It was this year’s big workplace health story. A major Oxford University study published online on 6 October 2016 concluded the classification of night work as a cause of breast cancer ‘is no longer justified’.
It was good news. A BBC headline blared “Breast cancer risk 'not increased' by night shifts”; the Daily Mail chipped in with "Working night shifts does NOT raise the risk of breast cancer." And Cancer Research UK (CRUK), which co-financed the study, said it hoped the conclusion “reassures women who work night shifts.”
The Health and Safety Executive (HSE), another co-funder, also welcomed the ‘vital’ study. It was one occupational health priority that could be shoved down the to-do list.
BIG DEAL Getting it right on health impacts of night working is important. The number of UK workers who regularly work through the night has risen dramatically, with the rise far more marked in women than men. more
Backed by the Medical Research Council, as well as HSE and CRUK, the research combined new results from three large studies, including nearly 800,000 women from the Million Women Study, EPIC-Oxford and UK Biobank cohorts, with data from seven already published studies from the USA, China, Sweden and the Netherlands.
The resulting ‘meta-analysis’, published in the prestigious Journal of the National Cancer Institute (JNCI), was on an impressive scale, using data from 1.4m women in total.
Big numbers count for a lot in epidemiology. The study, co-authored by some big names in the field, was certain to have an impact when it concluded: “The totality of the current prospective evidence suggests that night shift work, including long-term night shift work, has little or no effect on breast cancer incidence.”
Lead author Ruth Travis (right), a CRUK-funded scientist at the University of Oxford’s Cancer Epidemiology Unit, confirmed: “We found that women who had worked night shifts, including long-term night shifts, were not more likely to develop breast cancer, either in the three new UK studies or when we combined results from all 10 studies that had published relevant data.”
The researchers concluded the decision in 2007 by the World Health Organisation’s International Agency for Research on Cancer (IARC) to rank night work as a ‘probable’ cause of breast cancer in women was “necessarily based on limited epidemiological evidence” but had, it turned out, got it wrong.
Little or no effect?
This was a game-changing departure from the previously accepted position.
IARC’s classification had prompted a wave of new research, regulatory efforts and union calls for action to protect night workers.
The Danish authorities had recognised breast cancer caused by night work as a genuine work-related condition in some jobs that was eligible for employers' insurance payouts. Nurses, airline crew and soldiers were among those compensated.
An HSE-commissioned study, commissioned from the Institute of Environmental Health, Institute of Medicine and Imperial College London and published in 2012, concluded each year there are 555 deaths and 1,969 new cases of breast cancer in Great Britain attributable to shift work.
But the JNCI study made it crystal clear that IARC, HSE and the research that had gone before had got it wrong.
It was a big deal. Each year in the UK over 50,000 women are diagnosed with breast cancer and around 11,500 die.
HSE estimates 2 million women – about one in every six female workers – are currently working shifts, with over half a million of these involving night work. The numbers are rising fast. Professor Andrew Curran, HSE’s chief scientific adviser, said: “Breast cancer is the most common cancer in women so it was vital for us to fund work in this area to establish if there is a link to night work.”
He added: “This study has shown that night shift work, including long-term shift work, has little or no effect on breast cancer incidence in women.”
CRUK’s Sarah Williams said: “This study is the largest of its kind and has found no link between breast cancer and working night shifts. Research over the past years suggesting there was a link has made big headlines, and we hope that today’s news reassures women who work night shifts.”
The message was reinforced in a 14 October follow-up blog post from CRUK, which explored why “science had seemingly changed its mind”, concluding “this study is the largest of its kind, which suggests the results are solid.” It added it “found the link just wasn’t there.”
Women should concentrate on sorting out their wayward lifestyles, CRUK indicated.
“Breast cancer is the most common cancer in the UK and research to fully understand the different risk factors is vital so that we can give women clear health advice,” CRUK told Hazards. “Women can reduce their risk of breast cancer by keeping a healthy weight, drinking less alcohol and being active.”
But there were problems. The research may not have been as definitive as the authors claimed. And statements absolving nightwork of blame for breast cancer may have been more than premature.
Good night worries
The three new cohorts investigated by the Oxford University team found “no association of night shift work with breast cancer incidence in any of these UK studies.”
Fold in the other seven studies, including ones that had found an association, and the overall picture became one of “no evidence of any association of breast cancer incidence with shiftwork and, in particular, no evidence of any increase in incidence with 20 or more years of night shift work.”
