Killer O

O

Taggert Brooks points to this excellent news article by George Johnson, who reports:

Epidemiologists have long been puzzled by a strange pattern in their data: People living at higher altitudes appear less likely to get lung cancer. . . . The higher you live, the thinner the air, so maybe oxygen is a cause of lung cancer. . . .

But the hypothesis is not as crazy as it may sound. Oxygen is what energizes the cells of our bodies. Like any fuel, it inevitably spews out waste — a corrosive exhaust of substances called “free radicals,” or “reactive oxygen species,” that can mutate DNA and nudge a cell closer to malignancy.

Back to the epidemiology. Researchers Kamen Simeonov and Daniel Himmelstein adjusted for a bunch of demographic and medical variables, and then:

After an examination of all these numbers for the residents of 260 counties in the Western United States, situated from sea level to nearly 11,400 feet, one pattern stood out: a correlation between the concentration of oxygen in the air and the incidence of lung cancer. For each 1,000-meter rise in elevation, there were 7.23 fewer lung cancer cases per 100,000 people.

“7.23” . . . that’s a bit overprecise, there’s no way you could know it to this level of accuracy. But I get the general idea.

As Brooks notes, this idea is not new. He links to a 1987 paper by Clarice Weinberg, Kenneth Brown, and David Hoel, who discussed “recent evidence implicating reactive forms of oxygen in carcinogenesis and atherosclerosis” and wrote that “reduced oxygen pressure of inspired air may be protective against certain causes of death.”

The idea has also hit the mass media. For example, from a 2012 article by Michael Corvinus in Cracked (yes, Cracked):

One of the disadvantages of living at higher altitudes is that there’s less oxygen in the air, which can suck for those with respiratory problems. One of the advantages of those places, however, is that … there’s less oxygen in the air. A lack of oxygen makes people’s bodies more efficient, which makes them live longer. . . . Dr. Benjamin Honigman at the University of Colorado School of Medicine theorized that the lower levels of oxygen force the body to become more efficient at distributing that oxygen, activating certain genes that enhance heart function and create new blood vessels for bringing blood to and from the heart, greatly lowering the chances of heart disease.

31 thoughts on “Killer O

  1. On the history of this idea, I hereby refer a passage from Erik Pontoppidan, written in 1752, cited in Eilert Sundt, “Om dødeligheten i Norge” (“On Mortality in Norway”), published in 1855:

    “In the middle of the country I hold the air to be cleanest and healthiest, in particular in the mountain villages, where most people do not seem particularly affected by disease, unless it is inherited or obtained by vice. People say, though I would not vouch for this, that in Gudbrandsdalen, which has serious and healthy winds, there are, in particular in the parish of Lesje, several examples of very old people, who by boredom of living so long have travelled elsewhere, in hope of a sooner end.”

    (translated from: “Midt i landet holder jeg luften for allerrenest og sundest, særdeles på fjeld-bygderne, hvor de fleste ikke vide synderlig af sygdom at sige, hvis den ikke er arvelig eller og sanket ved umådelighed. Man siger, dog vil jeg ikke just give min forsikring derfor, at i Gudbrandsdalen, som gjennemblæses af meget ordentlige og sunde vinde, findes, særdeles i Lesje præstegjæld, adskillige exempler af ældgamle folk, som af kjedsommelighed ved at leve længere ere dragne andensteds bort, i håb om at få desto snarere ende derpå.”)

    • Actually, I don’t know about cancer treatments, but before antibiotics, tuberculosis was often treated by going to a sanatorium, which were usually in the mountains, “as health professionals believed that clean, cold mountain air was the best treatment for lung diseases”. See:

      https://en.wikipedia.org/wiki/Sanatorium

      Seems the same ideas keep coming around in a different guise.

  2. The relationship reported in this paper could be explained by residual confounding of elevation with smoking prevalence, due to measurement error in the covariate representing smoking prevalence.

    First, note that there seems to be a substantial relationship (at least at the state level, and so presumably at the county level, where this analysis takes place) between elevation and smoking prevalence:
    http://online.wsj.com/media/info-SMOKE20091113_chart.gif
    https://upload.wikimedia.org/wikipedia/commons/7/7e/US_states_mean_elevation_feet.PNG

    And obviously there is a strong relationship between smoking prevalence and lung cancer rates.

    The main model in the paper includes smoking prevalence as a covariate. But this smoking prevalence variable is measured by “% smoked in lifetime”. This is most likely a pretty noisy indicator of the underlying (not directly observed) covariate that is doing the real causal work, which true covariate might be something like, say, “average # of grams tobacco smoked during lifetime”. Due to this high degree of measurement error in the covariate, it’s very unlikely that controlling for “% smoked in lifetime” statistically equates the observations on the underlying average grams smoked covariate.
    http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0152719

      • >”Exactly! Lung cancer in non-smokers is a very rare disease, so the data are probably dominated by smoking-related lung cancers.”

