This is the final article in our BMR series. You can find the rest of the articles here, exploring how things like age, sex, weight gain, weight loss, and athletic status influence BMR. All of these factors are accounted for in our BMR calculator, but for reasons that will soon become clear, the calculator doesn’t account for PCOS status.
For many years, there’s been an ongoing conversation about the challenges women with Polycystic Ovary Syndrome (PCOS) face in managing their weight, since women with PCOS have higher rates of overweight and obesity than women without PCOS. A significant part of this discussion centers on the claim that women with PCOS have a lower Basal Metabolic Rate (BMR). This claim is mainly supported by a single study. Because of these BMR claims, many women feel a sense of discouragement when they believe that their biology is working against them, no matter how diligently they stick to diet and training. Should they? What does the collective research say? How valid is that single study people always cite?
This article explores these claims and the research to determine if women with PCOS have a lower BMR.
Let’s dig in.
A brief primer on PCOS
PCOS stands for polycystic ovary syndrome. There are actually two terms: PCO and PCOS. PCO denotes the presence of extra follicles, whereas PCOS is a collection of the various metabolic alterations from “normal” (specifically elevated androgen and insulin hormones). The “S” adds the syndrome portion to the equation. In reality, polycystic ovaries can be an anatomical condition. The presence of polycystic ovaries alone doesn’t automatically mean a person has the syndrome. As with most pathologies, the collection of symptoms and many heterogeneous factors lead to the actual “syndrome” part.
PCOS does not have a single standard diagnostic criterion. Different doctors, hospitals, research universities, and gynecologists may use varying criteria to determine if your collection of symptoms qualifies as PCOS. Each criterion has different symptoms that must be present.
In short, there are multiple diagnostic criteria for PCOS. But, all diagnoses require the presence of either clinical or biochemical hyperandrogenism, or chronic oligo-anovulation. In other words, there must be evidence that your body either produces different levels of hormones than women without PCOS, that it responds differently to those hormones, or both. Concerns about the metabolic effects of PCOS largely stem from these hormonal effects of the condition.
So, why do so many people think that women with PCOS have a low BMR?
At this time, if you Google “PCOS and BMR,” you’ll immediately be hit with multiple links to a 2009 study by Georgopoulos et al titled, “Basal metabolic rate is decreased in women with polycystic ovary syndrome and biochemical hyperandrogenemia and is associated with insulin resistance.” It’s cited by over 100 studies via Google Scholar metrics, making it the most frequently cited study comparing BMRs in women with and without PCOS. It’s cited in all of the blog posts and lay articles that show up near the top of Google search results discussing the impact of PCOS on BMR (one, two, three, four, five). It’s discussed in online communities for women with PCOS. The findings of the study have even shown up in educational materials distributed by the British National Health Service. It’s not an exaggeration to state that the perception that women with PCOS have decreased BMRs is driven almost exclusively by this one study.
The study purports to show that women with PCOS but no insulin resistance have BMRs about 13% lower than women without PCOS. Additionally, it purports to show that women with both PCOS and insulin resistance have BMRs nearly 40% lower than women without PCOS.
The study states, “In women with PCOS in the present study, adjusted BMR was decreased compared with control subjects, independently of obesity and insulin resistance (IR). Adjusted BMR was significantly decreased both in women with PCOS with or without IR and particularly in women with PCOS and IR.”
That paints a pretty gloomy picture for women with PCOS, and particularly women with insulin resistance.
So, is the 2009 study an outlier?
To ensure we did a thorough accounting of the research, we performed a systematic search of the scientific literature using variations of “RMR,” “REE,” “BMR,” “resting energy,” “basal energy expenditure,” and other terms alongside “PCOS,” “polycystic,” and similar variations to find all of the studies assessing metabolic rates in women with PCOS. Some studies were found that weren’t focusing on BMR directly, but which still assessed and reported BMR during the course of the research. It’s always possible that we missed a study or two, but we’re confident that our search turned up all – or virtually all – of the research on the topic. We also examined the methods of BMR measurement and ensured that only studies using indirect calorimetry were included, since multiple studies on the topic measured BMR using other methods that aren’t valid or reliable.
