Exercise and Cancer: Does physical activity improve colon cancer survival?

[Regina] (0:00 - 0:19)
Death is a binary variable. It's yes or no, but the thing is, everyone is eventually a yes. The chance of death, Kristin, is 100%.

Sadly. So what matters is not whether you will die, but when you die, because I would rather my when be later rather than sooner.

[Kristin] (0:24 - 0:47)
Welcome to Normal Curves. This is a podcast for anyone who wants to learn about scientific studies and the statistics behind them. It's like a journal club, except we pick topics that are fun, relevant, and sometimes a little spicy.

We evaluate the evidence, and we also give you the tools that you need to evaluate scientific studies on your own. I'm Kristin Sainani. I'm a professor at Stanford University.

[Regina] (0:48 - 0:53)
And I'm Regina Nuzzo. I'm a professor at Gallaudet University and part-time lecturer at Stanford.

[Kristin] (0:54 - 0:59)
We are not medical doctors. We are PhDs. So nothing in this podcast should be construed as medical advice.

[Regina] (0:59 - 1:04)
Also, this podcast is separate from our day jobs at Stanford and Gallaudet University.

[Kristin] (1:05 - 3:14)
Regina, today we're going to talk about something that's very personal to me, exercise and cancer.

[Regina]
Very important, too.

[Kristin]
A lot of people assume that exercise is good for both preventing cancer and extending life after a cancer diagnosis.

But surprisingly, we don't have much gold standard evidence of that.

[Regina]
Really? That is surprising.

[Kristin]
That's why this episode is exciting. We're going to look at a landmark randomized trial that just came out in July in the New England Journal of Medicine. It looked at exercise in colon cancer patients.

It's called the CHALLENGE trial, which is short for Colon Health and Lifelong Exercise Change trial. That doesn't quite give you the acronym CHALLENGE, but I guess close enough.

[Regina]
Clinical trials always need to have a catchy acronym.

[Kristin]
Exactly. All right. So the claim we'll be evaluating today is the claim of the study that exercise extends the time that colon cancer patients remain alive and cancer free.

And Regina, I'm sorry, but I was not able to get sex into this episode. It's kind of a serious topic. But instead, we are going to cover some very sexy statistical topics.

[Regina]
Statistician aphrodisiacs, basically.

[Kristin]
Oh, yeah. We're talking phase three trials, intention to treat analysis, interim analyses, survival analysis, Kaplan-Meier curves and hazard ratios, and the difference between ratio and difference statistics.

[Regina]
Ooh, baby. Hot and bothered.

[Kristin]
Regina, this one is personal for me, so I do have some biases.

First, I am very pro exercise. I used to be a serious competitive distance runner. I still run most days.

I collect exercise equipment in my garage. However, I did have invasive breast cancer last year. Luckily, stage one.

But in my N-of-1 experience, exercise did not prevent cancer. I am still hoping that exercise reduces cancer recurrence. That would be good for me.

So I really want this study to be true. Now, the GHALLENGE trial was colon cancer, not breast cancer. But it still feels relevant to me.

And I am coming in with these biases.

[Regina] (3:14 - 3:26)
Yeah, you do exercise a lot and you are otherwise very healthy, extremely fit. And just for context, you run how many miles a week again?

[Kristin] (3:26 - 3:44)
Well, it varies. But my goal for years and years has been 28 miles a week because I'd really like to get 30. But sometimes that's just too much. So I give myself a little wiggle room.

And as we're going to see, this is way above the amount of exercise that was prescribed in that CHALLENGE trial.

[Regina] (3:45 - 3:59)
So the intervention in this study was not about running a marathon every week, which is good news for the rest of us. And Kristin, it does make sense to me that exercise might help keep cancer away.

[Kristin] (3:59 - 4:06)
Right. Regina, you've probably heard all over the news, exercise is good for cancer. It seems like it's already proven.

[Regina] (4:06 - 4:46)
We do hear it all the time. So I assumed it was well studied and you mean it's not?

[Kristin]
Well, it is well studied, but primarily with observational studies.

[Regina]
That is a problem. Observational studies can't give us that gold standard evidence because people who choose to exercise like you, Kristin, are different in important ways. They tend to be more educated, more financially comfortable, have better willpower, better access to health care.

So better outcomes in exercisers could be just due to all of those confounding factors, not exercise itself.

[Kristin] (4:47 - 5:10)
Exactly. And there's another issue. Exercise might be a marker of good health, like we talked about with vitamin D in the vitamin D2 episode. People who are already healthier have the ability to do more exercise and they are also less likely to get cancer.

So it could look like exercise is protective when really health is driving exercise and not the other way around.

[Regina] (5:10 - 5:22)
Kristin, OK, so lots of observational studies and we need randomized trials. Were there any good trials before this one?

[Kristin]
The short answer is surprisingly no.

[Regina] (5:23 - 5:42)
That's not good. What is the long answer?

[Kristin]
Well, Regina, first of all, there are two different types of cancer studies.

One type focuses on preventing cancer from occurring in the first place. And we don't have any major randomized trials that have shown that exercise prevents the occurrence of cancer. It's kind of shocking, really.

[Regina] (5:43 - 5:56)
So even though we all just kind of believe that exercise prevents cancer, there's no actual randomized trial telling us that is true. But why not? Is it because trials like this are just hard to do?

[Kristin] (5:56 - 6:07)
That's it. Exactly. So, basically, cancer takes years to develop and it's relatively rare. So to test prevention, you'd need to enroll a huge number of people and follow them for a really long time.

[Regina] (6:08 - 6:17)
And I imagine randomized trials of exercise are already hard because patients need to stay motivated and actually exercise.

