(This is a lengthy post, dealing with the often misunderstood topic of health screening)
We use diagnostic tests (blood tests, ECG recording, X-rays, scanning) in two different ways. Mostly we use them on patients who come to us with symptoms, in an attempt to find out what’s causing the problem. Screening, on the other hand, refers to the use of those same tests to pick up disease in a population with no symptoms. The theory is that early detection of disease will increase the chances of successful treatment. There are two main types of screening around at the moment: targeted programmes aimed at a specific disease, and general ‘health checks’ which include a battery of tests in an unfocused trawl for abnormalities in multiple organs. It’s difficult to argue with the assertion that prevention is better than cure, and because screening is seen as a key element in preventive care, there is fairly uncritical acceptance of the fact that screening for disease must always be a good thing. Unfortunately, that’s not the case, and there are a number of conditions that have to be fulfilled by any screening exercise if it is to be either feasible or beneficial.
Any screening programme starts by applying a test of some type ‒ it may be an imaging investigation such as mammography for breast cancer, or a biochemical test like the one used to detect small quantities of blood in the stools used in screening for colon cancer. Patients who have a positive screening test are recalled for further investigation to see if this is a true positive or a false positive result . These further tests will often be more invasive ‒ for example, a positive mammogram will usually lead to a fine needle biopsy of the suspicious area of breast tissue under local anaesthetic. If these tests confirm the presence of disease, the patient will be treated appropriately. If they show that the initial result was a false positive, and that there is no disease present, the patient returns to the normal screening programme. The best way to illustrate the principles of screening is in relation to targeted screening programmes, so I’ll deal with those first.
If you live in the UK, you will be familiar with the breast cancer and cervical cancer screening programmes, and a screening programme for colonic cancer has recently been introduced. There has also been some publicity about the possibility of screening for prostate cancer, and we’ll have a look at why this may or may not be a good idea a bit later. But first, I said there were some conditions that had to be met by any screening exercise, so let’s start by looking at those:
The disease should represent a significant threat to health : In other words, it must be serious enough and common enough to pose a significant population risk. To use a silly example, there would be no point instituting an expensive national screening programme for ingrowing toe nails (not a sufficiently serious threat to health), or for a disease which only affects a handful of people.
There must be an early, asymptomatic, stage detectable by the screening test: this is an important one. If you are going to detect disease at an early stage, it has to have an early stage, and that early stage has to last long enough for the programme to pick it up. To use an even sillier example than the ingrowing toe nail one, you can’t screen the population for broken legs, because with a broken leg, you are OK one minute and writhing in agony on the ground the next − there’s no window of opportunity during which you can employ your screening test.
The natural history of the disease has to be understood: this follows on directly from the previous condition. The ‘natural history’ of a disease refers to the way it behaves and the effect it has on our health. Part of it is knowing that there is the early, latent phase mentioned in the previous paragraph. But we also need to know how the disease progresses, and importantly, we need to know how often the early asymptomatic stages progress to the full-blown disease (this is an important factor in the often heated arguments over screening for prostatic cancer, as we’ll see later).
There must be a suitable test available: Well, obviously − but what do we mean by ‘suitable’? It clearly needs to be sensitive, otherwise it will miss so many cancers that the screening programme loses all credibility. Equally, it needs to be specific, because if it generates too many false positives, patients won’t want to attend for screening, and the programme would in any case become too expensive to be sustained (see here for explanations of sensitivity, specificity and other issues concerning the accuracy of tests). It also has to be reasonably cheap and readily available in all areas. And it needs to be safe ‒ there would be little point using a test that caused more cancers than it detected, for example. Finally, it needs to be acceptable to patients (i.e. non-painful and not too time-consuming) otherwise they just won’t turn up, and screening programmes need high levels of take-up if they are to achieve anything useful.
There must be effective treatment available: there’s little point diagnosing disease early if you can’t treat it (although there are exceptions to this in the case of inherited diseases where you may need to counsel carriers concerning the risks to any offspring).
The programme must be affordable: population screening is expensive. People tend to be uncomfortable about the application of cost/benefit principles to healthcare, but in any healthcare system which is cash-limited (and they all are) you need to be sure that by spending money on screening, you aren’t neglecting areas where it could be used more effectively.
We’ll use breast cancer screening as an example. The NHS programme offers free screening to women between the ages of 50-70. Every three years, they are invited to attend for mammography (X-ray examination of the breast). If there are any suspicious features, they are recalled for repeat mammogram and ultrasound examination of the breast, and a fine needle biopsy under local anaesthetic if the presence of an abnormality is confirmed. If the biopsy reveals cancer, they are treated promptly. If all is well, they return to the screening pool and wait for their next routine three yearly appointment, along with those women whose initial mammogram was normal. So let’s see how breast cancer screening stacks up against the criteria outlined above.
