A Scary Plastic Study Should Probably Be Recycled

Lately, the press has feasted on a new Lancet article that concludes that about 350,000 of you are going to die yearly from heart disease brought about by long-term ingestion of di-2-ethylhexylphthalate (DEHP), a chemical used to soften plastics. The good news is that the study's data are hardly convincing. Why? We need to look at the good and the bad – the numbers behind the study and how they were used. 

The claim

The authors claim that in 2018 about 356,000 people, aged 55-64, died from heart disease caused by DEHP, one of many plastic-softening chemicals that belong to the class of organic chemicals called phthalates. This number represents about 13.5% of all heart disease deaths. If true, this is pretty bad. But is it? Don't go hurling your Tupperware out the window just yet.

DEHP (and other phthalate analogs) have long been poster children [1] of the anti-plastic/anti-chemical movements. Countless studies of phthalates, many of them junk, have been published, raising concerns that they can potentially interfere with hormones and affect metabolism (aka "endocrine disruptors"). The studies — mostly from the U.S. — have postulated links between phthalate exposure and conditions like obesity, diabetes, and heart disease. 

The idea behind this new study was to estimate whether DEHP might be contributing to cardiovascular deaths globally, and if so, and if so, how much — especially in light of efforts like the Global Plastics Treaty, which aims to curb plastic pollution and its health effects. Keep the word "estimate" in mind. 

For the study the researchers built a computer model and used available data on DEHP exposure levels –based on urinary metabolites– from biomonitoring studies conducted U.S., Canada, Europe, and parts of Asia and Africa, as well as published meta-analyses. Where no data existed, they estimated exposure levels using regional averages, looking for associations between average cardiovascular death rates in 55–64-year-olds and estimated DEHP exposures in the population

Does this study mean anything?

This depends on whether you consider a hazard ratio (HR) of 1.10 meaningful. In English, an HR of 1.10 means that there is only a 10% difference between two groups that are being compared, a very small difference by any measure. In this case It means that there was a 10% increased risk of cardiovascular death associated with higher DEHP levels. But even this number may be overstated (or possibly understated) depending on how well the retrospective study [2] used appropriate controls to rule out confounders – uncorrected variables that can impact the HR significantly. More on this below.

This figure wasn’t newly calculated. It came from an earlier U.S. study of about 5,300 people aged 55–64 in the NHANES [3] database, who had their urinary phthalate levels measured and were followed over time to track mortality. The global study then applied this same 10% risk estimate across nearly 1 billion people worldwide in that age group in 2018 — regardless of country, health system, lifestyle, or environmental conditions, and came up with the number.

Positives: A very large study population

The study’s strength lies in its massive scale. It pulled together mortality and chemical exposure data from around 200 countries, helping spotlight areas — especially South Asia and the Middle East — with disproportionately high estimated exposures. This type of global modeling is useful for setting public health priorities, particularly where local data is scarce. It looks for trends, but not as good looking for specific numbers.

Negatives: Confounders and questionable assumptions

The negatives appear to badly outweigh the positives, so is the study's estimate meaningless? No, it's not — but it’s very far from definitive either. here are a few issues:

• The conclusion is built on a single modest sized (~5,000) U.S.-based study with a very low HR, and applied broadly using assumptions that may not hold across diverse populations.
• For example, smoking is far more prevalent in Asia than in the US. This could result in the overstatement of the impact of DEHP in Asia because the excess deaths that are attributed to DEHP could arise from more tobacco use instead. This is probably the most important cofounder and probably the biggest weakness of the study.
• There is no correction for the differences in public health and healthcare access in the US and other parts of the world. Much higher statin use in the US compared to other countries could result in an artificially high low CV death in here, but obesity and other metabolic diseases could push the numbers in the opposite direction.
• Adding to the uncertainty, the model included a 10-year time lag between exposure and outcome — comparing DEHP levels from 2008 to cardiovascular deaths in 2018. This also introduces more room for error, since many other factors — like diet, pollution, smoking rates, and access to care — also changed during that time and weren’t accounted for.
• DEHP exposure wasn’t measured directly. Instead, researchers used four urinary metabolites as proxies, introducing individual or group metabolism as another possible confounder.

Because the model doesn’t account for these differences across countries, it may be attributing cardiovascular deaths to DEHP that are actually caused by well-known risks like smoking or poor healthcare access.

More on hazard ratios

For a typical retrospective study the rule of thumb is that an HR of 2.0 (double the risk) is likely to be meaningful, but the closer the number is to 1.0 the less valid any conclusion will be. Although the confidence interval here excludes 1.0 — making the result technically statistically significant — the effect is still marginal.

Also, the HR is reported as 1.10 (95% CI 1.03–1.19). The fact that the HR resides within the confidence interval (CI) means that the results are statistically significant but quite possibly not clinically significant. (If the HR is not within the confidence interval the results are not significant.) Even though this is not the case in this study, the fact that some of the participants had an HR of 1.03 (a 3% difference) is most likely a measuring of nothing real. 

Bottom line

Don't get me wrong. I think that there is WAY too much plastic produced and used globally and this is a serious problem if only for environmental concerns, even if there were no health concerns at all. But this does not warrant (IMO) an unnecessarily alarmist news story which will succeed only to scare consumers, many of whom are already frightened about far too much.

NOTES:

[1] An even more "prominent" endocrine inhibitor is bisphenol-A (BPA), which is still a boogeyman despite an enormous FDA study that concluded the chemical is safe as used.

[2] A retrospective study looks back in time using existing data to find links between past exposures and later health outcomes. It’s quicker than long-term studies but more prone to bias and missing information.

[3] NHANES (National Health and Nutrition Examination Survey) is a long-running public health surveillance program conducted by the National Center for Health Statistics (NCHS), part of the CDC. It has been collecting detailed health information from the U.S. population since the early 1960s.