        I guess it depends on your definition of “very rare”, mortality from lung cancer in non-smokers is about the same as prostate cancer (~ 20k/year)[1,2] and number of cases is about the same as reported for stomach/ovarian/brain cancer (~25k/year)[2; table 1]:

        “More than 161,000 lung cancer deaths are projected to occur in the United States in 2008. Of these, an estimated 10 to 15% will be caused by factors other than active smoking, corresponding to 16,000 to 24,000 deaths annually. Thus lung cancer in never smokers would rank among the most common causes of cancer mortality in the United States if considered as a separate category.”
        http://clincancerres.aacrjournals.org/content/15/18/5626

        [1] http://seer.cancer.gov/statfacts/html/all.html
        [2] http://www.cancer.org/acs/groups/content/@research/documents/document/acspc-047079.pdf

        • OK, I erroneously exaggerated the proportion of lung cancers attributable to smoking. However, even if smoking causes only (!) 85 to 90% of all lung cancers, a difference in smoking propensity would be a good explanation of the observation. Good enough that it deserves prominence.

        • >”However, even if smoking causes only (!) 85 to 90% of all lung cancers, a difference in smoking propensity would be a good explanation of the observation. Good enough that it deserves prominence.”

          I agree, the lung cancer-smoking link is just a case where a lot of propaganda (not saying it was necessarily *bad* propaganda) has lead to a clouded perception of what is going on. Lung cancer is simply a very common cancer, so even if only 10% of cases are due to something other than smoking, that is still a lot.

          However, the paper under discussion here* includes a tsv file with data from 3142 counties. Here is a quick plot of elevation vs their measure of smoking: https://s31.postimg.org/fqhver4zv/Smoke_Elevation.jpg

          There doesn’t seem to be any very strong relationship. For the analysis in the paper they only used data from ~8% of the counties (260) due to various filtering that was performed, so maybe it would look different if only that data was plotted.

          *https://peerj.com/articles/705/

    • I was going to say something similar based on personal experience. I moved from Chicago (~600ft) to Albuquerque (~5k feet) about 3 years ago. In Chicago I smoked recreationally, with drinks, etc. In abq, however, this quickly lost its appeal. Too harsh. Not enjoyable. So based on this, I expect higher altitudes to have lower smoking rates. I haven’t looked at any data on it, though.

  3. The saddest thing about the NYT is that it is the best newspaper.

    Besides, the real danger of oxygen is that it is a fire hazard. And hey there we go–less oxygen, less fire, less smoke, less lung cancer.

  4. Can anyone explain where this data came from? Is it just straight from SEER? In the paper they reference this statecancerprofiles site:

    “Total county cancer incidences (‘All Races (incl. Hisp)’, ‘Both Sexes’, ‘All Ages’) were obtained from the National Cancer Institute (NCI) State Cancer Profiles for the following categories: ‘Lung & Bronchus’, ‘Breast’, ‘Prostate’, ‘Colon & Rectum’, and ‘All Cancer Sites’ (National Cancer Institute, 2013b).
    […]
    National Cancer Institute. 2013b. State Cancer Profiles. Available at http://www.statecancerprofiles.cancer.gov/map/map.noimage.php (accessed March 2013).”
    https://peerj.com/articles/705/

    From skimming the faq they seem to hint that it came from SEER but don’t come out and say it, so maybe it has been further processed in some way?
    https://www.statecancerprofiles.cancer.gov/faq.html

  5. >”Dr. Benjamin Honigman at the University of Colorado School of Medicine theorized that the lower levels of oxygen force the body to become more efficient at distributing that oxygen, activating certain genes that enhance heart function and create new blood vessels for bringing blood to and from the heart, greatly lowering the chances of heart disease.”

    Reminds me of this:
    “Biologists summarize their results with the help of all-too-well recognizable diagrams, in which a favorite protein is placed in the middle and connected to everything else with two-way arrows. Even if a diagram makes overall sense (Fig. 3a), it is usually useless for a quantitative analysis, which limits its predictive or investigative value to a very narrow range. The language used by biologists for verbal communications is not better and is not unlike that used by stock market analysts. Both are vague (e.g., “a balance between pro- and anti-apoptotic bcl-2 proteins appears to control the cell viability, and seems to correlate in long-term with the ability to form tumors”) and avoid clear predictions.”
    http://bml.bioe.uic.edu/BML/Stuff/Stuff_files/biologist%20fix%20radio.pdf

    There really should be a different word for these types of vague narratives. “Theory” should be reserved for something more concrete.

  6. Good heavens, a cause of cancer that I think the UK’s Daily Mail missed.

    Still, I am worried especially as where I live we not only have a good bit of O2 but a lot of dihydogen monoxide around. Things are looking dicey.

