Table 7: Risk of bias assessment of Zhang, et al.'s study.

A.     Selection bias

Evaluation

High risk

Justification

A simple web-based randomization technique without any stratification was adopted. Since there were many subtypes of breast cancers, lack of stratification might have affected the balance post-randomization. Allocation concealment focuses on preventing selection and confounding biases and safeguards the assignment sequence before and until allocation (Schulz, et al. [21])

B.     Performance bias

Evaluation

Unclear risk

Justification

The authors haven’t discussed about the care and support apart from intervention that was available to participants in both arms. Neither participants nor caregivers were kept blind to intervention allocation. Since the outcome did not involve subjectivity, participant or researcher has very little to no influence on measurement of outcome. So even though double-blind trials are often recommended, it doesn’t make much difference if the outcome is not subjective (Day & Altman, et al. [22]).

C.     Attrition bias

Evaluation

Low risk

Justification

Loss to follow up was none and all the patients reported their outcomes. Intention-to-treat analysis was performed which adds more credibility and rigour.

D.     Detection bias

Evaluation

Low risk

Justification

Tumour response was evaluated post every 2 treatment cycles in both arms. Tumour response was measured by RECIST version 1.1 and thus was uniform across the groups. Thus, there is a low risk of detection bias.

E.      Internal validity

Evaluation

+

 

Justification

Internal validity is compromised due to flawed randomization and allocation concealment technique.

F.      External validity

Evaluation

+

 

Justification

The overall generalizability is possible but only to certain extent due to high risk of bias and low internal validity.