[Kristin] (6:18 - 6:27)
Absolutely. Now, we do have randomized trials showing that exercise can reduce not cancer, but things we think are related to cancer, like weight or inflammation or stress.

[Regina] (6:28 - 6:51)
So they are perhaps connecting the dots. If exercise improves weight and healthier weight is linked to less cancer, then maybe exercise lowers cancer risk. But it's all kind of leaps of faith there.

[Kristin]
Leap of faith is a good way to describe it. Absolutely.

[Regina]
So there are not major trials about preventing cancer, but Kristin, you mentioned there were two types.

What about the other?

[Kristin] (6:51 - 7:20)
Right. The other type of trial focuses not on preventing cancer, but on keeping people alive and cancer-free after a cancer diagnosis. There are a lot of randomized trials, actually, of exercise in cancer patients, but they focus on outcomes like fatigue or quality of life or biomarkers. Outside of the CHALLENGE trial, no other major randomized exercise study has tested survival or cancer recurrence as its primary outcome.

[Regina] (7:20 - 7:29)
It feels weird to not study survival or recurrence because, ultimately, they seem like the most important things that we care about.

[Kristin] (7:29 - 7:46)
Yeah. Now, a few trials did tack these on as secondary outcomes, but the trials weren't large enough to give any definitive answers. That's what makes the CHALLENGE trial so important.

It's the first sufficiently large Phase III randomized trial to focus on survival outcomes.

[Regina] (7:46 - 7:54)
Yeah, this is important because if exercise helps even a tiny bit, this could be a big source of hope and empowerment for cancer patients.

[Kristin] (7:55 - 8:11)
Absolutely. Regina, before we get into this study, let's talk a little bit about the biology. The study we're looking at today doesn't tell us anything about mechanisms. They did collect blood samples, and so perhaps there'll be some follow-up papers in the future that are about biomarkers and potential mechanisms, but we don't have that in this New England Journal paper.

[Regina] (8:12 - 8:20)
Hmm. But don't we already have lots of theories just floating out there about how exercise might affect cancer?

[Kristin] (8:20 - 8:48)
We sure do. For example, exercise can lower estrogen and testosterone, and that might reduce the risk of hormone-fueled cancers like breast or prostate cancer. Exercise also improves insulin sensitivity, and it lowers circulating insulin, cuts down on inflammation, and strengthens immunity, and stronger immune cells may be better at spotting and destroying rogue cancer cells. And exercise may also reduce body fat, particularly the visceral abdominal fat that we talked about in our hookworm episode.

That's also been linked to cancer.

[Regina] (8:49 - 8:55)
You know, if I had to choose between hookworms and exercise, I might choose the worms.

[Kristin] (8:55 - 9:11)
No, really? I'll take exercise. Thank you very much.

Yeah. The authors of this paper also mentioned a few biological mechanisms that are specific to colon cancer. For example, they note that exercise may, quote, decrease gastrointestinal transit time.

[Regina] (9:12 - 9:22)
Transit time? What is that? Is it how long it takes for food to transition through your body before you poop it out?

[Kristin] (9:22 - 9:49)
That's it, exactly.

Yes. Yeah. The idea is that if the waste products sit in the colon for too long, that could increase cancer risk, and that actually ties in with what we already know about high-fiber diets being protective, since fiber also speeds transit.

The authors also mentioned that exercise may change your gut flora, maybe improve your microbiome. Everyone's talking about the microbiome these days, and maybe that's also linked to cancer. All right, Regina, let's now dive into the methods.

This was a phase three trial.

[Regina] (9:50 - 10:01)
Oh, cool. We've talked about phase one randomized trials before in our episode on hookworms. And in those phase one trials, the whole investigation is really exploratory and small.

[Kristin] (10:02 - 10:04)
Yes, hookworms, that's an episode that everyone should check out.

[Regina] (10:05 - 10:24)
But phase three trials are different. They are large, carefully planned, and they are meant to give us definitive answers before something is approved and put on the market. But now that I think about it, this is exercise, Kristin, not a product we are trying to get FDA approval for.

[Kristin] (10:25 - 10:36)
No, but that's a great point, Regina. In this trial, they were actually thinking about exercise as a prescription. The doctor would literally give you a prescription for exercise the same way they'd prescribe medicine.

[Regina] (10:36 - 10:40)
Hmm. Would that mean insurance could pay for my personal trainer?

[Kristin] (10:40 - 10:52)
Well, actually, you could imagine that. I'm not sure about in the U.S., but in somewhere like Canada, where the trial was run out of, you could imagine support for exercise being covered by insurance. Yes.

The way a rehab program might be.

[Regina] (10:52 - 11:18)
That's an important point, actually, Kristin, about phase three trials, because they are really expensive, and that means researchers need to carefully think about every detail of the intervention, because this is not the time to just throw together a half-thought-out plan. So, here, was it telling people try to take a walk each week, or was it more like sending trainers to drag you out of bed?

[Kristin] (11:19 - 11:32)
More like dragging you out of bed. Actually, just telling people to walk every week, that's actually closer to what the control condition looked like in this trial. Regina, they did put a lot of thought into the exercise intervention, and it included a lot of support, and we're going to talk about that in a few minutes.

[Regina] (11:32 - 11:35)
Good. So, you mentioned the trial was in Canada. Were there a lot of sites?

[Kristin] (11:35 - 12:43)
Yes. Fifty-five sites, mostly in Canada, but also in Australia, plus a handful in the U.S., including one at my alma mater, Dartmouth. This is interesting. Recruitment began in 2008, with the first randomization in 2009.

