Discover what's new
in nQuery

nQuery 9.7 Release Notes

nQuery Pro Tier

What's new in the PRO tier of nQuery 9.7?

1 new sample size + two major feature updates have been added to the Pro tier in nQuery 9.7

• Multiple Comparisons for Continuous Endpoint (1 Table)
• 3 Additional Custom Boundary Input Scales for Group Sequential
  
• Additional and Improved Plots for Survival Group Sequential

1. Multiple Comparisons

What is it?

Multi-arm trials which evaluate multiple treatment arms or doses simultaneously within a single protocol are common across multiple stages of clinical research. By avoiding the duplication of control arms, designs which include multiple arms can reduce the cost of evaluating multiple candidate treatment options significantly. However, appropriate statistical methods are required to ensure that Type I error is controlled when evaluating multiple treatments or doses. One class of statistical methods used for this purpose is Multiple Comparisons Procedures.

Multiple Comparisons Procedures share the common objective of controlling the Type I error when comparing multiple arms. However, a wide range of potential multiple comparison procedures are available depending on trial design and analysis aspects such as which comparisons will be made (e.g. all arms vs all arms, control arm vs treatment arms only), definition of error control (e.g. Family-Wise Error Rate (FWER), False Discovery Rate (FDR)) or arm testing sequence (e.g. simultaneous, step-up/step-down algorithms).

nQuery’s new multiple comparisons table features 14 multiple comparison procedures across five main types:

• Single-step (Bonferroni, Weighted Bonferroni, Sidak, Dunnetts, Unadjusted),

• Step-up (Dunnetts, Hochberg),

• Step-down (Dunnetts, Holm-Bonferroni, Holm-Sidak),

• Stepwise (Fixed Sequence, Fallback)

• FDR-based (Benjamini-Hochberg, FDR Benjamini-Yekutieli).
  
  

These methods can be used to find the sample size for three definitions of power for a multiple comparisons study: conjunctive power (all arms significant), disjunctive power (at least one arm significant) and marginal power (specific arm significant).

nQuery also contains seven algorithms for sample size allocation across the arms: equal allocation across all arms, equal allocation for treatment arms only (with shared sample size ratio to control arm), root K (sample size ratio in each treatment arm equals square root of number of treatment arms), three optimality criteria which minimize variance (A, D, E Optimality available) and custom (specify separate sample size ratio per treatment arm)

nQuery 9.7 adds a multiple comparisons table for a continuous endpoint using analytic formulae. Future updates will extend to additional endpoints (binary, count, survival) and simulation-based methods.


 Tables added: 

• Multiple Comparisons Procedure - Multiple Arms Single Stage Non-inferiority /Supersuperiority Tests for Means

Multiple Comparisons Summary:

• 14 Multiple Comparisons Procedures:

  • Unadjusted, Bonferroni, Weighted Bonferroni, Sidak, Single-Step Dunnett, Step-Up Dunnett, Step-Up Hochberg, Step-Down Dunnett, Step-Down Holm-Bonferroni, Step-Down Holm-Sidak, Stepwise Fixed Sequence Testing, Stepwise Fallback Procedures, FDR Benjamini-Hochberg, FDR Benjamini-Yekutieli

• 3 Target Power Definitions:

  • Conjunctive, Disjunctive, Marginal

• 7 Sample Size Allocation Methods:

  • Equal Across All Arms, Equal Across Treatment Arms, Root K, A-Optimality, D-Optimality, E-Optimality, Custom

2. Group Sequential Design Update: Additional Custom Boundaries

What is it?

Group Sequential Design is the most common adaptive design used in confirmatory clinical trials. This design allows trialists to stop a trial early at pre-specified interim analyses if there is sufficient evidence that treatment is effective (efficacy) or ineffective (futility). Group sequential designs capacity to stop trials early can lead to significant cost savings while also getting vital treatments into the hands of patients faster.

Over recent releases, nQuery has added an overhauled set of group sequential tables which provide access to a wide suite of boundary methods plus reporting and plotting options. Based on industry best practice and user feedback, we continue to improve our group sequential offering while also adding group sequential tables for additional design scenarios.

nQuery offers unparalleled ability to specify a custom group sequential design by providing the Custom Boundary method which allows users to enter their own custom Efficacy and/or Futility boundaries on a variety of parameterization scales including Z-statistic, p-value, Score Statistic, Effect Size and Conditional Power (futility only).

In nQuery 9.7, the ability to calculate Conditional Power for a custom “true” effect size assumption was added, in addition to the following three new custom boundary scales:

• B-value

• Predictive Power (futility only)

• Reverse Conditional Power (futility only)

The B-value is an independent increments statistic for monitoring (equal to the Z-statistic times the square root of information time) which maps to Brownian motion. The B-value simplifies conditional power calculations and has more stable behavior than the Z-statistic at early looks.

In addition, we also extend our custom conditional power boundary option where in addition to specifying the true effect size under the specified effect size (used for power calculation) or estimate effect size (at futility boundary), the user can now set a custom value for the true effect size directly.

Predictive Power is an extension of conditional power where instead of assuming a point estimate for the assumed “true” effect size, the conditional power function is averaged over the posterior distribution at an interim analysis, providing a more balanced assessment of the expected power. nQuery offers the option to calculate predictive power under an uninformative uniform prior or a user specified normal prior.

Reverse conditional power (also known as inverse conditional power) is an extension of conditional power which shifts from “expected power given current interim data” to “if trial will succeed, how likely is current interim data”. A major advantage is that reverse conditional power removes the requirement for an assumed “true” effect size by conditioning on the final known alpha threshold instead.

Features Added:

• 3 Additional Custom Boundary Input Parameterizations:
  • B-values, Predictive Power (Futility Only - Uniform or Normal Prior), Reverse Conditional Power (Futility Only)
• Custom Effect Size option for Conditional Power Custom Boundary
  

3. Group Sequential Design: Additional Survival Plots

What is it?

nQuery currently offers plots for the group sequential boundaries and error spending (i.e. exit probability) function across all of our group sequential tables. However, survival analysis is a more complex endpoint (due to having to model expected events, sample size and dropout) and therefore plots specifically for survival analysis can provide additional insights on important outcomes such as interim analysis timing.

In nQuery 9.7, we add two additional plots specifically for our survival analysis group sequential tables:

• Sample Size/Events vs Time

• Survival % vs Follow-up Time

These two plots are also added to the fixed term survival analysis table:
STT37 - Two Sample Log-Rank Test (Event Driven or Fixed Follow-up, Piecewise Accrual %, Hazard Rates + Dropout Rates).

Sample Size/Events vs Time plots the total sample size, total events and pre-group events expected under the alternative hypothesis at a given time after study start. This plot provides a practical insight into the expected trajectory of your trial and where you can expect its status to be at the interim and final analyses.

Survival % vs Follow-up Time plots the percentage of subjects expected to survive under the specified hazard rates after a given length of follow-up time after subject entry. This plot provides an insight into the survival pattern expected in your trial which is particularly valuable for more complex piecewise hazard rate patterns.

Tables Added:

• Sample Size/Events vs Time Plots for Survival Tables (GST3, GST5, STT37)
  
• Survival % vs Follow-up Time Plots for Survival Tables (GST3, GST5, STT37)
  

How to update?

If you have a subscription for the nQuery Pro Tier nQuery should automatically prompt you to update. You can manually update nQuery by clicking Help>Check for updates. 

To discuss any aspect of your subscription, click here.