Does it pose a significant threat to health?
Clearly breast cancer is a relatively common disease with serious implications for health, so it meets the first condition.
Is there a latent, asymptomatic stage?
We know that most breast cancers are relatively slow-growing, and that they have been present for quite a long time before they get big enough to feel as a lump, so we do have that important window of opportunity during which we can make the diagnosis.
Do all cancers progress to cause symptomatic disease?
Probably not all, but many/most do. I’ll say more about this in ‘the downside of screening’, below.
Do we have a suitable screening test?
The test used is mammography. It performs well enough in terms of accuracy, it is readily available and not prohibitively expensive. It seems to be acceptable to women, although it does cause some discomfort due to the need to compress the breast tissue. The safety requirement is interesting. Because mammography uses radiation, there is a small risk of causing the very cancers it seeks to detect. However, the dose used is so low that there is a very clear net benefit. That wasn’t always the case though. When breast screening was first mooted in the 1980s, the doses required by the mammographic equipment then available were considerably higher, and the equation setting risk against benefit was much more evenly balanced.
Is there suitable treatment available?
We can tick this box, because effective treatment for breast cancer is available, and it is continually being refined and improved.
Is the national programme affordable?
In the absence of a health economist, this probably isn’t the place to discuss it (although I will, below). Those who have had early tumours diagnosed and successfully treated, would say the programme is worth every penny, and as my wife is one of them, you won’t hear me arguing against breast screening either. Whether it is the best way to use a very large chunk of the NHS budget is a question for others.
So, with the possible exception of affordability, breast cancer screening ticks all the boxes. These conditions have also been satisfied by the cervical cancer screening programme (using cervical smears) and we also now have a national screening programme for colonic cancer using a test which detects small amounts of blood in the faeces.
The downside of screening
What downside? – surely screening must always be a good thing? Well, not necessarily:
the costs of screening
There are two closely-related costs related to screening. There’s the obvious, and very considerable, direct financial cost of running national programmes like those for breast and cervical cancer. Then there’s the ‘opportunity cost’‒ in any cash-limited health care system choices have to be made about where to spend the available resource. If you choose to devote many millions every year to a screening programme, that means that funding won’t be available for some other initiatives that might seem equally (or more) worthy. For example, would the money spent on breast screening be better spent on making the latest expensive drugs available to patients who already have breast cancer? I don’t know the answer to that one, but the problem has to be recognised.
The psychological impact of screening
We’ve already seen that for every patient with early breast cancer picked up and treated by screening, there will be one or two other women whose mammograms were suspicious, and who were recalled for further investigation and biopsy, but who proved not to have cancer. These are the unavoidable ‘false positives’ thrown up by any screening programme, no matter how efficiently it performs. Each of those women, although eventually assured that they don’t have cancer, will have had to live through the anxiety of being recalled and subjected to more tests. What’s more, for many of them there will be a degree of ongoing anxiety that ‘there’s no smoke without fire’, and that perhaps they really do have cancer, despite the benign biopsy result. Whether you think that’s a price worth paying will depend on who you ask: the woman whose early cancer was detected and treated, or one of the anxious false positives.
the over-treatment problem
This is a difficult one, and it’s becoming a hot topic in relation to breast screening. It all comes down to the natural history of breast tumours (see above). We know that breast cancer is a common disease and one which is frequently fatal, but once you start screening for it, you pick up a lot of very early tumours ‒ that’s the whole point of the exercise. What we don’t know is how many of those small tumours would have gone on to produce clinical disease, although we do know that if you look at the breast tissue of women dying of other natural causes, you do sometimes find breast cancers that they had been unaware of, and which had never caused them any problems in life. My wife went for her first screening mammogram when she was 51, and it revealed a small (14mm) lump that had all the appearances of malignancy, and this was confirmed on needle biopsy. She had the lump removed, along with twenty or so lymph nodes from her armpit, one of which had a few cancer cells in it. She then underwent six months of chemotherapy, losing all her hair in the process, followed by radiotherapy. Twenty years later she is (touch wood) fit and well. Was her life saved by screening, or was she subjected to months of unpleasant treatment and years of anxiety just to remove a tumour that would never have caused any trouble if it had been left alone? The answer, of course, is that we just don’t know ‒ there is currently no way of knowing which tumours will progress and which ones won’t, although a great deal of research is going on to solve this conundrum.
