Science does not support banning gas stoves

he US CPSC has proposed banning gas stoves. Few people believe there will be an outright ban, but it won't matter. It follows the standard pattern: the government comes up with a plan to ban something everybody uses. The opponents don't read the science and come up with weak arguments against it. The news media deny it's happening. And the chilling effect works as intended: companies pull out of the market. Et voilà: a de facto ban.

But is there any valid science behind this? According to the news media, it all started with an article by members of a political group called Rewiring America that was published in the International Journal of Environmental Research and Public Health ^[1]. Rewiring America is an organization dedicated to “electrifying everything in our communities.” The authors report odds ratios calculated by other studies and claim that 12.7% of current childhood asthma in the US is attributable to gas stove use.

This paper is junk science, but there are some real scientific papers on the topic. It is a teaching opportunity to help people understand what level of scientific rigor is needed to inform public policy.

What hazards come from gas stoves?

A gas stove can produce nitrogen oxides, water vapor, and CO₂. Depending on the skill (or absence thereof) of the chef, it can also produce fine particulate matter (i.e. smoke). It could also release carbon monoxide and methane if the stove was defective.

Although most studies focus on nitrogen oxides, when gas or electric stoves burn food they can also produce acrolein, a reactive aldehyde that is a known inhalation hazard^[4,5]. Some researchers speculate that burning food, which can happen on any kind of stove, could even create carcinogens.

Can we estimate illness or death among the population from an odds ratio?

No, the best that can be done is to show a correlation. An odds ratio (OR) is simply the ratio of probability / (1 − probability) of two phenomena for a given exposure. If the OR is below 1, the exposure protects. If it is above 1, the exposure makes it worse. As such, it is a crude estimate of correlation.

Let's take a crude example. Consider the number of people eaten by a shark each year for each million people. It's a rare “eating disorder” that's fairly easy to diagnose. If ten patients with the disorder lived in Florida and one lived in Iowa, the odds ratio would be 10. This means being diagnosed with the being-eaten disorder is ten times more likely when the patient lives in Florida. It does not necessarily mean that patients in Florida are 10 times more likely to have it, but it does suggest a correlation.^[2]

The research on stoves also uses ORs, but unlike a shark attack exposure to a chemical is not all or none. We can now measure vanishingly small levels of most chemicals. The days when someone could claim that something is a danger because it is “present” are long gone. If a paper doesn't measure the concentration, they are not doing science.

To properly interpret an OR, you must also look at the CI or confidence interval. A rule of thumb is if the CI straddles 1.00, the effect will not be statistically significant.

An odds ratio gives you only a correlation. A ratio of 1.18–1.48 reported by Lin et al.^[3] is a weak correlation. It gives you a clue that something might be different in that population, but doesn't tell you what that might be. An OR below 2.0 is routinely dismissed in science as inconsequential because it means the effect is indirect. Here are some examples of factors that do cause asthma:

All of these are much more strongly correlated with asthma than gas stoves. By the CPSC's argument, they must all be banned.

The one thing you can't conclude is that by a specific percentage of the population will not get the disorder if they stop using the gas stoves. Gruenwald et al. say that no one has ever calculated that before. The reason is not that they never thought of it. The reason is that doing it is not valid.

Asthma is a multifactorial disease that is thought to have an autoimmune component^[6]. The most common type is atopic (allergic) asthma, which is triggered by dusts, pollen, and foods. Germ-free mice are more susceptible to asthma,^[7] which supports the ‘hygiene’ hypothesis. The second most common type is nonatopic asthma, which is triggered by respiratory tract infection. Asthma can also be caused by aspirin and by repeated exposure to a variety of chemical fumes.

‘Multifactorial’ means that many factors work together in an unknown way to cause it. The factors could add together or be contingent on other factors. When we say something is multifactorial it means we don't understand the true chain of events—what scientists call the ‘mechanism.’

What do the papers show?

Since few people eat all their food raw, it must be in comparison to something else. This is not stated in the papers, and we must assume it would be electric, as the authors explicitly throw away comparisons between stoves that use gas vs. other combustibles.