[Regina]
2009, but this was a recent paper, so.

[Kristin]
Yeah, we're talking way back in the Obama era, it started.

[Regina]
Lady Gaga era, I think that was Poker Face time.

[Kristin]
I will always think of it wistfully as the Obama era, simpler times.

But yes, this trial took 15 years to complete, much longer than they originally planned.

[Regina]
What happened?

[Kristin]
It was supposed to end like a decade earlier, but it took them a lot longer than expected to recruit enough patients, as we're going to see.

[Regina]
Interesting.

[Kristin]
Okay, Regina, let's talk about who was actually in the study. They were all survivors of colon cancer.

Ninety percent had stage three cancer. The rest were high-risk stage two. This means that the cancer had not metastasized, it had not spread to distant sites, but these were people at pretty high risk of the cancer coming back.

[Regina] (12:43 - 12:50)
Ah, so they focused on higher-risk patients, probably because those are the people that we would expect to get the most benefit from the exercise?

[Kristin] (12:51 - 13:11)
Exactly. To be eligible, they had to have had their tumor surgically removed and to have finished chemotherapy within the past two to six months, so they couldn't be too far out from treatment. And they only enrolled people who were relatively sedentary, people who were not meeting physical activity guidelines of at least 150 minutes a week of moderate to vigorous activity.

[Regina] (13:11 - 13:21)
150 minutes a week? I am doing the math in my head. That is, what, 2.5 hours per week?

I think I got that. I think I got that.

[Kristin] (13:21 - 13:31)
I'm sure you're getting that, Regina, yeah.

Now, before they were randomized, patients underwent this extra battery of tests to make sure that they could handle the trial, and that's important because you don't want to lose people after you've randomized them.

[Regina] (13:31 - 13:37)
Talking about randomization, Kristin, did they do anything fancy here? Because it feels like a fancy trial.

[Kristin] (13:37 - 14:02)
It is. So for randomization, they did not just flip a coin. The patients were first stratified, meaning they were broken into groups based on whether they had stage two or stage three disease, whether they were very overweight or not, and whether they were fully active or had some limitations.

And the researchers applied what's called stratified randomization with minimization, which is a mouthful of jargon that you're going to unpack for us, Regina.

[Regina] (14:02 - 15:04)
Yeah, we talk about this in our clinical trials course, actually, on Stanford Online, and it is pretty cool. So the stratified part just means what you said before. Patients were first put into those categories you mentioned, you know, whether they were very overweight or not.

Then the minimization part kicked in, and this is the cool part. So for every new patient that's being enrolled, a computer would first look at the two groups that they already had, the exercise group and control group, and see how balanced they were according to all these characteristics. And then the computer would give a little nudge to the randomization algorithm to assign that patient to whichever of those two groups would minimize any imbalance between the groups.

So still a random assignment, but just with a little nudge to keep the groups even as possible. Really cool, really modern. You definitely could not have done this 20 years ago.

[Kristin] (15:05 - 15:18)
Yeah, you definitely need a computer for this. Regina, let's talk now about the interventions, starting with the control group. They were not given nothing.

They received standard health education, like some kind of a booklet about healthy eating, fiber, and a recommendation to exercise.

[Regina] (15:19 - 15:38)
It makes sense that they were not given nothing because exercise is still healthy. So you can't ethically put cancer patients into a group and then say nothing at all about exercise. Absolutely.

What about the exercise intervention? Is it like planks? Are they doing pushups?

They're doing jumping jacks. What?

[Kristin] (15:39 - 16:02)
Good question, Regina. Actually, the specific exercises were up to the participants. This was a three-year program.

The first six months were the most intensive. During that phase, participants had mandatory in-person bi-weekly sessions with a behavioral support person, kind of like a medically trained personal trainer. They also had to attend bi-weekly supervised exercise sessions.

[Regina] (16:02 - 16:17)
So behavioral support plus exercise. That makes sense. It's not enough just to say, hey, guys, exercise is good for you and then send them out.

You need a coach or something to keep you accountable. Well, at least most of us do. Maybe not you, Kristin.

[Kristin] (16:18 - 17:28)
Yes, the behavioral piece was really important. The behavioral sessions were mandatory the entire three years of the intervention, although you could do them over the telephone after six months. But after six months, the supervised exercise sessions were actually optional.

They were available to participants, but they were optional. Participants also got access to a gym or home equipment if they needed it and pedometers to help track their activity. And the intervention was individualized.

Everyone was encouraged to gradually add about 10 MET hours of exercise per week. But some people were nudged to do even more if they were handling that amount well.

[Regina]
Remind me again, 10 MET hours of exercise a week.

What is a MET hour?

[Kristin]
Yes, MET stands for Metabolic Equivalent of a Task. And it's a measure that combines both the duration and intensity of exercise.

So 10 MET hours a week is roughly equivalent to two and a half hours of brisk walking per week.

[Regina]
What about blinding, though, Kristin?

[Kristin]
So the trial was not blinded.

Everyone knew which group they were in and the investigators did, too. That introduces some potential bias, but it is very difficult to blind in a behavioral trial like this. Regina, let's talk outcomes now.

[Regina] (17:28 - 17:33)
I cannot wait to hear about the outcomes, but I think it's time for a short break first.

[Kristin] (17:41 - 17:52)
Regina, I've mentioned before on this podcast, our clinical trials course on Stanford Online. It's called Clinical Trials, Design, Strategy and Analysis. I want to give our listeners a little bit more information about that course.

[Regina] (17:52 - 18:01)
It's a self-paced course. We cover some really fun case studies designed for people who need to work with clinical trials, including interpreting, running and understanding them.