From there, we performed a meta-analysis – a statistical “study of studies” – on all of the research comparing BMRs in matched groups of women with and without PCOS. The meta-analysis was performed in JASP, using a restricted maximum likelihood model with inverse variance weighting.
We found 18 studies that assessed BMR in women with PCOS. Of these, four didn’t directly assess BMR using indirect calorimetry. Three reported predicted BMRs from body composition assessments (in other words, they didn’t actually measure BMR in the first place), and one reported predicted BMRs from wearable armbands (again, not an actual measurement of BMR). These four studies were excluded from all further analyses. Of the remaining 14 studies, 7 directly assessed BMR in women with PCOS, without any comparison to a control group of women without PCOS. These seven studies therefore couldn’t be used in our primary meta-analysis, but they’ll be discussed in secondary analyses to characterize the research on PCOS and BMR more broadly. So, seven studies ultimately met our inclusion criteria for the meta-analysis.
Results of the Studies that have Assessed Basal Metabolic Rate (BMR) in Women With and Without PCOS
| Author, Year | Group | BMR (Calories/day; Mean ± SD) | BMR difference between PCOS and non-PCOS women1 |
|---|---|---|---|
| Segal, 1990 | PCOS with obesity | 1508 ± 1682 | -50 Calories |
| non-PCOS with obesity | 1558 ± 1862 | ||
| Robinson, 1992 | PCOS | 1624 ± 1374 | -9 Calories |
| non-PCOS | 1633 ± 2154 | ||
| Cosar, 2008 | PCOS | 1167 ± 371 | 121 Calories |
| non-PCOS | 1046 ± 296 | ||
| Georgopoulos, 2009 | PCOS | 1446 ± 7252 | -395 Calories |
| non-PCOS | 1841 ± 3052 | ||
| Graff, 2013 | PCOS | 1469 ± 227 | 16 Calories |
| non-PCOS | 1453 ± 249 | ||
| Larsson, 2015 | PCOS | 1411 ± 229 | 86 Calories |
| non-PCOS | 1325 ± 193 | ||
| Doh, 2016 | Non-obese PCOS | 1272 ± 1673 | 32 Calories |
| Non-obese, non-PCOS | 1240 ± 2163 | ||
| 1 Positive values indicate higher BMRs in women with PCOS 2 Standard deviations were calculated from reported standard errors and sample sizes 3 Data was reported as median and IQR. Here, median is assumed to be sufficiently close to the mean, and SDs were estimated from the IQRs 4 Data was reported as median, minimum, and maximum. Median is assumed to be sufficiently close to the mean. SDs were estimated from reported range, on the assumption that the largest and smallest values were approximately 2SDs from the mean | |||
As a note, two of these studies assessed BMR in three groups of women. Segal and colleagues assessed BMR in obese women with PCOS, obese women without PCOS, and non-obese women without PCOS. The comparison between the two groups of women with obesity was used for this meta-analysis, to provide an apples-to-apples comparison. Similarly, Doh and colleagues assessed BMR in obese women with PCOS, non-obese women with PCOS, and non-obese women without PCOS. The comparison between the two groups of non-obese women was used for this meta-analysis. The other five studies only included one group of women with PCOS, and one group of women without PCOS. In all five of these studies, basic demographic and anthropometric characteristics were similar between groups. So, in total, this meta-analysis pools the data from 444 subjects in 14 groups across 7 studies.
The results of the meta-analysis can be seen in the forest plot below. Across these seven studies, there was virtually no difference between the BMRs in women with and without PCOS (g = -0.01, p = 0.925).
Visually, the Georgopoulos study appears to be a bit of an outlier. It found a negative effect that was more than twice as large as any other study. And, statistically, it was identified as an “influential” study, with a covariance ratio of <1, and a standardized residual, DFFITS value, and Cook’s distance that were ~2.5-4 times larger than any other study.
Re-running the meta-analysis with the Georgopoulos study excluded improved model fit (all residual heterogeneity statistics decreased), but the pooled effect was still non-significant (g = 0.12, p = 0.274).
