“Breast cancer screening could cause more harm than good,” The Daily Telegraph has today reported.
The newspaper says up to half of the benefit that some women get from living longer lives could be cancelled out by others having misdiagnoses or unnecessary treatments. For example, some women may have surgery to remove cancers that would not have gone on to cause them any problems. Diagnosis of these cancers is referred to as “overdiagnosis” and their treatment as “overtreatment”.
The news is based on a study that updated the Forrest report, the 1986 research that led to the start of the UK breast screening programme. This older analysis included the data available at the time, but did not examine the harms of overdiagnosis or false positives. The current analysis updated the original study’s calculations by adding in recent data, and taking into account these potential harms of screening.
Unsurprisingly, the inclusion of these harms reduced the benefits estimated for the screening programme. The updated model including harms suggested that the screening programme may not have yielded a net benefit until about 10 years into the programme, although the balance shifted towards benefit after this. However, no model is perfect, and the researchers acknowledge that their analysis has limitations. For example, the model is based on results from the available trials of mammography, some of which are decades old. Screening techniques and treatments may have improved since then.
Estimating the balance of benefits and harms of screening programmes is complex, and models such as this can help to estimate this balance. An independent review of all relevant evidence is currently ongoing and due for publication next year.
Where did the story come from?
The study was carried out by researchers from the Faculty of Medicine at the University of Southampton. It received no specific funding. The study was published in the peer-reviewed British Medical Journal.
In general, newspapers covered this study well. While news headlines generally suggest that breast cancer screening does more harm than good, the results are slightly more nuanced than this, with the study predicting there would be an overall benefit from screening, but only after 10 years.
What kind of research was this?
This was a modelling study that aimed to examine whether mammographic breast cancer screening could be doing more harm than good. It was performed in response to recent questioning of the benefit of mammography screening in a systematic review from the Cochrane collaboration.
To examine the issue, the current study used recent research figures to update the analysis in the 1986 Forrest report - the research that led to screening being offered in the UK. This original report had suggested that screening would reduce the death rate from breast cancer by almost a third, with few harms and low cost.
Of note, the model in the Forrest report used the data that was available at that time, which suggested that overdiagnosis might not be a problem. Overdiagnosis is where a woman is treated for a potential cancer identified by screening that would otherwise never have gone on to cause her any problems. However, as it can be difficult to tell which cancer will go on to cause problems and which will not, doctors may decide to treat it just in case it does. In addition, some women who undergo screening will have an abnormal mammogram, but on further investigation will be found not to have cancer (false positives). Some argue that screening could lead to more harm than good due to these potential harms, as a proportion of women will have to go through unnecessary stress and treatments, such as removal of some or all breast tissue.
In order to assess issues such as the impact of screening programmes, scientists turn to a technique called modelling. The technique takes a theoretical population, uses data about factors such as the risk of a disease or of particular outcomes, and then predicts what outcomes would occur among that population. Modelling is often used to help determine the balance of benefits and harms of an intervention by converting benefits and harms into a common unit, usually a “quality adjusted life year” or QALY. Quality adjusted life years are a measure representing not just how long people live, but also how healthy a person is during that time. Living for a year in perfect health gives a higher QALY score than living for a year in poor health. Harms tend to reduce a person’s QALY score, while benefits tend to increase it.
Models such as this are based on a number of assumptions and inputs. No model is perfect, and how accurate they are will depend on the validity of the underlying assumptions and inputs.
What did the research involve?
The researchers developed a model similar to the one used in the Forrest report in 1986. They confirmed that their model produced the same results as the original Forrest report when they used the same input data.
The model was based on women aged 50 and over invited for breast cancer screening in England. The updated model combined life years gained from screening with losses in quality of life from ‘false positive’ diagnoses and surgery. The model assumed that 73% of women invited for screening attend, and analysed the effects of screening over 20 years for a group of 100,000 women.
The researchers updated the inputs into this model by using the mortality rate for breast cancer in England and the chances for undergoing breast cancer surgery in the English NHS. They used data from 1985, before breast cancer screening was introduced, as their “baseline” data. Other data for their model was drawn from systematic reviews where possible, as they are the most robust source of evidence. If systematic reviews were not available, data from individual randomised controlled trials (RCTs) was used, or alternatively data from other published models or from observational studies was used.
The model included estimates of the false positive rate for screening and the reduction in quality of life for women who have false positive results on screening, which were based on available research. These estimates were that there would be a 6.4% false positive rate at the first invitation for screening, and about 3.1% for subsequent invitations. Women who were false positives were estimated to have a 5% reduction in quality of life over 0.2 years.
