Posted on


Few topics in medicine are filled with as much misinformation and misunderstanding as hormone replacement therapy in women. Millions of women each year make important decisions about using hormones based upon poor and/or nonfactual counseling from physicians or false claims by a huge industry that has risen up to victimize these patients.

In July of 2002, data from the estrogen plus progestin (E+P) arm of the Women’s Health Initiative Study (WHI) were published in JAMA. The WHI was begun in 1991 and focused on a variety of issues which affected women’s health, including hormone therapy, dietary modifications, and calcium plus vitamin D usage. Several large papers were eventually published, and more than 160,000 postmenopausal women were enrolled in some part of the study. Most people forget about the huge amount of non-hormone related data collected in WHI. Briefly, the WHI found that dietary modification did NOT reduce the risks for coronary heart disease (CHD), stroke, or cardiovascular disease (CVD), nor did it reduce the risk of colorectal or breast cancers. Similarly, the WHI found that calcium plus vitamin D did NOT reduce the risk of hip fracture or colorectal cancer. Interestingly, these findings did not cause widespread abandonment of the recommendations that women modify their diets or supplement with calcium and Vitamin D. But hormones were a different story.

Prior to 2002, estrogen replacement was almost universally recommended to menopausal women to reduce the risk of CHD. It was believed, based largely on observational data, that estrogen reduced the risk of CHD (since the rates of CHD for premenopausal women were well below age-matched males but after menopause women caught up quickly). The study was designed to study this hypothesis, and women with a uterus were randomized either to placebo or to receive estrogen plus progestin (E+P) in the form of the PremPro, and women without a uterus were randomized to receive estrogen alone (E only) in the form of Premarin, or placebo. The progestin was necessary only to protect the uterus from the risk of endometrial cancer associated with unopposed estrogen.

Women were to be followed for 9 years after enrollment. However, the E+P group was ended early, after 5.2 years, because it found that women in the treatment group had an increased risk of CHD, stroke, pulmonary embolism (PE), and breast cancer. The E only group continued until 2004. The results of the E+P arm were the only data published in July, 2002, but the results were sensationalized. Women felt betrayed because the study found that not only did E+P not prevent CHD, it made it worse, and it contributed to increased risks of stroke, PE and breast cancer. Women in the millions stopped their hormone replacement therapy, including women who were taking estrogen only (since they did not have a uterus). The results were lumped together and the media and most physicians blamed estrogen as the bad player, especially for breast cancer, which was dramatically over-played by the media. The fallout was tremendous, not just because millions of women stopped using their hormones, but also because women no longer trusted their physicians’ advice about hormones. The new experts were Oprah Winfrey, Suzanne Somers, and scores of snake-oil salesmen and women who sought to fill the hormone void with new products and scams. The book The Truth About Women’s Healthcare: How and Women are Lied to By The Medical Establishment (2004) made absurd claims, like 42,000 women per year were getting breast cancer from hormones, and the media and women unfortunately listened.

But what did this 2002 WHI paper actually state?

  • A 1.29 hazard ratio (HR) for CHD for women on E+P (an excess of 7 per 10,000);
  • A 1.26 HR for breast cancer (an excess of 8 per 10,000);
  • A 1.41 HR for stroke (an excess of 8 per 10,000);
  • A 2.13 HR for PE (an excess of 8 per 10,000);
  • A 0.63 HR for colorectal cancer (6 less women per 10,000);
  • A 0.83 HR for endometrial cancer (1 less women per 10,000);
  • A 0.66 HR for hip fracture (5 less women per 10,000);
  • A 0.92 HR for death due to all other causes;
  • Overall, a HR of 0.98 for composite death. In other words, total causes of death were no different (and insignificantly reduced) for women who took E+P.

Coronary Heart Disease

The results of the study were only that E+P should not be initiated or continued for primary prevention of CHD, which had been the standard of care up until that point. But the study did not indicate that women who wanted E+P for menopausal symptoms should not take it; the risk of mortality was the same with and without it. While 8 more cases of breast cancer occurred, there were 6 fewer cases of colorectal cancer. Around 16% of women who are diagnosed from breast cancer die from it, while about 37% of women who are diagnosed with colorectal cancer will die from it. So from a cancer perspective, women were better off staying on E+P. But the media and physicians focussed only on the increased risk of breast cancer.

There are few more problems with this initial WHI study and the conclusions drawn from it. Much has to do with confusion of absolute risk and relative risk. For example, the HR for CHD in the E+P group was 1.29; and while a woman was 29% more likely to have CHD, the absolute risk was very low, just 7 cases per 10,000 women, or 0.07% increased absolute risk. Stated this way, the increased risk seems negligible or, stated another way, not clinically significant.

Clark, in his excellent analysis of the data, has also pointed out that the increased risk of CHD is not statistically significant when adjusted confidence intervals are used. Even though the authors of the WHI paper acknowledged that this was the appropriate statistical method, the data they reported were unadjusted confidence intervals. Why the switch? Perhaps the desire to have something that appears statistically significant? The authors themselves in a 2003 paper admitted that the data were not statically significant. So the risk of CHD with E + P has been overplayed and has not been established.

