Langer's insights on the WHI study: unbiased evaluation and behind the scene facts

27 March, 2017:

I urge everyone carefully to read Robert Langer's article in the April issue of Climacteric [1]. Members can download the paper from the IMS website. Langer was one of the chief investigators of the WHI project, both clinician and epidemiologist, who was able to bring not only his unbiased view on the clinical data, but also some important anecdotes related to behind-the-scene administrative issues. Below are the main points that I found most significant, as quoted from the article.

1. 'The WHI set out to test whether (contemporary in 1993) HRT prevents CHD, fractures, and bowel cancer, in women well beyond menopause who are at greater risk of those diseases than the younger women represented in the prior studies. Enrollment was restricted within age groups so that no more than 10% of women would be 50–54 years old, and no more than 20% would be between 55 and 59 years old. Fully 70% of WHI women were to be 60–79 years old.'

2. 'The initial results paper was written by a small group from the coordinating center and program office and submitted to the journal without informing or consulting the clinical site principal investigators. . . . On June 27th, the entire investigator group consisting of the principal investigators for the 40 clinical sites, the coordinating center team, and the NIH program staff gathered for the semi-annual meeting in Chicago. After minor preliminaries, the investigator group was stunned by the announcement that the Data Safety and Monitoring Committee (DSMB) had recommended stopping the CEE + ‏MPA trial and that the Director had accepted their recommendation. Minutes later the group was shocked by the distribution of a typeset copy of the primary results paper soon to be published in JAMA. This was the first time that the vast majority of principal investigators had seen the paper. . . . Concerns were raised about the propriety of producing a paper on behalf of the entire study group in this manner. More importantly, concerns were raised about the tone, the analyses conducted and reported, and the interpretation of the results in the paper. After some discussion, it was agreed that the concerned investigators could quickly provide edits addressing the tone and interpretation. . . . Edits were produced in the brief time remaining before lunch and taken to JAMA. The courier returned shortly with the message that the journal issue had already been printed.'

3. 'The NIH press release announcing the stopping of the study began with the headline "NHLBI Stops Trial of Estrogen Plus Progestin Due to Increased Breast Cancer Risk, Lack of Overall Benefit". The draft press release was distributed to the investigator group after lunch on June 27th, following on the news that the journal was already printed and the paper could not be edited. There was heated discussion about the wording of the press release. But, in the end, the wording favored by the program office prevailed. That headline, pandering to women's greatest fear – the fear of breast cancer – ensured that word of the study would spread like wildfire. And it ensured that the conversation would be driven much more by emotion and politics than by science.'

4. 'Contrary to the usual procedure in clinical trials . . . , no covariate adjusted analyses were reported. . . . The only significant findings in the 'adjusted' results were for a reduction in total fracture and an increase in VTE. The nominal results were significant for benefits in colorectal cancer, hip fracture and total fracture, and significant for adverse outcomes in CHD, stroke and VTE. Even the nominal results were not statistically significant for breast cancer; although the hazard ratio (HR) was 1.26, the 95% confidence interval (CI) (1.0–1.59) included 1.0. The 'adjusted' 95% confidence interval for breast cancer was 0.83–1.92. Nonetheless, incredibly, the paper included the statistically unsupported statement 'The WHI is the first randomized controlled trial to confirm that combined estrogen plus progestin does increase the risk of incident breast cancer and to quantify the degree of risk'.'

5. 'Another critical problem with the CEE‏ + MPA paper was the failure to clearly acknowledge that the WHI was not designed to assess outcomes, particularly the stipulated primary outcome of CHD, in younger menopausal women who were the vast majority of 'real world' patients using HRT. It did not acknowledge that only 30% of participants were < 60 years old, with just 12% aged 50–54. Instead, it inappropriately generalized the findings in a predominantly older population that was not representative of typical users to the population of typical users.' . . . Another inappropriate generalization was 'embedded in the initial outcomes paper, the use of the term 'estrogen plus progestin' in describing the regimen tested in the WHI, rather than the more scientifically correct terminology naming the specific drugs used, CEE + ‏MPA. This quickly led to generalizing the WHI findings, with all the caveats and distortions noted above, to all forms of HRT. Data that existed at the beginning of the WHI in 1993, and that had grown considerably by 2002, demonstrated clearly that that class assumption was inappropriate.'

