Resources

In a pharmaceutical bioequivalence study, an oral generic antibiotic is compared to a proprietary brand-name formulation. Let μgeneric\mu_{\text{generic}} and μbrand\mu_{\text{brand}} denote the true population mean peak plasma concentrations (in μg/mL\mu\text{g/mL}). Regulatory guidelines establish a non-inferiority threshold boundary of δ=2.50μg/mL\delta = 2.50\mu\text{g/mL}, specifying that the generic drug is considered non-inferior only if its mean concentration is not lower than the brand mean by more than 2.50μg/mL2.50\mu\text{g/mL} (i.e., μgeneric−μbrand≥−2.50μg/mL\mu_{\text{generic}} - \mu_{\text{brand}} \ge -2.50\mu\text{g/mL}). (a) Formulate the null and alternative hypotheses symbolically for this non-inferiority boundary test. (b) Explain why the non-inferiority claim must be placed in the alternative hypothesis rather than the null hypothesis.

Answer and steps

Final answer

(a) H0:μgeneric−μbrand≤−2.50μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le -2.50\mu\text{g/mL} versus Ha:μgeneric−μbrand>−2.50μg/mLH_a: \mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\mu\text{g/mL}. (b) In standard statistical inference, the burden of proof rests on the research claim. Placing inferiority in H0H_0 ensures that non-inferiority is only concluded if there is statistically compelling sample evidence to reject inferiority, thereby protecting patients against approving an ineffective formulation (controlling Type I error).

Steps

Part (a): Formulating the Hypotheses

  1. Identify the parameter of interest:
    The parameter being evaluated is the difference in population means: Δ=μgeneric−μbrand\Delta = \mu_{\text{generic}} - \mu_{\text{brand}}
  2. Define the research objective (Non-inferiority):
    The non-inferiority criterion states that the generic antibiotic is acceptable if its mean concentration does not drop below the brand mean by more than 2.50 μg/mL2.50\ \mu\text{g/mL}, which corresponds to: μgeneric−μbrand≥−2.50 μg/mL\mu_{\text{generic}} - \mu_{\text{brand}} \ge -2.50\ \mu\text{g/mL} In standard hypothesis testing, the substantive research claim or condition being argued for (requiring strong evidence) is set as the alternative hypothesis (HaH_a). Thus, the strict inequality defining the region of demonstrated non-inferiority is: Ha:μgeneric−μbrand>−2.50 μg/mLH_a: \mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\ \mu\text{g/mL}
  3. Establish the complementary null hypothesis (H0H_0):
    The null hypothesis represents the opposing, conservative condition—namely, that the generic drug is unacceptably inferior: H0:μgeneric−μbrand≤−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le -2.50\ \mu\text{g/mL}

    (Note: Setting H0:μgeneric−μbrand=−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} = -2.50\ \mu\text{g/mL} is also commonly used at the testing boundary, but the directional composite form H0≤−2.50 μg/mLH_0 \le -2.50\ \mu\text{g/mL} fully covers the region of inferiority).


Part (b): Rationale for Hypothesis Placement

  1. Burden of Proof:
    In classical statistical hypothesis testing, the null hypothesis H0H_0 is assumed to be true until the data provide overwhelming evidence against it. The "burden of proof" falls on rejecting H0H_0 in favor of HaH_a.
  2. Controlling the Most Critical Error (Type I Error):

    • A Type I error occurs when a true null hypothesis is incorrectly rejected.
    • By setting H0H_0 as "inferior" (μgeneric−μbrand≤−2.50\mu_{\text{generic}} - \mu_{\text{brand}} \le -2.50) and HaH_a as "non-inferior", a Type I error corresponds to erroneously concluding that an inferior or substandard drug is non-inferior.
    • Hypothesis testing rigorously bounds the probability of a Type I error at a predefined significance level α\alpha (typically 0.050.05 or 0.0250.025). This safeguard prevents ineffective generic formulations from entering the market, ensuring patient safety.
    • If the roles were reversed (placing non-inferiority in H0H_0), an underpowered study with high noise or a small sample size would fail to reject H0H_0, improperly granting regulatory approval due to a mere lack of data rather than genuine bioequivalence.

