O agents at all and might say these are patients we should treat with another type of experimental therapy or chemotherapy in that setting.

So I share your enthusiasm for looking at that interaction between the 2 biomarkers, as we will look at the interaction among all the biomarkers that we look at in lung cancer.

Second point you raised, Jeff, was the question of rationalizing -- whether TMB will hold with chemo.

Again, as I mentioned, we're looking at this in more than 100 different scenarios.

And again, we just haven't seen data yet in that setting, and we look forward to seeing that because I think that will add more color to the patients and how patients benefit in that setting.

And then the final question of file-ability.

As you know, we don't comment on our relationships or actual discussions with the FDA, and we certainly don't comment on the strategy that other companies might have used in getting their trials approved.

What I will say is that we do look forward to sitting down with the FDA and the EMA and sharing with them this dataset because we think it has the potential to be highly significant for patients.

Thanks, Jeff.

Next question is coming from Umer Raffat from Evercore ISI.

John, I figured everyone else is asking multiple, so I'll ask multiple today as well.

First, I guess, the first and the main question I have is, I just wanted to understand how you guys went about changing the stat plan.

Like what informed that decision?

And specifically, what informed these specific sets of changes, that let's pool the Part 1a, Part 1b; let's put only PFS but not OS and TMB positives?

Just curious about that one.

Secondly, I know the slides say the CLTA-4's role is "validated." I was just curious if you could explain how.

Presumably, you've seen the PFS curves on combo versus mono and TMB high as well.

And then finally, I noticed Foundation Medicine defines high TMB as 20 mutations per megabase, I mean, granted all this is still in flux.

And I know, you guys are using 10 mutations per megabase.

So I guess my question is, how does your data look for combo versus chemo in 10 to 19 mutations per megabase versus 20-plus mutations per megabase?

And I only asked because EMA has been doing some of this in prior regulatory reviews, for [Nivo FBM], for Merck on Keynote-021G.

So Umer, thank you for your questions.

So I'll start off with the first one, which is how did we change the stat plan and why did we change the stat plan.

And so just to give you a sense, the data in lung cancer has been developing and emerging incredibly quickly, and this -- the way we -- our understanding of the biology of lung cancer has improved.

So we look to Study -026, and I don't know if you guys know this, but Study -026 didn't exactly hit what we are hoping for in that setting.

So we looked at -- in great detail about why didn't it work and what were the differences and let's do a whole exome sequencing of our patients on Study -026 to be able to look in great detail to understand what about our patient population might have been different.

And it was while we were doing that, that we began to understand, based on some data with Yervoy and tumor mutational burden in the past and other emerging data that was being generated by our translational medicine group, that TMB might turn out to be an important marker, okay?

And that's how we came upon TMB.

And then we looked at -568, which is a practice and forming study of about 300 patients in lung cancer, and that gave us more depth of data to look at TMB in lung cancer.

And the same thing with Study -012.

And so those 3 things said, aha, this looks like this might be an important biomarker.

So we changed the stat plan to enable us -- to give us the size -- remember, we nearly doubled the size of study -- of Part 1a of the trial.

That was really important to be able to do, which gives us the statistical optionality to be able to look at a marker like TMB while still preserving the ability to look at OS in a PD-L1-defined patient group.

So that's really, I think, the background for how we got there.

I think what I can say about CTLA-4 being validated in this setting is from our look at the data we've seen so far, we believe that Yervoy is a very important part of the story, and we think that Yervoy is clearly critical to the benefit in this TMB-high patient population as well, say, at this point.

And your last question is really interesting, regarding Foundation might say 20 is high, what happens between 10 and 19, what happens between 6 and 10.

Those are all excellent questions, Umer.

We just haven't yet gotten to the point where we've been able to look at all the variation across it.

What I can tell you is, when we look at the 3 datasets, I told you about -026, -012 and -568, we found 10 was an endpoint that distinguished in lung cancer, using the FoundationOne assay, outcome.

And then when we prospectively analyzed that in a patient sample using archived tissue, we were able to show this result.

