A recent systematic review and meta-analysis of dozens of studies suggests that more exposure to fluoride may be linked to lower intelligence scores in children, adding to the ongoing debates over water fluoridation.

The review, published in JAMA Pediatrics on January 6, analyzes 74 international studies published from 1989 to 2023. It finds that overall, higher levels of fluoride in water or in children’s urine are associated with lower IQ scores in children.

We’re highlighting this meta-analysis for two reasons.

First, it lands amid heightened scrutiny of water fluoridation, which has been a cornerstone strategy to prevent dental cavities in the U.S. since the 1940s. Robert F. Kennedy Jr., President Trump’s nominee for the secretary of the Department of Health and Human Services, who is having his Senate confirmation hearings this week, is an opponent of water fluoridation and has called fluoride an “industrial waste,” citing its potential health risks.

Opponents, including the Fluoride Action Network, which lobbies against water fluoridation, argue that emerging evidence, including the effect on IQ, warrants a reevaluation of fluoridation policies, especially for vulnerable populations like pregnant women and young children.

A handful of towns and cities across the nation have already removed or are debating the removal of fluoride from their water systems.

But proponents of community water fluoridation, including major medical organizations like the  American Dental Association and the American Academy of Pediatrics, continue to emphasize its role in preventing cavities in children, particularly in underserved communities. They also cite a lack of evidence for health harms when fluoride is at the current low levels used in most community water systems in the U.S.

Second, the two editorials that accompany the meta-analysis have vastly different and nearly opposite takes on its findings. This is a good reminder for journalists that even meta-analyses, which sit on top of the hierarchy of evidence pyramid and provide an overview and analysis of scientific topics, can include a lot of nuance.

The study

The majority of the analyzed studies are from China. None were from the U.S. The authors note that they’re unaware of any U.S. studies on the effects of fluoride exposure on children’s IQ, so their findings may not apply to the U.S. population.

Still, they conclude that their findings “may inform future comprehensive public health risk-benefit assessment of fluoride.”

Among their findings:

• Analyzing the results of 59 studies that looked at fluoride in drinking water, and included a total of 20,932 children, showed that children who were exposed to higher fluoride levels had lower IQ scores than children exposed to lower fluoride levels. This average difference was 6.75 IQ points.

The review also breaks down the studies into three groups: those that measure children’s exposure to up to 4 milligrams per liter of fluoride in water or when they had less 4 mg/L of fluoride in their urine; those that measure up to 2 mg/L of fluoride in water or urine; and those that measure up to 1.5 mg/L of fluoride in water or urine.

• The analysis of 31 studies shows that fluoride levels of up to 4 mg/L and 2 mg/L in drinking water are linked with lower IQ scores in children, with an average drop of about 2.25 IQ points. There was no such statistical association when the water fluoride levels were up to 1.5 mg/L.
• Meanwhile, the analysis of 20 studies shows that when fluoride levels were up to 4 mg/L, 2 mg/L and 1.5 mg/L as measured in urine, there was a consistent link with lower IQ in children, the review finds. The average IQ point drop here was also about 2.25 points.

The authors also analyzed the results of 13 studies, including 4,475 children, that looked at the individual-level measures of fluoride in urine.

• The analysis shows that each 1 mg/L increase of fluoride in urine was associated with an average drop of 1.63 IQ points.

None of the studies evaluate the impact of exposure to 0.7 mg/L of fluoride specifically, which is the level U.S. public health officials have recommended since 2015.

As the authors of the meta-analysis point out, citing a 2023 and a 2024 study, there are regions of the U.S. where water systems and private wells contain natural fluoride levels of 1.5 mg/L or higher. About 4.5% of community water systems, serving nearly 3 million U.S. residents, have fluoride levels above 1.5 mg/L. More than 23 million U.S. households rely on private wells for drinking water.

Dr. Regina Nuzzo, a statistician and professor at Gallaudet University, says this paper shows that “when the fluoride level is high, it could affect children’s IQ but when it gets lower, for instance, the 0.7 that we have in the U.S., there’s not as much conclusive evidence that it affects children’s IQ, which is comforting for most people who get water from their taps.”

A polarizing topic

There seems to be little middle ground in the fluoride debate, as the two editorials published alongside the meta-analysis in JAMA Pediatrics show.

The authors of a supporting editorial, “Time to Reassess Systemic Fluoride Exposure, Again,” write that the meta-analysis is “the largest and includes the most rigorous series of meta-analyses of fluoride ever conducted.”

Drs. Bruce P. Lanphear, Pamela Den Besten and Christine Till, write that the study’s finding “underscores the need to reassess the potential risks of fluoride during early brain development.”

They also note that the “absence of a statistically significant association of water fluoride up to 1.5 mg/L and children’s IQ scores … does not exonerate fluoride as a potential risk for lower IQ scores at levels found in fluoridated communities.”

In an opposing editorial, “Caution Needed in Interpreting the Evidence Base on Fluoride and IQ,” Dr. Steven M. Levy points out several limitations in the study, including the quality of the selected studies, the choice of studies and lack of discussion about the findings of other recent studies that show no association between low fluoride levels and children’s IQ, and the validity of studies that use urine to measure fluoride exposure.

“Public policy concerning the addition of fluoride to community water systems and recommendations concerning the use of topical fluoride in its many forms should not be affected by the study findings, and the widespread use of fluoride for caries prevention should continue,” Levy writes.

Measuring fluoride levels in urine

Measuring the level of fluoride in urine, or, as researchers call it, urinary fluoride, is another point of contention among researchers.

Levy writes that “there’s scientific consensus that the urinary sample collection approaches used in almost all included studies… are not valid measures of individuals’ long-term fluoride exposure, since fluoride has a short half-life and there is substantial variation within days and from day to day.”

However, several studies to date, including a high-profile 2019 study that finds an association between pregnant women’s exposure to fluoride and children’s IQ — and co-authored by Lanphear — have relied on fluoride levels in urine and have defended it as a reliable method.

In their JAMA Pediatrics editorial, Lanphear and his co-authors write, “The distinction between water fluoride and urinary fluoride levels is important because regulatory and public health agencies must consider total fluoride intake when assessing risks.”

They add that water fluoride concentration doesn’t capture the amount of water ingested or other sources of ingested fluoride (such as toothpaste, some foods and pollution from burning coal).

11 tips for covering a meta-analysis

To help journalists better understand and parse out a meta-analysis, we spoke with Nuzzo, and Karen Kaplan, a science reporter at the Los Angeles Times, who recently reported on this meta-analysis’ findings.

The tips are listed in order of complexity.

1. Look for accompanying editorials.

In this case, the two accompanying editorials provide great context for the main study.

“Editorials usually give you a shorthand of what the implications of a study are,” Kaplan says. “And it’s unusual to have more than one editorial with a study, and the fact that they were diametrically opposed was kind of interesting.”

2. Assess whether a study’s findings are valuable to your audience.

Kaplan doesn’t often write about meta-analyses, since they typically don’t have new data. But this one was different because it’s about fluoride, which is in the news, and it involves children’s IQ scores.

“So, I had a conversation with my editor asking if we are better off not giving this study any press and saying nothing about it, or do we feel like it’s likely that people are going to hear about it because fluoride is in the news,” she says. “And given that people are going to talk about it, we decided that we’d rather give them something responsible to read than just whatever people might say on social media.”

She advises journalists to “do the analysis that the average reader isn’t able to do.”

3. Look at where the studies were conducted and whether the findings apply to your geographic area.

One of the first things Kaplan noticed when she started paging through the meta-analysis was that none of the analyzed studies were conducted in the U.S.

Of the 74 studies included, 45 were conducted in China, 12 in India, four in Iran, four in Mexico, three in Canada, two in Pakistan, and one each in Denmark, New Zealand, Spain and Taiwan.

The authors are upfront about that. They note that they’re not aware of any U.S. studies on the effects of fluoride exposure on children’s IQ and that their findings may not apply to the U.S. population, although they write that the findings may inform future public health assessment of fluoride.

In her reporting, Kaplan focused on the current recommended water fluoride levels in the U.S., which is 0.7 mg/L. So she started to cross off all the studies that included fluoride levels at 1.5 mg/L and above.

“And that didn’t leave me with very much else,” she says.

Of the 74 studies, 12 investigate water or urinary fluoride levels below 1.5 mg/L. Seven of those studies included fluoridated water below 1.5 mg/L and found no association with an IQ drop in children. The five studies investigating urinary fluoride levels at up to 1.5 mg/L found a statistically significant association with lower IQ in children.

4. Look for and interrogate the sources of data.

Levy writes in his editorial that the authors don’t mention the fact that their study is part of a multiyear effort by the U.S. National Toxicology Program to study this topic. NTP is part of the National Institutes of Health. The reviews began in 2015 and culminated in the publication of the “NTP Monograph on the State of the Science Concerning Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review” in August 2024. (A monograph is a detailed written study of a single specialized subject.)

This is noteworthy for two reasons.

First, the NTP Monograph has received a fair share of criticism from an independent committee at the nonprofit National Academies of Science, Engineering, and Medicine. The NASEM committee reviewed the NTP Monograph in two rounds and both times found flaws in its methods. In 2020, the committee found deficiencies in the Monograph and suggested further analyses. In 2021, the NASEM committee wrote, “As documented in this letter report, the committee had difficulty in following various aspects of the reported methods, identified a few worrisome remaining inconsistencies, was not able to find some key data used in the meta-analysis, and had concern about the wording of some conclusions.”

Second, a federal district judge cited the Monograph’s findings in his ruling last year that fluoride in community water systems could pose an unreasonable risk to people. He asked the Environmental Protection Agency to take regulatory action in response.

In his editorial, Levy writes, “as EPA proceeds with an appeal of the ruling or to implement the court-ordered regulatory action, better understanding of the limitations of the article’s data, analyses, and interpretation are extremely important.”

5. Check for the quality of included studies.

“Risk of bias” in research refers to the likelihood of systematic errors or flaws in how a study was designed, conducted or analyzed — factors that could lead to incorrect results. Researchers use certain calculations to decide a study’s risk of bias. A study that’s considered to have a high risk of bias is considered lower quality and those with low risk of bias are considered better quality.

In most meta-analyses, researchers include the studies’ risk of bias.

In this meta-analysis, of the 74 studies, 52 were considered to have lower quality and 22 were higher quality, the authors report.

In addition to the overall analysis, researchers also conducted analyses based on the studies’ quality.

For instance, when all 31 studies on water fluoride were analyzed — a mix of higher and lower quality studies — they showed an association with lower IQ scores at up to 4 mg/L, but not at up to 2 mg/L and 1.5 mg/L levels. Meanwhile, the analysis of seven of those water fluoride studies that were rated higher quality, showed an association between lower IQ scores when water fluoride levels were up to 4 mg/L and 2 mg/L, but not at up to 1.5 mg/L.

In the analysis of the 20 studies that reported on the levels of fluoride in urine, there was an association between lower IQ scores at all three levels of up to 4 mg/L, 2 mg/L and 1.5 mg/L of fluoride in urine. Further analysis of 10 of those studies that were of higher quality also showed a similar association.

6. Look at the number of studies in the sub-analyses.

For all 31 studies involving fluoride in water, there was a statistically significant association between water fluoride levels of up to 4 mg/L with lower IQ scores in children (23 studies), but not at up to 2 mg/L (eight studies) and 1.5 mg/L (seven studies).

For the seven higher-quality studies involving fluoride in water, this negative association remained at up to 4 mg/L (seven studies), up to 2 mg/L (four studies), but not at up to 1.5 mg/L (three studies).

For all 20 studies involving fluoride in urine, there was a statistically significant association with children’s IQ at all three levels.

And when this pool was narrowed down to 10 higher-quality studies, the statistically significant association remained. Ten studies had assessed urinary fluoride of up to 4 mg/L, four studies looked at 2 mg/L and four looked at 1.5 mg/L or less.

“So I would look at the totality of it,” Nuzzo says.

There’s no magic number for what makes up a good number of studies, Nuzzo says. But fewer studies in analysis means fewer people — the number of children in this case — which can make it harder to reach statistical significance or draw broad conclusions.

7. Understand the standardized mean differences and standard deviations.

Standardized Mean Differences, which are used throughout the tables and figures in this meta-analysis, are a statistical measure to quantify the difference between two groups. They’re commonly used in meta-analyses to compare the results.

SMD is “a way of creating a standard unit,” Nuzzo says. “It’s a way that we can compare apples to apples, by bringing everything to a common currency.”

SMDs are expressed in standard deviations. A standard deviation of a dataset is a number that indicates how much variation there is within that dataset.

For instance, if SMD is zero — a standard deviation of zero — it means there’s no difference between the two groups that are being compared.

An SMD of 1 means that there’s one standard deviation difference between the two groups, and a -1 means the same thing in the opposite direction.

For instance, the calculated SMD for 31 studies on water fluoride is -0.15. That means the drop in children’s IQ is 15% of one standard deviation.

Scientists like to work with SMDs because standard deviations follow certain patterns in nature, explains Nuzzo. Also, “you’re not likely to see standardized mean differences more than three,” she says.

8. Get to know IQ scores so you can explain the numbers.

As Kaplan points out in her story, “assessing IQ in children is not a straightforward affair since measurements can be skewed” by a range of factors including socioeconomics, genetics, nutrition and environment.

IQ tests were designed in 1905 in France. The average IQ score is between 85 and 115, with a mean and median of 100.

“That’s the way they’ve set up the IQ scores because humans have created it and we’re able to force it to be this nice number,” Nuzzo says.

An increase or decrease of one standard deviation in IQ means a 15-point increase or decrease.

“If you go up two standard deviations, now you’re up 30 points, which is an IQ of 130, which is where we often talk about the gifted, genius level,” Nuzzo says. “You’d be an outlier.”