This issue of possible over-treatment resulting from screening is even more relevant in relation to prostatic cancer. Prostate cancer is a common problem for middle-aged and elderly men, and is responsible for many deaths every year. A screening blood test exists (measuring the blood level of prostate-specific antigen ‒ PSA), a relatively cheap test which is widely available, and one which is extensively used in the USA to screen men over the age of 50 or so for occult prostatic cancer. There has been pressure for a national screening programme in the UK, not least because there is a widespread perception that two of the currently available national programmes (breast and cervical cancer) are aimed at women, and it’s time men ‘had their turn’. However, there is currently no convincing evidence that it would achieve anything useful. Why is this? I said above that you can find small breast tumours in women dying of other causes, and this is even more true for prostatic cancer. When pathologists look at the prostate glands of elderly men dying of other diseases, they find areas of cancer in over 50% of them, and yet they were unaware of the disease in life. In other words, more men die with prostate cancer than die of it, so a screening programme would result in many men receiving unnecessary treatment; treatment which may produce significant side effects, not least incontinence and impotence. Having said all that in order to illustrate the concept of over-treatment, research aimed at developing less invasive methods of treatment, and at predicting which tumours are likely to progress, may well make prostate caner screening a more feasible proposition.
I’ll use breast screening as an example again here, but the same principles apply to all other screening tests. There is a tendency to assume that if a patient develops breast cancer despite being given the all-clear on screening, someone must have been negligent. This is based on the assumption that tests are either positive or negative, and that it is a simple matter to interpret them. I explained elsewhere why this is not necessarily true, using the examples of blood cholesterol levels and chest X-rays to illustrate the point. I’ll risk boring you by spending a paragraph or two explaining it again in relation to screening mammography, because this is a particularly good example of the difficulties underlying a lot of medical decision-making. It’s also an area which has generated more heat than light in the media. The breast consists of breast tissue (obviously) and a variable amount of fat. Unlike the chest, there’s no air or bone in there ‒ it’s all soft tissue, and consequently all looks pretty much the same on an X-ray image, although breast tissue ‒ normal or abnormal ‒ can be differentiated from fat, which is slightly less dense. The breast tissue in young women is denser than in older patients, and in addition to age-related changes in appearance, the breast responds to the cyclical changes in hormone levels in pre-menopausal women. In other words, the concept of a ‘normal’ mammogram is meaningless − there is a huge range of normal appearances which will change with time in any individual patient, and benign lumps or generalised ‘lumpiness’ of the breast tissue are so common as to be almost normal.
Fortunately, in screening mammography unlike many other imaging examinations, we are only interested in spotting one type of abnormality ‒ breast cancer. Unfortunately though, the only difference between a breast cancer and a benign, harmless breast lump is its shape and the frequent but far from universal presence of small flecks of calcium (chalk) within it. With some tumours, there may be no visible lump at all, just a distortion of the surrounding normal tissue or a collection of those little flecks of calcium. So, the screening radiologist is looking for subtle early signs of cancer against a background of ‘normal’ breast tissue which is unpredictable and constantly changing. What this means is that, far from a simple cancer/not cancer decision, the mammographic appearances of breast lumps form a spectrum from definitely benign at one end to almost definitely malignant at the other, with the majority of abnormalities falling somewhere between those extremes. Radiologists reporting the films will look for possible abnormalities and decide whether they fall towards the benign or malignant end of the spectrum, and a decision then has to be made on whether to recall the woman for further tests, or discharge her back to the screening programme to await her next routine examination in three years time.
So, the radiologist has to decide where to set the threshold between normal and abnormal, and he or she will do that on the basis of their training and experience. Set the threshold too low, and although sensitivity will increase (fewer cancers missed), the number of false positives will rise to levels which cannot be coped with; they will all need expensive fine needle biopsies and high levels of anxiety will be engendered, which may in turn discourage women from participating in screening. Set the cut-off level too high, and you will miss too many tumours, and the programme will not meet its objectives. So, it’s a compromise, and even the best of screening programmes will generate some false negatives, which means that some women will present with breast cancer having been given the all clear on earlier screening (so-called interval cancers ). Misunderstanding of these basic principles of screening leads to the common assumption that if some women develop breast cancer despite attending for mammographic screening, those cancers have clearly been ‘missed’, and this is intolerable. However, we know that in the best centres, with the best equipment and with radiologists who have undergone special training in mammographic interpretation and whose performance is subjected to regular audit, 10-15% of breast cancers will not be detected by mammography. Some were there, but not visible; some will have been seen, but judged to be non-cancerous; a few will be genuine ‘misses’ which,
with the benefit of hindsight, could have been picked up.
The same is true of cervical cancer screening. A lot of bad publicity was generated a few years ago by an audit of the cervical screening programme in Leicester. Over a 10-year period some women developed cervical cancer despite having had regular smears, and in some of these, with the benefit of hindsight, it was thought the abnormalities were missed first time around. The tabloids raised their customary indignant cry of ‘scandal’ and ‘blunders’. As a percentage of the tens of thousands of smears screened in this period the number of serious misses was very small, probably close to the irreducible minimum of false negatives found in screening programmes of this type anywhere in the world.
Sometimes, of course, doctors do genuinely miss cancer which they should have picked up, and the close audit of approved screening programmes tends to ensure that these errors are detected, and that lessons are learned (and actual negligence punished). However, my point in this section is to emphasise that most of the ‘misses’ trumpeted in the press are nothing of the sort, and are actually an inevitable product of a screening process which has to make deliberate choices concerning the borderline between ‘normal’ and ‘abnormal’.
does any of this matter?
We now find ourselves in a situation in which public expectations of health screening, based on the sort of misunderstanding outlined above and fuelled by unrealistic government targets and promises, exceed our capacity to deliver. As a result, doctors working in the screening programmes, despite providing a high quality service and meeting all their performance targets, risk being taken to court for missing cancers which were actually not missed at all. Consequently, it is getting more and more difficult to find people to take on this work. So yes, it matters.
Measuring the effectiveness of screening
But assuming we’ve ticked all the boxes and decided to set up a screening programme, it should be easy to decide whether it’s making a difference, surely? Well, as with most things, it’s not a simple as it might seem. For example, most of the diseases currently subjected to screening programmes are cancers of various kinds. We usually quote survival figures for cancer in terms of the five year survival – the percentage of sufferers still alive five years after the diagnosis was made. For aggressive tumours like late stage lung cancer, that’s a very low figure – less than 5%. For bladder cancer, it’s much higher, around 80%. So, why not just compare the five year survival of patients with breast cancer who presented with physical signs of a tumour and those with screen-detected (asymptomatic) breast cancers? If you did, you would be delighted to find that the patients with a screening-detected cancer had a much better five year survival rate than those who presented with symptomatic tumours. So that must mean that the screening has had a beneficial effect, right? Wrong.
Let’s take an imaginary tumour which is slow-growing but uniformly fatal, and where early diagnosis makes absolutely no difference to outcome. Then let’s suppose that we ignore all the principles outlined above, and introduce a screening programme anyway, which results in the detection of the disease several years before it would have produced symptoms.
Now look at the diagram: The horizontal line represents the time course of the disease from time zero when the tumour begins to form up to the patient’s death 4 years later. One year after the tumour arises, it is detected by the screening test (green line), and the (ineffective) treatment starts. The orange line at 3 years indicates the point at which the tumour becomes big enough to be detected. With screening, the tumour is detected one year after it begins growing; without screening it is only detected at three years, when a lump becomes detectable. You can see that screening increases survival time to 4 years from diagnosis compared with 2 years if the patient had not been screened, but she still dies on the same day, five years after the tumour began to develop. The only difference is that with screening, she spends an additional two years with the knowledge that she has cancer. This is an example of something called lead time bias, which is just one of several complicating factors that have to be taken into account when assessing the effectiveness of screening programmes. So although we know (or think we do) that screening for breast cancer saves lives, we can’t measure that effect by simply looking at survival times, because any beneficial effect of early treatment will be exaggerated by the lead time bias.
To get round this, researchers instead look at measures such as the death rates from breast cancer in screened and unscreened populations. And that ‘populations’ is important – you have to include in the assessment those people in the population who were offered screening, but failed to turn up. As I said earlier, screening programmes are only effective when there are high levels of uptake in the screened population, and health economists looking at the cost-benefit of screening programmes have to take non-compliance into account. Another factor which is important in assessing the effectiveness of screening is length bias. You’ll recall that I said how important it was that the disease being screened for should have a reasonably long asymptomatic period during which screening tests can pick it up (i.e. it should be fairly slow-growing). Well, for any particular tumour, and we’ll take breast cancer as the example again, there will be quite a wide variation in speed of growth between different tumours. There will therefore be quite a long window of opportunity to detect slow-growing tumours, and a shorter one for the more aggressive cancers. This means that screening programmes may detect more slow-growing tumours than fast-growing ones, with the latter group more often producing symptomatic disease before the screening test can be applied (or resulting in interval cancers that arise between screening tests). This may exaggerate the beneficial effects of screening if the tumours detected by screening programmes are less aggressive than those producing symptomatic disease. So, any assessment of screening effectiveness has to correct for any imbalance in the type of tumour detected.
