Chapter 2 · Starting a study

How do you turn a doubt into a research question?

Research problem, PICO, objectives, variables and the research hypothesis.

15 TNMGRMU questions answered hereset 17 times, 2014–2025

The problem first

You are a first-year Anaesthesiology resident. Your guide asks for a thesis topic, and you submit "Role of dexmedetomidine in anaesthesia". The departmental research committee returns it with one comment: “This is a topic, not a research question.”

You add "a prospective study" to the title and submit it again. At the protocol presentation, the examiner asks one question.

“In which patients, compared with what, and measured how?”

Start here

How does a doubt become a research question?

A research question starts as a doubt from the ward, the operating theatre or your reading. Each step below makes it narrower, until it states what you will measure, in whom, and against what.

Teaching example. A topic names an area. A research problem is the gap in knowledge, and why it matters. The research question asks one answerable question about that gap. The hypothesis predicts the answer, and the objectives state what the study will do to find it.

Read the literature before you fix the question. If a good trial or systematic review has already answered it, change the question. The literature search chapter shows how to check.

Before you write it

Is the question good enough to study?

Hulley and colleagues proposed five checks for a research question, known by their initials: FINER.

Each draft question stops at the first check it fails. Feasible: enough participants, the technical skill, affordable in time and money, and a manageable scope. Interesting to you, your peers and the community. Novel: it confirms, refutes or extends earlier findings. Ethical: an ethics committee would approve it. Relevant to scientific knowledge, to clinical and health policy, or to future research. Criteria as summarised by Farrugia et al. (2010).

Interactive

Find the FINER check that each draft question fails

Try this: read draft 1 and its context, choose the check it fails, and read why.

Draft 1 of 6

"Does a new dietary supplement taken from age 30 reduce deaths from heart disease by age 70?"

You have a 3-year MD thesis.

Choose one answer to see the explanation.

Teaching examples. Each draft fails no more than one check.

PICO

What does a well-built research question contain?

Four parts: the population, the intervention, the comparison and the outcome. Richardson and colleagues proposed this four-part structure for clinical questions in 1995.

The submitted title names a drug and a specialty. The PICO version names the patients, what they receive, what they are compared with, and what is measured. Each part then becomes part of the protocol. P gives the inclusion and exclusion criteria, I and C give the study groups, and O gives the primary outcome, the measuring tool and the basis for the sample size.

A study should have one primary research question. Secondary questions can be added, but they should not compromise the primary question, because the hypothesis and objectives are built on it (Farrugia et al., 2010).

Other formats

What if your study has no intervention?

PICO was designed for treatment questions. Other types of study change one or more of the letters.

Format	Stands for	Use it for	Teaching example
PICOT	PICO + Time	Outcomes measured at a set follow-up time	Morphine use in the first 24 hours after surgery
PECO	Population, Exposure, Comparator, Outcome	Exposures you do not assign: smoking, biomass smoke, a blood marker	Children in homes cooking with wood or dung vs LPG. Outcome: pneumonia episodes in one year
Diagnostic PICO	Patients, Index test, Comparator (reference standard), Outcome (accuracy)	How accurate a test is	Ultrasound vs surgical and histopathology findings in suspected appendicitis. Outcome: sensitivity and specificity
SPIDER	Sample, Phenomenon of Interest, Design, Evaluation, Research type	Qualitative and mixed-methods research	Patients who stopped antipsychotics (S), reasons for stopping (PI), interviews (D), views (E), qualitative (R)

PICOT: Farrugia et al. (2010). PECO: Morgan et al. (2018). SPIDER: Cooke, Smith and Booth (2012).

Interactive

Choose P, I, C and O for a scenario and watch the question sentence change

Try this: change P from "Not stated" to the specific population, and watch the question sentence and the list of parts to fix change.

Type of question

TreatmentExposureDiagnostic test

Teaching example: many patients still ask for morphine after laparoscopic cholecystectomy, and you wonder whether dexmedetomidine during surgery would reduce it.

P: population

I: intervention

C: comparison

O: outcome

In [population not stated], does the role of dexmedetomidine, compared with [comparison not stated], change [outcome not stated]?

P: Missing. Without a population there are no inclusion and exclusion criteria.

I: Vague. Words such as "role", "helps" or "efficacy" do not say what is given, how or when.

C: Missing. Without a comparison group you cannot separate the effect of the drug from natural recovery or other care.

O: Missing. The outcome decides the measuring tool and the sample size.

Still to fix: P, I, C and O.

Decide

Which question format fits your study?

Answer one or two questions. Each answer links to the section that explains it.

What will your study do?

