Cancer Survival Revisited

This week we’ll briefly revisit the Cancer Survival Rate study. We’re doing this in preparation for an article next week that will present a comprehensive look at all the measures discussed to date of the effectiveness of the US health care system. But first, let’s re-cap the earlier article. A study was recently published that examined 3 or 4 of the most deadly cancers. The study was a world-wide study and thus included both industrialized countries and some less advanced countries. The results showed that the US was tied for either first or second place for the 5 year cancer survival rate. A good showing to be sure but when examined on a cancer-by-cancer basis, the top showing by the US was somewhat less impressive mainly because it was statistically tied for first or second place with several other countries.

In this article we’ll make the point even more succinct. For the purposes of comparing the effectiveness of a health care system, the cancer survival rate of a nation is ideally shown as a single number, not one number per cancer. Can we get there? The short answer is yes.

How would you blend the survival rate of breast cancer with the survival rate of colon cancer to get a single survival rate? One way would be to simply take the average the survival rates. This is the normal average, and it assumes that breast cancer and colon cancer should be valued equally. But valuing them equally may not be fair. Each year 200,000 women die from breast cancer in the industrialized world while 165,000 die from colo-rectal cancer1. Also of the countries studied in the original report 145,000 women are diagnosed each year with breast cancer while, 62,000 are diagnosed with colo-rectal cancer2. Clearly breast cancer should be weighted more heavily in a blended average, but which weighting should be used? This is essentially an arbitrary call. Since we’ve already used the mortality rate to establish these cancers as being significant, for the weighted average we will weigh based on the incidence rate.

One of the tenants of News With Numbers is that graphs should make complicated things clearer. In the case where an effect clearly stands out then graphs make it easy to see but in the case where the effect is more moderate then even a graph may require the eye of an expert to discern the difference. But, given the hype surrounding the health care debate, either is news. Either the US has a clear lead in some health care fields or it does not. On a cancer-by-cancer basis the earlier study showed the US in the top, but tied with several other countries for top honors. Not clearly in the top alone.

Below is a table that shows the weighting effect based on incidence rate. The original study provided statistics both on the survival rate as well as the incidence rate. Ideally we’d weigh based on the global incidence rate or at least the incidence rate in the industrialized countries, however, the US was by far the largest country in the study and using the incidence rate data from the study would effectively mean using only the incidence rate in the US. Thus for this comparison we’ve used the incidence statistics for every country except the US.

Breast Prostate ColoRect-M ColoRect-W
weight 237,991 174,253 124,060 113,007
36.7% 26.8% 19.1% 17.4% Weighted Score
United_States 83.9 91.9 59.1 60.2 77.2
Canada 82.5 85.1 55.6 58.9 74.0
Australia 80.7 77.4 56.7 58.2 71.3
France 79.8 73.7 55.6 61.5 70.4
Austria 74.9 86.1 52.7 55.1 70.2
Sweden 82.0 66.0 52.8 56.2 67.6
Germany 75.5 76.4 50.1 55.0 67.3
Netherlands 77.6 69.5 53.6 55.1 66.9
Iceland 79.0 69.7 49.5 54.0 66.5
Finland 80.2 62.9 52.5 54.0 65.7
Italy 79.5 65.4 50.7 52.7 65.5
Japan 81.6 50.4 61.1 57.3 65.1
Spain 77.7 60.5 52.5 54.7 64.3
Norway 76.3 63.0 51.1 55.3 64.3
Ireland 69.6 62.8 46.0 50.0 69.9
Portugal 72.2 47.7 46.5 44.7 55.9
Malta 73.5 44.3 35.7 55.5 55.3
United_Kingdom 69.7 51.1 42.3 44.7 55.1
Denmark 73.6 38.4 44.2 47.7 54.0
Brazil 58.4 49.3 47.3 43.5 51.2
Estonia 61.3 56.5 36.4 35.5 50.8
Czech_Republic 62.9 50.7 33.8 38.3 49.8
Slovenia 66.3 43.7 35.7 37.7 49.4
Slovakia 57.9 45.7 34.0 38.7 46.7
Poland 62.9 37.1 28.6 30.6 43.8
Algeria 38.8 21.4 22.5 22.6 28.2

Below is a dot-plot of the above data. The US is significantly ahead of 2nd place Canada and 3rd place Australia. How did the US go from “tied for first place” to such a clear lead? The US didn’t have such a clear lead over the other countries when these data are examined on a per-cancer basis, but collectively the US pulls ahead because it was alone in having top marks across all the cancers studied. In a competition that’s scored by cumulative marks, it is possible that the overall first place winner could merely have placed 2nd in every event. As long as no one placed first in more than half the events, those who consistently place 2nd in all events can be the overall winner. That is essentially what’s going on here. The US was first in Breast cancer survival, with Canada #2. The US was first in Prostate cancer survival with Canada #3. The US was #2 in male ColoRectal cancer survival with Japan #1 and Canada #4. And finally the US was #2 in female ColoRectal cancer survival with France #1 and Canada #3. Any country with a higher survival rate than the US was ahead of the US only once.

IncidenceOnly

Below is a chart that effectively proves this point. It shows the same information, but this time also weighted as in a normal average and by the mortality statistics. While US’s lead is larger with the chosen weighting, it is still quite apparent in these alternative weightings. (The US and Canada are ranked #1 and #2 in each graph.)

ThreeAveragesConclusion

The previous article concluded saying while the US did have better cancer survival statistics than the other countries, that lead didn’t appear to be significant enough to make the case that it is a shining example of the US health care system. Here, though when combining all the scores, the US pulls even further ahead to the point where its survival rate is indeed noteworthy. Skeptics have to agree that the US scores unusually highly here, but subject to the same lead-time bias criticisms that the study’s original authors warned against.

However, of the 4 measures we’ve studied in earlier articles (Longevity, Infant Mortality, Preventable Deaths and Cancer Survival), this is the only one that shows the US in a favorable light. Next week we’ll combine all 4 of these measures into a single “health care effectiveness” number and see how the US fares there.

Notes

1. Mortality rates are on page 3 of this document

2. Incidence data is from Table 1 of the document available here. The total was computed by adding the incidence statistics for each country listed. We excluded Cuba for the same reason as the original article and Switzerland because it only had info on breast cancer. Also, some countries reported countrywide statistics and additionally city/region-wide registries. Where this occurred (notably England) we chose only the countrywide statistic. We did not adjust for the reporting interval differences in some countries as it will not affect the final result.

83.9
82.5
80.7
79.8
74.9
82.0
75.5
77.6
79.0
80.2
79.5
81.6
77.7
76.3
69.6
72.2
73.5
69.7
73.6
58.4
61.3
62.9
66.3
57.9
62.9
38.8

Leave a Reply

Your email address will not be published. Required fields are marked *