Does global warming cause cancer?

e're all getting tired of having to debunk one bad global warming paper after another—especially when many skeptics publish articles that are full of embarrassing mistakes, like claiming that 0.04 percent CO₂ is too low to have any effect. (In fact, the opposite is true: it's too high.)

We're neither pro- nor anti-global warming fanatics here. We're anti-bad-science. Reacting to pseudo­scientific claims by using weak arguments won't work. That's what political people do routinely, and it's how they got stuck with gay marriage, DEI, and noxious EPA rulings. Whether you're a climate skeptic or a believer, you must learn atmospheric physics and statistics or your arguments will be ignored.

A new paper ^[1] in Frontiers in Public Health by academics from the Institute of Global Health and Human Ecology at American University in Cairo would be good to practice on. It's like those puzzles in the UK tabloids but instead of finding the panda, you have to find the 344 flaws in 30 seconds.

This one is easy. There's no science in the paper, only statistics. It starts out with their list of assumptions:

Climate change results in rising temperatures, reduced air quality, threatened water supplies and safety, and increased food insecurity, which ultimately impact the health of populations. Rising tempera­tures contribute to the creation of favorable circum­stances for disease vector breeding and illness dissemina­tion and thus increase the risk of infectious diseases. Moreover, climate disasters disrupt a nation's infrastructure, particularly its health systems, which challenges and reduces access to health care.
. . .

Women, in particular, face heightened exposure to modifiable risk factors intensified by climate change, such as increased contact with environmental toxins and prolonged UV radiation exposure

The authors do some data mining to try to prove it. They know if you test something enough times it's bound to come out statistically significant sooner or later, especially if you do your statistics wrong. Drug companies use this strategy effectively, knowing that the FDA is composed mainly of bureaucrats who are told if the company gets a “significant” effect for a drug, the FDA must approve it.

The authors do multiple linear regression of temperature changes from 1998 to 2019 vs cancer in women. For a comparison, they use gross domestic product (GDP) per capita vs cancer as a function of time. The idea is that if their method is valid, GDP should not correlate with cancer.

Right away we see several gigantic red flags.

1. When the countries were all analyzed together, over 493,264,873 people in 17 countries, GDP correlated negatively with cancer and temperature correlated positively with cancer. The standard error was 0.000 for every correlation except one which had an SE of 0.002. But something is terribly wrong. There's no way you could get an R² of 0.123, which means a poor fit, and still have a p-value of 0.000. What this means is that the data points all fell on a perfectly straight horizontal line. (R² is the correlation coefficient squared and is always positive, so you need other information to interpret it.)

   Note that their N, the population size, is not 493 million but 17. Two of those have incomplete data and one is not even a country.

2. When the results were broken out by country, they got the opposite result. Suddenly almost nothing was statistically significant. Significant positive correlations of temperature with cancer incidence were found in only 3 out of 15 countries: Bahrain, Qatar, and UAE. The other 12 were not significant. Correlations with death were similarly noisy. This means the “perfect fit” in Table 1 was driven by only those three tiny countries.

   There are ways this can happen, and none of them are good. For instance, all the data points might be clustered into two closely packed regions—one near the origin and one somewhere else. We would need to see the correlation plots to decide what, if anything, is happening, but they're not shown (and probably for a good reason).

3. They used first-order regression, which is not appropriate because temperature could have a quadratic, exponential, or subexponential effect.

4. In the combined table (Table 1), there is no correction for multiple correlation across countries. In the country table (Table 2), there is no correction for multiple correlation across diseases. This means the results in both tables are invalid.

5. In only one country (#12, Qatar) were the incidence and death correlated to each other.

Their results also show that temperature changes from 1998 to 2019 are NOT correlated with cancer in most countries. They don't say how much warming, if any, actually occurred; only that by 2050, 4°C warming is “expected.” So, without looking it up, we can't even be sure there was a temperature change. Maybe it snowed there?

GDP causes cancer

In the combined table, cancer incidence correlated negatively with GDP in 5 countries and not significantly in 3 countries, suggesting that wealth reduces the incidence of cancer in women.

But the tables for individual countries once again show the opposite effect. They show that GDP in many countries correlated positively with incidence of several types of cancer (example: breast cancer = 8 positive, 1 negative, and 4 not significant). It's not due to improved diagnosis because death rates followed the same pattern.

According to the World Bank, real GDP per capita in Egypt increased from 9,245 to 15,027 in constant 2021 international dollars. In Bahrain, GDP per capita remained mostly unchanged (55,639 to 56,750). Even in Jordan, which showed the biggest increase percentage of deaths by uterine cancer (123%), the GDP increased by 18%.

In Egypt, incidence of breast, cervical, and ovarian cancer, and deaths from breast, cervical, ovarian, and uterine cancer all correlated strongly positively with GDP; none correlated with temperature (uterine incidence is not reported). In Morocco, where GDP increased by 74%, there was a strong positive correlation between GDP and cervical cancer deaths but none with temperature.

These results suggests that increasing GDP increases, not decreases, cancer incidence and death in women in the Middle East / North Africa region. (They include “Palestine” as if it were a country and ignore Israel, which suggests politics, cherry-picking, or both.)

This is all very confusing. Higher GDP per person generally correlates with improved health. How then can one claim that there is simultaneously an increase in health from GDP and a decrease from a supposed temperature increase? The answer seems to be that increased GDP or some other unidentified factor like diet, not a bump in temperature, is what is killing people in the Middle East. The solution is obvious: give them our politicians—they'll get that GDP down for you.

In a real science, you must hold all other variables constant before you can say anything. When that's not done, you get nonsensical, contradictory results as this paper shows. In fairness, the authors seem puzzled by their results. That doesn't stop them from claiming the findings “support the existence of a correlation between prolonged exposure to high temperature and the burden of women's cancers in the MENA region.” That statement is contradicted by their results. Which in retrospect is not surprising, as the results contradict each other as well.

When you peer review a paper, there's always a box that asks whether the abstract accurately represents what the paper says. If you click that box, it automatically means rejection in any reputable journal. I would have clicked it.

If your results show two mutually contradictory things, you can either do more work to resolve the contradiction; or you can pick whichever one supports your hypothesis and ignore the other. That's called cherry-picking, and it means we're justified in ignoring both this paper's arguments and its conclusions.

may 27 2025, 8:16 am. updated may 28 2025, 4:29 am. minor updates may 29 2025, 5:30 am