Aside from these statistical issues, evidence has emerged since the WHI study was published to suggest that the CHD risk (if it even exists) may only be relevant to older women, not women just entering menopause. Women in the WHI study were about 12 years removed from menopause (average age of 63), and many of them may have developed occult coronary disease in that time. Consider these newer studies:

  • Hulley et al., in 1998, found no difference in cardiovascular disease among some who took E + P compared to women who did not, among 2,763 women enrolled in the HERS study.
  • Grodstein et al., in 2001, reported on the outcomes of the Nurses’ Health Study, and found that E+P use or E use alone was associated with a possible slight increase in cardiovascular events at initiation but over the long term a reduction in total events, with a relative risk of 0.65. The short-term increase in events were likely due to preexisting coronary disease.
  • The E only arm of the WHI (published in 2004) found no increased risk of CHD.
  • Schierbeck et al., in 2012, reported on a trial of 1006 recently menopausal Danish women who were randomized to receive either E + P or placebo. They found that after ten years, the women who received E + P had a reduced risk of MI, heart failure, and mortality, without an increase in cancer, stroke, or thromboembolism. This effect was still present after 16 years of usage.
  • Hodis et al., in 2016, published results of the ELITE Trial, which found that estrogen therapy initiated within 6 years of menopause was associated with delay in progression of atherosclerotic heart disease (but not after ten years), supporting the idea that estrogen is beneficial if started at menopause, but not if started years later, as in the WHI trial.

So the available evidence today would indicate that women with pre-existing heart disease should not start HRT, but otherwise, E + P therapy does not increase and likely decreases their risk of death related to cardiovascular events.

Breast Cancer

Perhaps the most damaging conclusion from the 2002 WHI paper was that E + P increased the risk of breast cancer. This is why women actually stopped the medication and it was this fact that was sensationalized in the media. Today, it seems most women and most physicians believe that estrogen increases the risk of breast cancer, even though this could not be concluded from the E + P arm of the study. I personally believe that physician familiarity with the estrogen-blocking drug Tamoxifen biased doctors into (falsely) concluding that estrogen usage was associated with an increased risk of breast cancer. However, the E  only arm of the WHI wasn’t published until 2004. Anderson et al. found:

  • A 0.91 hazard ratio (HR) for CHD for women on E alone;
  • A 0.77 HR for breast cancer;
  • A 1.39 HR for stroke;
  • A 1.34 HR for PE;
  • A 1.08 HR for colorectal cancer;
  • A 0.61 HR for hip fracture;
  • There was no overall difference in mortality.

This (at the time) surprising finding for breast cancer risk was present in an extended follow-up of the same patients. In 2012, Anderson et al. reported that the reduced risk of breast cancer benefit persisted, and in fact, not only were there less invasive breast cancers among women who took E only (HR of 0.77) but women with breast cancer were less likely to die from it. Death from breast cancer was therefore significantly less likely in the estrogen users (HR of 0.37). It is incredible that the media (and many physicians) continue to perpetuate the estrogen-breast cancer link myth.

So was it the progesterone that increased women’s risk of breast cancer in the original WHI paper? No. The observed 26% relative risk increase in breast cancer among women who used E + P was not statistically significant using adjusted confidence intervals. Clark, again, has an excellent explanation of this. It has widely been pointed out that detectable breast cancers have a ten year or longer lead-time. If anything, there was an initial increased risk of detection of breast cancer in the first two years of progesterone usage (knowing that detected tumor was likely present for almost a decade before starting progesterone). This increased risk of detection with E + P use (8 per 10,000 women, assuming the numbers are even significant) was likely driven by increased breast complaints among women taking hormones known to cause breast tenderness for the first time in 12+ years. If anything, it might be argued that E + P led to earlier detection. The hazard ratios were revised downward by the WHI team in a 2003 paper.

So it is beyond absurd to argue that E replacement is associated with increased risk of breast cancer, nor is it valid to conclude that E + P causes invasive breast cancer, though it may aid in its earlier detection.

Thromboembolism and Stroke

I would encourage you to read Clark’s paper for more information about the WHI findings regarding thromboembolism and stroke, but he shows that these data too are not statistically significant. This meshes with what we already learned from the Danish study, which found reduced rates of cardiac disease without an increased risk of thromboembolism or stroke.

So what can we conclude?

Clark’s analysis is broad and cogent and includes references to others who have performed detailed, peer-reviewed take-downs of the WHI, but he concludes that the WHI authors should have concluded the following: “no significant risks were found for cardiovascular disease, invasive breast cancer, stroke and venous thromboembolism.” Wow. But this concluson, which was available to anyone in 2002 who did a fair analysis of WHI results, has been validated by subsequent studies. Of course, there were improved outcomes relating to risk of colorectal cancer, hip fracture, and total mortality, as well as better quality of life.