Which format: all possible answers

PICO: Population, Intervention, Comparison, Outcome. Write one question, then a hypothesis that predicts its answer.
PICOT: PICO plus Time: the follow-up period at which the outcome is measured.
PECO: Population, Exposure, Comparator, Outcome. Use it when you do not give the exposure yourself, such as smoking, biomass smoke or a blood marker. The design will be observational.
PICO adapted for a diagnostic test: P: patients suspected of having the condition. I: the test being evaluated (index test). C: the reference standard. O: sensitivity, specificity and predictive values.
Descriptive question with objectives, no hypothesis: State the population, place, period and what you will measure. Nothing is being compared, so write objectives but no hypothesis.
SPIDER: Sample, Phenomenon of Interest, Design, Evaluation, Research type. Proposed for qualitative and mixed-methods research.

Hypothesis

What is a research hypothesis, and how do you write one?

A research hypothesis is a statement, written before data collection, of the result you expect: which variable will differ or change, in whom, in which direction, and by what time.

The same PICO question gives three statements. The research hypothesis is the one you expect to be true. It can be non-directional (two-sided) or directional (one-sided). Use a two-sided hypothesis unless you can justify one direction in advance (Farrugia et al., 2010). The null hypothesis of no difference is the statement the statistical test examines.

A good hypothesis is

Stated before data collection, in the protocol. Presenting a hypothesis formed after seeing the results as if it had been planned is called HARKing (Kerr, 1998).
Testable: some possible result would show it to be false. Karl Popper used this property, falsifiability, as his test of whether a theory is scientific.
Specific: it names the population, the independent and dependent variables, and the time point.
Based on earlier evidence, a pilot study or a clear biological reason.
About one relationship. Write separate hypotheses for separate outcomes.
Consistent with the research question and the primary objective, using the same words.

Types of hypothesis

Simple (one independent and one dependent variable) or complex (more than one of either).
Directional (one-sided) or non-directional (two-sided).
Associative (two variables change together) or causal (one variable changes the other).
Research or alternative hypothesis (H₁) or null hypothesis (H₀).

A descriptive study, such as a prevalence survey, compares nothing, so it has objectives but no hypothesis. How the null hypothesis is tested, and what the p-value means, is covered in the hypothesis testing chapter and in the Stateazy article on the null hypothesis.

Interactive

Sort statements into null, non-directional and directional hypotheses

Try this: read statement 1, choose its type, then move to the next statement and compare the wording.

Statement 1 of 8

"Morphine use in the first 24 hours will differ between the dexmedetomidine and saline groups."

Null hypothesisNon-directionalDirectionalNot a hypothesis

Choose one answer to see the explanation.

Teaching examples from this chapter. They make no claim about the effect of any drug or marker.

Comparison

Research question or research hypothesis: what is the difference?

The question asks. The hypothesis predicts the answer.

Feature	Research question	Research hypothesis
Form	A question, ending in a question mark	A statement of the expected result
Direction	Not stated	Stated (directional) or not (non-directional)
Needed in	Every study	Analytical studies and trials, but not purely descriptive studies
Comes from	The research problem and the literature	The research question
Checked by	The study as a whole	A statistical test of its null hypothesis
Example	In adults having laparoscopic cholecystectomy, does IV dexmedetomidine, compared with saline, reduce 24-hour morphine use?	Adults given IV dexmedetomidine will use a different amount of morphine in 24 hours from those given saline.

Objectives

How do you turn the question into objectives?

The aim states the purpose of the study in one sentence. The specific objectives state what you will measure or compare to achieve it.

Teaching example. The primary objective matches the primary outcome and the hypothesis, and it decides the sample size. Secondary objectives add information, but the study is usually not large enough to answer them firmly.

Characteristics of good objectives

Specific: one outcome in one population per objective.
Measurable: names the variable and how it will be measured.
Achievable with your patients, time, money and skills.
Relevant: follows directly from the research question.
Time-bound: states when the outcome is measured.
Starts with an action verb that describes a measurable task: to estimate, to compare, to determine the association between. Avoid "to study", "to know" and "to understand", which do not say what will be measured.
In order of priority, primary first. Each objective must be answered in the results chapter of the thesis.

These points follow the SMART checklist, first published by George Doran in 1981 for management objectives. His letters stood for specific, measurable, assignable, realistic and time-related.

Feature	Aim	Specific objectives
Also called	General objective	Objectives
Scope	The overall purpose, in one sentence	The exact tasks the study performs to meet the aim
Number	One	One primary and a few secondary
Measurable	Not necessarily	Must name what is measured, in whom and when
Example	To evaluate dexmedetomidine for pain relief after laparoscopic cholecystectomy	To compare total morphine used in the first 24 hours between the two groups

Variables

Which variable is independent and which is dependent?

A variable is any characteristic that takes different values in different people: age, haemoglobin, treatment group, survival. The independent variable is the one you compare. The dependent variable is the outcome you measure.

