Specific Aims Examples: Five Annotated Pages

Below are five complete Specific Aims examples: an NIH R01, an R21, a K career development award, an SBIR Phase I, and a weak version of the SBIR page for contrast. Each is written to fit on one page at NIH formatting and is annotated to show which sentences carry weight. All five are fictional and illustrative: the projects, investigators, and preliminary results are invented, though the disease statistics in the openings come from public sources. Real funded aims pages are rarely published, which is why worked examples are hard to find.

If you want the method behind these pages, paragraph by paragraph, read how to write Specific Aims first. This page assumes you know the structure and want to see it executed.

Example 1: NIH R01, chronic disease

Opening. Type 2 diabetes affects about 38 million adults in the United States, and a substantial share of patients started on first-line metformin do not reach glycemic control within twelve months. Clinicians cannot predict which patients will fail, so escalation happens reactively, after months of uncontrolled hyperglycemia. The determinants of early metformin non-response remain poorly characterized, and no validated predictive biomarker exists.

Objective, hypothesis, rationale. Our long-term goal is a point-of-care test that guides first-line therapy selection. The objective of this application is to validate and explain a blood-based predictor of early non-response. In a retrospective cohort of 412 patients, a three-gene expression signature in peripheral blood at treatment initiation distinguished responders from non-responders with an area under the curve of 0.79. We hypothesize that early metformin non-response is driven by variation in hepatic organic cation transporter expression, and that this variation is detectable in peripheral blood before treatment begins.

Aim 1. Validate the three-gene signature prospectively. We will enroll 600 treatment-naive patients and measure the signature at baseline, with glycemic control at twelve months as the primary endpoint. Expected outcome: a prospectively validated model with a pre-specified area-under-the-curve target of 0.75.

Aim 2. Determine the mechanism linking transporter expression to non-response. Using primary hepatocytes from 40 donors stratified by signature status, we will measure metformin uptake and downstream AMPK activation. Expected outcome: a mechanistic account of why the signature predicts, which regulators will expect before accepting any eventual test.

Aim 3. Establish the assay's performance in a point-of-care format. We will adapt the assay to a dried blood spot platform and compare it against the laboratory reference in 200 paired samples. Expected outcome: concordance sufficient to justify an implementation trial.

Impact. Together these aims deliver a validated, mechanistically grounded, deployable predictor of metformin non-response. Success would allow treatment escalation at initiation rather than after a year of uncontrolled disease and provides the evidence base for an implementation trial.

What is working

• The gap arrives in sentence three. No paragraph about the laboratory. Reviewers decide quickly whether a problem matters; spend those sentences on why it does.
• The hypothesis is bolded and could be wrong. Contrast the common weak form, "We will investigate the relationship between transporter expression and response," which cannot fail.
• Preliminary data come with numbers. An area under the curve of 0.79 in 412 patients establishes, in one clause, that this is not speculation.
• The aims are independent. Aim 2 is still worth funding if Aim 1's model underperforms, and Aim 3 is a technical translation that does not need Aim 2.
• Every aim has a pre-specified threshold. That tells the reviewer you know what success looks like before you start.

Example 2: NIH R21, exploratory

Opening. Adolescents discharged from psychiatric inpatient care face their highest risk of suicide attempt in the weeks after discharge, yet follow-up contact during this window is inconsistent and depends on families keeping outpatient appointments. Passive smartphone data (sleep timing, mobility, and phone use) change measurably before mood deterioration in adults, but whether these signals are acceptable to adolescents and families, and whether they carry any warning value in this population, is unknown.

Objective and rationale. The objective of this exploratory application is to determine whether passive sensing is feasible and potentially informative in the first eight weeks after adolescent psychiatric discharge. This is a high-risk question: adolescents may decline monitoring, data may be too sparse, or signals may not track clinical change. If feasible, it would open a new, low-burden route to post-discharge monitoring. Our team has run an adult passive-sensing study and has an established discharge pathway at a 40-bed adolescent unit.

Aim 1. Establish the feasibility and acceptability of passive sensing after discharge. We will offer a sensing app to 80 consecutive eligible adolescents at discharge and follow them for eight weeks. Benchmarks: at least 60 percent enrollment, at least 70 percent of enrolled participants providing data on four or more days a week, and acceptability ratings from adolescents and parents.

Aim 2. Explore whether sensing features track clinician-rated symptom change. We will compare weekly sensing features with standardized symptom ratings collected at weeks 2, 4, and 8. Expected outcome: effect-size estimates and the most promising candidate features to power a future R01; this aim is not designed to test a prediction model.

Impact. Meeting the Aim 1 benchmarks would show that passive monitoring is possible in this high-risk window, and Aim 2 would identify which signals merit a fully powered study.

What is working

• It says plainly that the idea is risky, and why it is worth the risk. That framing is what R21 reviewers are asked to reward; preliminary data are not required.
• Feasibility benchmarks are numeric. Enrollment and data-completeness thresholds give the reviewer a clear pass or fail for the main question.
• Aim 2 is honest about its limits. It promises effect sizes for a future study, not a validated predictor that 80 participants cannot deliver. Overclaiming in an R21 reads as poor judgment.
• Two aims fit two years. An exploratory award with three ambitious aims signals that the applicant has not scoped the risk.