Common Pitfall

  • Assuming H0H_0 always represents "no difference" (μ1−μ2=0\mu_1 - \mu_2 = 0): While typical superiority tests place zero difference in H0H_0, clinical non-inferiority and equivalence tests use a non-zero margin (−δ-\delta) determined by clinical significance to define the boundary between clinically acceptable and unacceptable performance.

Typical errors

OverviewACEpath Teacher's Diagnosis

01 The student reverses the direction of the inequality, testing whether the generic drug is inferior rather than non-inferior.

Sample Student Work

Reviewed by ACEpath Teacher
Step-by-Step Review
Walk Through It with Your Teacher
Partially correct
  1. Step 1

    H0:μgeneric−μbrand≥−2.50 μg/mL  ⟹  H1:μgeneric−μbrand<−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \ge -2.50\,\mu\text{g/mL} \implies H_1: \mu_{\text{generic}} - \mu_{\text{brand}} < -2.50\,\mu\text{g/mL}

    Correct

    H0:μgeneric−μbrand≤−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le -2.50\,\mu\text{g/mL} versus Ha:μgeneric−μbrand>−2.50 μg/mLH_a: \mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\,\mu\text{g/mL}

    You placed the non-inferiority condition (μgeneric−μbrand≥−2.50 μg/mL\mu_{\text{generic}} - \mu_{\text{brand}} \ge -2.50\,\mu\text{g/mL}) into H0H_0 and inferiority into H1H_1. Because non-inferiority is the claim you are seeking to demonstrate, it must be formulated as the alternative hypothesis HaH_a.

  2. Step 2

    "The research claim to be demonstrated with strong statistical evidence must be placed in the alternative hypothesis to control the probability of a false positive."

WHAT YOU DID WELL
Compliments from Your Teacher

You correctly identified the margin threshold as −2.50 μg/mL-2.50\,\mu\text{g/mL}.

You stated the correct methodological justification that the research claim belongs in HaH_a to control false positives.

THE ERROR
Where Your ACEpath Teacher Spotted the Mistake
Reversed Null and Alternative Hypotheses in Non-Inferiority Test

You correctly explained in words that the research claim belongs in HaH_a, but you assigned the inequality representing that claim to H0H_0 instead.

WHY IT HAPPENED
ACEpath Teacher's Diagnosis

Confusing how to translate a non-inferiority inequality into symbolic hypotheses: in non-inferiority testing, the desired claim ('not worse than brand by more than the margin') is the alternative hypothesis HaH_a, while the null hypothesis H0H_0 represents the unacceptable or status-quo condition (inferiority).

REASONING PATH
Your ACEpath teacher’s suggested approach
Define the Research Goal
  • The alternative hypothesis HaH_a always holds the claim being tested for or the condition requiring affirmative proof.

  • Identify the non-inferiority criterion μgeneric−μbrand>−δ\mu_{\text{generic}} - \mu_{\text{brand}} > -\delta and write it directly as HaH_a.

Formulate the Null Hypothesis
  • The null hypothesis H0H_0 represents the complement of HaH_a (the unacceptable outcome of inferiority) and contains the equality boundary.

  • Write H0:μgeneric−μbrand≤−δH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le -\delta (or =−δ= -\delta) so that rejecting H0H_0 provides statistical proof of non-inferiority.

CORRECTED PATH
ACEpath Teacher's Demonstration
  1. Identify the research claim to prove: The generic is non-inferior, so μgeneric−μbrand>−2.50 μg/mL\mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\,\mu\text{g/mL}. Assign this condition to HaH_a.

  2. Form the conservative null hypothesis: Assume the generic is unacceptably inferior until proven otherwise: H0:μgeneric−μbrand≤−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le -2.50\,\mu\text{g/mL}.

Final answer

(a) H0:μgeneric−μbrand≤−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le -2.50\,\mu\text{g/mL} vs. Ha:μgeneric−μbrand>−2.50 μg/mLH_a: \mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\,\mu\text{g/mL}. (b) Placing non-inferiority in HaH_a puts the burden of proof on the generic manufacturer, ensuring an inferior drug is not approved by chance (controlling Type I error).