Teaching example with a hypothetical marker. A confounding variable is linked to both the exposure and the outcome, and can make a link appear or disappear. The bias and validity chapter covers how to control it: randomisation, restriction, matching and adjustment in the analysis.

Feature	Independent variable	Dependent variable
Also called	Exposure, predictor, intervention, explanatory variable	Outcome, response variable
Role	What is compared or changed	What is measured to see the effect
In PICO	I and C (E in PECO)	O
Who sets it	The researcher in a trial. In an observational study it is already present.	Nobody. It is recorded.
On a graph	Usually the x-axis	Usually the y-axis
Example	Study group: dexmedetomidine or saline	Morphine used in 24 hours

Other variables to list in your protocol

Confounding: linked to both the exposure and the outcome, and not on the path between them.
Extraneous: any other factor that affects the outcome, such as the time of day a test is done. Keep it constant or record it.
Intervening (mediating): lies on the path from the independent to the dependent variable.
Effect modifier: changes the size of the effect in different subgroups, for example in men and women.
Background (demographic): age, sex and socioeconomic status, recorded to describe the sample.

Give each variable an operational definition: exactly how, with which instrument and when it will be measured. For example, "pain: score on a 0 to 10 numeric rating scale at 24 hours after surgery". Classification by scale (nominal, ordinal, interval, ratio) is covered in the descriptive statistics chapter.

Worked example

One doubt, from the ward to the protocol

The dexmedetomidine thesis from the top of this page, written out in full. This is a teaching example and makes no claim about the drug's effect.

Step	What you write
Doubt	Many patients still ask for morphine after laparoscopic cholecystectomy.
Literature	Search PubMed and the Cochrane Library for trials of dexmedetomidine in this operation. Note what is known and what is not.
Research problem	Whether adding dexmedetomidine during surgery reduces morphine use after laparoscopic cholecystectomy in our patients is uncertain.
FINER check	Enough operations each month in your unit (feasible). Both groups receive standard analgesia (ethical).
PICOT	P: adults aged 18–65, ASA I–II, elective laparoscopic cholecystectomy. I: dexmedetomidine infusion during surgery. C: saline infusion of the same volume. O: total morphine used. T: first 24 hours.
Research question	In adults having elective laparoscopic cholecystectomy, does a dexmedetomidine infusion during surgery, compared with saline, reduce morphine use in the first 24 hours?
Hypothesis	Research (two-sided): morphine use in the first 24 hours will differ between the groups. Null: there will be no difference.
Aim	To evaluate dexmedetomidine during surgery for pain relief after laparoscopic cholecystectomy.
Primary objective	To compare total morphine used in the first 24 hours between the two groups.
Secondary objectives	To compare pain scores at 1, 6, 12 and 24 hours. To compare the rates of bradycardia and hypotension.
Variables	Independent: study group. Dependent: morphine use, pain score, bradycardia, hypotension. Record as possible confounders: duration of surgery and opioid dose during surgery.
Design	Double-blind randomised controlled trial. Sample size calculated from the primary outcome.
Title	Effect of dexmedetomidine during surgery on 24-hour morphine consumption after laparoscopic cholecystectomy: a randomised controlled trial.

The title, question, hypothesis, primary objective and primary outcome all use the same words. The sample size chapter shows how the primary outcome sets the number of patients.

By specialty

Research question problems by specialty

A resident wonders whether patients with type 2 diabetes on long-term metformin have low vitamin B12. Metformin was prescribed for clinical reasons, not by the researcher.

Metformin use is an exposure, not an intervention you assign. Write a PECO question and use a cross-sectional or cohort design.

Common myths

Common mistaken beliefs

Myth: A topic and a research question are the same thing.
A topic names an area, such as "dexmedetomidine in anaesthesia". A question names the population, intervention, comparison and outcome, and can be answered with data.
Myth: Every study needs a hypothesis.
Analytical studies and trials test a hypothesis. A purely descriptive study, such as a prevalence survey, compares nothing, so it has objectives but no hypothesis.
Myth: The research hypothesis is the null hypothesis.
The research hypothesis states the relationship you expect. The null hypothesis states no difference, and it is the statement the statistical test examines.
Myth: More objectives make a stronger thesis.
The primary objective sets the sample size. Each extra objective adds data collection and analyses that the study is usually not large enough to answer.
Myth: The hypothesis can be written after looking at the results.
Presenting a hypothesis formed after seeing the results as if it had been planned is called HARKing (Kerr, 1998). Write hypotheses in the protocol, before data collection.
Myth: PICO is only for drug trials.
PICO has versions for other questions. PICOT adds time, PECO covers exposures, and diagnostic questions use the index test and reference standard.

