Diagnostics Health Care Public Health Screening Uncategorized

The case against early diagnosis of cancer

This blog may upset some. It’s quite uncompromising. It’s also a non referenced blog, not a PhD thesis.
Early diagnosis of lung cancer is a good thing. Undoubtedly and indisputably.
The problem is we don’t know how to do it reliably at population level.
There are two main spins


Screening for lung cancer. Mostly in cohorts with known xxxx pack years (>30 I think is where we settled)

Screening rarely saves lives….It sometimes changes, slightly, cause specific mortality.

As Muir once said – all screening does harm, some does good, v little does more good than harm, is cost effective and affordable.

The lung cancer screening studies have been concluded largely. They were good, well conducted RCTs – USA and U.K.

Spiral CT not XR.

From memory – happy to stand corrected – there was a reduction in lung cancer mortality but not all cause mortality – i.e. It changes the cause, but not the time of death.

We don’t really know whether it is cost effective

There are many interesting and quite nuanced debates going on around harms of this screening proposition.

I’m also aware that some parts of England – Liverpool? Manchester? have got on with it. Maybe they know better than NSC. They certainly know better than me.

We are awaiting a NSC view on these.

Whilst we wait, consider the Lung cancer screening data in Manchester pilot

Crosbie et al

Letters to 16.5k

Risk assessed 2541

Scanned (CT) 1.4k

81 assessed in lung cancer clinic

Of the 81, 42 were confirmed to have lung cancer.

false positive rate was 48.1% (n=39/81)

Thoughts –

  • Economics. I’m not expert but seems hopelessly optimistic to suggest this would pass muster in economic analysi
  • Quantity harms both to individuals, and to those denied interventions by rote of opportunity costs.
  • QA criteria?
  • What is the NSC view.





Early diagnosis initiatives.
The various spins of this:

get a chest XR if you’ve had a cough for longer than X weeks are basically unorganised screening – maybe we should call it fishing.

There are various spins – ads on back of bus – direct access X-ray

Coughing bus shelter, worried? go to GP, assessment, X-ray

National media campaign – go to GP, assessment, X ray

Some points

  • Given that the cohort is far less well defined than in a structured screening cohort the yield, this population level effectiveness, thus cost effectiveness is going to be considerably lower than in a structured screening
  • All the same biases apply to this as per screening – lead and length time. Both important and both of uncertain impact, but v v likely to make the proposition less likely to be effective
  • We have a poor understanding of and haven’t really modelled the demand this type of activity creates and the capacity required to meet it.
  • We have no idea of the specificity & sensitivity of this screening test
  • We don’t know the cost effectiveness. Well actually we don’t have a great handle on effectiveness….

From memory the back of bus advert to open assess x ray model basically irradiated lots of people and didn’t find much additional cancer and the cancer it did find wasn’t early enough, staging wise, for surgery. So net harm? Esp if you take into account the lost opp cost

From memory the evaluation of other models have led to lots of GP appointments, lots of XR (maybe meaning that those with more pressing need need to wait longer?) and not found much additional cancer early enough to undertake surgery.
Some evidence………..

Campbell et al gave some data on the open access XR model.

Enabling patients with respiratory symptoms to access chest X-rays on demand: the experience of the walk-in service in Corby, UK. Journal of Public Health | Vol. 36, No. 3, pp. 511 – 516 | doi:10.1093/pubmed/fdt104

  • Open access CXR for people worried about symptoms. Corby. n=22k 63% increase in CXR for open access. 47% increase in GP requested in same time.
  • 463 additional CXR in walk in service.
  • All clinically indicated according to local criteria
  • TWO cases of Ca Lung found.
  • Both too late to operate.
  • Similar to what Leeds found (Cheyne et al, Thorax, 2012 67(suppl 2) A44-A5

O’Dowd tested which characteristics of primary care and patients are associated with early death in patients with lung cancer in the UK?

  • Of 20,142 people with lung cancer, those who died early consult ed with primary care more frequently pre diagnosis. Individual factors associated with early death were set out.
  • The study concluded that patients who die early from lung cancer are interacting wit h primary care prediagnosis, suggesting potential missed opport unities to identify them earlier. A general increase in CXR requests may not improve survival; rather, a more timely and appropriate targeting of this investigation using risk assessment tools needs further assessment.