NEXT STEPS
Final Takeaways from Your Teacher
  • Before writing down symbols, label the inequality representing the 'claim to prove' explicitly as HaH_a.

  • Check that rejecting H0H_0 corresponds to the desired scientific or regulatory success.

OverviewACEpath Teacher's Diagnosis

02 The student misinterprets the non-inferiority margin as positive 2.50, requiring the generic drug to exceed the brand formulation.

Sample Student Work

Reviewed by ACEpath Teacher
Step-by-Step Review
Walk Through It with Your Teacher
Partially correct
  1. Step 1

    H0:μgeneric−μbrand≤2.50 μg/mL  ⟹  H1:μgeneric−μbrand>2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le 2.50\,\mu\text{g/mL} \implies H_1: \mu_{\text{generic}} - \mu_{\text{brand}} > 2.50\,\mu\text{g/mL}

    Correct

    H0:μgeneric−μbrand≤−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le -2.50\,\mu\text{g/mL} versus Ha:μgeneric−μbrand>−2.50 μg/mLH_a: \mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\,\mu\text{g/mL}

    You set the threshold at +2.50 μg/mL+2.50\,\mu\text{g/mL}, which requires the generic drug to outperform the brand by more than 2.50 μg/mL2.50\,\mu\text{g/mL} (a superiority test). Because non-inferiority allows the generic to be at most 2.50 μg/mL2.50\,\mu\text{g/mL} lower than the brand, the boundary is −δ=−2.50 μg/mL-\delta = -2.50\,\mu\text{g/mL}.

  2. Step 2

    The alternative hypothesis contains the claim that requires proof so that the regulatory body only accepts non-inferiority when the data provide overwhelming evidence.

WHAT YOU DID WELL
Compliments from Your Teacher

You correctly identified the general inequality direction for a non-inferiority alternative hypothesis (>>).

You provided an accurate, well-reasoned justification for placing the research claim in the alternative hypothesis to protect against false claims.

THE ERROR
Where Your ACEpath Teacher Spotted the Mistake
Sign error on non-inferiority margin

You wrote +2.50+2.50 instead of −2.50-2.50 for the non-inferiority margin.

WHY IT HAPPENED
ACEpath Teacher's Diagnosis

You interpreted the margin magnitude δ=2.50\delta = 2.50 as a positive cutoff rather than a allowable deficit below the reference group (−δ=−2.50-\delta = -2.50), inadvertently converting a non-inferiority test into a strict superiority test.

REASONING PATH
Your ACEpath teacher’s suggested approach
Direction of Allowable Deficit
  • When defining difference as μnew−μstandard\mu_{\text{new}} - \mu_{\text{standard}}, non-inferiority allows the new treatment to be slightly worse up to a margin δ>0\delta > 0, giving the condition μnew−μstandard>−δ\mu_{\text{new}} - \mu_{\text{standard}} > -\delta.

  • Always check whether the difference is allowed to drop below zero: if the new treatment can be slightly lower than the control, the boundary on μnew−μstandard\mu_{\text{new}} - \mu_{\text{standard}} must be negative (−δ-\delta).

CORRECTED PATH
ACEpath Teacher's Demonstration
  1. Identify the non-inferiority condition: μgeneric\mu_{\text{generic}} is not lower than μbrand\mu_{\text{brand}} by more than 2.50 μg/mL2.50\,\mu\text{g/mL}, which translates to μgeneric−μbrand≥−2.50 μg/mL\mu_{\text{generic}} - \mu_{\text{brand}} \ge -2.50\,\mu\text{g/mL}.

  2. Assign the strict inequality corresponding to the claim to the alternative hypothesis: Ha:μgeneric−μbrand>−2.50 μg/mLH_a: \mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\,\mu\text{g/mL}.

  3. Set the complementary condition representing unacceptable inferiority as the null hypothesis: H0:μgeneric−μbrand≤−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le -2.50\,\mu\text{g/mL}.