Why this matters to you

Five consequences of a poorly framed research question

Protocol returned by the committee
A title such as "Role of X in Y" does not state the population, comparison or outcome, so the committee cannot judge the design or the sample size.
A literature search that finds too much
Without P, I, C and O you have no search terms, and the search returns thousands of unrelated papers. See the literature search chapter.
Objectives the data cannot answer
An objective that names an outcome you never recorded cannot be reported in the thesis.
Questions you cannot answer at the viva
Examiners ask for the hypothesis, the primary objective and the independent and dependent variables. Each should be one sentence in your protocol.
Research that does not help patients
Choosing questions that do not matter to patients and clinicians is the first of four stages of avoidable research waste described by Chalmers and Glasziou (2009).

History

History of the research question and hypothesis

1938
John Dewey
In Logic: The Theory of Inquiry he wrote that "a problem well put is half-solved", and that finding out what the problem is means being well along in the inquiry.
1959
Karl Popper
In The Logic of Scientific Discovery he argued that a theory is scientific only if observations could, in principle, show it to be false. A research hypothesis must be testable in this sense.
1981
George T. Doran
Published the SMART checklist for writing objectives in Management Review: specific, measurable, assignable, realistic and time-related.
1995
Richardson, Wilson, Nishikawa and Hayward
In ACP Journal Club, proposed a four-part structure for "the well-built clinical question". It became known as PICO.
1998
Norbert Kerr
Named HARKing, "hypothesizing after the results are known": presenting a hypothesis formed from the results as if it had been stated in advance.
2009
Iain Chalmers and Paul Glasziou
Described avoidable waste in research at four stages, the first being the choice of research questions. They estimated that, if their figures applied generally, more than 85% of research investment was lost.
2012
Cooke, Smith and Booth
Proposed SPIDER as an alternative to PICO when searching for qualitative and mixed-methods research.
2018
Morgan, Whaley, Thayer and Schünemann
Published a framework for writing PECO questions (population, exposure, comparator, outcome) about exposures and health outcomes.
2025
SPIRIT 2025 and CONSORT 2025
Both require specific objectives related to benefits and harms: item 10 in trial protocols and item 7 in trial reports.

Guidelines

Research questions and objectives in guidelines

SPIRIT 2025, item 10 (trial protocols)

“Specific objectives related to benefits and harms”.

CONSORT 2025, item 7 (trial reports)

“Specific objectives related to benefits and harms”.

STROBE (2007), item 3

Observational studies must “state specific objectives, including any prespecified hypotheses”.

PRISMA 2020, item 4

Systematic reviews must give “an explicit statement of the objective(s) or question(s) the review addresses”.

ICH E9 (1998), section 2.2.2

The primary variable should give “the most clinically relevant and convincing evidence” related to the primary objective. There should generally be only one, and it is generally the one used to estimate the sample size.

Steps

Writing the question, hypothesis and objectives, step by step

Write down the doubt as it arose in practice or in your reading.
Search the literature to find what is already known.
State the research problem: the gap, and why it matters to patients.
Check the idea against FINER.
Write one research question in PICO form, or PICOT, PECO or SPIDER as the study requires.
For an analytical study or trial, write the research hypothesis and its null hypothesis.
Write the aim, one primary objective and a few secondary objectives, each with an action verb.
List the independent, dependent and confounding variables, with an operational definition for each.
Check that the title, question, hypothesis, primary objective and primary outcome use the same words.

Reading a paper

Five checks on any study's research question

Is the research question or objective stated clearly in the introduction?
Can you identify the population, intervention or exposure, comparison and outcome?
Was one primary outcome named in advance, in the protocol or trial registration?
Were the hypotheses prespecified, or do they appear only after subgroup results?
Does the conclusion answer the question that was asked, in the population that was studied?

Why is it called that?

Word surgery

“Hypothesis”

In everyday speech

A guess or a hunch.

In research

A specific, testable statement of the expected relationship between variables, written before data are collected.

“Problem”

In everyday speech

Trouble, or something that has gone wrong.

In research

A gap or uncertainty in knowledge or practice that a study can address.

“Objective”

In everyday speech

Unbiased, based on facts.

In research

A specific statement of what the study will measure or compare.

Memory trick

The dependent variable depends on the independent variable. You compare groups of the independent variable and measure the dependent variable.

Problem

Greek problēma: "thing put forward"

From proballein, "to propose". A research problem is the gap put forward for study.

Question

Latin quaestio: "a seeking, an inquiry"

From quaerere, "to ask, to seek". The research question states what the study seeks to find out.

Hypothesis

Greek hypothesis: "a placing under", from hypo (under) + thesis (a placing)

In Greek, a base or supposition. In research, the statement a study starts from and then tests with data.

Variable

Late Latin variabilis: "changeable"

From variare, "to change". A characteristic whose value changes from person to person.