But the new evidence dominating the analysis came from general health studies on a grand scale, not studies designed specifically to study the impact of working hours patterns on cancer rates.
Other equally eminent researchers, whose work over decades included several purpose-designed studies on night work and cancer risks, were seriously concerned about both the design of the JNCI study and the reliability of its conclusions.
The mean age of the women in the three UK studies was worryingly old, and the follow up period unusually short. The information on working patterns was at times less than optimum.
Hazards approached three researchers – Eva Schernhammer, Richard Stevens, and Johnni Hansen – who between them have contributed to the lion’s share of the more influential original research on night work and breast cancer.
How Oxford messed up
Two of Eva Schernhammer’s studies were among the seven previously published papers included in the new JNCI meta-analysis. Both had found an association between rotating night work and breast cancer risk.
The paper published in JNCI in 2001 had found an 8 per cent increase in breast cancer risk among nurses working rotating night shifts for up to 29 years. The women who had worked these shifts for more than 30 years had a 36 per cent increase in breast cancer risk.
A second, published in the journal Epidemiology in 2006 and which also looked at rates of the disease in nurses, had reported “a modestly elevated risk of breast cancer after longer periods of rotating night work.”
These weren’t though Schernhammer’s most recent papers on the topic. The 2001 Nurses’ Health Study, particularly, had been updated and extended, examining the causes of far more cancer cases than the Travis paper and confirming the original clear association.
The Harvard Medical School researcher (right) said a forthcoming paper on the nurses’ study, and whose findings were known to the JNCI editors and had been presented at a series of medical conferences, included 5,841 cases of breast cancer in night workers.
Take out the breast cancers in night workers in the two Schernhammer papers in the meta-analysis and Travis was looking at 3,001 cases overall.
The three new studies in Travis’ paper – which declared it was based on the ‘totality’ of the prospective evidence – involved just 768 breast cancer cases involving any history of night work.
The Travis paper acknowledged “the total number of cases of breast cancer arising in women in prospective studies who had reported long-term night shift work is still less than 1,000; this will increase with longer follow-up and publication of further studies, but it is already more than four times as many as the number available for the 2007 IARC review.”
Low numbers notwithstanding, it was these prospective studies that provided the reliable evidence, the Travis paper said. “Restriction to prospective studies is important when trying to detect or refute moderate hazards as it avoids the moderate biases that can result from retrospective methodology,” it noted.
STUDY THE DIFFERENCE
A case-control study compares individual who have a condition with those who do not and compared retrospectively how frequently the exposure to a risk factor is present in each group to determine the relationship between a risk factor and a disease.
A prospective, or cohort, study takes one or more samples (called cohorts), follows them and makes ongoing evaluations to determine which participants develop a disease and which risk factors are associated with it.
A meta-analysis takes data from several selected studies to develop a single conclusion that has greater statistical power.
Study design 101, George Washington University.
The Travis paper made big claims on the back of these low numbers. But the authors made confusing statements that appear to undermine their headline claim of ‘little or no evidence’ of any cancer risk.
In one sentence the Travis paper notes its findings “exclude a moderate association of breast cancer incidence with long duration night shift work.”
In the next it concedes the study’s “main limitation” – the low numbers in the prospective studies – means “an increase in the relative risk of breast cancer incidence of only a few per cent cannot be ruled out.”
This ‘few per cent’ could translate to hundreds if not thousands of new night work-related diagnoses in the UK each year.
It wasn’t just the numbers that fell short. Travis’ data sources were also problematic.
The mean age of the 500,000 plus Million Women Study participants was over 68, and late-50s for the 22,000 plus EPIC-Oxford group.
These studies, split between ‘never’ and ‘ever’ worked nights, meant “as little as a single shift at some time in their entire life” could classify a person as a night worker, said Schernhammer.
OLD PROBLEM The Oxford University study giving night work a cancer all-clear included many retired workers who’d not worked nights for years. But you need to look at younger workers with a recent long-term history of night work if you want to spot raised breast cancer rates.
“In the Nurses’ Health Study, ever versus never night shift work is too crude to show any associations with cancer - it really never did,” she said. “You tend to need longer exposures to show an effect on risk.”
The short follow-up period on this group, at around three years, was a major drawback. It gave only a snapshot of the night work risk to a group of retired women in a very narrow time window.