So, for another example from the study’s findings, a -0.15 drop in IQ is a drop equal to 15% of one standard deviation. One standard deviation is 15 points. If you multiply -0.15 by 15 IQ points, you get a drop of 2.25 IQ points.

“So that is telling us that the difference between the mean IQ of kids in high-fluoride regions versus low-fluoride regions is an average of 2.25 IQ points,” Nuzzo says.

In their editorial, Lanphear and co-authors also explain this, but in a more technical language.

“For standardized IQ tests with a mean (SD) of 100 (15), an SMD of 0.5 is equivalent to 7.5 IQ points; an SMD of -0.19 is a 2.8-1Q point decrement,” they write.

More examples: For children exposed to water fluoride levels at up to 4 mg/L, the change in standard deviation was -0.22, or 22% of one standard deviation. That’s equal to a 3.3 drop in IQ points, according to the study.

And the IQ difference for children exposed to up to 1.5 mg/L of fluoride as measured in their urine is 1.2 points, the study finds.

In another sub-analysis, the authors find that at an individual level, a 1 mg/L increase in urinary fluoride was associated with a decrease in IQ by 1.63 points.

Lanphear and co-authors write that a 1.63 drop in IQ may seem very small, but according to studies on other neurotoxins such as lead, subtle IQ changes can affect people who are in the high and low ranges of the population’s IQ distribution.

But also keep in mind that some researchers, including Levy, don’t agree.

“Despite the presentation of some evidence of a possible association between IQ and high fluoride levels in the water, there is no evidence of an adverse effect at the lower water fluoride levels commonly used in [community water fluoridation] systems,” Levy writes in his editorial.

9. Look at the confidence interval (usually noted in parentheses).

A confidence interval shows the range of values that are expected for the study’s results to fall between if the study were repeated many times. They show the most plausible range of values, Nuzzo explains. The smaller the confidence interval the more reliable the reported number.

In academic research, confidence intervals usually appear in parentheses next to a reported number.

For instance, the reported confidence interval for the IQ drop of -0.15 in water fluoride studies is between -0.20 and -0.11. The range is negative, which means fluoride doesn’t seem to be helping IQ scores.

But, for the batch of three higher-quality studies that looked at IQ scores when water fluoridation was up to 1.5 mg/L, the reported confidence interval ranges from -0.91 to 0.26. (The negative shows a negative impact on IQ and the positive side shows a positive impact on IQ.)

“It covers zero and this is a huge range, which just tells me we don’t have enough information really here,” Nuzzo says.

Also, when the confidence interval ranges from negative to positive, the statistical significance, known as the p-value, becomes greater than 0.05, which means there’s no statistical significance.

“Too bad statisticians put confidence intervals in parentheses because it makes them seem like something that you just overlook,” Nuzzo says.

10. Take a look at the forest plots.

A forest plot is a graph where researchers plot and compare the results of multiple individual studies on the same axis.

The plots feature a vertical line that lands on a zero on the horizontal (x) axis.

Another vertical line, marked by a diamond on the x-axis shows the pooled result of all studies, in the case of this study, it’s the pooled standardized mean difference of -0.45, which falls to the left of zero (average IQ drop of 6.75).

The results of individual studies are then plotted and can be compared to the pooled result.

In one of the forest plots in this meta-analysis, you can quickly spot one outlier, a 2015 paper by Khan et al., which has an SMD of -5.3.

“That’s telling us the IQ difference between high fluoridated water areas and low fluoridated water areas is 80 IQ points,” Nuzzo says. “I’m sorry, no.”

Sometimes researchers stack the studies of the forest plot in chronological order, which can help show trends over time.

“As a statistician, I look to see how much the results are bouncing around” the zero line, Nuzzo says. Putting aside the 2015 study, “and looking at this, I think the results here are fairly consistent, most of them are small, but they’re slightly to the left of zero,” which points to a negative link between IQ scores and fluoride levels.

11. Look for other systematic reviews and meta-analyses on the topic.

For instance, a meta-analysis of 28 international studies, published in the journal Public Health in April 2023, finds that there’s no association between exposure to water fluoride levels below 1.5 mg/L and lower IQ scores. However, “the reported association observed at higher fluoride levels in endemic areas requires further investigation,” the authors write.

Another systematic review and meta-analysis of 33 international studies, published in the journal Environmental Research in January 2023, finds that overall, there are indications that more exposure to fluoride can have a negative impact on children’s cognitive development, even when fluoride levels are rather low. But, the authors note, that most of the studies have limitations, which raises uncertainties about a cause-and-effect link and the exact threshold of fluoride in water that may have negative effects.

As with many studies on fluoridation, the authors call for more, high-quality, long-term research.