Objective

Medieval Latin objectum: "thing put before"

The sense "goal" comes from the military term "objective point". In research, what the study sets out to achieve, stated so that it can be measured.

PICO

Population · Intervention · Comparison · Outcome

The four parts of a well-built question. PICOT adds Time. PECO replaces Intervention with Exposure.

Exam answers

Model answer outlines

Formulating a research question, and the PICO format with examples

Ph.D. RM 2023, 2024, 2025

Research question: a clear, focused, answerable question about a gap in knowledge, stated before the study and answered by its data.
Sources: clinical practice, systematic literature search, recommendations of earlier studies, interviews with patients and with experts.
Sequence: doubt → topic → review of literature → research problem → research question → hypothesis → objectives.
Check with FINER (Hulley and colleagues): feasible, interesting, novel, ethical, relevant.
Structure with PICO (Richardson et al., 1995): Population, Intervention, Comparison, Outcome. PICOT adds Time.
Versions for other designs: PECO for exposures, index test and reference standard for diagnostic accuracy, SPIDER for qualitative research.
Uses of PICO: sets the eligibility criteria (P), the study groups (I, C), the primary outcome and sample size (O), and the search terms for the literature review.
One primary question. Secondary questions must not compromise it (Farrugia et al., 2010).

Question type	P	I or E	C	O
Therapy	Adults having laparoscopic cholecystectomy	IV dexmedetomidine	Saline	Morphine use in 24 hours
Exposure (PECO)	Children under 5 in rural homes	Cooking with wood or dung	Cooking with LPG	Pneumonia episodes in 1 year
Diagnosis	Children with suspected appendicitis	Ultrasound	Surgical and histopathology findings	Sensitivity and specificity

Part (b) of the January 2023 question, data collection methods: interview, self-administered questionnaire, observation, clinical examination and measurement, laboratory and imaging investigations, and existing records (secondary data). See the data collection chapter.

Criteria of a good research problem, and "A research problem well stated is research half done"

Ph.D. RM 2022, 2023

Research problem: a gap, uncertainty or difficulty in knowledge or practice that can be studied by collecting data.

Clear and specific: names the population, variables and setting.
Researchable: can be answered by collecting and analysing data, not by opinion.
Feasible: enough participants, time, money, skills and equipment.
Novel: not already answered, as shown by a literature review.
Significant and relevant to patients, clinical practice, health policy or future research.
Interesting to the researcher, peers and the community.
Ethical: can be studied with consent, acceptable risk and ethics committee approval.
Within the researcher's competence, with access to data and supervision.

Why a well-stated problem is half the research: John Dewey wrote in 1938 that "a problem well put is half-solved". Once the problem is stated precisely:

The population defines the eligibility criteria and the sampling method.
The variables define the tools and the data to collect.
The relationship to be examined defines the design: descriptive, analytical or experimental.
The expected result gives the hypothesis, and the primary outcome gives the sample size.
The type of variables and design define the statistical analysis.
The literature review stays focused, and unnecessary data are not collected.

Contrast: "Anaemia in women" leaves every decision open. "Prevalence of anaemia among pregnant women in the first trimester attending the antenatal clinic of a district hospital over 6 months" fixes the population, setting, period, variable and design (cross-sectional).

Characteristics of good research objectives

Ph.D. RM 2023

Objective: a specific statement of what the study will do to answer the research question.

General objective (aim): the overall purpose, in one sentence.
Specific objectives: the measurable tasks, usually one primary and a few secondary.

Specific: one outcome in one population per objective.
Measurable: names the variable and the method of measurement.
Achievable within the available patients, time, money and skills.
Relevant: follows logically from the research problem and question.
Time-bound: states when the outcome is measured.
Action verbs: to estimate, to compare, to determine, to assess. Not "to study", "to know" or "to understand".
Linked to the hypothesis: the primary objective matches the hypothesis and the primary outcome, and it is used for the sample size calculation.
Few, and in order of priority, primary first.
Clear and unambiguous: another researcher would carry out the same task.

SMART comes from Doran (1981), whose letters stood for specific, measurable, assignable, realistic and time-related.

Poor: "To study dexmedetomidine in laparoscopic surgery." Good: "To compare total morphine used in the first 24 hours after laparoscopic cholecystectomy between patients given dexmedetomidine and patients given saline."