But the media convinced millions of women to abandon their hormones, opened the door to the quasi-scientific quackery of bioidentical hormone replacement, and ultimately harmed women. Half of the women in the US who used hormones stopped in the 18 months following the 2002 WHI publication. Sarrel et al., in 2013, reported that only 1/3 of post-hysterectomy patients today use estrogen replacement therapy. Based on our current understanding of the mortality benefit of estrogen-alone use (a decreased risk of mortality of 13 / 10,000), they estimate that between 18,601 and 91,610 women have died prematurely due to estrogen-avoidance in the ten years following 2002, spurred on by fear generated by the media and ignorance among doctors. This just among women aged 50-59! This is another excellent paper for the unconvinced, and can be read free of charge.

I have no idea how many women have died prematurely due to poor understanding of the evidence-based medicine by the media and physicians, but I don’t doubt that the number is in the many tens of thousands. Ignorance of evidence-based medicine and the cognitive tools necessary to interpret available scientific evidence (Bayesian methods in particular) is a public health crisis. Cognitive biases which lead us to reach premature conclusions and draw inappropriate inferences are certainly one of the most serious issues facing patients today. Physicians need to be at the fore of educating the public about these issues, but, unfortunately, they lag far behind. Shameless non-scientific garbage like this from Oprah is usually willingly supported by physicians eager to make a buck.

It is time to move well beyond the WHI’s initial paper as a source of our understanding of the risks and benefits of hormone replacement therapy.

A quick word about “Bioidentical hormone replacement” therapy (BHRT). Those making money off of  BHRT make many claims, including that,

  • BHRT is safer or more effective than normal prescription hormones;
  • BHRT prevents or cures heart disease, Alzheimers, or breast cancer;
  • Estriol is safer than estradiol;
  • Hormone therapy should be customized based on hormones levels, often from saliva testing;
  • FDA-approved hormones cause cancer, strokes, blood clots, etc.;
  • Charlatans like Suzanne Somers claim almost every imaginable benefit from taking hormones, treating many more symptoms than just hot flashes.  (I should note that Somers continues to take estrogen after having an Estrogen-receptor positive breast cancer. She also declined Tamoxifen. Then she had a hysterectomy secondary to complex endometrial hyperplasia likely secondary to inadequate endometrial protection from her “natural” progesterone.)
  • BHRT prescribers often give progesterone to women who don’t have a uterus and frequently add testosterone.

These and other myths are answered here by the FDA and here by Harvard Medical School. The Wikipedia article on bioidentical hormones is quite excellent and analyzes many false claims in some detail. Some proponents of BHRT believe the word bioidentical refers to the types of hormones used, like estradiol instead of conjugated equine estrogens. However, most FDA-approved estrogen products used by physicians contain the very same estradiol. The other, more pervasive definition of BHRT is that the hormones are prescribed to women who are deficient based on hormone level testing, and the doses of the hormones are tailored to make the hormone levels normal. I will make just a few points:

  • Salivary testing of hormones has never been shown to accurately reflect serum levels of hormones. What’s more, serum levels of hormones should not be used to guide replacement.
  • Sex hormones don’t have defined levels. The amounts of estrogen, progesterone, and testosterone in a woman’s body vary by the time of life, time of the month, and time of day. Progesterone levels, for example, are less than 1 ng/ml, but soar above 12 one week after ovulation. Which level is normal? Estrogen levels vary between 10 and 123 pg/ml throughout the month. What level is normal? And the ovaries have different functions in the young girl, reproductive aged woman, and menopausal aged woman. Which time of life should we imitate? Sex hormones are not like thyroid levels, which have a “normal range.” So replacing sex hormones to match a “normal range” is nonsensical.
  • Testosterone is potentially dangerous. It may cause irreversible androgenization of women if used inappropriately. We do not have large studies about the potential increased risk of cardiovascular disease, etc. The same providers who misuse the WHI study results to scare women into trying BHRT will often freely give women testosterone, which may make them feel “peppier” but come with largely unknown and potentially considerable risks. Testosterone replacement therapy has a very narrow indication for treatment of some sexual disorders in postmenopausal women.
  • Progesterone is given to women only to reduce the risk of endometrial cancer, that is, it should only be given to women with uteruses. There is no evidence of benefit otherwise, and, as we have seen, there are less risks of taking estrogen alone compared to estrogen and progesterone.
  • Any risk that might be attributable to FDA-approved estrogen and progesterone products can also be attributed to BHRT products; likewise, any benefit that might be attributable to BHRT products can also be attributed to FDA-approved products.
  • The practice of BHRT is considered a form of alternative medicine, just like homeopathy or magnetic therapy. Billions of dollars are being made in this industry, promoted by folks like Suzanne Somers, Oprah Winfrey, and Dr. Phil’s wife.

And one last trivia item. The first estrogen product marketed for treatment of menopausal symptoms was an orally active estrogen extracted from the urine of pregnant women by James Collip, who famously perfected Banting and Best’s method of extracting insulin (the three men shared the patent for insulin). Collip called his estrogen product Emmenin, which was widely used until the emergence of DES and later Premarin. I wonder if Marshall Mathers realizes this?