Hacker News Reader

Comment by DenisM | Hacker News Reader

>A single CT scan exposes a patient to the amount of radiation that epidemiologic evidence shows can be cancer-causing

That does not look right to me. A single CT scan is 2-7 millisieverts, and exposures under 13 can not be linked to health effects - look for UNSCEAR in http://en.m.wikipedia.org/wiki/Sievert

If you do the math by ICRP formulae, you end up with increase chance of cancer of 1 in 2600 over the next 20 years from a single 7ms CT scan today. This compares favorably with "natural" risk that is 1 in 10 without any extra exposure above background radiation.

Now if you do a scan every year for the next 20 years your chances go up to 1 in 130. So there is something to worry about, but article overstates the cause.

View on HN

The American Association of Physicists in Medicine seems to have your back:

http://www.usnews.com/science/articles/2012/01/17/phsycists-...

A full-body CT scan results in 12 mSv; a mammogram 0.13 mSv—a hundred times less. The risks from these procedures, according to AAPM, are too low to have been determined reliably, and may be "nonexistent."

Looks like this debate has been going on for a while now.

Here[1] is a very recent study from Austrialia which concludes CT scans during childhood and adolescent increase the risk of cancer and should be avoided where possible.

[1] http://www.bmj.com/content/346/bmj.f2360

Interestingly Stewart Brand was out this year arguing http://edge.org/response-detail/25392 against of the the Linear No-Threshold model of radiation risk:

"The "Linear" part of the LNT is true and well documented. Based on long-term studies of survivors of the atomic bombs in Japan and of nuclear industry workers, the incidence of eventual cancer increases with increasing exposure to radiation at levels above 100 millisieverts/year. The effect is linear. Below 100 millisieverts/year, however, no increased cancer incidence has been detected, either because it doesn't exist or because the numbers are so low that any signal gets lost in the epidemiological noise."

IDK if his case if convincing or not. In any case, even if say 10, 50 or 100 millisieverts a year is something the human body can shrug off with no increase in risk, I imagine that might not apply to 10-30 millisieverts delivered in a very short interval of time.

Taking your numbers at face value, it seems the problem you're having is the epidemiological interpretation of risk as it pertains to patient populations. You can decide for yourself if a 1/2600 risk is acceptable to you, but doctors are in a position to decide for a far larger number of people.

If 10m people are exposed per year to the given risk you outline, the expected number of additional cancers is ~4k people over 20y. In 10 years of scanning, that's 40k additional cancers over the next 30 years. For elderly people over 70, this may be less of an issue, but for everyone else, this is definitely a problem!

Epidemiological studies likely overstate the risk. If you read the Australian study for example, they certainly could not determine with certainty how many CT scans an individual had, because they did not include CT scans performed in hospitals. I would say that MOST CT scans performed in children would happen in hospitals. Therefore, a person having one CT may simply be a marker of them having many more CTs which are not captured by the study.

There were also other unusual correlations in the data. For example, CT scans of the brain seemed to increase the risk of cancer elsewhere in the body, and scans of the abdomen increased the risk of brain cancer. This seems implausible, and implies to me that at least some of the correlation between CTs and cancer is likely to be explained by some third factor whereby people with an increased risk for cancer have more CT scans. The most telling correlations are probably for rare tumours, like soft tissue cancers eg sarcoma, which show an increased risk post CT. These cancers are rare and have few known risk factors, so this correlation is more believable. But these events were much rarer than all cancers in the cohort.

But doesn't this effectively reduce to "we don't know?"

FWIW I have had one CT scan in my life (in childhood I had a seizure which was probably caused by medicines I was taking) and they did one to rule out brain tumors. Probably a reasonable choice given that all the other tests for other stuff kept coming back negative. But I wouldn't suggest doing one without either a clear immediate need or else a significant concern and no materialized alternative causes for a symptom.

>If you do the math by ICRP formulae, you end up with increase chance of cancer of 1 in 2600 over the next 20 years from a single 7ms CT scan today. This compares favorably with "natural" risk that is 1 in 10 without any extra exposure above background radiation.