Design a study of music therapy for anxiety in patients posted for CABG, and formulate its objectives

Ph.D. RM 2022

Title (teaching example): Effectiveness of music therapy on preoperative anxiety among patients posted for elective coronary artery bypass grafting: a randomised controlled trial.
Problem statement: whether listening to music before CABG reduces preoperative anxiety among patients in our hospital is not known.
P: adults posted for elective CABG who can hear and give consent. Exclude emergency surgery, diagnosed psychiatric illness and hearing impairment.
I: 30 minutes of recorded instrumental music through headphones, on the evening before and the morning of surgery, with routine care.
C: routine preoperative care.
O and T: state anxiety score on a validated anxiety scale, from baseline to the morning of surgery. Secondary: heart rate and blood pressure.
Hypothesis: H₁, the mean anxiety score after the intervention will differ between the music and routine care groups. H₀, there will be no difference.
Design: parallel-group randomised controlled trial, computer-generated 1:1 allocation, concealed in sealed opaque envelopes, with a blinded outcome assessor.
Variables: independent, music therapy or routine care. Dependent, anxiety score, heart rate, blood pressure. Background, age, sex, previous surgery, baseline anxiety.
Sample size: from the SD of the anxiety score in earlier studies and the smallest difference that matters, increased for dropouts.
Analysis: compare the change in scores between groups with an independent t-test, or ANCOVA adjusted for baseline score. See the choosing a test chapter.
Ethics: ethics committee approval, written informed consent, trial registration before enrolment, freedom to withdraw.

General objective: to evaluate the effectiveness of music therapy on preoperative anxiety among patients posted for CABG.

Specific objectives:

To assess the preoperative anxiety score in the music therapy and routine care groups before the intervention.
To compare the anxiety score between the two groups after the intervention.
To compare heart rate and blood pressure between the two groups after the intervention.
To find the association between the post-intervention anxiety score and selected demographic variables.

Research hypothesis: definition, characteristics, types and how to frame it

Ph.D. RM 2014, 2016, 2018, 2025

Definition: a specific, testable statement, made before data collection, of the expected relationship between two or more variables in a defined population.

Sources: clinical observation, literature and earlier studies, theory, pilot data.

Characteristics:

Testable and falsifiable (Popper).
Stated in advance, not after the results (HARKing, Kerr 1998).
Specific: population, independent and dependent variables, direction and time.
Based on evidence or a clear rationale.
Simple and clear, one relationship per hypothesis.
Consistent with the research question and the primary objective.
Measurable with available tools, and ethical to test.

Type	Meaning	Example
Simple	One independent and one dependent variable	Dexmedetomidine changes 24-hour morphine use
Complex	More than one independent or dependent variable	Dexmedetomidine changes morphine use and pain scores
Directional (one-sided)	States the direction of the difference	The dexmedetomidine group will use less morphine
Non-directional (two-sided)	States a difference without a direction	Morphine use will differ between groups
Associative	Two variables change together	Higher Protein-A levels go with more nephropathy
Causal	One variable changes the other	Dexmedetomidine reduces morphine use
Null (H₀)	No difference or no association, examined by the statistical test	No difference in morphine use
Alternative (H₁)	Accepted if H₀ is rejected	Morphine use differs

How to frame it:

Start from the PICO question.
Identify the population and the independent and dependent variables.
Decide the direction from earlier evidence. Use two-sided unless one direction can be justified.
Write one statement: "In [population], [intervention] compared with [comparison] will change [outcome] at [time]."
Write the matching null hypothesis.
Check that it is testable, measurable with your tools, about one relationship, and written before data collection.

Functions: gives direction to the design, identifies the variables to measure, sets the primary outcome and sample size, and decides the statistical test. Testing is covered in the hypothesis testing chapter.

Independent and dependent variables, and other types of variables

Ph.D. RM 2014, 2023

Variable: a characteristic that takes different values in different people, places or times.

Type	Definition	Example
Independent	Compared or manipulated, the presumed cause (exposure, intervention, predictor)	Study group, or serum Protein-A level
Dependent	Measured, the presumed effect (outcome, response)	Morphine used in 24 hours, or new nephropathy
Confounding	Linked to both independent and dependent variables, not on the path between them	Duration of diabetes
Extraneous	Any other factor that affects the outcome, kept constant or recorded	Time of day of blood sampling
Intervening (mediating)	On the path from independent to dependent variable	Hypothetical: Protein-A → kidney inflammation → nephropathy
Effect modifier	Changes the size of the effect across subgroups	An effect that differs between men and women
Background	Describes the sample	Age, sex, socioeconomic status

By scale: qualitative, which is nominal (blood group) or ordinal (mild, moderate or severe pain), and quantitative, which is discrete (number of seizures) or continuous (haemoglobin).
Operational definition: states how, with what instrument and when each variable is measured.
In PICO: I and C are the independent variable, and O is the dependent variable.
Control of confounders: randomisation, restriction, matching, stratification and multivariable analysis.

Frame a hypothesis and design a study to test the association of "Protein-A" with the progression of diabetes mellitus to its complications

Ph.D. RM 2017

Protein-A is treated here as the hypothetical biomarker named in the question.

Research hypothesis: among adults with type 2 diabetes and no complications at baseline, those with high serum Protein-A levels will develop nephropathy, retinopathy or neuropathy more often over 3 years than those with low levels.