[Kristin] (18:01 - 18:14)
You can get a Stanford professional certificate as well as CME credit. You can find a link to that course on our website, normalcurves.com. And our listeners get a discount.

The discount code is normalcurves10. That's all lowercase.

[Regina] (18:23 - 18:35)
Welcome back to Normal Curves. Today, we're talking about a randomized clinical trial on exercise for colon cancer patients. And Kristin, I think you're about to tell us about the outcomes.

[Kristin] (18:35 - 18:42)
Yeah, this is super important. In a phase three trial like this, it is critical to pre-specify the primary outcome before you start the study.

[Regina] (18:42 - 19:06)
Yeah, that's why the investigators can't do what we call outcome switching, which is a bad thing, which is choosing a different outcome to focus on after you've peeked at the data. It's a little too tempting for researchers to do that. And if you let them do that, it bumps up the odds of getting false positives.

We talked about this back in our alcohol episode, I believe.

[Kristin]
That's right. With the wine trial.

[Regina]
Yes.

[Kristin] (19:06 - 20:29)
And Regina, it's not good enough to specify a vague primary outcome. You have to be very specific and detailed about it. In this study, the primary outcome was disease-free survival.

[Regina]
Disease-free survival means what? Let's clarify that.

[Kristin]
Well, disease-free survival can actually be defined in different ways in different studies.

In this trial, it was a composite outcome that tracked when a participant had any one of the following events. They died from any cause, they had a recurrence of the original colon cancer, or they developed a new cancer altogether, a new primary cancer. And that last part, the new primary cancer, that's an important detail because disease-free survival doesn't always include that.

Sometimes it's just recurrence or death. So when I first looked at this study, I was concerned that maybe the authors had tacked on those new primaries after the fact to maybe nudge themselves into statistical significance. I read the protocol that they published originally in a journal in 2008, and all they said was disease-free survival.

They did not define it specifically, but the great thing here was that in one of the appendices of that New England Journal paper, they actually published the full 100-page detailed protocol. It was timestamped from 2008. So I was able to see that, yes, when I read their detailed protocol, they did indeed define disease-free survival as death, colon cancer recurrence, or any new primary cancer.

[Regina] (20:30 - 20:32)
These people are very thorough, are they not?

[Kristin] (20:33 - 20:45)
It was thorough and there was no outcome switching. So we can check that off of our worry list, Regina.

[Regina]
Oh, good.

[Kristin]
They also had some secondary outcomes. One was overall survival, which is just overall death, not whether or not you got cancer.

[Regina] (20:46 - 21:05)
That is important, right? Because sometimes a treatment can maybe lower cancer recurrence, but raise the risk of other things like heart disease. Maybe we save them from cancer, but we make them exercise so much they have a heart attack and die anyway.

We want to make sure we're not just trading one kind of death for another.

[Kristin] (21:05 - 21:33)
Absolutely. There were a few other secondary outcomes like did people exercise more? Did they feel better? I'm going to focus less on those outcomes since they've been studied in other trials.

All right, Regina, I want to talk about something important in phase three trials. They typically have what's called a Data and Safety Monitoring Board or DSMB committee for short, which includes outside experts who watch for safety issues, monitor the statistics, and kind of act as the statistics police to keep the trial honest and fair.

[Regina] (21:33 - 21:48)
You're right. We have not talked about DSMB yet. Kristin, have you ever served on a DSMB?

[Kristin]
I have.

[Regina]
Oh, wow. So you have actually been one of the statistics police.

That is awesome. Do you get a badge? Can you give out like bad data analysis tickets?

[Kristin] (21:49 - 22:23)
Not quite. You know, Regina, I'm still waiting for my establishment statistician badge, but that's a whole other story that we'll save for our p-value episode. One of the big roles of a DSMB is to do interim analyses, basically peeking at the data while the trial is still running.

These are pre-planned and highly structured to avoid false positives. But one of the things a DSMB can do is to stop a trial early if the results look so dramatically good or bad that it would be unethical to keep the trial going.

[Regina] (22:24 - 22:48)
I remember a famous example of that is the Women's Health Initiative trial, right, of hormone therapy. They stopped early because they saw higher risks of heart disease and breast cancer, if I remember correctly, in women on the hormones, that was what they were testing. And the DSMB, the stats police, decided it would be unethical not to stop and warn people early.

[Kristin] (22:49 - 23:23)
That was a famous case, Regina. In the case of CHALLENGE, remember I mentioned that they had trouble enrolling enough patients quickly enough? It turned out that the patients were also healthier than they were expecting, which means that the bad events, the cancer and death, occurred at a slower rate than expected.

[Regina]
Oh, good for the patients, but not so good for the study.

[Kristin]
Right. That's why the trial took so long.

They had to run it longer to get enough participants and enough events. But the decision to extend the trial was one of the things that the DSMB committee for CHALLENGE had to approve.

[Regina] (23:23 - 23:40)
Ah, yes, right, because you do not want to leave it to researchers to play fast and loose with study end dates, because this is one way that they could cheat. They might peek at the data and then decide to finagle the end date to make the data look better.

[Kristin] (23:41 - 24:07)
Exactly. That's why you need the stats police. All right, Regina, let's talk about the results. The trial ultimately enrolled eight hundred and eighty nine patients, evenly split 445 in the exercise group, 444 in the health education group.

The median age was 61. It ranged from 19 to 84. 19 is really young to get high risk colon cancer.

Yeah, 51 percent were women and the median follow up was almost eight years.

[Regina] (24:08 - 24:26)
So we are talking about several hundred sedentary cancer patients taking on a new exercise routine after chemo. This is ambitious and not just for a few weeks, but for years. What was the adherence like?

