Is Exercise Addictive?

Athletes, particularly those that engage in the sustained-activity aerobic sports such as running, cycling and Nordic skiiing, are occasionally to be found describing their chosen physical activity as addictive. Some of them talk about symptoms of anxiety and depression and discomfort when not permitted to engage in their usual level of activity due to injury or other life circumstances. These can sound suspiciously like syndromes associated with addictive drugs, to those of us who are familiar with the type and aware that drug "withdrawal" is not limited to the dramatic symptoms associated with withdrawal from a substantial intravenous heroin addiction.

Is this real, though? Is there something neurobiologically similar to drug addiction about what can happen to the brain in the course of a sustained aerobic exercise habit? To answer these questions it helps, of course, to have an animal model, preferably one that has a lot of similarity to our drug-abuse models.

ResearchBlogging.orgEikelboom and Lattanzio published two papers in 2003 that proposed a possible model of exercise-dependence using activity wheels in rats. You will be familiar with the notion that hamsters, rats and mice  will run on a treadwheel under single-housed conditions from your local pet store. Drop by one on your way home today if you don't know what I mean. There is also an older behavioral literature that shows that wheel access can act as a reinforcer in laboratory rats- they will press a lever to get a brief interval of wheel running. Manipulating the length of time they can run on the wheel acts, to a first approximation, like manipulating the number of food pellets delivered in a traditional setup.

But just because something is chosen voluntarily and acts as a reinforcer does not necessarily mean that it can model compulsive, repetitive behavioral patterns. It does not necessarily mean that it will tap into the disruption of brain reward pathways and mechanisms that is the hallmark of substance dependency. For this we need a little more evidence, starting with behavior and moving into neurobiology as the evidence mounts.

The background for the Lattanzio and Eikelboom work is  the Ahmed and Koob 1998 paper which has become hugely influential in substance abuse models. The short version was that instead of permitting rats an hour of access to intravenous cocaine each day, they permitted the animals 6 hours of access per day. They observed that the 6-hr (termed "Long Access") group took more cocaine than animals run in the traditional 1-hour sessions, perhaps unsurprisingly given the increased opportunity. More importantly, as the sessions of longer access continued, the rats took an ever increasing amount including in the first hour of access. Not only that, but the Long Access group took more cocaine in the very first 10 minutes of the session.  This seminal paper has been followed by a lot of additional evidence that this change in the Long Access group is brought about by lasting disruptions of common reward mechanisms.

[Sidebar: You will recall from my posts trying to work out the conditional probability of dependence, that I am not a fan of simple, drug-feels-good models of drug reinforcement; even though they have a place. The short version of my thinking is that we already know from the human epidemiology that a large fraction of the individuals who find that drugs make them feel good do not go on to develop dependence, addiction or what we might term a drug problem. Animal models that move on from the simple feels-good stage of drug taking resonate more strongly with me.]

Lattanzio and Eikelboom set out to provide rats with longer and shorter access to wheel running and see if there was behavioral evidence of the sort of "escalation" that was reported by Ahmed and Koob.

In the first paper the authors compared wheel running in rats who had 24/7 access to the wheel to rats who had only a two hours of access for 24 days. As depicted in the first figure, wheel activity gradually increased in the 24/7 group across the three weeks of study. This is pretty consistent with my reading of the circadian literature and has some interesting implications with respect to the development of aerobic fitness. Activity in the 2-hour group remained stable.

So far so good. They also show that the 24/7 animals ran more in a comparable 2-hr interval, reminiscent of the key first-hour comparison in the drug self-administration paradigm.

Or, they sort of show that.

When they selected the same two-hour interval of the day in which the 2-hr group was exposed to the wheel, the 24-hr animals ran less. A lot less. Because. they. ran. the. 2-hr. group. in. the. light. cycle. Rats are nocturnal and more active in the dark. So when the authors (cherry) picked the most-active 2-hr interval in the 24/7 access group, then yes, it looked like an escalation across the days of training.

Frustrating. A hint of an escalation type of effect with longer access to the wheel but confounded by an inexplicable choice to run the short-access group in the light part of the cycle. Luckily, the authors did not leave off at this one study.

The second paper is more interesting because they run both the longer and shorter sessions in the dark, when rats are most active. In addition they are pitting 1 hr access against only 4 hours of wheel access, instead of a full 24 hrs. So it makes it a little more comparable to the typical drug self-administration experiment. These results are again consistent with escalation of wheel running. In Figure 2B they show that the 4-hour group's wheel running in the first two hours of access increased substantially more with sequential training sessions in comparison with the 1-hour group's running.

These papers are, to my eye, a good first attempt at a model. This is not the answer to whether exercise is addictive or becomes a compulsive behavior similar to drug self-administration. However it shows that we can now go on to ask additional questions which might answer the question. Are the brains of the longer-access rats changing in the same way that the long-access to cocaine rats' brains change? Are they in a state of reward deficit (disrupted allostasis in the Ahmed/Koob handwaving) that generalizes across reinforcers?