The Travis study examined whether – sometimes minimal – exposure to night work could increase the risk of breast cancer during these two to three arbitrary years in the participants’ often post-retirement life.
But the risk of women developing breast cancer appears to wane in the years after night working ends. And, says Schernhammer, the higher risk is seen in women with long exposures – at least 15 years – early in their careers.
Looking at retired workers with no recent exposures misses the point and the cancers.
Absurd claims, troubling study
Richard Stevens (right), an epidemiologist at the Department of Community Medicine and Health Care at the University of Connecticut, is also a prominent researcher on occupational breast cancer risks.
One of his papers, on the suppression of melatonin secretion in night workers as a possible mechanism behind an increased breast cancer risk, is cited in the Travis paper.
Stevens told Hazards it was “absurd” that the night work association with breast cancer was being dismissed on the back of a “troubling” paper by “a distinguished group of experienced researchers who should have known better.”
Stevens said three of the most senior authors, Richard Peto, Valerie Beral and Timothy Key, had published “deservedly influential” studies on both hormone replacement therapies and oral contraceptives and breast cancer risk.
Like these studies, the “primary proposed mechanism for a possible shift work effect on breast cancer risk is also by hormonal perturbations,” Stevens said.
“The lesson from both analyses was that time since cessation of exposure to these hormonal preparations rapidly attenuates the excess risk; risk becomes virtually normal within five years of quitting no matter how long they had been used before that. And both reports used primarily case-control studies to reach their conclusions.”
Unlike prospective studies, they give a valuable insight into why and where you might look for an association. Understanding the process, something integral to his own research, was important, he indicated.
Some authorities, including IARC and the US National Institute for Environmental Health Sciences (NIEHS), stress understanding how an exposure causes cancers is an important supplement to studies determining the extent of the risk, and can aid their interpretation.
The centrepiece of the Travis research, the prospective Million Women Study, had critical failings, he added, with “virtually no ability to examine recent night work due to the age of the women (average 68 years) when asked about past night work experience.”
The journal that published the Travis study, JNCI, had blundered in not seeking an external review of the paper from researchers with substantial experience on the topic, Stevens said.
He and Schernhammer, together with Danish researcher Johnni Hansen and Scott Davis, a professor of epidemiology in the University of Washington’s School of Public Health, are the stand-out epidemiologists on night work and breast cancer.
Not one of them was asked to review the paper. “We are the four epidemiologists who have been working for by far the longest on the epidemiology of night work and breast cancer,” said Stevens.
Counting the survivors
Johnni Hansen, (left) too, believes what he sees as the study’s inherent flaws – an “unusually old” cohort with no recent exposures and an “exceptionally short” follow up – invalidates its conclusion that night work doesn’t cause breast cancer and that the IARC cancer designation should be withdrawn.
This is “misleading and destructive for further studies of night work and the risk of cancer,” the Danish Cancer Society researcher said.
His 2012 case-control study, co-authored with Stevens, found nurses who worked rotating shifts after midnight had a “significantly increased” risk for breast cancer compared to nurses with permanent day work. No association was found in a small group of nurses with evening work and no night work.
The paper, published in the European Journal of Cancer, concluded: “The results provide further evidence that night shiftwork may increase the risk for breast cancer and suggest that the largest impact on risk is associated with the most disruptive shifts.”
It was part of a large body of important evidence overlooked in the Oxford University study, Stevens said.
Travis’ meta-analysis “excluded case-control studies, of which there are many, for no good reason,” he said.
The International Agency for Research on Cancer (IARC) is standing by its ranking of night work as a ‘probable’ cause of female breast cancer.
The World Health Organisation cancer agency said the 2007 Group 2A designation had stimulated extensive new studies, and it would undertake a re-evaluation towards the end of the 2015-19 research cycle.
Kurt Straif, who heads up IARC’s monographs section, told Hazards: “Many new studies, positive and negative and with different strengths and limitations have already been published, and until any update the current IARC evaluation of shiftwork involving circadian disruption as probably carcinogenic to humans stands and remains the official classification.”
Hansen said neither the Million Women Study nor the EPIC-Oxford study were designed to investigate the problem, with a single night work question only recently added.
For the Million Women Study, “most, if not all,” had retired by the time they answered the shiftwork question. Echoing a concern raised separately by both Schernhammer and Stevens, Hansen said in the years after cessation of exposure, the passage of time “attenuates or fully eliminates” previously observed risks.
