Servier Laboratories Ltd v National Institute for Health and Clinical Excellence & Anor

A public decision-maker may reject post-hoc subgroup evidence, but it must explain why the particular analysis is unreliable. Generic weaknesses in post-hoc analysis are insufficient where such evidence may be reliable in an individual case. The reasons must address the material explanation advanced for the subgroup’s selection and any stated objection to its reliability. A decision-maker is not bound by a different conclusion reached by an equally authoritative expert body, but divergence reinforces the need for proper reasoning. Where the evidence is central to the decision, failure to give intelligible, study-specific reasons is unlawful.

NICE appraised drugs for preventing osteoporotic fractures in post-menopausal women. It recommended alendronate as the treatment of choice and limited the use of strontium ranelate. Servier sought judicial review, challenging NICE’s rejection of a post-hoc subgroup analysis from the TROPOS randomised controlled trial, which indicated reduced hip fractures.

Holman J granted review on one ground but rejected Servier’s challenge concerning the subgroup analysis. Servier appealed, arguing that NICE had failed to give adequate reasons and had acted irrationally, particularly because the European Medicines Agency had accepted the analysis.

Appeal allowed. The Court of Appeal unanimously quashed NICE’s decision concerning the assessment of strontium ranelate and required a fresh decision.

1. Adequacy of reasons. Smith LJ, whose reasoning was adopted by Wilson LJ and Pill LJ, held that the post-hoc subgroup analysis was central to Servier’s case. Such evidence may be weaker than evidence from a pre-specified randomised controlled trial, but it is not necessarily unreliable. NICE therefore had to explain why this particular analysis was rejected.
2. A generic reference to the inherent weaknesses of post-hoc subgroup evidence did not suffice. NICE neither adopted ScHARR’s stated concern about randomisation nor addressed Servier’s detailed rebuttal. The later evidence relied upon in the proceedings explained only why post-hoc analyses may be unreliable generally, not why this analysis was unreliable.
3. NICE was not bound by the European Medicines Agency’s assessment, since the bodies performed different functions. However, a decision that departed from the reasoned conclusion of an equally eminent body required proper explanation, particularly where the evidence was central to the decision.
4. Rationality. Smith LJ treated the following observations as obiter. It might be rational to reject the analysis for data mining, inadequate randomisation or biological implausibility, but the decision-maker would need to consider the explanations and information advanced on those issues. The NICE Methods Guide did not impose a blanket prohibition on subgroup analysis.
5. It might also be rational to prefer a statistically non-significant result from a high-quality randomised controlled trial to a statistically significant result from a fundamentally unreliable study. That conclusion depended on NICE first being entitled, for adequately explained reasons, to regard the subgroup analysis as unreliable.

• Court of Appeal (Civil Division) [2010] EWCA Civ 346: allowed the appeal and quashed NICE’s decision concerning the post-hoc subgroup analysis.
• Administrative Court, Queen’s Bench Division: Holman J, by order dated 19 February 2009, granted judicial review on one ground but rejected the challenge to NICE’s treatment of the subgroup analysis.