Did the participants actually get out and exercise?

[Kristin] (24:27 - 24:51)
They did. The adherence was pretty good for an exercise trial in the first six months, that intensive period. Participants attended about 83 percent of the behavioral sessions and about 80 percent of the mandatory exercise sessions.

Not perfect, but not bad. By the end of the three years, it dwindled a bit. Attendance at the behavioral sessions was down to 59 percent and attendance at those optional exercise sessions was 38 percent.

[Regina] (24:52 - 25:02)
Wow, that is still really high, though. So we're talking about more than a third of people still showing up for optional exercise after three years.

[Kristin] (25:02 - 25:47)
Yeah, so not bad. Interestingly, both groups, exercise and control, increased their exercise.

[Regina]
Wait a minute, the control group exercised more just from the pamphlet?

[Kristin]
Yeah, they got some educational materials and some encouragement to exercise because all colon cancer patients should get that. And of course, as cancer survivors, they had motivation, too. But the exercise group did more exercise than the control group.

About five to seven MET hours more per week, on average, throughout the three year intervention period. Regina, that's about an hour and a half more of brisk walking per week or like an hour of slow swimming. One thing is the exercise levels were self-reported.

That's not ideal because you worry that the exercise group, you know, maybe they were trying to impress the investigators. Maybe they fudged upwards a bit.

[Regina] (25:49 - 25:57)
We talked about social desirability bias in the male equipment episode where people also maybe fudged a little bit upwards.

[Kristin] (25:58 - 27:29)
I recommend that everyone listen to the male equipment size episode. Great episode. But there was some objective confirmation.

So they measured something close to VO2 max, and that's a marker of aerobic fitness. And during the three years, the exercise group had a higher VO2 max than controls on average. Not a massive difference, but enough to show us that they were exercising more and they were fitter.

[Regina]
Good for them.

[Kristin]
One additional result that I want to put in here before we get to the survival outcomes. The study looked at weight over the three years of the intervention, and they did not see a difference between the two groups.

This is important because you can imagine that one way that exercise might reduce the risk of cancer is through weight loss. But if they had lost weight here, then we wouldn't know whether the effects were mediated through weight loss or through exercise. So it appears that there was no difference in weight loss here.

So that's good for interpretation. All right, Regina, now let's dive into the statistics. Since this was a randomized trial, they used what's called an intention to treat analysis.

That means once randomized, always analyzed. Every person who is randomized gets included in the final analysis no matter what happens after. So in this trial, for example, some people were randomized and then after that, they figured out that they already had metastasis, unfortunately.

And so they actually couldn't do any of the trial. Some people also dropped out of the trial just for logistical reasons. Under intention to treat, all of those people are still counted in their original group.

[Regina] (27:30 - 27:41)
This is really confusing for a lot of people. It's like, wait a minute. They didn't actually exercise, but you're counting them in the analysis as if they exercise.

So, Kristin, let's talk about why we use this approach.

[Kristin] (27:42 - 28:14)
Yeah, I actually have a whole column on this, Regina, that we can put in the show notes. One of the reasons is that the whole point of randomization is balance. And if you start excluding people after randomization, you lose that balance.

So imagine that the exercise group has more dropouts because exercise is hard. What if those dropouts happen to be the sicker patients with less family support? If you leave them out of the analysis, the exercise group may suddenly artificially look healthier, right?

So intention to treat prevents that kind of bias.

[Regina] (28:14 - 28:45)
Yes, exactly. I also think about how it shows effectiveness in the real world because doctors can prescribe exercise, for example, but they cannot force patients to actually do the exercise. So what we're really testing in a randomized trial is not perfect adherence to exercise.

It's what happens if we prescribe exercise as a whole package. And that's the clinically relevant question, because it's the only thing we can actually control. Right.

[Kristin] (28:46 - 29:03)
We didn't actually randomize people to exercise. Technically, we randomized them to a prescription for exercise. So we're testing how well that prescription works.

Yes. The researchers apply survival analysis to analyze the data. It's a set of statistical tools that statisticians find very sexy.

[Regina] (29:04 - 29:15)
I'm getting hot and bothered right now, Kristin. Maybe we can take a little statistical detour and explain survival analysis in juicy detail.

[Kristin] (29:15 - 29:43)
Absolutely. We use survival analysis when we have what is called a time to event outcome. This is a two part outcome. There is a binary part, whether or not you had the outcome, in this case, the death, recurrence or new cancer, because not everybody has that.

But then there is also a time part. If you had the outcome, when did you have it? Time matters.

If a cancer recurs after 15 years, that's a lot different than if a cancer recurs after one year.

[Regina] (29:43 - 30:03)
Yes, exactly. Death is a binary variable. It's yes or no.

But the thing is, everyone is eventually a yes. The chance of death, Kristin, is 100 percent. So what matters is not whether you will die, but when you die, because I would rather my when be later rather than sooner.

[Kristin] (30:04 - 30:43)
Exactly, exactly. Survival analysis was actually developed around death. So everybody eventually has it and therefore it's the time that matters.

Survival analysis tools are great because they account for time. This is super important in CHALLENGE, because remember, some patients joined in 2009, others joined in 2023, and the end date was 2024, which means we have very variable lengths of follow up. Survival analysis handles that beautifully.

Also, some people in CHALLENGE dropped out of the study before they had an event, but also before the study ended in 2024. Those people are called censored, and survival analysis handles these censored people well.

[Regina] (30:43 - 30:56)
I should point out that they're censored not because they are R or X rated or triple X rated, but because they are censored. All of their information, what happened to them afterwards, is censored from us.

