Sharp declines in mortality rates for non-small cell lung cancer in recent years are driven primarily by advances in treatment, NCI researchers concluded in a study published Aug. 12 in the New England Journal of Medicine.
The NCI study brings clarity to a debate about the contribution of lung cancer treatments to the rapid decline in overall cancer mortality. In January, the American Cancer Society declared that the latest available data—from 2016 to 2017—show the “largest ever single-year drop in overall cancer mortality of 2.2%.” (The Cancer Letter, Feb. 7, 2020).
While it was clear that the overall drop was driven by a reduction in lung cancer mortality, it wasn’t immediately apparent whether improvements in survival resulting from new drugs were significant enough to be detected.
Overall cancer mortality rates have been falling since 1990 due to a consistent decline in incidence of lung cancer. Is it possible that the sharper decline in recent years is caused by advances in drug treatments? Some experts pondered the impact of targeted therapies and immunotherapies for NSCLC, and whether that signal could be observed in population-level data (The Cancer Letter, Jan. 8, 2020).
There is now sufficient data to elucidate that link, said Nadia Howlader, lead author of the NCI study, and a mathematical statistician in the Data Analytics Branch within the Division of Cancer Control and Population Sciences at NCI.
“For non-small cell [lung cancer], we saw a steady decline in mortality, initially, followed by a rapid decline in 2013 through 2016—primarily explained by dissemination of targeted therapies that were approved in 2013 for stage IV EGFR-positive non-small cell lung cancer as the first line of therapy,” Howlader said at a virtual meeting of the National Cancer Advisory Board Sept. 2. “Our estimates suggest possible population-level impact with targeted therapies.”
In this study, researchers looked at data for NSCLC, which accounts for 76% of lung cancer in the U.S., as well as small-cell lung cancer, which accounts for 13%. Other subtypes of lung cancer that constitute the remaining share of cases were not covered in this study.
The researchers found that, in recent years, deaths from NSCLC decreased faster than the decrease in NSCLC incidence, and the decrease in deaths was associated with a substantial improvement in survival. Among men, for example, deaths from NSCLC decreased 3.2% annually from 2006 to 2013 and 6.3% annually from 2013 to 2016, whereas incidence decreased 1.9% annually during 2001 to 2008 and 3.1% annually from 2008 to 2016.
Two-year survival for men with NSCLC improved over this time, from 26% for patients diagnosed in 2001 to 35% for those diagnosed in 2014. Similar improvement was observed for women. In addition, improvements in two-year survival were seen for all races/ethnicities, despite concerns that the newer cancer treatments, many of which are expensive, might increase disparities.
In the past decade, new treatments for NSCLC have become available, including those that target genetic changes seen in some NSCLC tumors, as well as immune checkpoint inhibitors that help the immune system better attack NSCLC. In contrast, there have been limited treatment advancements for SCLC.
More time is needed for researchers to observe the effect of immunotherapeutic interventions, Howlader said.
“I think the effect of immunotherapy on our data, it would be too early to see,” Howlader said. “So actually, it’s a very good question, we are tracking the mortality trends by subtype. And we’re hoping to see it sort of show up in the data in the next year or two.”
Senior authors on the study include Douglas R. Lowy, NCI principal deputy director, and Eric J. Feuer, a branch chief in the Surveillance Research Program, and the NCI project scientist for the Cancer Intervention and Surveillance Modeling Network.
“Reduced tobacco consumption in the U.S. has been associated with a progressive decrease in lung cancer deaths that started around 1990 in men and around 2000 in women. Until now, however, we have not known whether newer treatments might contribute to some of the recent improvement,” Lowy said in a statement.
“This analysis shows for the first time that nationwide mortality rates for the most common category of lung cancer, non-small cell lung cancer, are declining faster than its incidence, an advance that correlates with the [FDA] approval of several targeted therapies for this cancer in recent years.”
Although death records do not distinguish between lung cancer deaths attributable to NSCLC versus SCLC, the cancer diagnosis records compiled by NCI’s Surveillance, Epidemiology, and End Results cancer registry program do distinguish between these two subtypes of lung cancer. As a result, researchers were able to estimate lung cancer mortality trends for these specific lung cancer subtypes by linking the lung cancer death records for each patient to the incidence data for these patients in the SEER cancer database.
“I want to highlight that this was an important analysis that used really novel methodologies to us, and NCI-level data to really show that the therapeutic benefits that have been developed over decades through elegant basic science investigations are now affecting the national cancer survival statistics in a dramatic and important way—with a very measurable benefit for patients,” NCI Director Ned Sharpless said at the Sept. 2 NCAB meeting.