  7. I’d like to see more research into altitude and health in old age. There are a lot of nice places to retire to in the American west at altitude. For example, my backpacking uncle built his retirement dream home at 9000 feet. I would be concerned, however, about buying a retirement home at altitude. You may feel great in, say, Aspen at age 60, but will you feel as chipper at 75?

    • I recall reading a study about 25-30 years ago that was done in Colorado that showed downhill migration of retired folks which was postulated to reflect respiratory needs. They did not control for the well known migration of retired folks away from snow. The easy answers tend to reflect superficial questions.

      • You can trade-off altitude and latitude to get less snow by moving further south: e.g., Ruidoso, New Mexico at 6900 feet gets 35 inches of snow, while Aspen, Colorado at 7900 feet gets 135 inches of snow. Sedona, AZ at 4300 feet gets little snow, but averages mid-90s highs in summer (although pleasantly cool at night).

        I’d love to pore over a database of altitude, latitude, and climate, but I’ve never seen one. I’ve always been pretty obsessed with elevation, but most people don’t pay it much attention until they wonder why they’re not feeling so well.

  8. Several people have commented on smoking and altitude. Procrastinating, I poked around and found this, “Smoking, acute mountain sickness and altitude acclimatisation: a cohort study”, from the National Key Laboratory of High Altitude Medicine, (China), where presumably all your altitude/health related questions can be answered. Not commenting on its data and conclusions, I’ll just note that it’s amazing that the study design was to recruit several hundred construction workers and get them jobs at a site 4000m above sea level. I don’t think this scenario ever came up in the human subjects training I’ve done.

    • Raghuveer:

      Procrastinating, I followed your link and read the summary parts of this study. It does not seem to answer all our attitude/health related questions. Amazingly, despite it involving the intervention of moving people to a new location, it does not appear to be an experimental study comparing people who live at low and high altitudes. Rather, it seems to be a purely observational study of smokers vs. nonsmokers, all of whom were moved. That is, I did not see in the summary any mention of a control group who stayed at low altitude, even though that would seem like a natural thing to do in such a study, given the trouble it must have taken to move those people.

      In retrospect, I guess it makes sense. They studied a sample of 400 people, which should be enough to detect some effects of smoking, because smoking has such huge effects on health (and in particular on respiration). But health effects of living at high or low altitude must be much smaller, and you’d need a much bigger sample to study these. An experiment on 400 people would not be enough to compare effects of high or low altitude living; it would be just be enough for clever researchers to find statistically significant comparisons and, if they’re lucky, get into the tabloids. If you want to study the effects on cancer risk of living at altitude, you’ll need a large cohort study, or very good intermediate measurements on individuals.

      • Good points! About “all our questions,” looking back at my comment, I was unclear. I meant “where presumably all your altitude/health related questions can be answered” to refer to the National Key Laboratory of High Altitude Medicine, not this particular paper. The lab is part of the “High Altitude Medical Research Institute,” by the way. I can’t find much information about this, but presumably it’s full of people studying effects of high altitude.

  9. I was telling my wife and son (10) about this news story, and my son immediately remarked that the number of people living up on the mountains must be much fewer than on ground level, so the comparison’s not fair. And then my wife said that the group living up in the mountains must be self-selecting; the unfit people would probably move to lower levels, leaving a baseline level of fitness much higher in the mountains.

  10. I agree with the smoke confounding idea. But what about confounding due to pollution levels? My first thought was, the higher the altitude the less likely you are to live in a place with a lot of traffic congestion, factories, etc. Could this explain the finding? Any thoughts?

    • There sure might be some positive effect of higher altitudes also due to lower atmospheric pressure, so you could assume that–even if the number of factories and volume of traffic is held constant–people are exposed to less air pollution. Poor air quality is associated with increased cancer risk.
      It would still be quite interesting to see whether smoking prevalence is lower at higher altitudes (i.e. perhaps the buzz you get from smoking is greater with less O, so you don’t smoke so much).

  11. This whole idea (low oxygen -> lower cancer risk) seems like a very straightforward thing to test in animals, which one can move to high and low oxygen environments more easily than construction workers, and in which smoking rates and exposure to traffic congestion are, I would guess, rather low. This is, in fact, noted in the Simeonov and Himmelstein paper, though only briefly, and I didn’t dig further.

  12. These studies puzzle me — some great comments above. But,
    1. How is this information useful exactly?
    2. How is this at all predictive? Even if the relationship is real and causative, there is no real reason is there to expect a *change* in altitude will have any more impact than say getting fitter?
    3. Higher => colder; is it the cold or the O? I’m not totally sure but if O and Temp are in a direct relationship, would the results be the same?

    • I guess your right, but not all science has to be actionable to be valuable; you could still get a better base rate for other, more actionable, research.

Leave a Reply to Raghuveer Parthasarathy Cancel reply

Your email address will not be published. Required fields are marked *