It is a problem known to researchers as “truncation bias”, said Hansen.
“Travis’ cohorts represent survivor populations, which means that if night work increases the risk of breast cancer then the cohort members would most probably have had their breast cancer prior to the recently included question on night work.”
All four studies with a follow-up time of eight or more years included in the Travis meta-analysis, did in fact find an association between night work and breast cancer. All those with a shorter follow-up time did not.
Bad science, bad news
Hazards approached Ruth Travis, lead author of the JNCI paper, and its funders, HSE and CRUK, for comments on the questions raised about both the methodology and conclusions of the research.
Travis said she would not respond until “correspondence about the many issues you raise” had been published in the scientific literature “in the next few months.” Hazards understands JNCI has received strongly worded rebuttals from a number of researchers, including Schernhammer, Stevens and Hansen.
Travis did say, though, that findings of the in-press update to the Nurses’ Health Study, whose authors include Schernhammer and Stevens, were included in her paper’s supplementary analyses, whose findings were “materially unaltered” by their inclusion.
This might have been expected as with fewer than 200,000 participants, a significant risk in this purpose designed nurses’ study could still be swamped by the “null association” conclusion of the near 800,000 mostly retired, elderly women including in Travis’ three UK cohorts.
Faced with direct questions about what some see as serious flaws in the paper it financed, HSE distanced itself. The research was “independent”, it said, adding: “As declared in the publication: ‘The sponsors had no role in the design of the study; the collection, analysis, or interpretation of the data; the writing of the manuscript; or the decision to submit the manuscript for publication’.”
The HSE spokesperson said the regulator “will continue to monitor developments in this area”, adding: “If you have any questions about the research methodology these would need to be raised with the University of Oxford.”
Hazards invited Travis again by email to defend the JNCI paper’s methods and conclusions but she did not respond.
CRUK did respond. A spokesperson told Hazards: “Given that such a large study, which has managed to look at the effects of working night shifts for long periods of time (eg. 20 years or more), has found no link with breast cancer, we are reassured that night shift working is unlikely to increase the risk of breast cancer.
“But, no study is perfect. And there are limitations to this study. There were a variety of ways in which the data on night shift working was collected, and the individual studies will have defined ‘night shift working’ in slightly different ways.”
The 2007 IARC classification of night work as a ‘probable’ cause of breast cancer in women was “based on the best available evidence at the time”, she said, but the new study changed things as “it is such a big improvement on the evidence we have to date.”
The cancer charity added: “We stand by our original comments and press release, and reiterate that, as with all research, we can’t categorically say that night shift working doesn’t increase the risk of breast cancer – but that this is a very solid piece of evidence that suggests that there is no link.”
It’s a line that dismays seasoned researchers who have spent decades studying the issue.
“They base their conclusion on a poor study, but even worse is that their conclusion may hinder preventive initiatives for night workers,” said Johnni Hansen.
Eva Schernhammer concurred. “It seems like bad science”, she said. Given the limitations of the Travis study it was “not surprising” the study found no effect. She said these “many shortcomings – as unanimously asserted by my colleagues and myself – preclude it from making the conclusion that there is no association between night work and breast cancer risk.”
Richard Stevens was blunt. “Why was the paper written in the first place?” he asks. “And why was it published in a high visibility journal like JNCI?”
Ruth C Travis and others. Night shift work and breast cancer incidence: Three prospective studies and meta-analysis of published studies, Journal of the National Cancer Institute, volume 108, number 12, published online 6 October 2016.
Kurt Straif and others. Carcinogenicity of shift-work, painting, and fire-fighting, The Lancet Oncology, volume 8, number 12, pages 1065-1066, December 2007.
IARC. IARC Working Group on the Evaluation of Carcinogenic Risks to Humans: Volume 98 Painting, firefighting and shiftwork, 2010.
The burden of occupational cancer in Great Britain, RR800, HSE, 2010.
The burden of occupational cancer in Great Britain – Breast cancer, RR852, HSE, December 2012.
Schernhammer ES and others. Rotating shifts and risk of breast cancer in women participating in the Nurses’ Health Study, Journal of the National Cancer Institute, volume 93, number 20, pages 1563-8, 2001. Related editorial: Hansen J. Editorial: Light at night, shiftwork, and breast cancer risk, Journal of the National Cancer Institute, volume 93, number 20, pages 1513-5, 2001.