There was less data on the effects of breast cancer surgery on quality of life, and researchers had to make an assumption about this, based on recent RCTs. They estimated that women who had surgery (necessary or unnecessary) would have a 6% reduction in quality of life over the rest of their lives. This differed from the original Forrest report, which had only assumed a reduction in quality of life with treatment in the additional years of life years gained from screening. This approach only adjusts the quality of life in those who benefit from screening, and essentially assumes that there is no unnecessary surgery.
The researchers looked at what happened if they varied their input data and assumptions. This is called “sensitivity analysis” and shows how robust the model is to these changes.
What were the basic results?
The researchers’ model predicted results for a group of 100,000 women aged 50 being invited for screening, over a 20-year period.
The researchers first updated the original Forrest report using the mortality data from the recent Cochrane review. This review had pooled eight existing mammography screening RCTs and found that after 13 years, deaths from breast cancer were reduced by 19%. This analysis did not separate women by age group.
If the model was updated using this 19% reduction in mortality, but not including harms, it suggested that across the 100,000 women mammography screening increased QALYs gained by a total of 195 after five years of screening. After 20 years, screening produced a 3,145 rise in QALYs.
Adding harms to this updated model (false positives and surgery) reduced the QALYs gained to 12 QALYs gained at five years and 1,536 QALYs gained at 20 years.
However, based on the quality of the trials, the Cochrane reviewers felt that their best estimate was that screening would reduce breast cancer deaths by 15% rather than 19%. Running the model using this lower figure and with harms suggested that at five years screening actually reduced QALYs by 31. Screening only became of net benefit at seven years - at 10 years the benefit had increased by 70 QALYs, and at 20 years QALYs increased by 834.
A separate systematic review on behalf of the US Preventive Services Task Force did an independent analysis of the eight existing mammography screening RCTs featured in the Cochrane review, and split the results by age group. It suggested that screening reduced breast cancer deaths by 14% in women aged 50-59, and by 32% in women aged 60-69. Using these figures in the model along with harms also suggested that screening reduced QALYs at five years by 42. By 10 years screening had increased QALYs gained by 27, and at 20 years QALYs increased by 1,685.
Varying the inputs into these models gave similar results, particularly for the first 10 years.
How did the researchers interpret the results?
The researchers concluded that their analysis “supports the claim that the introduction of breast cancer screening might have caused net harm for up to 10 years after the start of screening”. They say that “from a public perspective, the meaning and implications of overdiagnosis and overtreatment need to be much better explained and communicated to any woman considering screening”. They also call for further research to assess the extent of unnecessary treatment and its impact on quality of life.
This study has updated the analyses of the Forrest report, the 1986 report that led to the introduction of screening in the UK. The updated model includes more recent estimates of the effect of mammographic screening on breast cancer deaths, and has added data on some of the potential harms of screening, (effects on quality of life of false positives and surgery).
Unsurprisingly, the inclusion of additional harms in the model reduced the estimated benefits for the screening programme. Overall, the updated model including harms suggested that the screening programme may not have yielded a net benefit until about 10 years into the programme, although the balance did tip in favour of screening after this point.
Balancing the benefits and harms of screening programmes is complex. Models such as the one used in this study are a way of attempting to place benefits and harms on the same scale so that they can be weighed up against each other. Inevitably, modelling relies on assumptions, and no model is perfect. However, models can help researchers and policy makers to visualise these complex scenarios.
The researchers acknowledge that their analysis has limitations, and discuss these in their article. These include that:
The trials on mammography were mostly carried out outside the UK, and several decades ago.
The model centres on the use of mammography breast scans, but the screening methods have developed and changed over time.
They applied a single rate of quality of life loss to all surgeries, but the effect is likely to vary depending on the type of the surgery performed. Also, no effect on quality of life of non-surgical treatments was included.
When assessing the effects of screening today, researchers may need to disentangle the effects of screening from those of improved treatment, which is difficult.
Trials provide vastly different estimates of how common overdiagnosis is within screening programmes, reportedly ranging between less than 1% and 30%. This uncertainty could be addressed in future research.
There has been a lot of discussion about the balance of benefits and harms of breast cancer screening. As a result, a review of the effects of breast screening was announced by Professor Sir Mike Richards (National Cancer Director) earlier this year. Professor Richards is carrying out this review with Harpal Kumar, Chief Executive of Cancer Research UK. This review will analyse all relevant research. Independent advisers who have never previously published on breast screening will carry out the review to maintain distance from the current differences of opinion. The review report is expected in early 2012.
- Raftery J, Chorozoglou M. Possible net harms of breast cancer screening: updated modelling of Forrest report. BMJ 2011; 343.