If evidence develops for this, we can only then move on to the larger issues. Does a substantial history of exercise leave individuals at risk for reward-related disorders when they stop exercising? Are they at increased liability for compulsive eating or drug abuse? If so, what is the threshold? Etc. Really, there is a lot of fascinating research ahead on this topic.

I have a few current questions about the exercise physiology angle because I know there is a blogger or two around and about that might have some information. Are the physiological changes brought about by wheel running in laboratory rats similar to those we might expect from a human in aerobic conditioning training? The circadian literature shows pretty consistently, to my eye, that daily running in rats given 24/7 access to wheels increases over a several week interval to reach a sustained plateau of daily activity. This suggests that there are perhaps cardiovascular and muscular adaptations at play, in a word "fitness". But then again the sort of exercise that results in human conditioning is sort of aversive at the start, isn't it? We force ourselves to do it because we want to be fit or to race or whatever. We don't do it because every step of the 6 mi run is pure bliss right off the couch, right? So why would rats voluntarily run themselves into this sort of conditioning effect?

Those of us who are looking at this from a perspective of reward mechanisms will eventually need to show that wheel "escalation" is not just a result of a conditioning effect which permits the rats to run for longer distances. Or for that matter a motor skill effect which permits them to tread the wheel bars more efficiently.

Disclaimer: See the Disclaimer page for the usual about my conflicts when it comes to drug abuse research topics. Also, I am professionally acquainted with some of the authors of the work under discussion in this post.

Literature Cited:
Ahmed SH, & Koob GF (1998). Transition from moderate to excessive drug intake: change in hedonic set point. Science (New York, N.Y.), 282 (5387), 298-300 PMID: 9765157
Lattanzio SB, & Eikelboom R (2003). Wheel access duration in rats: I. Effects on feeding and running. Behavioral neuroscience, 117 (3), 496-504 PMID: 12802878
Eikelboom R, & Lattanzio SB (2003). Wheel access duration in rats: II. Day-night and within-session changes. Behavioral neuroscience, 117 (4), 825-32 PMID: 12931966

17 responses so far

  • [...] This post was Twitted by ScientopiaBlogs [...]

  • D. C. Sessions says:

    Referring to the high-exercise reward processes as "disrupted" presumes that low-exercise is baseline. Given evolutionary history, and for that matter the inclinations o f children, one could put forth the contrary hypothesis that sedentary behavior is "addictive" in the same sense and likewise "disrupts" reward processes.

    Either way, we appear to be somewhat bistable.

  • CoR says:

    THe short/long groups are definitely different in both papers you present. I'm curious about those error bars of the longer group tho -- they consistently get bigger across days in the long group. Why? How long does a rat live? Why would the variability increase over time and only in the longer group?

  • drugmonkey says:

    Cackle, I suspect it is because of individual differences becoming more pronounced.

    This is similar to what you see in the drug self-administration models.

    As I've alluded to on the blog now and again, the animal models in drug abuse fields have included a major limitation of approach in that the models are mostly focused on all of the animals expressing the effect of interest. And yet human substance dependence occurs in a minority. I do think we are, as a field, shifting to be interested in individual differences. So for data like these "escalation" models, I'd like to see median splits or quartiles included as a matter of course. that requires bigger experiments, though.

  • palmd says:

    This may get interesting. We are fairly certain exercise affects mood and often recommend it for a variety of medical and neuropsychological problems, always with a "it probably helps and couldn't hurt" bias---except in one particular realm, that of eating disorders, where one of the compulsive behaviors frequently seen is exercise, to the point where, when hospitalized, patients often have their access to it limited.

  • [...] a post over at Drugmonkey yesterday got Sci to thinking about drug self-administration in rats, and more particularly, HOW we use it to [...]

  • i am more inclined, as DC Sessions suggests, that sedentary behavior ("blob-ation" as i call it) is the alternative addiction and physical activity is the norm. certainly, for rodents, it is normal for them to run 2-5 miles per 24 hour period of time, gathering food and spreading their genes. in the wild, those rodents that stayed home watching TV didn't live long enough to spread their genes throughout the population, so there was strong selective pressure to maintain a high level of physical activity -- they are "driven to run." in humans, it's possible that a certain subset of the population were rewarded by their family/community/society for not being physically active (thus, an alternative behavioral strategy arose early and was not strongly selected against).

    or so the argument goes.

  • drugmonkey says:


    While you have good points to keep in mind, there are some additional considerations. First, you aren't really addressing the *difference* in running between the shorter and longer access individuals. Remember, we're looking at an increase across several weeks of daily 24-hr access in the normal circadian type of wheel access and a difference associated with 1 vs 4 hr access in the second paper.