[Kristin] (30:57 - 31:42)
Yes. Right. We don't know what happened to them after we lost them to follow up.

One of the sexiest tools in the survival analysis toolkit is the Kaplan-Meier curve, which some listeners may have seen before. That curve starts at 100 percent, meaning everyone starts alive and cancer free. And then the curve drops every time someone experiences a bad event.

And the steeper the drop, the faster bad outcomes are occurring. And the reason Kaplan-Meier curves are sexy is that they're visual and they can account for those censored people that we talked about. They account for them in a clever way that allows us to use their information up until the time that we've last seen them.

And so we're not losing their information altogether. Kaplan-Meier curves are often paired with a time to event regression model called Cox regression.

[Regina] (31:44 - 31:49)
Kristin, you are just making this a little too easy. Did you just say Cox?

[Kristin] (31:50 - 32:05)
Oh, we can get the sex in, I guess, here. No, Regina, this is spelled C-O-X. C-O-X.

It's also called proportional hazard regression. The model produces something called hazard ratios, which we're going to talk about in a minute.

[Regina] (32:05 - 32:16)
OK, back to the study away from the Cox. What did the sexy Kaplan-Meier curve show for this study, for the exercise and control groups?

[Kristin] (32:16 - 32:37)
For the outcome of disease-free survival, the exercise and control groups look almost identical for the first year. Both curves drop at about the same rate. But after a year, they begin to separate.

The control group dips more steeply, meaning more patients died, had recurrences or developed new cancers, while the exercise group held up better.

[Regina] (32:38 - 32:58)
Sounds like we're about to get deep into the results, Kristin. So I am looking forward to this, but it's a good time for a short break. Kristin, we've talked about your course, Writing in the Sciences on Coursera.

Maybe you could tell people a little bit more about it.

[Kristin] (32:58 - 33:14)
It's a self-paced course for anyone who needs to write scientific papers. And I give a lot of practical demonstrations for how to improve your writing to make it much more clear and concise. And you can earn a certificate from Coursera.

You can find a link to that course on our website, NormalCurves.com.

[Regina] (33:22 - 33:33)
Welcome back to Normal Curves. Today, we're talking about a randomized clinical trial of exercise for colon cancer patients. And Kristin, you were about to present the results for us.

Right.

[Kristin] (33:33 - 33:53)
There are actually a lot of different statistical ways to present the results. I like that the authors were transparent and they gave all the relevant numbers. And I want to make an analogy to help us weed through all of those numbers.

I'm thinking of the show Survivor. I've actually never seen it, but I think the premise is people get voted off the island.

Have you ever seen it, Regina?

[Regina] (33:53 - 34:00)
I have. I watched the first season, I think, when I was at Stanford, like 30 years ago.

[Kristin] (34:00 - 34:08)
Oh, wow. All right. So we're going to imagine this trial as two survivor islands. On one island, we drop off all the exercise group patients. And on the other island, we drop off all the control group patients.

[Regina] (34:09 - 34:11)
And what happens on these islands?

[Kristin] (34:12 - 34:28)
Over time, people occasionally get voted off the island. And that's our stand in for the bad outcomes. Cancer recurrence, a new cancer or death.

Also, a few people are leaving the island for other reasons. These are the people who are censored. And most of our math incorporates them, too.

[Regina] (34:29 - 34:34)
Oh, Kristin, good analogy. I love it. I love a good TV reality show analogy.

[Kristin] (34:35 - 35:04)
Me, too. Let's start with the easiest measurement to understand. Since both islands started with about the same number of people, 444 versus 445. And since they were followed for about the same amount of time, it's actually fair to compare the raw counts.

That's only usually true in well-done randomized trials. But I like counts because they're simple. By the end of the study, using our island analogy, 93 people were voted off the exercise island compared to 131 on the control island.

[Regina] (35:04 - 35:11)
Oh, that's a big difference, because if they were the same kind of island, we would expect equal numbers in both.

[Kristin] (35:11 - 36:26)
Yeah, it's a lot less bad events on the exercise island. Moving from raw counts, we can do something a little more sophisticated and we can calculate rates. Rates are more precise because they account for time.

A rate tells us how fast events are occurring in each group. On the control island, about five to six out of every 100 people were voted off the island each year, whereas it was only three to four out of 100 on the exercise island. That means every year about two fewer people per 100 were having bad events.

That's big in cancer.

[Regina]
Because two out of 100 per year, it adds up over the years.

[Kristin]
It adds up a lot.

Yeah. Regina, that roughly two out of 100 per year, that is called an absolute rate difference. It's called a difference statistic.

We got that by subtracting the rates between the groups. But we also have this other weird way we like to compare groups in statistics. We like to divide the rates to give a ratio.

We call that a ratio statistic or sometimes a relative risk. So here we can divide the rate of getting voted off the island in the exercise group by the rate of getting voted off the island in the control group. That's going to give us a value below one because there were fewer events in the exercise group.

The precise number we get here is a value of 0.69.

[Regina] (36:26 - 36:33)
You know, ratio statistics like this are so confusing. 0.69. What does that even mean?

[Kristin] (36:34 - 36:47)
Yeah, it means that people were being voted off the exercise island 31 percent more slowly than off of the control island. One minus zero point six nine gives that 31 percent decrease because we're always comparing to one.

[Regina] (36:48 - 36:55)
But 31 percent decrease sounds a lot more dramatic than two per 100 per year.

[Kristin] (36:56 - 37:01)
It does. Yeah. Ratio statistics like this can be misleading because they hide information.