I can't tell you that patients who got 25 don't do even better, they might, or that patients who have 7 might also do well.

And again, those are all things, I think, we're going to learn with more time.

Umer, thanks for the questions.

Next person is Geoff Meacham from Barclays.

I just have a couple.

On TMB, what work have you guys done in other tumor types beyond lung?

I guess, at this point, what do you think is the next step in tumor types that you already have formal approval?

Do you think of redoing that?

Or do you think the more logical path are indications where you haven't seen more I-O activity?

And then on -227, Tom, as the data for Part A and -- Part 1a and b mature, how important do you think overall PFS and OS are on an ITT basis, not the TMB segment?

Geoff, just a first comment on your first question.

As Tom mentioned, we are measuring TMB in over 100 trials.

You will remember that one of the first trials in which we started stratifying for TMB status was 9LA.

That's included in the study design.

We're looking at TMB across multiple types of tumors, lines of therapy.

And obviously, as our clinical development program evolves, we will learn much more about the role of TMB in selecting patient populations outside of lung cancer.

And Geoff, thank you for the question on -227.

I think a couple of things I want to just emphasize on what we have found and what we haven't looked at yet.

So in a group of patients with high TMB, regardless of PD-L1 expression, and remember these were also across histology, so including both squamous and non-squamous, the addition of low-dose Yervoy to Opdivo improved PFS in that setting, okay?

What we haven't reported is the overall survival from a PD-L1 selected group, and you'll be seeing that data down the road.

And there's really not much more we can say about the survival data.

I'd love to have had the answer now as well.

I'm sure you would as well.

We just don't have that data yet, and we look forward to having that dataset.

Again, I keep coming back to the fact that there's a -- that we're still early in our understanding and there's a lot of data that's going to emerge not just from our studies in lung cancer, but from other investigators working with other agents in this disease.

And we're going to keep learning and we'll keep following that biology as we did in this circumstance to improve outcome.

Thanks, Geoff.

Next question comes from Vamil Divan from Crédit Suisse.

Just 2 more around -227.

So just following up, I think, on prior comments, and I'm still not clear on these.

So one, it sounds like very positive comments and data around TMB and PFS.

I'm still wondering why you decided to stick with PD-L1 then as a driver for the OS analysis for Part 1. Why -- when you were making the changes, why did you not change that also to TMB?

I know you can't talk about secondary endpoints right now, but maybe if you can just clarify why the decision was made the way it was.

And then the second part, you mentioned Part 1b is now finished.

And I'm curious as to why is that not continuing to see the OS data, why are we stopping that one -- why are you stopping that one now?

Was it something that you saw in your analysis?

Or was that always the plan, as you adjusted your stats plan, to not look for OS benefit in that population?

So Vamil, thank you for your questions.

I'd say a couple of things to answer them.

The first question is -- Part 1b is finished because we used patients from Part 1b that were PD-L1 negative but high TMB there in the TMB analysis, okay?

And the second thing is we then designed Part 2, which is asking the Opdivo, chemo question across a broad group of patients regardless of their PD-L1 status, and that's the way we designed the statistical analysis.

I think your question on -- the first part of your question on overall survival for TMB versus overall survival in the PD-L1 selected patient population, because this study is so large, that's one of the things -- again, one of -- when people asked earlier how did we change it, why did we change it, we wanted to have optionality to look at a couple of important questions.

And while we reported PFS today on this, we just haven't seen the OS data for TMB.

And so it's entirely possible that we will see OS data from TMB, and we look forward to seeing that data, but we just have not seen that data at this point.

We're waiting for that dataset to mature.

At the same time, we are -- we have this preserved ability to look at overall survival in a PD-L1 population.

So I think that those are really important points to get across.

And Part 1b was the smallest of all the 3 parts of Study -227, and Part 2 has almost 750 patients.

So if you're wondering about the chemo question with Opdivo, that will be answered in Part 2 of our study.

Thanks, Vamil.

Next person is Marc Goodman from UBS.

A couple.

So first of all, I just want to make sure I understand.