But that is a meaningless comparison. The doctor should be comparing this additional risk, to the benefit of the scan.

The "natural" risk of developing cancer is totally irrelevant to the cost-benefit analysis.

Because we operate in likelihood ratios, the unexposed risk is important so the risk can be computed (at least, in theory; in practice, people go more by indication than by formal risk prob/benefit prob).

That still makes no sense.

If I want to know whether to do a CT scan, the only relevant fact is the increase in probability (in an absolute sense) of cancer that it causes.

It seems like either you, or the entire field of medicine, is obfuscating this basic fact using some field-specific concepts.

EDIT: It seems like you are referring to relative risk (http://en.wikipedia.org/wiki/Relative_risk). But as I said, there is no reason that relative risk should be considered in this case.

I see your point.

I wonder though if there should be a cutoff somewhere, below which you stop caring? I used background cancer rates as an anchor to judge if the cutoff has been reached, without getting specific about benefits of treatment in each particular case. Would you suggest another cutoff level that is broadly applicable?

I think the only comparison that makes sense is the benefit from the scan.

So if you are certain that the benefit of the scan is, no matter what the particular case, going to be worth much more than a 1/1000 chance of getting cancer, then that is a reason to ignore the risk.

> Now if you do a scan every year for the next 20 years your chances go up to 1 in 130

Unless the article is wrong about the number of CT scans many people are getting, that's going to result is a huge number of avoidable cancer cases.

The distribution of these cases is likely to skew towards the Elderly and over 50 demographic. I'd be curious to see the impact of this on the analysis.

I'm not very good at math, but I'm reading that as a 0.7% increase in your chance of getting cancer.

From 'Physics for Future Presidents', they say it is commonly accepted that we have a 20% chance of DYING from cancer before calculating for outside influences. http://books.google.com.au/books?id=6DBnS2g-KrQC&pg=PA98&lpg...

Let's say the average person does not get a CT scan every year for 20 years, but rather gets 5 scans over their lifetime. Those scans may detect an early stage of cancer, therefore saving the persons life.

They're risk of dying from cancer is approximately 20.2%. I suggest with the CT scan, their still coming out on top.

that's an extra 0.7% for you. For a population of 300m (about the pop of the usa) that'd be an extra 2.1m people getting cancer.

Any idea what the error bars are on that .7% figure?

The Wikipedia article lists a full body CT scan at 10-30 mSv.

I was going off the list http://en.m.wikipedia.org/wiki/X-ray_computed_tomography and in haste I stopped after head and chest. You are right, I should have gone further down to the full-body as well. Thanks for that.

Incidentally, this is exactly why I included the numbers - so you can check them, and why I don't like the headline article - no numbers and no references, so there is nothing to check, only vaguely scary claims.

Newer machines are at least reducing the radiation. This is due to better compute power and to better sensors (essentially a mineral which converts x-rays into flashes of light which then get captured by a photodiode).

It's been huge in dental x-rays, going from film to CR/DR (although dental isn't huge, it's routine, and involves a large patient population including kids.)

We probably will be <5 mSv for whole-body pretty soon. Ultimately there's no reason it can't be well under 1 mSv.

Fluoroscopy on the other hand is easily 50 mSv/min and multi-hour procedures (!!!). People get acute effects from that.

The article is terrible and you've done a good job of explaining why. But there are two ways to look at this:

1. From an individual's perspective

2. From society's perspective

Let's make up some numbers for illustration. If getting a CT scan gives someone 0.01% chance of getting cancer, the risk to any individual is negligible. But to society, that means 100 extra cases of cancer.

100 extra cases of cancer out of how many already...? Yes, it's negligible either way.

Yes, it's negligible either way

But doctors have that pesky "do no harm" credo that makes them consider whether the benefit of CT scans is worth the harm of 100 people getting cancer that wouldn't have otherwise gotten it.

It is proper for society to examine this without dismissing 100 people getting cancer as "negligible".

you are probably right. I know several people who fought or died of cancer. None ever had a CT scan. Or that many X-rays to begin with.

Reply on news.ycombinator.com