Example 3: K award, career development

Opening. Chronic kidney disease affects an estimated one in seven U.S. adults, and most do not know they have it. Primary care clinics serving rural populations detect it late, in part because guideline-recommended urine albumin testing is ordered for a minority of eligible patients. Clinic-level interventions to raise testing exist, but none has been designed for small rural practices with no on-site laboratory, and the reasons testing is skipped in these settings have not been studied.

Objective and career goal. My long-term goal is to become an independent investigator in implementation science for chronic disease detection in rural primary care. The objective of this award is to identify why albumin testing is missed in rural clinics and to develop and pilot a clinic-ready intervention. This project requires skills I do not yet have, in qualitative methods, intervention design, and pragmatic trial design, which my training plan provides under the mentorship of established investigators in each area.

Aim 1. Identify the barriers to albumin testing in rural primary care. Through interviews with 30 clinicians and staff across six clinics and analysis of ordering data, we will map where and why testing is skipped. Expected outcome: a prioritized set of modifiable barriers. (Training: qualitative methods.)

Aim 2. Co-design an intervention with clinic teams. Using the Aim 1 barriers, we will run structured design sessions with two clinics to develop a point-of-care testing workflow. Expected outcome: a specified intervention and implementation manual. (Training: intervention design.)

Aim 3. Pilot the intervention in two clinics. We will measure testing rates for six months before and after implementation and assess feasibility and fidelity. Expected outcome: preliminary effect and feasibility data to support an R01 cluster-randomized trial. (Training: pragmatic trial design.)

Impact. This award will produce a deployable intervention and the training needed to test it at scale in my first R01.

What is working

• The research and the training are visibly linked. Each aim names the skill it builds, which answers the reviewer's central K-award question: will this person be independent at the end?
• The candidate states the skills gap directly. K reviewers expect a candidate who needs training; pretending otherwise weakens the case for the award.
• The aims build, but each delivers on its own. Aim 1's barrier map is publishable even if the pilot slips, and the design work in Aim 2 can proceed with partial Aim 1 results.
• The impact names the next award. An R01 is the expected destination of a K award, so say so.

Example 4: SBIR Phase I, medical device

Opening. Hospital-acquired pressure injuries affect more than 2.5 million U.S. patients each year at an estimated annual cost of up to $11 billion. Prevention depends on repositioning immobile patients every two hours, but compliance is low on understaffed units. Existing monitoring systems require wearable sensors that are dislodged, uncharged, or refused by patients, and none has achieved sustained adoption.

Innovation and objective. We have developed a pressure-sensing mattress overlay that infers patient position with no device attached to the patient, using a sparse sensor array and a classification model. Bench testing on a 30-subject dataset achieves 91 percent position-class accuracy. Our Phase I objective is to demonstrate that overlay-derived repositioning data match clinician observation closely enough to serve as a compliance record on a live ward.

Aim 1. Establish accuracy against direct observation in a clinical environment. In two 20-bed units over eight weeks, we will compare overlay position classification against timed observation by trained research staff across 1,200 observation events. Success criterion: at least 85 percent agreement, weighted kappa of at least 0.70.

Aim 2. Determine whether nursing staff can use the system without disrupting workflow. We will measure setup time, alert response, and false-alarm rate, and administer a structured usability instrument to 25 nurses. Success criterion: median setup under 4 minutes and fewer than 1 false alarm per bed per shift.

Commercial path. Meeting both criteria retires the primary technical and adoption risks and supports a Phase II multi-site study and a 510(k) submission. Two health systems have provided letters of intent to host Phase II sites.

What is working

• Two aims, one question. Phase I exists to answer whether the core idea works. A third aim usually means the applicant has not identified the real risk.
• Success criteria are numeric and set in advance. That is what makes it a feasibility study rather than an exploration.
• Aim 2 tests adoption, not technology. The opening says existing products failed on adoption, so the aims test the thing that actually kills products in this market.
• The commercial path is concrete: named regulatory route, letters of intent. How far this must go in the full application is covered in our guide to the SBIR commercialization plan, and what each stage must prove is in SBIR Phase I vs Phase II.

Example 5: the same device project, written weakly

Pressure injuries are a serious problem in hospitals and have been studied extensively. Our team has significant experience in sensor technology and machine learning, with over 30 publications in the field. We propose to develop an innovative new mattress system to address this important problem.

Aim 1. Develop the sensor array hardware. Aim 2. Train a machine learning model on data collected in Aim 1. Aim 3. Validate the model in a clinical setting using the system from Aims 1 and 2.

This work will have a major impact on patient safety.

Four things have gone wrong, all of them common:

1. It opens with the team, not the problem. Credentials belong in the biosketch; our annotated NIH biosketch example shows where they go.
2. There is no gap and no hypothesis. "Has been studied extensively" argues against novelty.
3. The aims are a chain. Aim 2 cannot start until Aim 1 finishes, and Aim 3 needs both. One hardware delay ends the project.
4. Nothing can be measured. No numbers, no success criteria, and "major impact" in place of a payoff.

Check your own page against the examples

• Does sentence one describe a problem, and is the gap stated by sentence four?
• Is there one central hypothesis or feasibility question, stated so it could be wrong?
• If Aim 1 returns a null result, is the rest still worth funding?
• Does every aim have an expected outcome or success criterion?
• Does the emphasis match the mechanism: preliminary data for an R01, risk and payoff for an R21, training for a K award, feasibility criteria for SBIR?
• Does the final paragraph say what becomes possible next, rather than restating the aims?

To draft against a scaffold, use our Specific Aims template. For the application around these pages, see our SBIR application guide and our overview of NIH grants and activity codes.