Skip to content

“Nightshifts Linked to Increased Risk for Ovarian Cancer”

Zosia Chustecka writes:

Much of the previous work on the link between cancer and nightshifts has focused on breast cancer . . . The latest report, focusing on ovarian cancer, was published in the April issue of Occupational and Environmental Medicine.

This increase in the risk for ovarian cancer with nightshift work is consistent with, and of similar magnitude to, the risk for breast cancer, say lead author Parveen Bhatti, PhD, and colleagues from the epidemiology program at the Fred Hutchinson Cancer Research Center in Seattle, Washington.

The researchers examined data from a local population-based cancer registry that is part of the Surveillance Epidemiology and End Results (SEER) Program. They identified 1101 women with advanced epithelial ovarian cancer, 389 with borderline disease, and 1832 without ovarian cancer (control group).

The women, who were 35 to 74 years of age, were asked about the hours they worked, and specifically whether they had ever worked the nightshift.

The researchers found that 26.6% of the women with invasive cancer had worked nights at some point, as had 32.4% of those with borderline disease and 22.5% of those in the control group.

In the entire cohort, the median duration of nightshift work was 2.7 to 3.5 years. The most common types of nightshift jobs were in healthcare, food preparation and service, and office and admin support.

I hadn’t known so many people worked night shifts, but I guess these numbers make sense given that they’re asking people whether they’d ever worked nights. I wonder if I’d count? I taught a night class for a couple of semesters. [From comments: No, my evening class wouldn’t count, they define nightshift as “the hours between midnight and 04:00.” In that case, I didn’t realize so many people had worked in such jobs.]

Here’s the punchline:

The researchers conclude that working nights is associated with an increased risk for both invasive ovarian cancer (odds ratio [OR], 1.24, 95% confidence interval [CI], 1.04 – 1.49) and borderline disease (OR, 1.48; 95% CI, 1.15 – 1.90).

The Bayesian in me suspects the true population odds ratios are on the low end of this range, and the Uri Simonsohn in me is suspicious that the low ends of these confidence intervals are so close to 1.0. Indeed, a look at the statistical analysis section of the article suggests the researchers had various degrees of freedom which could induce small changes in the p-value. I also worry about the sensitivity of their results to their choice of adjustments for pre-treatment variables. They say they did their analysis in Stata, which is fine, but I’m not clear what exactly they did in their adjustments.

This is not to say I think the published findings are wrong. Any particular study is a brick in the wall, it provides some information and future studies can give more.

P.S. I found the above news article via Google after reading this summary which sent me to a not-so-detailed unsigned news report that was not so detailed and had no link to the original study. The unsigned report was attributed to HealthDay: “Daily Health News and Medical News for Licensing & Syndication.” Creepy.


  1. dmk38 says:

    But if methods & inferences aren’t valid, it is not even a brick….

  2. Fernando says:

    I can’t comment on the details but it always strikes me that it is things many people want to believe are bad turn out to be bad in observational studies.

    Confirmation bias? I mean, what if they’d found the opposite. Maybe it would have been shot down.

    You might argue that there are priors, but if the prior is informative, the data weak, and scientists have to publish, I suppose we can fit the data to the prior. I would call that confirmation bias.

    But again, I’ve not read the paper only Andrew’s description.

    PS I’d say specification searches are informative priors. E.g. instead of putting priors on the parameters you play with the data until you get the parameters you would have put in the prior to begin with. That would be much more honest.

  3. Rahul says:

    “whether they had ever worked the nightshift.”

    Isn’t that a particularly uninformative question to ask?! The body presumable reacts differently to one night shift a month as opposed to 5 a week?!

    • Andrew says:


      As can be inferred from the above post, they also asked duration and they did some duration analyses, which I think might not have been statistically significant. And, as I’m sure you’re aware, when results of a study are reported, it is the comparisons that reach the 5% level of significant that get emphasized.

      • Rahul says:

        Zing. I’m suitably chastised. :) My bad. My sloppy reading didn’t realize they did use duration.

  4. Rahul says:

    Could there be some partial reverse causation here?

    Could cancer treatment itself or the attendant high medical-bills nudge patients to adjust / supplement work-schedules to make-up or add work-time?

    • Anonymous says:

      Not an area I know anything about but there was a study re breast cancer and night-shift work with what appears to be better controls on the definition of night shift (IIRC from reading it quickly last year) that probably does not suffer from the health insurance problems since it was in Denmark.

    • Why would there? It would make sense for induced sleep problems etc. to cause health problems (disrupting the natural sleep cycle, the potential oncostatic properties of melatonin), whereas the reverse causation (opting-into night shifts for work reasons etc.) doesn’t hold for the entire population, cf. the example from Denmark below.

      More generally, this is a field in which we have strong prior knowledge about the adverse health effects of night shifts, which should aid our identification.

  5. BMGM says:

    > I wonder if I’d count? I taught a night class for a couple of semesters.
    No, unless you taught after midnight, you worked evening shift, not night shift.