In effect, even with the Georgopoulos study included in the analysis, the research suggests that women with and without PCOS have pretty similar BMRs. When the Georgopoulos study is excluded as an outlier, the other six studies suggest that women with PCOS may actually have slightly higher BMRs than women without PCOS, but the difference is still trivial in magnitude and not statistically significant. Additional analyses were performed to confirm the robustness of this result (testing a fixed-effect model, excluding the studies by Doh and Robinson that didn’t report means and SDs or SEs, and adjusting weighting parameters), and none of those analyses materially changed the result.
Moving on, let’s pull back in the other seven studies that only assessed BMR in women with PCOS, without a comparison to a non-PCOS control group. This leaves us with a total pool of 14 studies that characterize the BMRs of women with PCOS. From these studies, we can draw comparisons to research that has assessed BMRs of women in the general population.
These 14 studies assessed BMR in a total of 642 women with PCOS. Below, you can see the sample size, and average height, weight, age, and BMR in all of these studies.
Characteristics of the studies reporting BMRs in 642 women with PCOS
| Author, Year | Sample size | Age | Height (cm) | Weight (kg) | BMI | Average BMR |
|---|---|---|---|---|---|---|
| Segal, 1990 | 10 | 25 | 163 | 84.1 | 31.71 | 1507.7 |
| Robinson, 1992 | 14 | 27 | 161.22 | 70.3 | 27.1 | 1624.3 |
| Bruner, 2006 | 12 | 30.7 | 163.82 | 98.1 | 36.6 | 1531.3 |
| Kritikou, 20064 | 63 | 24 | 164 | 72.5 | 27.4 | 1505.6 |
| Moran, 2006 | 13 | 32.6 | 165.92 | 96 | 34.9 | 1840.3 |
| Saltamavros, 20074 | 73 | 24 | 164 | 70.9 | 26.7 | 1381.5 |
| Cosar, 2008 | 31 | 25.9 | 164.4 | 72.93 | 27 | 1166.9 |
| Georgopoulos, 20094 | 91 | 23.9 | 164.4 | 72.13 | 26.7 | 1445.6 |
| Koika, 20094 | 156 | 22.8 | 164.3 | 69.23 | 25.6 | 1415.7 |
| Graff, 2013 | 61 | 22.7 | 164.5 | 78.23 | 28.9 | 1469 |
| Larsson, 2015 | 46 | 30.2 | 167.3 | 79.6 | 28.5 | 1411 |
| Broksey, 2017 | 28 | 28.6 | 161.52 | 104.1 | 39.9 | 1689 |
| Rodrigues, 2017 | 30 | 30.8 | 161.82 | 85.3 | 32.6 | 1677 |
| Doh, 2016 | 14 | 26.6 | 169.32 | 86 | 29.7 | 1331.2 |
| Totals or weighted averages | 642 | 25.1 | 164.3 | 75.9 | 28.2 | 1456.6 |
| 1This study reported height and weight but not BMI, so average BMI was calculated via the standard formula: BMI = Weight/Height 2These studies reported BMI and weight, but not height, so height was estimated via this formula: height = √(weight/BMI) 3These studies reported BMI, but neither height nor weight. Weight was estimated via regression-based imputation. BMI was strongly linearly associated with weight in the 10 studies that reported both values (r = 0.93). So, the regression equation describing that relationship was computed, and BMI values from the four remaining studies were used to estimate the missing weight values 4These studies came from the same lab that conducted the Georgopoulos study, which is discussed more below | ||||||
Given the average height, weight, and age of these subjects, we can calculate their expected average BMR, using equations developed from large samples of women in the general population. Based on a thorough analysis of the research on the topic, we tend to think that the Oxford/Henry and Mifflin-St Jeor equations are the two best “off the shelf” equations for calculating BMR from height, weight, and age. Both of these equations would calculate an estimated BMR just north of 1500 Calories (1501 for Mifflin-St Jeor, and 1512 for Oxford/Henry) for women matching the characteristics observed in the PCOS research. So, on average, the women with PCOS in these studies had BMRs that were about 50 Calories or 3% lower than would be predicted. This is a pretty trivial difference, bolstering the findings of our meta-analysis.