[Regina] (37:01 - 37:57)
Yes. These ratio statistics, they hide information. And what they are hiding is the actual rate of disease in the two groups.

So let's talk about, you know, a situation where we have a treatment and it's cutting the rate of disease from two hundred out of one thousand to one hundred out of one thousand. That means we have saved one hundred people out of a thousand. Two hundred minus one hundred.

And that's a 50 percent decrease. But now let's talk about another situation. And we've got a treatment and it is cutting the rate of disease from two out of one thousand to one out of one thousand.

Now we've only saved one person out of one thousand, but that's a 50 percent drop, too. So we hide that information. We hide how rare or common the disease is in the first place.

And we just say, look, 31 percent decrease. How impressive.

[Kristin] (37:57 - 38:30)
Exactly. And people love those ratio statistics, relative risks, because they sound more exciting. Also, you see them a lot in the literature just because certain regression models produce ratio statistics. And these models are useful for doing some sophisticated things.

In this case, the authors use that Cox regression because they wanted to account for some additional factors like cancer stage. Cox regression gives hazard ratios. And hazard ratio is very similar to the rate ratio we just calculated by dividing the raw rates in the two groups.

[Regina] (38:30 - 38:35)
Yeah. Hazard ratio is just slightly more sophisticated and fancy. Exactly.

[Kristin] (38:36 - 39:03)
The hazard ratio here was 0.72, which is very close to 0.69 because they are similar measures. 0.72 means people were being voted off the exercise island about 28 percent more slowly than on the control island. Regina, let's go over one more number that the authors gave.

They reported the five year survival probability. And this is basically imagine that we pause the TV show at five years and we just counted up what percent of people were still on each island.

[Regina] (39:04 - 39:12)
Yet we should point out that five years is completely arbitrary, though. We could do four years, we could do six years. There's nothing magical about five years.

[Kristin] (39:13 - 39:37)
Yeah, it's just a snapshot. And it's the common one that is often used as a benchmark in cancer statistics.

[Regina]
OK, so five years.

What happened on our exercise island and control island?

[Kristin]
On the control island, the five year survival probability was 74 percent. On the exercise island, it was 80 percent.

So that's six extra people out of every hundred who avoided being voted off the island in those first five years.

[Regina] (39:37 - 39:58)
Oh, that's a good number. So no matter how you look at it, you've given us all these different statistics. We've got raw counts.

We've got the yearly rates. We've got the ratio statistic and the hazard ratio statistic. And now we have that five year snapshot everywhere.

And the exercise island was the better place to be.

[Kristin] (39:59 - 40:54)
Yeah. And that's why the results made headlines. The effect was statistically significant.

The P-value was 0.02. And the size of the difference was large enough to feel clinically meaningful. I'll add that overall survival was also different. At eight years, 83 percent of control patients were still alive versus 90 percent in the exercise group.

So that's seven extra survivors out of every hundred or about one in 14 over eight years.

[Regina]
That's a big effect.

[Kristin]
Yeah.

So this did get a lot of news coverage for good reason, Regina. Some of the news coverage, I think, was a little misleading. Some people said exercise is better than a drug, and that could be misread.

[Regina]
Oh, yeah, that's no good.

[Kristin]
Exercise was not a substitute for treatment. Every patient in this trial got surgery and chemo.

And had they not, they probably would have died. So exercise is on top of all of our standard treatments. It's improving upon what we already have, not a substitute.

[Regina] (40:55 - 41:21)
Yeah. Maybe a better way to look at it is that exercise had an effect that was about as big as adding another chemo drug. So it's the sort of effect size that oncologists would be thrilled to get if we were talking about an added chemo drug.

And here it's even more exciting because it's just exercise. So it's in addition to, not instead of, your proven therapies do not stop chemo.

[Kristin] (41:21 - 42:04)
Exactly. Yes. Another great thing about this paper, Regina, is they broke down all those raw counts into more specifics. So we can look for interesting patterns. And a few things jumped out at me.

In the exercise group, 16 people had metastasis to the liver. That was 29 in the control group.

[Regina]
Oh, that's a big difference.

[Kristin]
That feels very big. Yeah. It was also interesting to look at the new primary cancers.

Remember, this means totally different cancers, not a recurrence of the original colon cancer. In the exercise group, there were two cases of new breast cancers, but there were 12 new breast cancers in the control group, which feels really big. There were also zero new colon cancers in the exercise group compared with five in the control group, like totally new cancers in the colon.

[Regina] (42:05 - 42:20)
Wow. So these are interesting. Now, this could all be just random fluctuations that we get from doing our small talk, you know, subgroup analyses.

We talked about this in our review episode. But still, they're very intriguing.

[Kristin] (42:20 - 42:46)
They're very interesting. Yeah, it starts to get a little bit at potential mechanisms, for sure. Regina, before we wrap up, I want to talk about one more thing. As we said, the CHALLENGE trial made a lot of headlines, but there was another study on colon cancer in exercise that also grabbed media attention very recently.

[Regina]
I think I know the one you're talking about.

[Kristin]
It was in The New York Times. The headline was ultramarathoners may be at higher risk of colon cancer.

[Regina] (42:46 - 42:50)
That makes it sound like maybe you should not do ultramarathons.

[Kristin] (42:52 - 44:11)
Yeah, it at first glance sounds kind of contradictory to the results of the CHALLENGE trial, like exercise is bad, right? I want to break it down a little to show it's really not contradictory. So first of all, this new study is a very low level of evidence, unlike the CHALLENGE trial.

So it was just a conference presentation. So it hasn't gone through peer review. At first, I thought maybe we can do an episode on this, but I have no details other than the little abstract from the conference.

