Yet another “promising” novel antidepressant has failed to actually treat depression.
That’s not an uncommon occurrence these days, but this time, the paper reporting the findings is almost as rubbish as the drug: Translational evaluation of JNJ-18038683, a 5-HT7 receptor antagonist, on REM sleep and in major depressive disorder
So, Pharma giant Janssen invented JNJ-18038683. It’s a selective antagonist at serotonin 5HT-7 receptors, making it pharmacologically rather unusual. They hoped it would work as an antidepressant. It didn’t – in a multicentre randomized controlled trial of 230 depressed people, it had absolutely no benefits over placebo. A popular existing drug, citalopram, failed as well:
About the only thing JNJ-18038683 did do in humans was to reduce the amount of dreaming REM sleep per night. This REM suppressing effect is also seen with other antidepressants and this is evidence that the drug does do something – just not what it’s meant to. Being charitable you could call this a failed trial.
Ouch! But it gets better. Unhappy that JNJ-18038683 bombed, Janssen reached for their copy of the Cherrypicker’s Manifesto. This is a new statistical method, proposed by fellow Pharma company GSK in a 2010 paper, which consists of excluding data from study centres with a very high (or very low) placebo response rate.
Anyway, after applying this “filter” JNJ-18038683 seemed to do a bit better than placebo, but the benefit over placebo still wasn’t statistically significant – with a p value of 0.057, the wrong side of the sacred p=0.05 line (on page 33).
Yet Page 33’s “trend towards statistical significance” magically becomes “significant” – in the Abstract:
[with] a post hoc analyses (sic) using an enrichment window strategy… there was a clinically meaningful and statistically significant difference between JNJ-18038683 and placebo.
Well, no, there wasn’t actually. It was only a trend. Look it up.
That aside, the problem with the whole filter idea is that it could end up biasing your analysis in favour of the drug, leading to misleading results. The original authors warned that “data enrichment is often perceived as a way of improperly introducing a source of bias… In conventional RCTs, to overcome the bias risk, the enrichment strategy should be accounted for and pre-planned in the study protocol.“ They should know, as they invented it, but Janssen rather oddly say the exact opposite: “This methodology cannot be included in a protocol prospectively as it will introduce operational bias in that scheme.”
Anyway, even after the filter technique, citalopram didn’t work either… bad news for citalopram, except, was it citalopram at all? This is really unbelievable: Janssen don’t seem clear on whether they compared their drug to citalopram, or to escitalopram – a quite different drug.
They say “citalopram” in most cases, but they have “escitalopram” instead, in three places, including, mysteriously, in a “hidden” text box in that graph I showed earlier:
I’m not making this up: I stumbled upon a text box which is invisible, but if you select it with the cursor, you find it contains “escitalopram”! I have no idea what the story behind that is, but at best it is seriously sloppy.
Come on Janssen. Raise your game. In the glory days of dodgy antidepressant research, your rivals were (allegedly) concealing data on suicides and brushing whole studies under the carpet, to make their drugs look better. Despicable, but at least it had a certain grandeur to it.
Bonaventure, P., Dugovic, C., Kramer, M., De Boer, P., Singh, J., Wilson, S., Bertelsen, K., Di, J., Shelton, J., Aluisio, L., Dvorak, L., Fraser, I., Lord, B., Nepomuceno, D., Ahnaou, A., Drinkenburg, W., Chai, W., Dvorak, C., Carruthers, N., Sands, S., and Lovenberg, T. (2012). Translational evaluation of JNJ-18038683, a 5-HT7 receptor antagonist, on REM sleep and in major depressive disorder Journal of Pharmacology and Experimental Therapeutics DOI: 10.1124/jpet.112.193995