O ’Dowd EL,et al . Thorax.doi:10.1136/thoraxjnl-2014-205692. What characteristics of primary care and patients are associated with early death in patients with lung cancer in the UK?

There are no control group studies of any model yet as far as I’m aware. Ditto we don’t know the cost effectiveness of such models, nor the opportunity cost, nor the unanticipated impact on GP and chest clinic / XR dept.
It’s also worth saying that chest XR is a pretty poor test for early lung cancer.





The money – this will all save mega cash

The notion it will save money is worth exploring. One of the better economic analysis in this area of recent years was from Incisive  Health

Saving lives, averting costs. An analysis of the financial implications of achieving earlier diagnosis of colorectal, lung and ovarian cancer

Click to access Saving%20lives%20averting%20costs.pdf

It is excellent and well worth reading. Key points from the summary here:-
Early stage cancer treatment is significantly less expensive than treatment for advanced disease:

  • For colon cancer, stage 1 treatment costs £3,373, whereas stage 4 treatment costs £12,519
  • For rectal cancer, stage 1 treatment costs £4,449, whereas stage 4 tre atment costs £11,815
  • For lung cancer, stage 1 treatment costs £7,952, whereas stage 4 treatment costs £13,078
  • For ovarian cancer, stage 1 treatment costs £5,328, whereas stage 4 treatment costs £15,081

However, the costs of recurrence can be significant and should be taken into account when modelling overall cancer treatment costs.

late diagnosis is a major driver of NHS cancer treatment costs. Treatment for stage 3 and 4 colon, rectal, lung and ovarian cancer costs the NHS nearly two and a half times the amount spent on stage 1 and 2 services.

Significant savings could be realised if all CCGs were able to achieve the level of early diagnosis of the best.

For lung cancer, over 3,400 patients would benefit. Due to the higher level of recurrence that occurs in lung cancer, achieving this level of earlier diagnosis would incur a cost of £ 6.4 million.

Although delivering earlier diagnosis for lung cancer would not be cost saving, it would be highly cost – effective. Achieving the stage distribution of the best CCG in England could generate an additional 4,275 years of life. This equates to a cost of £1,515 per year of life gained.
(NB – this is fab, though still not cost saving, it’s pretty good value in the scheme of things…..but……it’s not at all clear whether Incisive included in their modelling the cost of programmes to achieve the early diagnosis, all the diagnostic cost, GP time and false positives etc etc….that’s important!

If I’m right in my assumption that this wasn’t factored in, it would send the economic model southwards, how far south is unknown).

Taking the four cancers together, achieving the level of early diagnosis comparable with the best in England could deliver savings of over £44million, benefitting nearly 1 11,000 patients.
(NB 2 see above caveat re programme and diagnostic / GP costs)
Lesson of incisive – it’s a good thing to do, but don’t sssumr there’s mega dollars to be saved.


Pushing beyond lung cancer per se….

This, from Prof Michael Baum in the Times (28 11 18) about nails the problem re using 5 yr survival as a measure on which to base either comparisons of “how we do” around cancer or, pertinently, as a premise for early diagnosis initiative. This is worth reading in full, and taking the time to understand the concepts behind it. If you don’t want to understand them, go talk to a public health consultant – they are well trained in these matters

I was in a very interesting twitter conversation following this one. Key points

1) Li et al – ‘Five-Year Survival is Not a Useful Measure for Cancer Control in the Population: an Analysis Based on UK Data

“Our findings suggest that there are no reliable relationships between changes in 5-year survival and cancer incidence or mortality. Increases in 5-year survival might therefore represent poor indicators of progress in cancer control at the population level. In the absence of over-diagnosis, 5-year survival might only indicate improved diagnosis and treatment in clinical practice”

2) Explaining Europe’s survival gaps.

This is an astoundingly good dose of cancer epidemiology.

Despite what some papers tell you, on mortality in Breast cancer, We are doing the second best out of 36 European countries for the predicted fall in breast cancer mortality by 2020

3) Sarfati – Measuring cancer survival in populations: relative survival vs cancer-specific survival

“In this article we investigate the assumptions underlying both methods of survival analysis. We provide simulations relating to the impact of misclassification of death and non-comparability of expected survival for cause-specific and relative survival approaches, respectively.