Final answer

(a) H0:μgeneric−μbrand≤−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le -2.50\,\mu\text{g/mL} vs. Ha:μgeneric−μbrand>−2.50 μg/mLH_a: \mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\,\mu\text{g/mL}; (b) placing the claim in HaH_a ensures the burden of proof rests on demonstrating non-inferiority, controlling Type I error.

NEXT STEPS
Final Takeaways from Your Teacher
  • Review the distinction between superiority tests (μ1−μ2>0\mu_1 - \mu_2 > 0 or >+δ> +\delta) and non-inferiority tests (μ1−μ2>−δ\mu_1 - \mu_2 > -\delta).

  • Practice drawing the difference parameter axis with 00, −δ-\delta, and +δ+\delta to visually confirm the acceptable region.

OverviewACEpath Teacher's Diagnosis

03 The student mistakenly places the non-inferiority claim in the null hypothesis because of the conventional belief that the null represents equality or an acceptable status quo.

Sample Student Work

Reviewed by ACEpath Teacher
Step-by-Step Review
Walk Through It with Your Teacher
Incorrect
  1. Step 1

    H0:μgeneric−μbrand≥−2.50 μg/mL  ⟹  H1:μgeneric−μbrand<−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \ge -2.50\ \mu\text{g/mL} \implies H_1: \mu_{\text{generic}} - \mu_{\text{brand}} < -2.50\ \mu\text{g/mL}

    Correct

    H0:μgeneric−μbrand≤−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le -2.50\ \mu\text{g/mL} versus Ha:μgeneric−μbrand>−2.50 μg/mLH_a: \mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\ \mu\text{g/mL}

    You reversed the hypothesis directions. In non-inferiority testing, the claim being tested for regulatory approval (non-inferiority: μgeneric−μbrand>−2.50 μg/mL\mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\ \mu\text{g/mL}) belongs in the alternative hypothesis HaH_a, while the null hypothesis H0H_0 represents the conservative condition of inferiority (≤−2.50 μg/mL\le -2.50\ \mu\text{g/mL}).

  2. Step 2

    The non-inferiority claim belongs in the null hypothesis because the null hypothesis typically represents the desired standard that the generic is good enough, while the alternative represents falling short.

    Correct

    The burden of proof rests on demonstrating the research claim. Placing non-inferiority in HaH_a ensures the generic drug is approved only when strong sample evidence rejects inferiority, thereby controlling Type I error and protecting public safety.

    You argued that H0H_0 should represent the desired standard. In statistical hypothesis testing, the null hypothesis represents the skeptical default position, not the desired outcome. If non-inferiority were placed in H0H_0, a poorly designed or underpowered study would simply fail to reject H0H_0, falsely concluding the drug is acceptable without sufficient evidence.

WHAT YOU DID WELL
Compliments from Your Teacher

You correctly identified the boundary margin of −2.50 μg/mL-2.50\ \mu\text{g/mL} for the mean difference μgeneric−μbrand\mu_{\text{generic}} - \mu_{\text{brand}}.

You used complementary mathematical inequalities spanning the full parameter space.

THE ERROR
Where Your ACEpath Teacher Spotted the Mistake
Reversed Null and Alternative Hypotheses in Non-Inferiority Testing

No arithmetic slip was made; the error stems from a conceptual misunderstanding of how the burden of proof is assigned in hypothesis testing.

WHY IT HAPPENED
ACEpath Teacher's Diagnosis

You treated the null hypothesis as the desired or acceptable benchmark rather than the skeptical baseline that must be rejected with convincing statistical evidence.

REASONING PATH
Your ACEpath teacher’s suggested approach
Burden of Proof Rule
  • The research claim that requires verification or regulatory approval is always assigned to the alternative hypothesis HaH_a.

  • Place the desired conclusion (non-inferiority, superiority, or effect) in HaH_a, not H0H_0.

Type I Error Safeguard
  • A Type I error represents a false positive (approving an ineffective or inferior drug).

  • Set H0H_0 as the undesirable or unsafe condition (inferiority) so that rejecting it requires strong evidence beyond a controlled risk level α\alpha.