    My mom worked as a nurse at the VA, who required every nurse (even those who work day shift) to work at least 5 graveyard shifts per month. The sleep disruption was terrible and lasted half the month as she shifted her schedule to try to prepare for it, worked graveyard, and then recovered from it.

    Since the body performs many cellular repairs and lymphocytes search and destroy malignant cells while we sleep, these findings are not surprising.

    My mom quit the VA after one year. (She was a single mom of two and it was just too much.) Now that she is in her 70s, we are watching the body count of those who stayed mount up.

    The VA is making changes now that it is known that, though working the night shift increases risk, forcing people to switch back and forth appears to be the most risky. My mom was just an RN, but she knew that after a few months.

    • Andrew says:

      Aaahhhh, yes, reading the article more carefully I see they defined nightshift as “the hours between midnight and 04:00.”

    • Fernando says:

      In Spain much of the young (and not so young) population stays up until 5-6am every Friday and Saturday night. International pilots routinely experience severe time changes. Most business men and women who regularly travel internationally ditto, security guards also, many lawyers and investment bankers do all nighters at least once a month.

      If the theory is about removal of malignant cells during sleep (and I know nothing about this) then: (a) it would seem that sleep deprivation, rather than when one sleeps, or changing sleeping patterns, is the problem (though the latter could cause insomnia); and (b) other types of cancers should also be affected.

      This study has multiple implications that can be tested.

      • Most people don’t adjust well to sleep pattern disruptions, so a sleep pattern disruption is in effect sleep deprivation. This might not affect young people nearly as hard as older people for a number of reasons: young people can generally sustain sleep better than older persons, an uniform increase in risk wouldn’t affect the young as much, etc.

        As far as I know, investment bankers have pretty high morbidity and mortality, compared to their peers. I wouldn’t know about international pilots, but it was my impression that flying intercontinentals was mostly a thing for relatively young pilots.

        • Fernando says:

          Good to know age might be a moderator (uff!). So what is the age distribution of nurses?

          I don’t know about the morbidity and mortality of I. bankers. I can only presume it is confounded by a whole bunch of other things like stress, sedentary behavior, etc… Very hard to pin it down to sleep patterns.

          My guess is it’s older pilots that fly intercontinental. The pay is better, the responsibility greater, and the plane bigger, which for whatever the reason adds to prestige.

          BTW not sure the debate is whether sleep is necessary or not. My strong prior is it’s necessary. Rather, the question is whether staying up until 4am once a month, for however many months, has an effect on health of the magnitude reported.

          • Fernando says:

            PS I would have found the evidence presented in the paper a whole lot more convincing if they had pre-registered a study protocol before carrying out the analysis.

            That would assuage many of the worries expressed in this thread.

    • “Since the body performs many cellular repairs and lymphocytes search and destroy malignant cells while we sleep, these findings are not surprising.”

      I’m extremely skeptical of this claim, particularly the latter half.

      • Skeptical about what? That sleep is important in securing immune system function?

        • What is known, and pretty much all that is known, is that there’s a correlation between sleep deprivation and immune system suppression. That is not quite equivalent to “sleep is important in securing immune system function” and is very much not equivalent to “the body performs many cellular repairs and lymphocytes search and destroy malignant cells while we sleep”.

          • There are some experiments wherein the researchers induce sleep deprivation and study the effect on different immunological reactions, such as blastogenesis.
            See for example Moldofsky H, Lue F, Davidson J, Jephthah-Ochola J, Carayanniotis K, Gorczynski R. The effect of 64 hours of wakefulness on
            immune functions and plasma cortisol in humans. Horne J (ed.): Sleep 88. Stuttgart: Fischer 185-7.

            • Fernando says:

              Having a spoonful of sugar shoots up my insulin, and insulin resistance is related to many chronic diseases.

              The above does not imply that one spoonful of sugar a month will give me diabetes. Moreover, given my preference for chocolate, I am willing to make some tradeoffs at low dosage.

              • “The above does not imply that one spoonful of sugar a month will give me diabetes.”

                It most certainly doesn’t, but exposure to sugar increases the risk of diabetes. We’re not talking about whether or not sleep disturbances will give you ovarian cancer, but whether or not it’ll increase the chance.

                “Moreover, given my preference for chocolate, I am willing to make some tradeoffs at low dosage.”

                Very understandable! But – that’s a totally different matter.

              • Fernando says:


                OK, let me be more precise: “it will not increase my risk in any substantive way”?

                Again the question is not so much whether sleeplessness is bad, I am sure it is, but rather whether at low dossages it is something to be concerned with.

                In this regard I am not convinced that a single observational study, without a pre registered study protocol, and given the perverse (and getting worse ) incentives faced by most scientists, is enough to make claims about substantive increases in risk. Let alone suggest regulatory changes to working practices.

                I wish they had pre-registered their theory, methods, and priors.

  6. This provides a good overview of potential causal mechanisms: