In the previous chapters I described the major cancer types, and briefly talked about some of the mimics of skin cancer. You might now be thinking that you have got the hang of things, ready to move on. Sadly, I have yet more bad news for you. Let me sketch out a simple thought experiment, a simple example that reveals how messy the real world of clinical medicine is, and why dermatologists matter.
This chapter in one minute
Acres of skin
Take one thousand patients, the sort of number that might be on a (small) GPs list, or the practice of a successful office dermatologist. Each patient has near enough two square meters of skin, and I have therefore represented the total area of skin (2000 square metres)in the figure below.
Now let me add in some skin cancer data. For simplicity I will use melanoma because, although it is not the commonest skin cancer, melanoma accounts for over 75% of deaths from skin cancer.
The incidence of melanoma in many countries is close to 20/100,000. So in our target area of skin, for one thousand people over one year, we would expect to see 0.2 melanomas. I have shown the red arrow pointing to a single melanoma, but in reality this is five times more than we could expect over a year.
If this was all there was to skin cancer diagnosis, many dermatologists would be out of work. Simply, because if all you have to do was detect a single lesion — any lesion — it would require little skill. But now look at the figure below.
The arrow still points to the melanoma, but the whole area of skin is covered with other lesions — the estimated more than 250,000 other lesions for every melanoma. This is the clinician’s problem. No matter that skin cancer is the commonest humans cancer, non-cancer lesions — moles, seborrhoeic keratoses, solar lentigines etc — are, in total, much more common.
The clinicians skill is therefore not demarcating cancer from normal (skin), but demarcating cancer from the benign lesions that are orders of magnitude more common. This is the reason we spent so long discussing the mimics of cancer in the earlier chapters.
What is the chance of a mole becoming a melanoma?
If the above presentation is not stark enough, I can represent the problem in a different way using some simple arithmetic and figures I have used already.
The average adult in the UK has close to 25 melanocytic nevi. We believe that fewer than a half of all melanomas are derived from benign melanocytic nevi. Given the incidence of melanoma is approximately 20/100,000, simple arithmetic shows that the chance of any single melanocytic nevus becoming a melanoma in one year is close to 1:250,000.
If you have ever thought that ‘wholesale removal of moles’ was the way to go, then you will have to remove 250,000 per year to catch one melanoma — and you will still miss at least half of all melanomas.
And in this simple estimate, I have ignored all those mimics such as solar lentigines, seborrhoeic keratoses etc.
Signal to noise
The visual example above, and the simple arithmetic in the preceding paragraph, highlight that diagnosis in this domain of medical practice can be viewed as a signal to noise problem. Even if skin cancer is so common, how can we distinguish cancer from the myriad of mimics that are orders of magnitude more common?
In many domains of medicine we could rely on tests or machines and simple algorithms. We might measure blood pressure with an automated tool, and decide on action if the machine tells us that the diastolic is over 100 (or some other figure). We might imagine similar approaches with say renal failure: a blood test, and a simple rule that all can understand.
The problem in dermatology is that we have no machines, and only to a very limited extent can we use rule based approaches. So how do we proceed?
The main method of distinguishing skin cancers from their mimics, and for distinguishing between different skin cancers, is a process called non-analytical reasoning. Do not worry about the term, because you use this process every minute of your life. It is how you recognise faces, how you distinguish cats from dogs, and how you know the difference between a bicycle and a lawnmower. Once you have seen enough examples, and have received feedback — just like your parents gave you when you learned the differences between cats and dogs — the process appear effortless and is usually very fast. The expert recognises a melanoma simply by noting that the index lesion resembles earlier examples he or she has seen. But there are some counterintuitive aspects to this skill. Although experts can demonstrate their expertise, they are usually unaware of how they accomplish the task in hand but, despite this, they can coach others such that novices become experts if provided with the right learning environment. After all, you couldn’t distinguish between a bicyle and a lawnmower at birth.
What about the novice?
Everybody has to start somewhere, and one way to start acquiring these skills in simply to start looking at lots of images (hence the reason I produced this image rich web site and atlas). But there are tricks and simple heuristics that we can use to get you going, and I am going to talk about some of the ones that are used for melanoma.
As easy as the ABCD(E)
One approach to melanoma diagnosis relies on the following nmenonic, ABCD(E)
- A for asymmetry
- B for border irregularity
- C for colour
- D for diameter
- E for evolution or elevation
Melanomas are often asymmetric in terms of pattern, colour or shape. The border of melanomas is often irregular; there are often multiple colours within a melanoma; and melanomas are often greater than 1cm across. In distinguishing between melanocytic nevi and melanomas, many experts claim these simple ‘rules’ help.
However, a sceptic might question how these terms are operationalised. How irregular? How are colours defined? How asymmetrical? My own view is that the term may be useful as a reminder of features to consider or examine (i.e a mnemonic), but that its role is limited. In any case, mimics such as many seborrhoeic keratoses score highly on these criteria, too. Still, you should know the term, because others are not so sceptical as me, and you will come across clinicians who claim to use it.
