Thanks Paul!

Whoops, thanks for catching that! You're right. Really wish BlueSky had an edit function now...

Thanks a lot Jonathan!!

10/ Our work provides an illustrative example of how statistical methods can obtain adjusted effect estimates in the setting of non-adherence. Investigators should consider the potential impact of protocol non-adherence on their trial results and conduct relevant analyses specific to their context.

9/ Our analyses show that non-adherence did not affect the internal validity of BALANCE's results, and that 7-day antibiotic therapy should be the standard of care for most patients with non-S. aureus BSI.

8/ We applied inverse probability of weighting (IPW) and instrumental variable (IV) approaches to account for the potential bias introduced by non-adherence. All analyses favored the 14-day group, and met the 4% NI margin for a conclusion of non-inferiority consistent with primary BALANCE results.

7/ Meanwhile, vascular catheter source and presence of antimicrobial resistance were associated with higher odds of treatment shortening in the 14-day arm. This latter finding was surprising; and may be due to the lack of available oral treatment options and preference to limit broad-spectrum abx.

6/ Patients receiving protocol-adherent treatment durations were significantly different from those who received non-adherent durations; multivariable models showed that higher disease severity, persistent fever/bacteremia, and lower age <70y were associated with higher odds of prolonged treatment.

5/ This histogram shows the wide spread of actual antibiotic durations received in both treatment arms, and illustrates the clear pattern of treatment crossover between arms. Median duration was 8 days (IQR 7-11) in the 7-day arm and 14 days (IQR 14-15) in the 14-day arm.

4/ In the BALANCE trial, 432/1802 (24%) of patients in the 7-day arm and 296/1779 (16.6%) of patients in the 14-day arm received treatment durations that were protocol non-adherent (>2 days longer or shorter than the assigned treatment duration).

3/ However, intention-to-treat and per-protocol approaches are both not ideal and do not adequately address this bias. There is no good consensus in guidelines or reporting standards on how best to address this non-adherence; though a variety of statistical techniques have been described.

2/ Non-adherence, especially with the treatment crossover pattern, is a threat to internal validity in non-inferiority trials, since it biases results towards no difference between groups, increasing the probability of a false conclusion of non-inferiority.

I can't say enough how incredible my experience with this joint PhD program has been. The coursework at @ihpmeuoft.bsky.social is truly world-class, and this is also a great opportunity to work with some amazing clinical trialists and build an international collaborative network.

"Addressing these challenges requires rethinking funding models, fostering equitable collaborations, and strengthening LMIC research leadership, trial capacity and infrastructure —not just as a matter of justice and equity, but as a necessity for global health security."

"Reduction in funding is likely to lead to reduced trial capacity, weakened surveillance, and delayed access to vaccines and therapeutics, and ultimately poorer global health outcomes."

High-income countries fund a large majority of clinical trials in ID, with the NIH in particular funding 20% of trials in our review. Current changes in funding (especially the de-prioritisation of global health amongst some funders) has major implications on ID clinical trials worldwide.