Mainstream media continues to act as the marketing arm of Big Pharma

Why bite the hand that feeds you?

Many, many people I know are on ozempik (or similar medication), or otherwise talking about trying it.

To me, the rapid rise of this product to blockbuster status epitomises everything that is wrong about our relationship to health, and Big Pharma in particular.

In this article - which I wrote for

HART’s Substack

a few months ago - I outline how:

• A pivotal trial (for ozempik) published in NEJM fails to report all-cause effects.

• It uses a composite endpoint of fatal and non-fatal events without giving a breakdown.

• It suggests that around 8% of subjects withdraw due to side effects.

• It fails to describe the benefits - such as they are - in a meaningful way, twisting a “number needed to treat” of ~66 (for the 4 years of the study) into a 20% reduction¹.

I then go on to show that the trial data is not being made publicly available, and some of the key medical promotors of these products are in receipt of cash from the manufacturers.

Mainstream media continues to act as the marketing arm of Big Pharma

UPDATE 10 Oct 2024:

Re-reading the above on HART’s substack, I came across an interesting comment by

Satan's Doorknob

which mentioned his own article on this study.

In that, he points out that he’d found the full version of the study mentioned above - here, and in that they DO cite all-cause mortality data. He expresses surprise that an apparent 19% reduction in ACM (termed “death from any cause” in the NEJM paper) was not being shouted from the rooftops.

I have had a look at the full paper. He is right that the ACM suggests a 19% reduction (expressed as a hazard ratio of 0.81):

The first point to make about this is that even if the data is taken at its highest with no suspicions of jiggery-pokery (for more on that see below) is that the absolute risk reduction is 0.9%, meaning that over 100 people would need to be treated (for the 4 years of the trial) to save one life.

Given that death is a “hard” enpoint, this might be regarded as a still rather worthwhile benefit.

However, it must be noted that this result is NOT actually reported as statistically significant. “NA” appears in the right hand column and the relevant footnote states that:

‡ Confirmatory secondary end points were analyzed under multiplicity control through a stagewise hierarchical testing scheme in which all P values after the first nonsignificant P value are not reported. The P values (unadjusted) for the primary and confirmatory secondary end points were to be compared with the nominal significance level derived from the relevant alpha spending function for the end point; if the P value was below the nominal limit, superiority would be shown. The nominal two-sided significance level was 0.023 for death from car‑ diovascular causes.

So, given that the reduction in death from cardiovascular causes did not reach statistical significance (p=0.07), that meant that all P values after that are not reported.

I must admit that I do not fully understand what the implications of this are to the usefulness of the ACM datapoint, but it’s likely that the reason that the ACM reduction has not been widely reported is because it cannot be touted as one which is statistically significant.

However, I don’t think that’s the end of the story with this.

Take a look at the graph for deaths from any cause:

This is a weird graph. Unless manipulated in some way, with such large treatment groups mortality really should rise gradually in a straight or straightish line, gradually diverging (if there’s a real difference between the groups).

Here, the lines are not straight at all, the most striking anomaly being a marked and non-natural looking flattening of mortality in the semaglutide group at 3 years.

The footnote to this gives some further insights into the issue discussed above (the statistical significance of the ACM reduction reported):

Because the between-group difference in death from cardiovascular causes did not meet the required P value for hierarchical testing, results for the two subsequent end points in the testing hierarchy are reported as point estimates and 95% confidence intervals. The widths of these confidence intervals have not been adjusted for multiplicity and therefore should not be used to infer definitive treatment effects for these secondary end points.

So in summary:

• They aren’t shouting about ACM reductions as they weren’t statistically significant.

• The reductions for deaths from cardiovascular deaths did not reach statistical significance either.

• There are some non-linear elements to the mortality curves which are suspicious and unexplained.

1. technically true - 8% down to 6.5% - but misleading as it’s a flavour of relative risk reduction with no relation to the absolute risk reduction.