    Species differences in selective pressures for certain behaviors are actually an endorsement of the point. It matters little which particular behavior we are talking about, species are going to be reinforced by many of them. These very natural behavioral processes can become expressed to a degree that is far out on the tail of the spontaneous distribution whether because of individual genetic variation, experience or the interaction of the two. At some point these behaviors become expressed to a pathological degree (and yes, the definition of pathology is not absolute, see my prior comments about the distinction between a diagnosis of substance dependence under DSM criteria and our colloquial understanding of what represents a drug problem). Our question for this area of work is whether wheel running can be used as a model of pathological disruption of reward processes.

    when it comes to humans, well our sources of reinforcement have become stylized, let us say, to an almost cartoonish degree so reference to selection pressures may be beside the point. But then again, wasn't there that paper a few years ago demonstrating how humans were designed to be superior distance runners and could run down, over time, just about any other species?

    With respect to where DC's and Grrl's comments overlap, it is *also* important to keep in mind that relative impoverishment of the naturalistic reinforcers may be a risk factor for substance abuse. Here you may think of many factors including interpersonal, vocational and recreational opportunities. Not just the opportunities, either, but how a given individual interacts with them- being shy or risk averse or whatever may put an individual who is in the presence of a normal set of opportunities into a defacto impoverished category, for example.

  • becca says:

    "Male Homo Sapiens desires company of female Homo Sapiens for mutual talk and grooming behaviors, possibly mating and reproduction. Must be happy, run fast."
    -Kim Stanley Robinson, Forty Signs of Rain

  • mjlaye says:

    I can answer some of the exercise adaptations.

    As said above, a rat running on a wheel (Fischer/Norway 344 cross) for 6 weeks will plateau at about 8km/night of running. This amount of running takes place in about 2 hrs, mostly over the 12 hr dark cycle as you pointed out. It is a sufficient exercise stimulus for some adaptations. For instance, increased glucose uptake in skeletal muscle, enlarged left ventricle mass relative to total heart weight, and a modest 20-25% in skeletal muscle mitochondria (in a mixed fiber type muscle).

    Now compare to a rat undergoing forced treadmill running for 2 hours continuously (PMID: 165725). This amount of activity results, in addition to the increased glucose uptake and physiological cardiac hypertrophy, in a DOUBLING of mitochondria in mixed fiber type skeletal muscle. Same amount of running as the voluntary wheel running rats, but much better oxidative capacity adaptation.

    The likely reason is the intermittent nature of wheel running. Their running behavior is not like you or me going out for a run. It is much more akin to us jumping out of chair running down the hall back and forth a couple times and then sitting down again. Repeat hundreds of time during the dark cycle and you are rat. (Side note: in mice with real time wheel running tracking there is anticipatory voluntary wheel running that consistently occurs right before the lights go off, never been able to figure out why this might be happening)

    It is important to recognize that they call it forced treadmill running for a reason. My former mentor, an experiment rat runner, said that about 25% of the rats just will not run on the treadmill, while others are much more inclined. Whether those are the same or different rats driving those error bars up is an interesting question.

    I am not a neuroscientist, so I would be interested to hear whether others believe that the brain benefits from exercise (reduced depression, anxiety, increased cognitive function) operates on a dose response relationship similar to skeletal muscle oxidative capacity or if you just need to meet minimum threshold (be it time, distance, or intensity).

    Lastly, I'd be interested in request of a future research review I would love to hear your opinion of this newer paper: Selection for intrinsic endurance modifies endocrine stress responsiveness PMID: 20682296 where the authors look at the stress responses in rats artificially selected for running ability (and likely shock avoidance).

  • [...] my prior post, I overviewed a pair of papers which suggested the possibility that rats provided with running [...]

  • drugmonkey says:

    Interesting, mjlaye. Thanks for the analysis and paper links...

  • David says:

    To offer an anecdote from the human perspective - a couple of years ago I was racing triathlons, and training quite hard. I had noticed that a ~3 day break from training left me quite noticably grumpy and irritable, but these mood effects disappeared immediately if I did any sort of endurance type exercise.

    I then crashed my bike and broke my collarbone. Obviously a break in training then ensued. The thing that I found interesting was that I was taking codiene for pain in the week immediately after the crash, and noticed no mood effects whatsoever. About two days after stopping taking the codiene, I found myself quite grumpy and irritable - a state that lasted a couple of days then subsided, but felt identical to my previous 'training withdrawal' experiences.

    It would be interesting to see if blocking post-exercise endorphin production in athletes would produce the same wothdrawal effects as suspension of exercise.

  • [...] in rats. Drugmonkey has also been posting some interesting stuff about addiction and…EXERCISE. And of course, Sci can’t see that and leave it alone. So I’m going back to a paper that [...]

  • [...] Monkey, Is Exercise Addictive? Review of some seminal works that modeled this question using wheel running in rats, and provided [...]

Leave a Reply