A history of change
In the ABCD mnemonic, E is often added, representing either elevation (some melanomas are raised), or for evolution, meaning change. History of change in a pigmented lesion is a key history point: most melanomas have arisen due to ‘change’ and with time, will continue to change. So, eliciting the history is key. However, as you might have guessed there is a caveat: lots of non-melanomas change too, and there are many more non-melanomas than there are melanomas. Melanocytic nevi are more common in 20 year olds than 10 year olds. They are even more common at thirty, but then seem to fade with advancing age. As you get older, more and more seborrhoeic keratoses appear, too. In fact studies show that a significant proportion of the normal population report changes in lesions on their skin in any time period. And of course normal lesions greatly outweigh the numbers of melanomas.
All the above not withstanding, once you focus in on a particular lesion, history can be key to how you decide what to do next. A history of change raises the stakes. Do not ignore it. No single factor tells you something is a melanoma — you have to be suspicious.
Background and demographics
Risk factors may provide important insights into disease pathogenesis in many areas of medicine, but their role in the diagnostic process in dermatology is often more marginal than most non-experts realise. Yes, melanoma is say three times more common in those with red hair. Yes, melanoma is more common in 70 year olds than 7 year olds. Melanoma is perhaps 4 times more common in Australians that British people. But all this data may only have a faint influence on your judgment about a particular lesion. Let me use some simple arithmetic again.
I have said that the chance of a single mole becoming a melanoma in any one year is approximately 1:250,000. If the patient has red hair and we argue that the risk is now 3:250,000, this still means that the chances of it not being a melanoma are still 99.9988% (249,997/250,000) as compared with 99.999% (249,999/250,000).
So, as a general rule, demographics and history of sun exposure and sunny holidays etc only play a minor role in the clinical assessment. What you see with your eyes, and a history of change — and a history of previous skin cancers — count for much more.
There once was an ugly duckling...
Dermatologists — skin watchers — will often tell you that the moles on a particular person often have a common ‘character‘, however hard it is to describe exactly what this character is. To put this in terms a statistician would use: between person variance is much larger than within person variance.
Sometimes, you see a pigmented lesion on a patient, that stands out from all the other melanocytic nevi, even though if you saw this lesion on another person it might not attract attention. This is the so-called ugly duckling sign. Look at the back of the individual below. Notice how one mole stands out (white arrow) like an ugly duckling.
Hard though it is make explicit, many of us view this as a red flag sign — get expert advice. So, a person may have ten moles that all look similar — we would probably rarely get suspicious about this because, it is very unlikely that a patient will present with 10 primary melanomas. But one of these visually identical moles on the background of a different patient, might arouse intense clinical suspicion.
More than melanoma
The above sections have concentrated on melanoma, simply because that is the skin cancer where early diagnosis is both critical and so hard, and because melanoma accounts for most skin cancer deaths. Similar thoughts apply to the other skin cancers such as basal cell carcinoma and squamous cell carcinoma, and the premalignant lesions, too. Most clinical expertise relies on having seen many similar lesions (and patients) before, rather than the application of formal rules. The distinguished epidemiologist and biostatistician Alvan Feinstein in a book entitled ‘Clinical Judgement’ described a clinician as somebody who:
‘depends not on a knowledge of causes, mechanisms, or names for diseases but on a knowledge of patients’
Just for the record, do I think machines will reduce the need for these clinical skills. Yes, but that is another story.
Continued exposure and practice as the basis for expertise
One point that is implicit in what I have said above relates to the fact that perceptual expertise in dermatology (or pathology or radiology, for that matter) relies on exposure to large training sets (i.e. seeing lots of examples). This is not the only basis for expertise, but it is key.
However, as all medical students know, you tend to forget what you do not use. So, it is not enough to have seen lots of lesions, if you do not continue to see lots of lesions. And you must see both positive and negative examples frequently enough to maintain any skills you once had. So, for instance in primary care, an average GP will only see a patient with a primary melanoma once every 5 to 10 years. This makes maintenance of skills problematic whatever skills were gained in more intensive periods at an earlier stage.
Is Dermatoscopy / dermoscopy the solution?
A dermatoscope (dermoscope) is a simple lower power lens that if placed on the skin’s surface with suitable liquids (e.g. oil, alcohol), alters the optics such that specular reflection is reduced. Non-touch polarising light devices can produce a similar but not identical effect. You feel as thought you can ‘see into the skin‘. The major role for dermatoscopy (dermoscopy) is in melanoma diagnosis.
Dermatoscopy has become very popular in the last 20 years, but we do not expect our students to use it. In expert hands, most of us feel happier with the knowledge it brings. However, available studies suggest it may worsen diagnostic performance in the hands of non-experts. In any case, if you are not seeing a high throughput of melanomas, the criticisms made in the previous paragraphs still stand.
Skincancer909 by Jonathan Rees is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. Where different rights apply for any figures, this is indicated in the text.