Table 2. Controls used in DR-TB trials.

Choice of control	Examples	Strengths	Limitations
Placebo, added to optimized background regimen	Delamanid phase II and III trials [28,29] Bedaquiline phase II trial [30,31], Opti-Q [32]	Permits blinding of healthcare providers and participants, yielding unbiased estimates of the efficacy and safety of the individual drug. This design was used to inform regulatory approval of new drugs.	Yields little or no information on how to use the drug in a regimen, which is essential for programmatic implementation; effect of drug can be masked if background regimen is highly effective.
External control (historical or concurrent)	NiX-TB (NCT02333799), ZeNiX-TB (NCT03086486)	Smaller sample size and operational efficiencies due to absence of randomization and use of only one regimen. Considered the only option if there is no accepted standard of care. The justification for use of a historic control can only be used in the first successful trial in that patient population; subsequent trials could use the previous intervention as internal control.	Highly dependent on choice of external control, differences between patient populations and secular trends (with a historical control) affect interpretation of results. Challenging to quantify how much “supportive care” in the trial affected outcomes relative to control outside trial [33].
Randomized comparison in DS-TB, parallel uncontrolled DR-TB cohort with same regimen	STAND (NCT02342886), SimpliciTB (NCT03338621)	Randomized comparison in DS-TB provides strong evidence for safety of regimen in TB patients and efficacy in DS-TB. The parallel DR-TB cohort informs whether results differ between the two TB patient populations.	Only appropriate for regimens that are targeted for both DS- and DR-TB. Extrapolation from DS-TB comparison to DR-TB population requires assumptions.
Local standard of care (varying by site)	STREAM Stage 1 [34], endTB [35]	Better external validity because of randomization to genuine standard of care, operational efficiencies because sites use local standard for control arm participants.	Control regimen may differ by site and over time. This will increase variability in results and may need to be accounted for by increasing sample size.
Prescriptive regimen	NEXT (NCT02454205), STREAM Stage 2 [36]	Better internal validity because of clear randomized comparison of two regimens.	Limited external validity since choice of control regimen may not reflect standard in many places. This would change if a standardized regimen is widely adopted; there are currently variations in how the short regimen is used (for example, choice of fluoroquinolone and bedaquiline in South Africa).

Abbreviations: DR-TB, drug-resistant TB; DS-TB, drug-sensitive TB; TB, tuberculosis.