Read the Latest from Our Blog​

Explore practical insights and research experiences all in one place.

Click the + to start reading.

Starting Your Research Journey: The Roadmap From Idea to Submission

  • Danyal Bakht
  • Muhammad Arham

Forest plot is the visual summary of a meta-analysis, and understanding its parts is not optional if the underlying evidence is to be interpreted correctly. Each horizontal line on the plot represents one included study. The point on that line marks the study’s own effect estimate, whether that is an odds ratio, a risk ratio, a mean difference, or a hazard ratio, and the length of the line marks its confidence interval. A short line reflects a precise estimate, usually from a larger sample. A long line reflects an imprecise one, usually from a smaller trial, and this is often the first thing worth noticing before reading any numbers at all.

The size of the box drawn at each point is not decorative. It represents the weight that study contributes to the pooled estimate, and weight is generally driven by sample size and the precision of the study’s own result. A visually large box from a modestly sized headline trial can look reassuring, yet a forest plot correctly gives more visual weight to studies that reduce statistical uncertainty rather than to studies that are simply well known. Learning to notice which boxes are large, rather than assuming importance from a familiar author or journal name, is part of reading the plot honestly.

Beneath all the individual study lines sits a diamond, and this is the pooled estimate, the actual result of the meta-analysis. Its horizontal position along the axis reflects the combined effect, and its width reflects the confidence interval around that combined effect. A narrow diamond suggests a precise summary; a wide one suggests that pooling the studies has not resolved much uncertainty despite the added numbers.Running vertically through the plot is the line of no effect, positioned at one for ratio measures or at zero for difference measures. Any study whose confidence interval crosses this line has not, on its own, reached statistical significance. What matters more, though, is where the pooled diamond sits relative to this line, since a meta-analysis exists precisely to answer what individual small trials often cannot.

None of this is complete without checking heterogeneity, typically reported alongside the plot as an I squared value. A high I² indicates that the individual studies disagree with each other more than chance alone would explain, and a pooled estimate calculated across genuinely disagreeing studies deserves caution regardless of how narrow its diamond appears. This is where a forest plot connects directly back to how the PICO was defined at the outset, since much of the disagreement visible in a plot traces back to populations, interventions, or outcomes that were never quite the same across the pooled studies to begin with.Reading a forest plot well means resisting the pull toward the diamond alone. The individual lines, their lengths, their weights, and their scatter around the vertical line together tell a fuller and often more honest story than the summary point by itself.

  • Maaz Amir
  • Muhammad Arham

Reporting guidelines exist to standardize how research is described, so that readers can judge its quality and, ideally, replicate it. Each guideline is matched to a specific study design, and using the wrong one, or none at all, is one of the most avoidable mistakes in academic writing.

For randomized controlled trials, the standard is CONSORT, which specifies a 25-item checklist covering everything from randomization procedures to participant flow diagrams. For observational studies, cohort, case-control, and cross-sectional designs, the guideline is STROBE. Systematic reviews and meta-analyses follow PRISMA, which now includes specific extensions for network meta-analyses and scoping reviews. Diagnostic accuracy studies use STARD, while studies developing or validating prediction models follow TRIPOD. Qualitative research has its own standard in COREQ, and case reports follow CARE.

Choosing the right guideline starts with correctly identifying your study design, which sounds obvious but often isn’t. Researchers sometimes misclassify a retrospective cohort study as a case series, or treat a diagnostic accuracy study like a standard observational study, and end up applying the wrong checklist entirely.

The EQUATOR Network is the most reliable resource for this. It maintains a searchable library of reporting guidelines organized by study type, and most journals now require authors to submit a completed checklist alongside their manuscript. Skipping this step doesn’t just risk a desk rejection — it signals to reviewers that the methodology itself may not have been rigorously planned.

A practical habit worth building: consult the relevant reporting guideline before you start writing, not after. CONSORT and STROBE, for instance, are useful not only for reporting but for confirming that you collected and documented the information you’ll actually need to report. Retrofitting a guideline onto a completed study often exposes gaps that can’t be fixed retroactively.

Reporting guidelines aren’t bureaucratic box-ticking. They exist because inconsistent, incomplete reporting has historically made it difficult to trust or reproduce published research. Before you submit your next manuscript, ask yourself: have I matched my study design to the correct guideline, and does my write-up actually meet it?