So not a lot of details published for people to critique. It's an observational study and it had just 100 people and there was no control group. It was also about cancer prevention rather than survival after cancer.

And it focused on extreme exercise levels. These are ultramarathoners, not people walking briskly, you know, several hours a week. It also didn't measure cancer.

It just measured colon polyps, which are growths that can sometimes turn into cancer. So the study gave colonoscopies to 100 ultramarathoners. These are people running more than 26.2 miles at a time. They were pretty young, 35 to 50 years old. They found that 39 percent of them had at least one polyp and 15 percent had advanced polyps, ones that are more likely to turn into cancer. That's quite a bit higher than you would expect for this age group.

But again, no control group. It's really just anecdotal.

[Regina] (44:12 - 44:27)
Anecdotal. And maybe people volunteered for this study because they knew they were at higher risk for colon cancer. Maybe they have colon cancer in their family because getting a colonoscopy when you're 35 is not something you do for fun.

[Kristin] (44:28 - 44:37)
Right. Might be some volunteer bias here. It also could be confounding. So maybe ultramarathoners don't have a perfect diet.

I mean, Regina, you've run an ultramarathon. What did you eat?

[Regina] (44:37 - 45:04)
Hmm. I've run a couple and I ate junk food the entire way. I have also volunteered at 100 milers and there are aid stations and pretzels, candy and soda were among the healthier options here.

Because if you're running 100 miles straight through, you know, 24 hours, whatever, you've got to just pack in calories. Right.

[Kristin] (45:04 - 45:23)
You're looking for calories and not fiber. I mean, when I trained for marathons, when I would get back from a long run, I would eat a big chunk of Ghirardelli chocolate for the calories. It's not exactly colon healthy.

I also wonder if maybe alcohol could be a factor because alcohol has been related to colon polyps. Do ultramarathoners party a lot, Regina?

[Regina] (45:24 - 45:36)
They absolutely party a lot. So the thing is, you do your ultramarathon in the mountains and then you hang out at the trailhead or in the parking lot and you tailgate and you pass around your bottle of whiskey and beer.

[Kristin] (45:38 - 45:58)
Right. So, you know, there could be something here, extremes of anything. Even good things like exercise can be harmful.

And there is some evidence that very intense training can suppress the immune system. But if we were handing out smooches for evidence quality, this is a one smooch finding for me. That's our lowest rating, basically anecdotal.

[Regina] (45:58 - 46:05)
I would call this like a hug or maybe a handshake, like come back when you've got something, you know, more than just an abstract.

[Kristin] (46:06 - 46:08)
Yeah. We'll keep following it, but not too worried yet.

[Regina] (46:09 - 46:18)
Yep. OK, Kristin, I think we are now ready to wrap this all up and rate the strength of evidence for the claim. So remind us of what the claim is.

It was very specific.

[Kristin] (46:19 - 46:25)
That exercise extends the time that colon cancer patients remain alive and cancer free.

[Regina] (46:26 - 46:39)
OK, so we are going to rate this using our smooch rating scale. One to five smooches where one means little to no evidence, five means extremely strong evidence. So, Kristin, what do you say?

Kiss it or diss it?

[Kristin] (46:40 - 47:14)
I'm going with four smooches on this one, Regina. I think this was a rigorous, very well done trial. Honestly, I was a little bit surprised by how big these effects were.

The exercise group, as we said, did maybe one and a half hours more of risk walking per week than the control group. It's kind of striking that such a modest bump could have such a big impact. I went through the paper looking for red flags, but the trial was well done.

So it could be that the effect is really that big. Could have been that they got a little bit lucky in the trial and maybe the real effect is a little bit smaller. So, Regina, I'm giving it four smooches.

How about you?

[Regina] (47:14 - 47:57)
Yeah, I'm going to give it four smooches, too. I was also surprised by the size of the effect. I'm a little cautious about this, but I wonder if it was less about the exercise.

Well, no, definitely the exercise, but also about all that behavioral support. Right. They had these mandatory check-ins.

And I feel like you're a cancer patient. You're fresh off of chemo. This is a chance to really just have a whole new life ahead of you.

And I think that can be really powerful. So I wonder if that had something to do with it. So four smooches for me.

[Kristin]
Right. Maybe more than just the exercise involved here.

[Regina]
Yeah.

Right. Right. Friends. Friends help.

[Kristin]
Yeah. OK.

[Regina]
Methodological morals. Do you have one for us?

[Kristin] (47:57 - 48:03)
I do. Ratio statistics sell headlines. Absolute differences sell truth.

[Regina] (48:04 - 48:14)
Oh, I like it. That's a nice dig on journalism as well.

[Kristin]
Oh, whoops.

[Regina]
No, in a good way. In a good way. You need to hold people accountable.

[Kristin] (48:14 - 48:24)
Yeah.

How about you, Regina?

[Regina]
Yeah. All right.

Mine is survival analysis is this sexy stats tool that makes every moment and every Cox count.

[Kristin] (48:24 - 48:29)
Oh, I love it. I love how you got the sex in there, Regina.

[Regina] (48:29 - 48:33)
Thank you. And the Cox. And the Cox.

C-O-X. C-O-X.

[Kristin] (48:33 - 48:34)
We're talking. Yeah.

[Regina] (48:35 - 48:54)
All right. Kristin, this is really surprising, really exciting. And I think it's going to motivate me to just get out and exercise a little bit more.

Oh, nice. Yeah. Yeah.

Not as much as you. Because that would be crazy, but just a little bit. So thank you, Kristin.

Thanks, Regina.

[Kristin] (48:54 - 48:55)
And thanks, everyone.