CORRECTED PATH
ACEpath Teacher's Demonstration
  1. Identify the research claim: The generic manufacturer wants to prove that μgeneric−μbrand>−2.50 μg/mL\mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\ \mu\text{g/mL}.

  2. Assign this claim to the alternative hypothesis: Ha:μgeneric−μbrand>−2.50 μg/mLH_a: \mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\ \mu\text{g/mL}.

  3. Set the complementary condition as the null hypothesis: H0:μgeneric−μbrand≤−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le -2.50\ \mu\text{g/mL}.

Final answer

(a) H0:μgeneric−μbrand≤−2.50 μg/mLH_0: \mu_{\text{generic}} - \mu_{\text{brand}} \le -2.50\ \mu\text{g/mL} vs. Ha:μgeneric−μbrand>−2.50 μg/mLH_a: \mu_{\text{generic}} - \mu_{\text{brand}} > -2.50\ \mu\text{g/mL}; (b) The claim must be in HaH_a because the burden of proof is on proving efficacy/safety, which controls the Type I error rate against approving an inferior generic.

NEXT STEPS
Final Takeaways from Your Teacher
  • Review the connection between the research objective, the burden of proof, and HaH_a.

  • Practice identifying what constitutes a Type I error (false positive) in clinical and regulatory testing contexts.

Lock it in

Similar question

01

A new generic pain reliever is tested against a standard pain reliever. Let μnew\mu_{\text{new}} and μstandard\mu_{\text{standard}} be the mean hours of pain relief. Non-inferiority is defined as the new drug not being worse by more than 0.50.5 hours (μnew−μstandard≥−0.5\mu_{\text{new}} - \mu_{\text{standard}} \ge -0.5). State the null and alternative hypotheses.

Answer

H0:μnew−μstandard≤−0.5H_0: \mu_{\text{new}} - \mu_{\text{standard}} \le -0.5 hours vs. Ha:μnew−μstandard>−0.5H_a: \mu_{\text{new}} - \mu_{\text{standard}} > -0.5 hours. Key step: The research claim of non-inferiority belongs in HaH_a, while the inferiority region belongs in H0H_0.

02

An online retailer tests a new checkout interface against the current one. Let pnewp_{\text{new}} and poldp_{\text{old}} denote conversion rates. Management will adopt the new interface if its conversion rate is not lower than the old one by more than 1%1\% (pnew−pold≥−0.01p_{\text{new}} - p_{\text{old}} \ge -0.01). (a) Formulate H0H_0 and HaH_a. (b) What constitutes a Type I error in this business context?

Answer

(a) H0:pnew−pold≤−0.01H_0: p_{\text{new}} - p_{\text{old}} \le -0.01 vs. Ha:pnew−pold>−0.01H_a: p_{\text{new}} - p_{\text{old}} > -0.01. (b) A Type I error is rejecting H0H_0 when it is true—falsely concluding the new checkout is non-inferior when it is actually unacceptably worse by more than 1%1\%, leading to the costly rollout of an inferior checkout.

03

In a biosimilar equivalence study, regulators require demonstrating equivalence within a symmetric margin [−δ,+δ][-\delta, +\delta]: −δ<μT−μR<δ-\delta < \mu_T - \mu_R < \delta. Express this equivalence test as two one-sided hypothesis testing (TOST) pairs: (H01,Ha1)(H_{01}, H_{a1}) for testing against the lower boundary −δ-\delta, and (H02,Ha2)(H_{02}, H_{a2}) for testing against the upper boundary +δ+\delta.

Answer

Lower boundary pair: H01:μT−μR≤−δH_{01}: \mu_T - \mu_R \le -\delta vs. Ha1:μT−μR>−δH_{a1}: \mu_T - \mu_R > -\delta. Upper boundary pair: H02:μT−μR≥δH_{02}: \mu_T - \mu_R \ge \delta vs. Ha2:μT−μR<δH_{a2}: \mu_T - \mu_R < \delta. Equivalence is established only if both null hypotheses are rejected simultaneously.

Loading workspace…