  • Maaz Amir
  • Muhammad Arham

According to ICMJE, authorship should be based on four criteria, and a person needs to meet all four to qualify. First, they must have made a substantial contribution to the conception or design of the work, or to the acquisition, analysis, or interpretation of data. Second, they must have drafted the work or revised it critically for important intellectual content. Third, they must have given final approval of the version to be published. Fourth, they must agree to be accountable for all aspects of the work, ensuring that questions about accuracy or integrity are appropriately investigated and resolved.

Why does this matter so much? Because authorship is not just a credit line — it is a claim of responsibility. If a paper is later found to contain errors, fabricated data, or ethical violations, every listed author is expected to answer for it. That’s why simply providing funding, supervising a lab, collecting data without further involvement, or offering general administrative support does not meet the bar for authorship. These contributions matter, but they belong in the acknowledgments, not the byline.

A common mistake among early-career researchers is assuming that seniority automatically earns a spot on the author list. It does not. A lab head who did not contribute intellectually to the analysis or writing has no ICMJE-based claim to authorship, regardless of their position. On the other side, a junior researcher who designed the methodology and wrote significant portions of the manuscript has earned it, regardless of title.

Disputes over authorship are one of the most common sources of conflict in academic teams, and they are largely preventable. The best practice is to discuss authorship early, ideally before data collection begins, and revisit it as the project evolves. Journals increasingly require a contribution statement clarifying who did what, which makes vague or inflated authorship claims harder to sustain.

Understanding ICMJE criteria protects both the integrity of the scientific record and the researchers who put their names on it. Before you add someone to a manuscript, or accept a spot yourself, ask: does this contribution meet all four criteria? If not, acknowledgment is the honest choice, not authorship.

  • Danyal Bakht
  • Muhammad Arham

Every meta-analysis begins with a question, and that question is only as sound as its PICO. Population, Intervention, Comparison, and Outcome are not administrative labels added after the idea is formed. They are the criteria used to decide, before any search is run, which studies may honestly be combined and which may not. Procuring a workable PICO is therefore the actual starting task of a meta-analysis, not a formality that follows it.

Building one begins with the population, and this is also where most early attempts fail. A population statement needs enough precision that two researchers reading it would select the same trials. Consider a proposed review on venous thromboembolism following hepatic resection. Left broad, the term “liver surgery patients” would pull in living donor hepatectomies, oncological resections, and cirrhotic liver operations together. These groups carry different baseline clotting risk and different operative exposure, so averaging them produces a number that is statistically calculable yet clinically meaningless. Narrowing the population to adults undergoing elective resection for malignancy is what makes the eventual pooled estimate usable.

Intervention and comparison must be defined with the same discipline. Loose terminology tends to surface only once data extraction has already started, by which point months of work may need to be redone. The literature on deferred versus delayed stenting in ST elevation MI shows this clearly. Unless both terms are tied to explicit time windows, one study calling anything past twenty four hours delayed and another using forty eight hours as its cutoff, the two cannot be merged under a single label without quietly changing what is being measured.

Outcome is usually the most neglected element, and often the one a statistician questions first. A review targeting something as vague as clinical improvement in epiglottitis management will struggle to pool anything, since one trial may define improvement as resolution of stridor and another as readiness for discharge. These are different endpoints regardless of the shared wording, and combining them is not defensible. Specifying the outcome precisely, for example time to symptom resolution in hours, before the search begins is what keeps the pooled result honest.

A carefully built PICO also answers a question researchers often skip until too late, which is whether the topic is genuinely available. Two teams starting from what looks like the same broad subject can end up with entirely separate reviews simply because one defined the intervention window differently. Subscribing to table of contents alerts from relevant journals, most offer this without charge, along with saved PubMed search alerts built around specific PICO terms, is a practical way to notice a gap before it closes.

A meta-analysis with an unclear population produces a result describing no one in particular. One with an inconsistent outcome produces a result measuring nothing consistent across its own studies. Time spent settling these four elements before screening begins is rarely wasted, and skipping it is usually paid for later, during revision.

  • Danyal Bakht
  • Muhammad Arham

The answer to all three questions sits in the same place: the search strategy built before a single article is ever opened.

What PubMed Actually Offers: PubMed is a free database maintained by the National Center for Biotechnology Information, built largely on MEDLINE along with additional life science and biomedical literature. It is usually the first stop for clinical and health research because of one feature in particular: every article indexed in MEDLINE is tagged with standardized subject headings, which makes searching far more precise than a plain keyword search on a general search engine. Beyond the search bar itself, PubMed offers an advanced search builder that lets each concept be searched separately before combining them, a set of filters that narrow results by date, species, language, or article type, and the ability to save a search and receive alerts whenever new matching articles are indexed. None of this is decorative. Each feature exists because searching the literature well is genuinely difficult, and PubMed was built around that difficulty rather than around simplicity alone.