6. 'Another key and reassuring fact regarding the breast cancer outcomes emerged relatively soon after the initial publications. It has been largely ignored in reporting and interpreting the study. This is the observation that the apparent increase in the breast cancer rate in the CEE‏ – MPA group was due to an unexplained lower rate in the women randomized to placebo who had previously used HRT, NOT an increased rate in women randomized to CEE‏ + MPA. Among women with no prior use of HRT before entering the WHI, there was no difference in breast cancer rates over time between the women assigned to placebo or CEE‏ + MPA. This HRT-naive subgroup likely represents the best population for assessing HRT effects. The breast cancer trend in women with prior use of HRT who were assigned to active CEE‏ + MPA was similar to that in the active and placebo HRT-naive groups. In contrast, the breast cancer rate in women assigned to placebo who had previously used HRT was much lower than the rates in all three other groups. That unexpected and unexplained low rate, different from the rate in the other placebo group, was the basis for the apparent increased hazard.' . . . In fact, the annualized breast cancer rate associated with CEE‏ + MPA in women with prior HRT (0.46%) was essentially equal to the breast cancer rate in control women in the WHI DM [diabetes mellitus] trial (0.45%). . . . The outlier group was the women with prior HRT randomized to placebo in the CEE‏ + MPA trial; the stunningly low annualized rate in that group was 0.25%.

7. 'Also, buried and downplayed in the paper stratifying on prior HRT use is an appropriately adjusted Cox model for breast cancer in the CEE‏ + MPA trial. Specifically, the model was adjusted for age, race/ethnicity, body mass index, physical activity, smoking, alcohol use, parity, oral contraceptives use, family history of breast cancer, family history of fractures, mammography and presence of moderate to severe vasomotor symptoms. The resulting non-significant HR was 1.20 with a 95% CI of 0.94–1.53. In the text, this key finding follows the marginally significant unadjusted result (HR 1.24;95% CI 1.02–1.50), but is said to "not substantially alter this [unadjusted] risk estimate".'

8. 'The CEE-alone trial testing CEE 0.625 mg daily versus placebo in women with prior hysterectomy was also stopped early. The decision was driven by an increase in stroke. However, the number of strokes in women 50–59 years old at enrollment was identical in the active and placebo groups, so that association was found only in women at or above 60 years old. There were striking contrasts between the results with CEE-alone and the results with CEE‏ + MPA. There was a trend toward reduced breast cancer that was on the cusp of statistical significance with a HR of 0.77 and 95% CI of 0.59–1.01. There was also a trend toward reduced CHD (HR 0.91; 95% CI 0.75–1.12). Subsequent papers showed that CEE-alone was associated with statistically significant reduced rates in three key outcomes: breast cancer in adherent women; CHD in women aged 50–59 when revascularization was included; and a lower degree of coronary artery calcium in women aged 50–59.'

9. 'Where does that leave us in 2016? It is time to get past the misinformation and hysteria generated by the highly irregular circumstances of the WHI and stop denying potential benefits (control of vasomotor symptoms, prevention of fractures, prevention of CHD) to women who have indications and may be helped. HRT is appropriate for symptomatic women within 10 years of menopause who have no major contraindication. Good evidence from over 50 years of observational studies and clinical trials suggests that the benefits outweigh the risks for most women when started early. The International Menopause Society has recently published updated recommendations for HRT in a new format that highlights key messages and clinical pearls. It is a well documented and authoritative guide for contemporary clinical practice.'

Amos Pines
Sackler School of Medicine, Tel-Aviv University, Tel-Aviv, Israel