A good paper, not cited by Gruenwald et al., can be found in International Journal of Epidemiology ^[3]. These authors focused on nitrogen dioxide (NO₂), which is produced by high-temperature combustion. They found a small increase in asthma that correlated with the use of gas stoves. They found that NO₂ was not enough to account for it. They also found that the effect did not depend on the amount.

We did not find an increase in asthma (random effects meta-OR = 1.10, 95% CI, 0.35–3.40, I² = 49.5%, heterogeneity P-value = 0.159) and in wheeze (random effects meta-OR = 0.81, 95% CI, 0.59–1.12, I² = 0.0%, heterogeneity P-value = 0.715) among children with the highest compared with the lowest NO₂.

In other words, the effect was not dose-dependent. We may conclude that whatever is causing asthma, it is not NO₂ but something else. More research is needed, but the finding doesn't look particularly robust.

If the results are valid, does it mean gas stoves are unsafe?

No, and here's why: if you find that gas stoves are correlated with asthma without understanding why, all you have is a correlation between two things that might be related in some way and might not.

Do gas stoves produce more airborne oil droplets than electric stoves? Do their owners cook different kinds of food? Do these stoves oxidize more of the cooking oils or Teflon than electric? Are the children different (for instance, are their parents more likely to worry about the children turning on the gas and causing a fire, and buy an electric stove instead)? Are patients with gas burners more likely to burn their food and generate particulates? Are there socioeconomic or cultural differences? These are just some of the actual relevant factors. Researchers don't have to study all these factors at once, but papers that ignore them altogether are essentially junk science.

Authors of these meta-analyses often claim there's no publication bias, but there's always publication bias. Indeed, if they said they found publication bias they probably wouldn't get published. That in itself is a form of publication bias.

Another factor is that these types of studies are merely observational in nature. As such the measurements are easily biased by contamination with traffic-related air pollution, inappropriate study selection, and reporting bias. Indeed, Fig. 2a in Ref. 3 shows that only 3 of the 21 studies on asthma had CIs that did not cross the 1.0 line. Five showed a benefit and 16 showed some amount of risk.

Lifetime asthma is only one of dozens of possible health effects. What we really want to know is whether stoves affect overall health. But another question concerns meta-analyses: is it valid to average two studies that come to qualitatively opposite conclusions, as often happens, and conclude that if the average is above 1 there must be a risk?

What scientific results are needed?

To show that stoves are a hazard, you would have to measure the suspect molecules and show them to be present in the lung at toxic levels. You'd have to show that the severity of the disorder depends on the dose (or at least the concentration in the air). If a hazard is actually found, the CPSC could, if it was interested in product safety, provide guidance on how the public could avoid it. Manufac­turers could easily redesign exhaust hoods to inactivate acrolein and other oxidants, and the public would buy them if they were concerned.

I would applaud the CPSC if they recommended the use of venting exhaust hoods. The trend among vendors and installers has been to claim that recirculating hoods, which don't exhaust to the outside, are better. What they really mean is that they're cheaper to install. The drawbacks are obvious: removal of cooking fumes is only as good as the filter, which is essentially a metal screen that must be periodically cleaned. Toxic gases are unaffected and are recirculated back into the interior space.

As a former owner of one such hood, I can also attest that they only trap a small percentage of the cooking oils. The remainder ends up on your walls, on your ceiling, on your Venetian blinds, and in your lungs. I once had to spend half an hour cleaning my kitchen blinds with acetone to get it off.

Catalytic converters to eliminate potentially harmful gases might help, but they'd be overkill—and probably useless until we know what molecule is causing the problem. If the CPSC is really interested in safety they'll start nagging people to get venting hoods for all types of stoves.

But really, who are we kidding? Asthma is just a red herring. The real reason they want to ban gas ovens is the belief that electricity is “greener.”

jan 15 2023, 6:23 am. updated January 18, 2023, 4:32 am