Remind us about -026.

And when you described that you hit on PFS but not on OS, why, in this case, you feel confident that -- what are the key differences so that we can have more confidence that you'll hit on OS eventually here?

And then second, each quarter, you guys are pretty good about giving us Opdivo sales and the breakdown of where it comes from.

So if you could that this for the quarter, that would be helpful.

And then third, you mentioned of FGF, I know we've been waiting -- did you actually talk to the FDA?

Have you signed off on Phase III?

When will you begin?

And what should we be looking for as endpoints there?

So Marc, let me start with FGF21 maybe.

As you know, we had discussed and we were in conversation with the FDA to start the next phase of development for FGF21.

One of the things we discussed with the FDA was the fact that we had -- in our original studies, we had not included a posttreatment biopsy of different doses.

So we've decided to go ahead with a study that will be looking at posttreatment biopsies before we move into a large Phase III study, and that study is starting right now.

So Tom?

Just to quickly address the question on the -026.

I assume you're talking about the -026 retrospective look that we did, that we presented at AACR last year that showed a PFS.

But you could separate them out by PFS based on TMB in that setting.

There was a lot of crossover, remember, in -026, and it was getting used to generate hypothesis-generating data.

It was not a predefined, prespecified prospective analysis like -227 was.

Murdo?

Yes.

Thanks for the question, Marc.

For Opdivo sales by major tumor type, and I'll give you U.S. numbers and I can give you worldwide after that, but we currently enjoy roughly 44% to 50% of our business from lung.

That's all lines because we continue to get a bit of off-label first-line lung cancer usage, particularly in squamous; in renal, about 18% to 22%; head and neck, around 5% to 10%; melanoma, 14% to 20%; and then there's a kind of all others.

And you should think of this as probably early uptake in leading indications because of the approvals now in HCC and adjuvant melanoma.

We've got about 9% to 15% there.

And the reason I give ranges is it's very hard obviously to get specific point estimates by tumor type.

I'll just give you the lung number.

When you go worldwide, that climbs to between 50% and 60%, and that's because of the strength we have in ex-U.

S. markets in lung cancer because of a leading position we've been able to establish through earlier access approval, particularly strength in markets like Japan and France.

So very strong performance in lung outside the U.S.

Thanks.

The last person is Jason Gerberry from Bank of America Merrill Lynch.

So just one quick regulatory question on -227.

So is the thought that you'd need in TMB high a favorable OS as a secondary endpoint for it to gain full approval and European approval or that the co-primary endpoints in different subpopulations would be registration enabling?

I just wanted to get clarification of that in light of Merck's comments earlier in their 3Q about meeting OS to drive approval in the target population they're going after.

And then just one other question.

Just can you provide the proportion of how prevalent TMB high is in PD-L1-negative patients?

Jason, let me just maybe answer that question.

We're not going to comment more on our regulatory interactions, but I just want to stress the TMB part of the study met its primary endpoint.

And the primary endpoint is progression-free survival in that study, and we look forward to discussing that with the regulatory authorities around the world.

I think that Tom has said earlier that we have just received the top line data.

There are a significant number of secondary analysis that we will be conducting, and we look forward to presenting that data together with other analysis at a future meeting.

Obviously, as we've mentioned, the study looked at TMB regardless of PD-L1 expression, and so it included patients across the full spectrum of PD-1.

So let me just close and thank everybody for participating in the call.

And in closing, I'd like to reaffirm my enthusiasm for the results we've shared with you today.

And during my career and many, many years in oncology, I can't remember a time in which cancer research was advancing at the speed that we are experiencing today.

And I'm very proud that we at BMS are at the front -- forefront of incredible scientific advances.

We had a great year in 2017, and I'm very confident in the multiple opportunities that we are pursuing to build a strong future for the company in '18 and beyond.

Thank you.

Thanks, everybody.

Appreciate your time.

So ladies and gentlemen, that will conclude today's Bristol-Myers Squibb 2017 Fourth Quarter Results Conference Call.

Thank you for your participation.

You may now disconnect.