The researchers had originally considered the possibility that lung cancer screening might help explain the decreases in NSCLC mortality, but their findings suggest that lung cancer screening rates, which remained low and stable, do not explain the mortality declines. Instead, the rapid decline in deaths reflects both declines in incidence—due in large part to reductions in smoking—and improvement in treatment.
In contrast, the decrease in deaths from SCLC corresponded with the decrease in incidence, and two-year survival was largely unchanged. Among men, for example, deaths declined 4.3% annually and incidence 3.6% annually. Findings were similar among women. The reduced mortality from SCLC over time, therefore, primarily reflects declines in incidence—again, due largely to reduced smoking.
The researchers note that the accelerating decline in NSCLC mortality that began in 2013 corresponds with the time when clinicians began routinely testing patients for genetic alterations targeted by newly-approved drugs.
In 2012, the National Comprehensive Cancer Network recommended that all patients with nonsquamous NSCLC undergo genetic testing. Subsequently, genetic testing for EGFR (epidermal growth factor receptor) mutations and ALK (anaplastic lymphoma kinase) gene rearrangements—which are targeted by the newer treatments—increased substantially.
Considering that immune checkpoint inhibitors were not in widespread use over the period of the analysis, the authors hypothesize that most of the survival benefit was attributable to effective EGFR or ALK inhibitors or other advances in therapy. The effect of immune checkpoint inhibitors on NSCLC survival is significant, which suggests that this improving trend in survival should continue beyond 2016.
“The survival benefit for patients with non-small cell lung cancer treated with targeted therapies has been demonstrated in clinical trials, but this study highlights the impact of these treatments at the population level,” Howlader said in a statement. “We can now see the impact of advances in lung cancer treatment on survival.”
An excerpt of Howlader’s remarks at the Sept. 2 virtual NCAB meeting follows:
We first wanted to understand, what is the relative contribution of the subtype? How do, particularly, the two major subtypes of lung cancer, namely small cell and non-small cell lung cancer, contribute to the overall lung cancer mortality decline?
And then, the second aim was to understand, is the decline in mortality more related to incidence or survival? Because, as we all know, mortality is influenced by both incidence and survival over time. So, we wanted to understand, what is the relative contribution of these two quantities on the mortality decline we observed?
So, there are three possible scenarios which could lead to the mortality decline that we see in the population:
In scenario one, mortality could decline if incidence rates remain flat, but survival improves.
In scenario two, mortality could decline if incidence declines and survival remains flat—which is a bit counterintuitive, but that could arise.
And scenario three, which is the most optimistic scenario, mortality could decline both if incidence is declining and survival is improving.
And we will see that two other scenarios play out when we take the overall lung cancer death rates and break them up by subtype.
So, for this particular study, we identified lung and bronchus cancer cases in SEER 18 areas during 2001 through 2016. SEER 18 covers about 28% of the U.S. population. Some small cell and non-small cell lung cancer subtypes were defined using coding schema developed by Lewis et al. in the Cancer 2014 publication. Because of some coding challenges, we were not able to go back beyond 2001, when we started the analysis using SEER data.
We use this technique called incidence-based mortality, taking the overall mortality and partitioning it by subtype. This is because regular death certificate data do not have subtype information. And I will talk a little bit more about this technique in a few slides. And then we use Joinpoint to assess IBM trend changes over time.
As I have alluded to earlier, we also wanted to understand how incidence and survival sort of explain this mortality decline that we are seeing. So, we also estimated age-adjusted incidence rate by subtype, by further adjusting for reporting delay. And we use Joinpoint to assess if and when there is a trend change over time. And finally, we estimated two-year lung cancer-specific survival by subtype using relative survival approach.
So, why did we need to use incidence-based mortality or briefly IBM? Because information on lung cancer subtype are not directly available for death certificate mortality data, but they’re available from SEER data on incident cases. So, IBM provides a resource to address the limitation in death certificate mortality data by linking SEER incident cases to mortality records. What it does is it allows us to use information on deaths in SEER incident cases to reconstruct the mortality curves using IBM.
What is IBM? It is a rate where the numerator of the rates are deaths among incidence cases by a subtype in a given calendar year, and the denominator is the general population in SEER in a given calendar year. And you will see in the next slide that IBM rates are valid for a shorter period of time than the death certificate data, this is because we would have to allow sufficient follow-back time to observe all the deaths that are happening, using the IBM approach.