Null hypothesis: the incidence of complications does not differ between the high and low Protein-A groups.

Design: prospective cohort study, because progression needs the exposure to be measured before the outcome. A nested case-control study using stored baseline samples is an alternative. A cross-sectional study can show association but not progression.
P: adults with type 2 diabetes, screened at baseline to exclude existing nephropathy (urine albumin), retinopathy (fundus examination) and neuropathy (clinical and monofilament examination).
E and C: high vs low serum Protein-A at baseline, with the cut-off defined in the protocol before data collection.
O and T: new nephropathy, retinopathy or neuropathy by standard definitions, assessed every 6 months for 3 years.
Measurement: a validated assay. Laboratory staff do not see clinical data, and outcome assessors do not know Protein-A levels.
Confounders to record: age, duration of diabetes, HbA1c, blood pressure, lipids, smoking, body mass index and treatment.
Sample size: from the expected incidence of complications in the low and high groups, increased for loss to follow-up.
Analysis: incidence in each group, relative risk, Kaplan–Meier curves, and Cox regression adjusted for confounders to give a hazard ratio. See the risk and survival chapter.
Ethics: ethics committee approval, and informed consent that covers storage of blood samples.

PICO statement of the PALACE trials

DM Clinical Immunology and Rheumatology 2021

PALACE (Psoriatic Arthritis Long-term Assessment of Clinical Efficacy) is a phase III programme of randomised, double-blind, placebo-controlled trials of apremilast, an oral phosphodiesterase 4 inhibitor, in psoriatic arthritis.

P: adults with active psoriatic arthritis meeting CASPAR criteria, with at least 3 swollen and 3 tender joints, despite prior conventional DMARDs and/or biologics (PALACE 1–3). PALACE 3 also required active skin psoriasis. PALACE 4 enrolled DMARD-naive patients.
I: oral apremilast 20 mg or 30 mg twice daily. Stable background methotrexate, leflunomide or sulfasalazine was allowed in PALACE 1–3.
C: placebo, with 1:1:1 randomisation to placebo, 20 mg or 30 mg.
O: proportion of patients achieving ACR20 (20% improvement in modified American College of Rheumatology response criteria).
T: week 16.

PICO statement (PALACE 1): in adults with active psoriatic arthritis despite prior DMARD and/or biologic therapy, does oral apremilast 20 mg or 30 mg twice daily, compared with placebo, increase the proportion achieving ACR20 at week 16?

Trial	Patients	Population	ACR20 at week 16: 20 mg, 30 mg vs placebo
PALACE 1 (Kavanaugh 2014)	504	Prior DMARD and/or biologic	31%, 40% vs 19%
PALACE 2 (Cutolo 2016)	484	Prior DMARD and/or biologic	37.4%, 32.1% vs 18.9%
PALACE 3 (Edwards 2016)	505	Prior DMARD and/or biologic, with active skin psoriasis	28%, 41% vs 18%
PALACE 4 (Wells 2018)	527	DMARD-naive, apremilast as monotherapy	28.0%, 30.7% vs 15.9%

Design points: patients whose swollen and tender joint counts had not improved by at least 20% at week 16 entered early escape, and placebo patients were re-randomised to apremilast by week 24. The key secondary outcome in PALACE 3 and 4 was the change in HAQ-DI score at week 16.

The one thing to remember

Write your research question in PICO form before you write the title, the hypothesis or the objectives.

Worked examples and figures use teaching examples. They make no claim about the effect of any drug or marker named.

Sources: Dewey J. Logic: The Theory of Inquiry. 1938. Popper KR. The Logic of Scientific Discovery. 1959. Doran GT. There's a S.M.A.R.T. way to write management's goals and objectives. Management Review 1981;70(11):35–6. Richardson WS, Wilson MC, Nishikawa J, Hayward RS. ACP J Club 1995;123:A12–3 (doi:10.7326/ACPJC-1995-123-3-A12; PMID 7582737). Kerr NL. Pers Soc Psychol Rev 1998;2:196–217 (doi:10.1207/s15327957pspr0203_4). Hulley S, Cummings S, Browner W, et al. Designing Clinical Research, 3rd ed. 2007, as cited by Farrugia et al. Chalmers I, Glasziou P. Lancet 2009;374:86–9 (doi:10.1016/S0140-6736(09)60329-9). Farrugia P, Petrisor BA, Farrokhyar F, Bhandari M. Can J Surg 2010;53:278–81 (PMID 20646403). Cooke A, Smith D, Booth A. Qual Health Res 2012;22:1435–43 (doi:10.1177/1049732312452938). Morgan RL, Whaley P, Thayer KA, Schünemann HJ. Environ Int 2018;121:1027–31 (doi:10.1016/j.envint.2018.07.015). Kavanaugh A et al. Ann Rheum Dis 2014;73:1020–6 (doi:10.1136/annrheumdis-2013-205056). Cutolo M et al. J Rheumatol 2016;43:1724–34 (doi:10.3899/jrheum.151376). Edwards CJ et al. Ann Rheum Dis 2016;75:1065–73 (doi:10.1136/annrheumdis-2015-207963). Wells AF et al. Rheumatology (Oxford) 2018;57:1253–63 (doi:10.1093/rheumatology/key032). ICH E9 Statistical Principles for Clinical Trials, 1998. von Elm E et al. STROBE statement. Lancet 2007;370:1453–7 (doi:10.1016/S0140-6736(07)61602-X). Page MJ et al. PRISMA 2020 statement. BMJ 2021;372:n71 (doi:10.1136/bmj.n71). Chan AW et al. SPIRIT 2025 statement. BMJ 2025;389:e081477 (doi:10.1136/bmj-2024-081477). Hopewell S et al. CONSORT 2025 statement. BMJ 2025;389:e081123 (doi:10.1136/bmj-2024-081123).

