Kirk Week 1: Reading A Scientific Paper

Parts of a scientific paper

Abstract

Introduction

Methods

Results

Discussion

Abstract

Summary

Limited number of words

Is this article relevant?

NOT used for decision making

Introduction

Background information

Literature review

Rationale for the study

Gaps in knowledge

Clearly defined research objectives

Methods

Detailed description

Study population/location

Randomization/controls

Treatment description

Outcome measurements

Statistical analysis

Results

Presentation of data

Key findings

Population characteristics

Secondary or subgroup analyses

Efficacy and safety

Discussion

Data Interpretation

Findings on primary objective

Comparison to baseline

Limitations/discrepancies

Clinical translations

Future research

What types of studies can be used to understand a diagnosis or therapy?

Randomized controlled trials (RCTs)

Systematic reviews of RCTs

What types of studies can be used to provide a prognosis?

Cohort studies

Case controls

Case studies

What types of studies can be used to identify level of harm or prevention?

Randomized controlled trials

Cohort studies

Case controls

What is the main drawback of randomized controlled trials?

They are usually shorter than other types of studies, which means they cannot collect data about long term safety/efficacy

What does PICOT stand for?

Population and Problem

Intervention (what is changed)

Comparison (baseline)

Outcome

Time

What types of studies contain the highest quality and most filtered information?

Systematic review and meta-analysis

What are signs that a journal is credible?

Peer-reviewed

Indexed in secondary literature

Good impact factor

Impact factor

Number of citations in the past 2 years/Number of citable papers published in the past 2 years

~3 = good

>10 = remarkable

What conflicts of interest may be found in scientific papers?

Author affiliations

Funding source

Mitigation

What factors determine the clinical importance of a study?

Effect size

• Tiny differences can be statistically significant but clinically irrelevant

• “Is this actually a big deal?”

Absolute risk reduction

Number needed to treat

• High NNT = minimal real-world benefit

Patient-centered outcomes

Benefit vs. harm

What are the best uses for randomized controlled trials?

Therapy, prevention, and diagnosis

What are the best uses for cohort studies?

Prognosis and harm

What are the best uses for case-control studies?

Rare outcomes or conditions with long latency periods

What are the best uses for cross-sectional studies?

Prevalence studies and diagnostic accuracy assessments

What are the best uses for qualitative studies?

Exploring experiences, perceptions, behaviors, and complex social phenomena

Selection Bias

Picking participants in a way that makes the study look better

How can selection bias be prevented?

Randomized selection

Performance bias

Changing treatment (other than the thing you are studying) for the group you want to “win”

How can performance bias be prevented?

Blinding everyone involved in the study

Examples of inappropriate statistical methods

Wrong test

Incorrect assumptions

Examples of flawed data

Sample size

Missing data

Untrusted data

Examples of misinterpreted data

Correlation vs. causation

Incomplete data

Ignoring outliers

Clinical Endpoints

Direct measures of how a patient feels, functions, or survives after some intervention

Composite Endpoints

Include two or more outcomes

Surrogate Endpoints

Indirect measures of clinical endpoints

Internal Integrity

Study integrity

How can internal integrity be increased?

Blinding

Randomization

Strict study protocolHow

What factors decrease internal integrity?

Attrition (decrease in study participants over time)

Confounding variables

Study variation

External Validity

Applicability

How can external validity be increased?

Defined population

Diversity

Replication

• ex. Systematic reviews

What factors decrease external validity?

Selection bias

Low sample size

Strict study protocol