So, in this figure, I’m showing you the blue line, which is the lung cancer mortality from regular death certificate data. And the red line, which shows the IBM trend. Note that, here, the blue line, which is the lung cancer mortality on the basis of regular death certificate data, looks to be higher compared to the red line. And I will talk about why that is.
Also, as I have said, because IBM is derived based on death certificate records linked to cancer cases, we need to allow sufficient followup time to see all the deaths. So, for that reason, we were able to look at subtype-specific trends from 2006 onward. Because we felt the five-year follow-back period to observe all deaths from lung cancer was enough time to see all the deaths, given the lethality of the cancer.
So, why does the blue line—using the regular death certificate mortality data—seem to be higher, using the IBM approach? Well, given that the lungs are a common site of metastases for other cancers at diagnosis, misattribution or over-attribution of deaths of lung cancer is possible in the death certificate data.
In fact, using the IBM approach, we showed and identified that the number of lung cancer deaths is actually linked to non-lung cancer diagnoses in the SEER area, which is shown by this green curve. The implication is that deaths from lung cancer are actually somewhat lower than currently reported using regular death certificate mortality data.
So, moving on to results: First, I would like to show the results for non-small cell lung cancer. In this graph, I’m showing you the IBM trends, the mortality trends for non-small cell lung cancer for males.
Here, the death rates from non-small cell lung cancer were declining 3.2% annually from 2006 to 2013, and then started to decline at a doubling rate, 6% annually from 2013 onward. The concurrent incidence was also declining, but at a much slower rate—as one can see from these graphs, going down 1.9% annually from 2001 through 2008, and then at 3% annually from 2008 onward.
When we look at survival, there has been impressive improvement in two-year lung cancer-specific survival for non-small cell cancer patients—going from 26%, if diagnosed in 2001, to 35% if diagnosed in 2014. Similarly, the results for females show a similar trend in the sense that we do see a rapid drop in death rates among females for non-small cell lung cancer, and a very improved, impressive survival trend going from 35% in 2001 to 44% in 2014.
We further looked at the survival improvements to see if we saw similar types of improvement across all races and ethnicities. So, here is the data showing two-year lung cancer survival for non-small cell cancer patients among four major race/ethnic groups. And one can see from the graph that survival improvements persisted across all these racial groups, even though there is disparity in the sense that non-Hispanic Blacks still experience worse survival compared to the other racial/ethnic groups in the population.
Moving on to the results for small cell lung cancer, the story here is actually quite interesting, in the sense that there is not much improvement in survival, as you can see, the survival trend is fairly flat. And then, we still see some decreasing mortality and it’s parallel to the incidence, mostly being driven by reduction in incidence as a result of reduced tobacco use.
So, I would like to put a few interpretations of the trends. How did we conclude that the sharp drop we saw in non small cell lung cancer mortality could not be explained by some other factors?
To interrogate the data further, we considered: could lung cancer screening or changes in risk factors or advances in treatments explain this rapid or sharper drop in non-small cell lung cancer mortality?
So, when we came to screening data, the answer was No. Because the screening rates remained low and fairly stable throughout our study period. That did not or could not explain this drop in death rates.
Declining smoking rates; could that explain this sharp drop in death rates?
Undoubtedly, the declining smoking rates contribute to the declining incidence and mortality rates for lung cancer over time. But given the timing and the magnitude of the drop, smoking alone did not explain this for us.
The third factor that really stood out was that, could targeted therapies explain this? Well, it actually correlated with several targeted therapies that were approved by the FDA around 2013 in the population.
In conclusion, I would say that what we showed is that for small cell lung cancer, there was steady decline in mortality, which was explained almost entirely by lower incidence, potentially attributable to reduced tobacco use.
This analysis shows for the first time that nationwide mortality rates for the most common category of lung cancer, non-small cell lung cancer, are declining faster than its incidence, an advance that correlates with the [FDA] approval of several targeted therapies for this cancer in recent years.
Douglas Lowy
For non-small cell, we saw a steady decline in mortality, initially, followed by a rapid decline in 2013 through 2016—primarily explained by dissemination of targeted therapies that were approved in 2013 for stage IV EGFR-positive non-small cell lung cancer as the first line of therapy.
Our estimates suggest possible population-level impact with targeted therapies.
Currently, SEER does not have data on individual level of drug use. But looking to the future, the SEER program is doing a lot of collaborations with the Department of Energy, and collaborating with industry, and trying to get to multiple data sources from insurance claims, pharmacy records,
EHRs—enabling collection of cancer surveillance data from all these multiple sources to populate detailed treatment, biomarkers, and also decrease the interval for reporting and create these detailed longitudinal patient trajectories.