Past papers

Every TNMGRMU question on research question and hypothesis

Each one is answered in this chapter. Tap a sitting to open the university's own question paper.

×3Research HypothesisPh.D. RM · Jan 2025 Ph.D. RM · Jan 2018 Ph.D. RM · Jul 2014
×1How do you formulate a research question? Discuss the PICO approach with examplesPh.D. RM · Jan 2025
×1PICO format in Research QuestionPh.D. RM · Jul 2025
×1Formulation of Research Question and PICO formatPh.D. RM · Jul 2024
×1a) Construction of Research question. b) Discuss various data collection methods with suitable examplesPh.D. RM · Jan 2023
×1Independent and Dependent variablePh.D. RM · Jan 2023
×1Formulation of Research QuestionPh.D. RM · Jul 2023
×1Criteria of a good research problemPh.D. RM · Jul 2023
×1Characteristics of good research objectivesPh.D. RM · Jul 2023
×1a) Design a research work to study to evaluate the effectiveness of ‘Music Therapy on relieving Anxiety among patients posted for CABG’. Also, formulate the objectives for the above problem statement. b) “A Research Problem well stated is Research half done” – Explain in detailPh.D. RM · Jul 2022
×1‘PICO’ statement of PALACE trialsDM Clinical Immunology and Rheumatology · Aug 2021
×1How should a hypothesis be framed?Ph.D. RM · Jul 2018
×1Frame a hypothesis and design a research work to test association of “Protein-A” in the progression of a Diabetes Mellitus to its complicationsPh.D. RM · Jul 2017
×1HypothesisPh.D. RM · Jan 2016
×1Variables & HypothesisPh.D. RM · Jan 2014

How does a doubt become a research question?

Is the question good enough to study?

Find the FINER check that each draft question fails

What does a well-built research question contain?

What if your study has no intervention?

Choose P, I, C and O for a scenario and watch the question sentence change

Which question format fits your study?

What is a research hypothesis, and how do you write one?

A good hypothesis is

Types of hypothesis

Sort statements into null, non-directional and directional hypotheses

Research question or research hypothesis: what is the difference?

How do you turn the question into objectives?

Characteristics of good objectives

Which variable is independent and which is dependent?

Other variables to list in your protocol

One doubt, from the ward to the protocol

Research question problems by specialty

Common mistaken beliefs

Five consequences of a poorly framed research question

Protocol returned by the committee

A literature search that finds too much

Objectives the data cannot answer

Questions you cannot answer at the viva

Research that does not help patients

History of the research question and hypothesis

John Dewey

Karl Popper

George T. Doran

Richardson, Wilson, Nishikawa and Hayward

Norbert Kerr

Iain Chalmers and Paul Glasziou

Cooke, Smith and Booth

Morgan, Whaley, Thayer and Schünemann

SPIRIT 2025 and CONSORT 2025

Research questions and objectives in guidelines

Writing the question, hypothesis and objectives, step by step

Five checks on any study's research question

Word surgery

“Hypothesis”

“Problem”

“Objective”

Problem

Question

Hypothesis

Variable

Objective

PICO

Model answer outlines

Formulating a research question, and the PICO format with examples

Criteria of a good research problem, and "A research problem well stated is research half done"

Characteristics of good research objectives

Design a study of music therapy for anxiety in patients posted for CABG, and formulate its objectives

Research hypothesis: definition, characteristics, types and how to frame it

Independent and dependent variables, and other types of variables

Frame a hypothesis and design a study to test the association of "Protein-A" with the progression of diabetes mellitus to its complications

PICO statement of the PALACE trials

Every TNMGRMU question on research question and hypothesis

Related Stateazy articles