Richard Stevens and others. Electric power, pineal function, and the risk of breast cancer, FASEB Journal, volume 6, number 3, pages 853-860, 1992.
Johnni Hansen and Richard Stevens. Case-control study of shift-work and breast cancer risk in Danish nurses: Impact of shift systems, European Journal of Cancer, volume 48.11, pages 1722-1729, July 2012.
Why night working and health is a big deal
Getting it right on health impacts of night working is important. The number of UK workers who regularly work through the night has risen dramatically, with the rise far more marked in women than men. An analysis by the TUC found that the number of people who work night shifts increased by 275,000 (9 per cent) between 2011 and 2016 to 3,135,000. Britain’s large army of night workers now accounts for one in eight (12 per cent) employees.
The TUC analysis released on 29 October 2016 noted that while most night workers used to be men working in manufacturing plants, this had ‘changed drastically’. In 2016 one in seven male employees (14 per cent) were night workers and one in 11 (9 per cent) female employees. However, women account for more than two-thirds (69 per cent) of the growth in night working over the past five years. Between 2011 and 2016 the number of women regularly doing night work increased by 190,000, while for men it increased by 86,000.
The two most common professions for female night workers are care working and nursing. Male night workers are most likely to work in protective service occupations (military, security, policing) and road transport.
TUC general secretary Frances O’Grady commented: “Whether its nurses looking after patients, or police officers keeping our streets safe, we all depend on Britain’s army of night workers.” She said these essential workers must be properly protected. “Employers must play fair and play safe, or public safety will be put at risk and the families of night workers will suffer.”
In addition to evidence linking night work in certain professions, notably nursing, to a higher risk of breast and other cancers, negative health impacts of night work include heightened risks of cardiovascular disease, diabetes and depression. Night work has also been linked to an increased injury risk. The journey home from a night shift has also been linked to an increased risk of a road traffic accident.
The Health and Safety Executive estimates over half a million women in Great Britain are currently working shifts that involve night work.
|Richard Stevens’ critique of the Travis paper|
|“There are two disturbing aspects of the Travis et al publication: Why was the paper written in the first place? And why was it published in a high visibility journal like JNCI?
“First, the authors are a distinguished group of experienced researchers who should have known better. The three most senior authors on Travis et al., Drs. Beral, Key, and Peto, were also collaborators on two very large pooled analyses of breast cancer risk in women and use of Hormone Replacement Therapy and of Oral Contraceptives both published in the Lancet in 1997 and 1996, respectively. These were deservedly influential publications.
“The lesson from both analyses was that time since cessation of exposure to these hormonal preparations rapidly attenuates the excess risk; risk becomes virtually normal within five years of quitting no matter how long they had been used before that. And both reports used primarily case-control studies to reach their conclusions.
“Yet the Travis analysis of the Million Women Cohort had virtually no ability to examine recent night work due to the age of women (average 68 years) when asked about past night work experience. The primary proposed mechanism for a possible shift work effect on breast cancer risk is also by hormonal perturbations. In addition, the Travis et al. meta-analysis excluded case-control studies, of which there are many, for no good reason.
“The second troubling issue is how it got published in an otherwise good journal. Typically, a paper of this potential magnitude has three outsides reviewers, at least one of whom should have had long experience in the specific field covered by the submitted manuscript. I was not invited to be one of the three reviewers by JNCI; I contacted Eva Schernhammer, Johnni Hansen, and Scott Davis, and none of them was asked by the journal to review the paper either.
“We are the four epidemiologists who have been working for by far the longest on the epidemiology of night work and breast cancer. Any of the four of us would have quickly noticed the severe flaws of the Travis paper and pointed them out to the editors at JNCI. The Travis et al. paper did not get adequate review by JNCI on this very important topic, and the question is why not?”
Richard Stevens, UConn Medical School, 1 December 2016.
An Oxford University study concluded the classification of night work as a cause of breast cancer ‘is no longer justified’. The media trumpeted the good news, and women were told they should be ‘reassured’. No one demurred. Then Hazards editor Rory O’Neill discovered the research may have got it seriously wrong.
|•||Little or no effect?|
|•||Good night worries|
|•||How Oxford messed up|
|•||Absurd claims, troubling study|
|•||Counting the survivors|
|•||Bad science, bad news|
|•||Why night working and health is a big deal|
|•||Richard Stevens’ critique of the Travis paper|