At its most extreme, length bias can result in the detection of tumours that would never actually have gone on to produce symptomatic disease at all – the over-treatment problem mentioned earlier. Decisions concerning the introduction of screening programmes have to take into account all the issues we have considered here. However, health economists and epidemiologists (doctors who study disease patterns in populations) are well aware of the various forms of bias which bedevil the screening process– and those we have described here are just two of the more important ones. You can therefore be fairly certain that if you are invited to go along for screening for bowel cancer, breast cancer or cervical cancer, the evidence for benefit is compelling. Having said all that, though, some experts are now querying the value even of well-established schemes such as the UK breast screening programme. As with so much in the field of health care, it’s not easy.
Untargeted ‘health checks’
So far, we’ve looked at national programmes targeting specific diseases where there is good evidence that the scheme will be cost-effective, and yet we’ve seen that even there, debate continues over decisions to screen or not to screen. There’s even more controversy concerning the advisability of the more random and untargeted health checks being advertised by health care providers. ‘Going for a check-up’ just seems like a good idea, particularly when you get to my age and things start to wear out. There’s good evidence, for example, that picking up raised blood pressure or occult diabetes and treating it can be an effective intervention, and GPs in the UK now have well-developed schemes for screening their older patients, so you could put this type of screening in the same category as the national breast, colonic and cervical cancer programmes.
But there’s another sort of screening now being heavily touted by private providers, and like many of these things, it started in the USA. If you Google ‘CT + MOT’ you will, not surprisingly, find an advert for CT Autos of Reading, who I am sure provide an excellent mechanical once-over for your car. You will also find a number of websites for ‘whole-body screening’ centres that use CT (or occasionally MR) scanning to image the whole body and look for early signs of cancer or heart disease or any of the other ills that flesh is heir to. Must be a good thing, surely? Well, it’s a bit like the arguments over prostate screening mentioned earlier in this chapter, and there are at least three reasons why this indiscriminate whole-body examination might not be a good thing:
You may do more harm than good, particularly with CT scanning. The radiation dose from a whole body scan is considerable, and in patients with a low prior probability of disease, the likelihood that the adverse effects of the radiation can outweigh any small benefit is significant.
Scans reveal all sorts of minor abnormalities and variations of normal that are of no consequence, but which require further investigation just to confirm that they can be ignored. One of the main gripes (in the UK, at least) about these screening examinations is that, although the client pays for the whole body scan, the NHS is left to pick up the pieces at public expense. GPs are justifiably irritated when patients attend with print-outs from their whole body scan listing a whole series of minor findings which require them to spend a considerable amount of time arranging further outpatient appointments and tests, at significant expense to the NHS and inconvenience to the patient, simply to confirm that the ‘abnormalities’ don’t matter.
They give spurious reassurance that ‘everything’s OK’, and this can result in the client feeling justified in continuing an unhealthy lifestyle, because it clearly isn’t doing them any harm, but the screening only provides a snapshot of the situation at one point in time, and there are plenty of causes of serious illness which are not revealed by a scan at all.
There are some specific scans which may be helpful and cost-effective, for example a low dose CT scan of the coronary arteries to look for calcification, which is a predictor of coronary artery disease, but there is no evidence that the benefits of indiscriminate whole-body scanning outweigh their drawbacks. And there was certainly no need to send my 32 year old son a seductive pamphlet suggesting he should spend his hard-earned cash on an ultrasound of his carotid arteries in order to exclude atheroma and avoid the risk of a stroke. Nor is there much evidence of benefit from any of the other ‘check-ups’ that involve an hour chatting with a doctor in a private clinic and a battery of blood tests, X-rays and ECG recordings. One of the problems is that these type of checks attract the relatively wealthy ‘worried well’, who are generally all too aware of the need for a proper diet and regular exercise, and who, if they had any signs of illness, would have gone straight to their GP. Consequently, they have a low likelihood of harbouring undiagnosed disease, and the lifestyle counselling that is a part of these checks has little added value. This illustrates the catch 22 inherent in all health screening programmes, namely that those who are most likely to benefit are also the least likely to attend for screening.
So what’s a person to do? It’s clearly sensible to take advantage of the available national screening services that are provided, and for which there is a good evidence base, and to pay heed to the wise advice on lifestyle and health checks such as blood pressure, cholesterol and glucose levels which are freely available from your own GP (depending on your age). As far as the other random health checks and whole body scans that are advertised in the media are concerned, you might want to consider whether you’d get more benefit from spending the money on a nice holiday.