Results For cause-specific analyses, bias through misclassification of cause of death resulted in error in descriptive analyses particularly of cancers with moderate or poor survival, but had smaller impact in analyses involving group comparisons. Relative survival ratio (RSR) estimations were robust in relation to non-comparability of comparison populations for single RSR but were less so in group comparisons where there was large variation in survival.

Conclusions Both cause-specific survival and relative survival are potentially valid epidemiological methods in population-based cancer survival studies, and the choice of method is dependent on the likely magnitude and direction of the biases in the specific analyses to be conducted”

4) The classic is by Prof Bastian What’s so good about “early,” anyway?

5) Prof Richards article on the influence on survival of delay in the presentation and treatment of symptomatic breast cancer

“Multivariate analyses indicated that the adverse impact of delay in presentation on survival was attributable to an association between longer delays and more advanced stage. However, within individual stages, longer delay had no adverse impact on survival. Analyses based on ‘total delay’ (i.e. the interval between a patient first noticing symptoms and starting treatment) yielded very similar results in terms of survival to those based on delay to first hospital visit (delay in presentation).”

Prof Baum clarified that he was not saying that patients should delay presentation or that the system has no room for improvement, that is complaisant. I’m only saying that survival time is a flawed way of judging care in the NHS. Trends in mortality is what we should look at.






In conclusion 

So whilst I have no doubt that early diagnosis of lung cancer is a good thing, and that we must do better…..there isn’t much evidence that the various models to achieve the goal are effective, cost effective, affordable and we really understand the unanticipated impacts.

Others have documented that our push in this direction will  Increase The Number Of Cases Overdiagnosis

see also here Is “early cancer diagnosis” a meaningless concept Picks up on point that the same biases applied to screening also apply here. Those biases described with brilliance by Steve laitner

I’m not necessarily arguing against efforts to diagnose cancer earlier. It’s a worthwhile thing to try to do. I am arguing that we should enter into and programme armed with data, evidence and realistic assumptions about what will happen, and me mindful of the opportunity costs and unanticipated consequences.
We should also be mindful that efforts to achieve this goal may cost us more than the current model and not be the cost save we all want. This is in theory fine, but may take resources away from other areas such as heart disease, mental health, general practice, community nursing. Cancer is no more or less special than other areas we spend our health care dollar on.
Given the poor, or perhaps better expressed – developing, basis in science could this type of model be seen as the equivalent of marketting a drug that hasn’t finished clinical trials?

If we want improve lung cancer mortality, reducing incidence remains – by a country mile – the most effective, efficient and equitable thing we can do, and gives greatest return for taxpayers dollar. That means tobacco control and smoking cessation – something we are busy stripping money out of right now as the chancellor has told us to cut the public health budget. Go figure.

By the way, don’t just think I’m picking on lung cancer. I could say even more awkward things about bowel cancer campaigns……..maybe I’ll come back to that.

Nobody is saying that at individual level early diagnosis is not a good thing. It is not always a good thing, at either individual or at population level. There is a whole bunch of tricky stuff here that defies pithy headlines about “saving lives”, but few who are pushing hard on this seem to actually understand this stuff. (I am happy to be proven wrong in that one!)

One reply on “The case against early diagnosis of cancer”

[…] On being “an outcome based …….” Changing the language of “public health” – https://GREGFELLPUBLICHEALTH.WORDPRESS.COM/2016/10/02/CHANGING-THE-LANGUAGE-OF-PUBLIC-HEALTH/ Transforming public health. What does public health 3.0 look like. – to RCT or to not RCT. On the issue of “evidence” and public health - The glacial pace of change in the practice of public health. – what does the future of public health services look like?Old world meets new – “Determinants of health – if only local government took it seriously” – Ten grand challenges The individualisation of public health and health policy. – Public health in 15 years time. – Screening Is screening different to case finding in high risk groups https://GREGFELLPUBLICHEALTH.WORDPRESS.COM/2016/02/20/IS-SCREENING-DIFFERENT-TO-CASE-FINDING-IN-HIGH-RISK-GROUPS/ What would it take to get guideline bodies eg NSC to change their view on screening for AF – Why I argued against diabetes screening – The case against early diagnosis of cancer –… […]


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s