Finally, five of these studies also assessed both BMR and fat-free mass in women with PCOS. These 79 subjects had an average of 52.9kg of fat-free mass, and an average BMR of 1616.1 Calories. Using the 1991 version of the Cunningham equation, which estimates BMR from fat-free mass, these subjects would be predicted to have a BMR of 1512 Calories. So, on average, the women with PCOS in these studies had BMRs that were about 100 Calories or 6% higher than would be predicted. This is also a pretty trivial difference.
Studies Reporting Both Fat-Free Mass and BMR in Women with PCOS
| Author, Year | Sample size | Average Fat-Free Mass (kg) | Average BMR |
|---|---|---|---|
| Segal, 1990 | 10 | 48.7 | 1507.7 |
| Robinson, 1992 | 14 | 50.2 | 1624.3 |
| Moran, 2006 | 13 | 61.5 | 1840.3 |
| Broksey, 2017 | 28 | 52.5 | 1689 |
| Doh, 2016 | 14 | 51.2 | 1331.2 |
| Totals or weighted averages | 79 | 52.9 | 1616.1 |
All of these analyses paint a consistent picture: it doesn’t appear that PCOS has a notable impact on BMR. The idea that PCOS decreases BMR is driven by a single study that’s been widely cited, discussed, and shared, but the findings of that study are radically out of step with the rest of the research on the topic. A thorough analysis of this body of literature suggests that PCOS has no meaningful impact on BMR. To be clear, some women with PCOS will have low BMRs (and some will have high BMRs), because BMRs are much more variable than most people realize. But PCOS doesn’t appear to systematically and independently affect BMR.
What were some of the issues with the 2009 study?
We could leave it there, but the 2009 Georgopoulos study has been so influential that we think it’s worth pointing out a few other obvious issues with it. It departs from the rest of the literature to a much greater degree than can be explained by random chance. But, when you dig a bit deeper, an explanation for its eye-popping findings becomes much clearer.
Primarily, there’s a very good chance that the researchers ran into an equipment issue. The machine they used to measure BMR probably wasn’t very good.
This study assessed BMR using an indirect calorimeter called the PulmoLab EX-505. That may not mean much to most readers, but it was our first clue. We read a lot of metabolism research, and we see which devices and manufacturers are commonly used and trusted by researchers. The dominant manufacturers are Parvo Medics and Cosmed for stationary metabolic carts, and Breezing for portable units. DeltaTrac also pops up from time to time, especially in research conducted in hospitals. There are plenty of other players in the consumer market, but devices from those four manufacturers have the most research validating their accuracy. So, when a new name pops up, it’s good practice to try to find independent research validating the device.
We weren’t able to find any validation research on the PulmoLab EX-505. And, we were unable to find any research labs using this device other than the lab that conducted the outlier 2009 Georgopoulos study. We can’t claim that no such validation research exists, and that no other labs use the device, but we were unable to find them after a considerable amount of searching.
But, we did find a validation study on the higher-end sibling of the PulmoLab EX-505 – the PulmoLab EX-670. Unfortunately, the EX-670 performed quite poorly. It showed considerably more variability and less reliability in capturing accurate respiratory measurements than other systems. A low coefficient of variation is an indicator of high reliability, but this study reports that, “the coefficients of variation for … Douglas bags, Oxycon Pro and Oxycon Alpha were 3.3–5.1%, 4.7–7.0% and 4.5–6.3%, respectively, whilst that for the Pulmolab was highly variable (26.8–45.8%).” In other words, it was about 5-10 times less reliable than the other devices tested in that study.
To be clear, it’s possible that the EX-505 is a better device than the higher-end EX-670 from the same manufacturer. But, since we couldn’t find validation research for the EX-505, and since a (presumably better) device from the same manufacturer appears to be remarkably unreliable, we find it likely that the outlier findings of the 2009 Georgopoulos study may have been the result of simple measurement error, due to using an unreliable device to assess BMR.
Digging deeper into the results
Even if we couldn’t pinpoint a reason (like measurement error from using an unreliable device to assess BMR) for the eye-popping findings of the 2009 Georgopoulos study, a close examination of the data itself is enough to suggest that something went wrong when collecting the data.
To explain why, we need to discuss statistics a little bit.
When reading research, most people pay attention to the means (the averages). But, studies also report measures of variability, typically in the form of standard deviations or standard errors. The larger the standard deviation is in relation to the mean, the more variable the data is. If you see an average of 10 ± 1 (mean ± standard deviation), that means about two-thirds of values are between 9 and 11, 95% of values are between 8 and 12, and 99.9% of values are between 7 and 13. But, if an average is 10 ± 5, that means values between 5 and 15 are pretty common, values between 0 and 20 aren’t particularly rare, and values below 0 or above 20 should crop up about 5% of the time. The average is the same in both instances, but a larger standard deviation means the values are much more spread out.
I won’t bore you with the technical reasons someone might want to calculate a standard error instead of a standard deviation. All you need to know is that you calculate a standard error by dividing the standard deviation by the square root of the sample size. So, if your standard deviation is 100, and your sample size is 25, your standard error is: 100 25=20.
For certain types of data, there’s a certain amount of variability you expect to see. For example, if female subjects in a study are reported to be 165 ± 5 cm tall (mean ± standard deviation), you know you’re dealing with pretty typical data. The average woman in the study is 165cm (perfectly typical), and about 95% of the women in the study should be between 155cm and 175cm tall (also perfectly typical).
But, if female subjects are reported to be 165 ± 30 cm tall, you might start asking questions. Are one-third of your subjects really shorter than 135cm or taller than 195cm? Do you have a handful of enormous outliers dramatically increasing the variability in your data? Did you potentially make some errors when taking the measurement or transcribing your data into a spreadsheet? If you just paid attention to the averages, nothing about an average height of 165cm would seem strange. But, when you see a dramatically larger (or smaller) standard deviation than you’d expect from a particular type of data, that warrants further investigation.
How much variability do we tend to see in BMR data?
In all of the studies discussed above that didn’t come from the lab publishing the 2009 Georgopoulos study (and that didn’t use the PulmoLab EX-505 device to assess BMR), the average standard deviation for BMR values was 233 Calories. So, about two-thirds of subjects had BMRs within 233 calories of the group mean, and about 95% of subjects had BMRs within 466 calories of the group mean. In other words, if the average was 1500 Calories, most values should be between 1034 calories and 1966 calories.
Looking beyond this body of research, and using other large studies as a point of reference, the female subjects in the Mifflin-St Jeor study had an average BMR of 1349 Calories, with a standard deviation of 214. In a recent large study by Pavlidou and colleagues, the female subjects had an average BMR of 1533 Calories, with a standard deviation of 308. We won’t bore you with a dozen other examples, but most reasonably large studies (>50 subjects) that report female BMRs have standard deviations of about 200-350 calories.
So, turning our attention to the 2009 Georgopoulos study, there were 91 women with PCOS, who were reported to have an average BMR of 1445.57 Calories, with a standard error of 76 Calories. With a sample size of 91 subjects, a standard error of 76 means the standard deviation was 724 Calories. That’s way more variability than we tend to see in BMR research. Taken at face value, that would mean you’d expect about one-third of the subjects to have BMRs below 725 Calories, or above 2175 Calories. Furthermore, you should expect a decent number of subjects to have BMRs below 500 Calories or approaching 3000 Calories. For comparison, the lowest female BMR observed in Mifflin-St Jeor study was 927 Calories, and the highest was 2216 (in a sample of 247 women). In the Pavlidou study (with a sample of 549 subjects), the lowest female BMR was 908, and the highest was 2492.
Thankfully, we don’t just have to extrapolate and make assumptions about the extreme values implied by such large standard deviations. We can see the spread of reported BMR values when we turn to other research from the same lab, published in the same year, using the same device (and probably the same sample of women – This research group published four fairly large studies on BMR in women with PCOS in the span of four years. They almost certainly recruited a single sample and analyzed it multiple times, or reported on more data as they collected it).
In a sample of 156 women with PCOS, these researchers reported an average BMR of 1415.7 Calories, with a standard deviation of 672.9. Furthermore, they reported the range of values: the lowest reported BMR was 328.2 Calories, and the highest was 3969 Calories. Stated simply, those values are impossible in this population. Outside of research from this single group, the lowest adult BMRs I’ve encountered in the literature were from patients with severe anorexia; the very lowest BMRs in this population were just north of 2 kilojoules per minute (or about 700 Calories per day). The highest single BMR I’ve encountered was from an extremely muscular collegiate male athlete, with a BMR of around 3700-3800 Calories per day. There’s no other way to say this: the single research finding that led to the popular notion that women with PCOS have lower BMRs is based on obviously bad data.
To be clear, we’re not implying that anyone knowingly did anything unethical here. The data is bad, but we don’t think it’s bad as a result of ill intent. We suspect this research had the same lifecycle of most low-quality but highly cited research. People collected low-quality data. They found illusions that looked like clear patterns in the resulting noise (if you collect enough data, you should expect to find spurious “statistically significant” results purely by chance). Peer review is a poor system for catching these sorts of errors, thus allowing the results to get published. Then, due to a combination of excessive faith in results that have been peer reviewed, insufficient statistical training to spot these types of issues, and the constraints of writing under time pressure, journalists, bloggers, social media users, and even other scientists fail to recognize the problem and continue citing the research.
Stated another way, it would take less than a minute to skim the abstract of the study, see the reported finding that women with PCOS have BMRs that are dramatically lower than women without PCOS, and share an eye-popping finding that feels true (women with PCOS do have a range of metabolic hurdles to overcome, and do often have very understandable struggles with weight management as a result) and that has the imprimatur of scientific legitimacy. On the flip side, identifying the problems with the study required 1) a thorough knowledge of the rest of the research on this topic to first know that something might be up with it, 2) enough statistical knowledge and familiarity with other metabolism research for the large standard deviations to jump out, and 3) hours of legwork to learn more about the device used to assess BMR in the study, and to gather and fully analyze the rest of the research on the topic.
A small rant: Women with PCOS deserve better
Hi. This is Greg. Leigh did the vast majority of the work for this article. I helped out a bit with the statistical analysis. But, I did want to share a few personal thoughts after diving into this body of research to help out with this article.
I do a lot of deep dives into various bodies of research. I’m in the middle of one such deep dive for this BMR series. Most of the bodies of research I dig into relate to health and fitness in some way. Some are good (research on sex differences in muscle growth: surprisingly good!). Many are bad (a lot of the research on dietary supplements…boy howdy). Some are extremely bad. All of which is to say, this isn’t my first rodeo, and I have many points of reference to compare this body of research to.
Top to bottom, this body of research is particularly rough. This article has already said plenty about the Georgopoulos study that most conversation is centralized around, but that’s really just the tip of the iceberg.
The next most-discussed study online (which is the study used to generate the automatic snippet for Google, as of the time of writing) is a conference abstract, claiming that “Patients with PCOS have lower basal metabolic rate (BMR) even when controlled for BMI. … Therefore, PCOS patients may be at risk of lower metabolism that can lead to obesity related to PCOS.”
The most-discussed study that people will share as evidence that women with PCOS don’t have lower BMRs … is the exact same study. Between publishing the abstract and submitting the paper for peer review, I guess the researchers re-ran their analyses, and came away with the opposite result: “After adjusting for age and BMI, there was no significant difference in BMR between PCOS subjects and controls. BMR was also comparable in a secondary analysis comparing PCOS women with and without insulin resistance.”
The problem with both of these studies is that neither of them actually measured BMR. They estimated body composition using BIA (which isn’t even a great method of estimating body composition), and then estimated BMR from the body composition estimate. In other words, the study is actually saying, “After adjusting for age and BMI, women with and without PCOS had similar body composition. So, we’ll just assume they have similar BMRs as well.”
Three other studies claiming to assess differences in BMR between women with and without PCOS also didn’t actually assess differences in BMR. Two others used body composition as the same sort of extremely rough proxy for BMR (if you’re actually studying body composition, just say you’re studying body composition!). One used an armband that is known to do a poor job of estimating BMR – the armband overestimates energy expenditure when sedentary by a factor of two!
But, the hits keep coming. Three other studies came from the same lab that published the Georgopoulos study (one, two, three). They all have the same issues – they all measured BMR using the same device, and they all reported implausibly large standard deviations. And, they mostly have the same feel to them – find some factor to subdivide your study population (oftentimes based on genotype instead of insulin sensitivity), find an implausibly low BMR in a tiny sub-sample of subjects (that’s almost certainly just noise; I strongly suspect that a statistically significant result finally popped out after trying a dozen other ways to sub-divide the subjects that didn’t produce statistically significant results), publish it, collect more low quality data, rinse and repeat.
Oh, and there was a meta-analysis on BMR in women with and without PCOS that was only published as a conference poster, but it still shows up pretty high in the search results. It failed to find most of the research that had been published at the time (the authors should have turned up 10 studies, given the types of studies they were willing to include. They found 3). The three studies included were Georgopoulos (for which they dramatically miscalculated the effect size, probably because they couldn’t tell the difference between standard deviations and standard errors. The effect size should have been -0.46. Instead, it was -10.25), Koika (which doesn’t even contain a comparison between women with and without PCOS. This is a complete head scratcher), and Churchill (which is one of the studies that didn’t even measure BMR). Somehow, they found a way to exclude Churchill “after bias and weighting control.” This is very confusing, since the miscalculated Georgopoulos effect size is nearly 80x larger than the other two; I’m not aware of any statistical procedure that would justify keeping Georgopoulos and excluding Churchill. So, they wound up with a “meta-analysis” of two studies, which were probably both based on the same sample of subjects, and one of those studies didn’t even measure BMR in women without PCOS. It’s not an exaggeration to say that this is the most tragically flawed attempt at a meta-analysis I’ve ever seen (which is really saying something). I feel a little bad being this hard on a conference poster … but not that bad. It’s a real stinker.
To be clear, there is high-quality research on this topic. Segal, Rodrigues, Doh, Graff, Bruner, Robinson, Broskey, Moran, Larsson, and Cosar all deserve their flowers. But, approximately half of the research in this area is either hot garbage, or not even measuring what it claims to measure. That’s a really bad ratio.
And, that really bothers me, because women with PCOS do face a lot of unique challenges, and they need to sort through a lot of misinformation from health, fitness, and wellness influencers to find reliable information about their condition. Actual research on the topic should be a safe port in the storm, but it’s not. This body of research isn’t even “garden variety” bad – as a whole, it’s particularly awful. And that sucks. So, if I was a bit less tactful than normal when describing the shortcomings of this body of research, that’s why.
Last note: I re-ran the analyses of average height, weight, age, and BMR excluding all four studies from the lab that published the Georgopoulos study. The remaining subjects were a bit heavier, a bit older, and had slightly higher BMRs, but, the overall takeaway from that analysis didn’t change. None of the studies assessing fat-free mass came from that lab, and none of the other studies from the lab were included in the primary meta-analysis because they didn’t include comparisons to non-PCOS control groups.
Weighted average height, weight, age, and BMR of the remaining 249 women with PCOS
| Height | 164.7cm/64.9in |
| Weight | 84.8kg/184.7lb |
| Age | 27.4 |
| BMR (kcal/day) | 1494.4 |
What do other studies say regarding insulin resistance and BMR?
To close out this article, we should circle back to discuss one aspect of the Georgopoulos study that we haven’t addressed yet: whether increased insulin resistance resulting from PCOS reduces BMR.
So, let’s briefly look at the research examining the impact of insulin resistance on basal metabolic rate in women. Do other studies suggest that low insulin resistance in women decreases BMR?
For starters, a study by Drabsch et al found that in women, greater insulin resistance was associated with a slightly higher BMR. Furthermore, while not all type 2 diabetics have insulin resistance, the vast majority do, and there are studies that continue to show that BMR is not decreased (and is often slightly increased) in people with type 2 diabetes – here and here.
But, we don’t even need to look outside of the PCOS research. In all of the studies discussed above that measured at least one marker of insulin sensitivity, the women with PCOS had elevated markers of insulin resistance. Graff, Broskey, and Moran all reported elevated HOMA-IR values, Doh reported slower glucose disposal in a euglycemic-hyperinsulinemic clamp protocol, Robinson also reported slower glucose disposal after insulin infusions, and Cosar reported elevated fasting insulin levels. In the body of research discussed above, which found that women with PCOS have normal BMRs, most of the women with PCOS had poor insulin sensitivity.
Is reduced insulin sensitivity a factor for PCOS in general?
A 2016 systematic review and meta-analysis by Cassar et al examined insulin resistance in women with polycystic ovary syndrome (PCOS) using euglycemic–hyperinsulinemic clamp studies (considered the gold standard for assessing insulin sensitivity). They looked at data from 28 studies including 741 women with PCOS and 1,224 controls. The study found that women with PCOS had 27% lower insulin sensitivity compared to controls, independent of BMI – though elevated BMI further reduced insulin sensitivity by 15% in PCOS women compared to controls.
In the Doh et al study featured in this article’s meta-analysis, women with PCOS also had decreased insulin sensitivity. For clarity, insulin increases the body’s response to certain hormones, leading to higher levels of androgens, and this increased androgen activity is linked to insulin resistance. Women with PCOS show a stronger insulin response to glucose compared to healthy women, regardless of obesity status.
| M value (mg/kg/min) | Obese PCOS (n=6) | Non-obese PCOS (n=8) | Non-obese, non-PCOS (n=10) | p value |
|---|---|---|---|---|
| Unadjusted to lean body | 6.6 [5.5-7.3] | 9.1 [7.7-10] | 11.9 [9.4-14.5] | 0.002 |
| Adjusted to lean body mass | 11.2 [10.1-12.4] | 12.9 [12.1-13.8] | 16.6 [13.8-17.9] | 0.012 |
We also see in Graff et al and Larsson et al that caloric intake and glycemic load is higher. In general, it would appear that decreased insulin sensitivity, even regardless of obesity, is an issue for women with PCOS. However, in the end, that insulin resistance does not appear at this time to be a factor itself in decreased BMR.
Of course there are other metabolic factors at play than insulin resistance for women with PCOS. PCOS is associated with an increased risk of metabolic syndrome, which encompasses much more than just insulin resistance. But even the research on metabolic syndrome suggests people with metabolic syndrome may have slightly higher – not lower – metabolic rates.
When we take a broader view of different issues, it becomes clear that focusing on BMR might not be particularly fruitful when trying to understand why many women with PCOS struggle with weight management. We simply aren’t seeing in other studies that insulin resistance or trending metabolic factors in women with PCOS create the dramatic difference in BMR that the 2009 study shows.
That said, it does appear that insulin sensitivity (among other factors) is a valid problem with PCOS that is possibly present regardless of BMI, which brings me to my closing point.
What’s the goal of an article like this? How does it help?
Hi, this is Leigh. To close this article, I wanted to convey a clear sentiment. I always worry that articles like this might come across as belittling the concerns of those dealing with weight management challenges due to a disorder or syndrome. Not to come across as pulling a card, but I have PCOS — and boy, do I have it all. I put the ‘S’ in syndrome, fitting all three Rotterdam criteria (my ultrasound is a sea of bumpy follicles). My point is that this article aims to do the opposite of dismissing real issues.
I’m a big believer in not suffering the consequences of misplaced priorities. I think there are a lot of minefields in diet management and PCOS, and I think time would be better spent focusing on those issues. For example, while reduced BMR does not seem to be a significant factor, decreased daily physical activity in those with PCOS has shown to be a trend here, here, and here. Women with PCOS also have a higher chance of being affected by mental health issues. And as discussed, reduced insulin sensitivity mixed with a propensity for high glycemic intake is also a repeating occurrence.
All these things could be addressed with better education on these topics versus the distraction of BMR hyperbole. We could discuss the roles that medication and lifestyle play in the treatment of PCOS. Wouldn’t it be great if more attention was paid to mental health, dietary practices, or the benefits of exercise?
Ultimately, the data in this article gives a different level of clarity and understanding of the topic, allowing us to say, “There’s a lot of interesting things going on here, but BMR probably isn’t one of them. What should we focus on next?”