Why the Search Strategy Matters So Much: A search strategy sits between two opposite risks, and a good one is built to avoid both at once. Search too broadly, and the review is buried under thousands of results, most of them irrelevant, turning screening into an exercise that eats months without adding value. Search too narrowly, and the strategy quietly drops eligible studies before a reviewer ever sees them, which is far more dangerous, because a missed study does not announce itself. It simply never appears, and the final review reports a conclusion built on incomplete evidence without anyone realizing it. The goal, then, is not simply to get a small number of results. It is to get a number that is manageable and complete at the same time, and reaching that balance is what separates a well built strategy from a rushed one.

Building the Strategy: Concepts, Then Boolean Logic: A search strategy is built one concept at a time, not as a single long sentence. The first step is to break the research question into its separate ideas, usually following the PICO structure, and list every reasonable synonym, spelling variation, and abbreviation for each concept. Only once each concept has its own list of terms does the strategy get combined.

This is where the three Boolean operators do their work. OR is used within a single concept, connecting all its synonyms together, since an article only needs to mention one of them to be relevant. AND is used between different concepts, since a relevant article must touch on all of them at once, and this is the operator that narrows a broad set of results down to something usable. NOT is the operator to use with real caution. It excludes any article containing a given term, even if that same article also contains everything else you are looking for, which means a NOT applied carelessly can quietly remove eligible studies alongside the irrelevant ones. It is generally safer to exclude unwanted study types through filters than through a NOT statement inside the search itself.

Using Filters to Ease the Load: Once the core strategy is built with OR and AND, filters are what fine tune the result count without touching the underlying logic. PubMed allows filtering by publication date, article type such as randomized controlled trial or systematic review, species, language, and age group, among others. Applying filters after building the concept based strategy, rather than baking these restrictions into the search terms themselves, keeps the strategy transparent and easy to justify later in a manuscript’s methods section.

Using MeSH Terms to Sharpen the Search: Medical Subject Headings, or MeSH terms, are the controlled vocabulary that MEDLINE uses to index every article, regardless of the exact wording the original authors used. Searching by MeSH term catches articles that describe the same concept in different language, which a plain keyword search would miss entirely. The MeSH database also allows a term to be exploded, which automatically includes its narrower, more specific subheadings in the same search. That said, MeSH indexing takes time, so very recent articles may not yet carry the correct heading. For this reason, a well built strategy usually combines MeSH terms with free text keywords for the same concept, connected by OR, so that both older, properly indexed articles and newer, not yet indexed ones are captured together.

A Worked Example: SGLT2 Inhibitors Versus DPP4 Inhibitors in Heart Failure With Diabetes: Take a comparative clinical question: in adults with type 2 diabetes and heart failure, how do SGLT2 inhibitors compare with DPP4 inhibitors for cardiovascular outcomes? Breaking this into PICO gives a population of adults with type 2 diabetes and heart failure, an intervention of SGLT2 inhibitors, a comparator of DPP4 inhibitors, and an outcome centered on heart failure hospitalization and cardiovascular events.

Each concept then gets its own list of terms, mixing MeSH headings with free text synonyms:

Concept 1 (Population)

(“Diabetes Mellitus, Type 2″[Mesh] OR “type 2 diabetes” OR “T2DM”)

AND (“Heart Failure”[Mesh] OR “heart failure” OR “HF”)

Concept 2 (Intervention)

“Sodium-Glucose Transporter 2 Inhibitors” OR “SGLT2 inhibitor”

OR empagliflozin OR dapagliflozin OR canagliflozin

Concept 3 (Comparator)

“Dipeptidyl-Peptidase IV Inhibitors”[Mesh] OR “DPP4 inhibitor”

OR sitagliptin OR saxagliptin OR linagliptin

Concept 4 (Outcome)

“Heart Failure”[Mesh] OR “hospitalization for heart failure”

OR “cardiovascular events” OR “cardiovascular mortality”

These four concept blocks are then joined with AND, since a genuinely comparative article needs the population, both drug classes, and the outcome present together:

((“Diabetes Mellitus, Type 2″[Mesh] OR “type 2 diabetes” OR “T2DM”) AND (“Heart Failure”[Mesh] OR “heart failure” OR “HF”)) AND (“Sodium-Glucose Transporter 2 Inhibitors”[Mesh] OR “SGLT2 inhibitor” OR empagliflozin OR dapagliflozin OR canagliflozin) AND (“Dipeptidyl-Peptidase IV Inhibitors”[Mesh] OR “DPP4 inhibitor” OR sitagliptin OR saxagliptin OR linagliptin) AND (“Heart Failure”[Mesh] OR “hospitalization for heart failure” OR “cardiovascular events” OR “cardiovascular mortality”)

From here, filters narrow the result set without altering the logic above: limiting to humans, English language, and randomized controlled trials or comparative observational studies, depending on what the review is designed to include. Requiring both drug classes directly inside the search, rather than searching each one separately and merging later, is what keeps this strategy focused on head to head comparisons rather than single arm studies of either drug alone.

The same process used here, listing concepts, gathering synonyms, combining MeSH with free text, joining with OR within a concept and AND between concepts, then filtering rather than excluding, applies regardless of the clinical topic. The specific terms change every time. The underlying method does not.

  • Maaz Amir
  • Muhammad Arham

Bias in research isn’t the same as dishonesty. Most bias is unintentional, built into how a study was designed, who was included, or how outcomes were measured. Risk of Bias tools exist to make these weaknesses visible rather than hidden behind statistical significance.

There are several recognized categories of bias worth knowing. Selection bias occurs when the way participants are chosen or assigned to groups systematically favors a particular outcome, such as non-random allocation in a trial. Performance bias arises when participants or those delivering an intervention know which group they’re in, which can influence behavior or reporting. Detection bias happens when outcome assessors are aware of group assignment and this awareness affects how outcomes are measured or interpreted. Attrition bias reflects differences in how participants drop out of comparison groups, which can skew final results. Reporting bias occurs when only some outcomes, usually the favorable ones, are published or highlighted.

Tools like the Cochrane Risk of Bias tool for randomized trials, ROBINS-I for non-randomized studies, and QUADAS-2 for diagnostic accuracy studies give reviewers a structured way to rate these domains as low, high, or unclear risk. This matters enormously in evidence synthesis. A meta-analysis that pools a well-conducted randomized trial with a poorly controlled observational study, without accounting for differing risk of bias, can produce a misleading pooled estimate.

For researchers, understanding RoB isn’t just about critiquing other people’s work. It should shape how you design your own study from the outset. Are you randomizing properly? Is your outcome assessment blinded? Are you accounting for dropouts in your analysis? Thinking about Risk of Bias at the design stage prevents problems that are far harder to fix after data collection.

For readers of research, especially clinicians and policymakers, Risk of Bias assessment is what separates evidence you can act on from evidence that needs to be treated with caution. A study with a strong result but high risk of bias should carry less weight in decision-making than a modest result from a rigorously designed study.

Not all studies are created equal, and Risk of Bias is the framework that tells you why. Before accepting a study’s conclusions, ask what its risk of bias assessment actually shows.

  • Maaz Amir
  • Muhammad Arham

Predatory journals exist to collect publication fees while bypassing genuine peer review. They mimic the appearance of legitimate academic publishing but skip the substance. Recognizing them requires looking past the surface.

Start with the review timeline. Legitimate peer review, even when efficient, takes weeks at minimum for most fields, because it involves recruiting qualified reviewers and allowing time for genuine critique. A journal promising publication within 48 hours to a week is not conducting real peer review, regardless of what it claims.

Check the editorial board carefully. Predatory journals often list researchers as editors without their knowledge or consent. Search for the named editors independently; if they have no verifiable connection to the journal, or if the same names appear across dozens of unrelated journals from the same publisher, that’s a red flag.

Look at the fee structure. Legitimate open-access journals are transparent about author processing charges upfront, before submission. Predatory journals often hide fees until after a manuscript is accepted, when authors feel pressured to pay rather than withdraw.

Verify indexing claims. Predatory journals frequently claim to be indexed in databases like PubMed, Scopus, or Web of Science when they are not. Don’t take the journal’s word for it — search the database directly to confirm.

Examine the journal’s scope. A legitimate journal has a defined subject focus. Predatory journals often claim to cover an implausibly broad range of disciplines, from oncology to civil engineering to linguistics, because their goal is maximizing submissions, not maintaining expertise.

Useful tools for verification include the Directory of Open Access Journals, which vets journals against clear criteria, and Think.Check.Submit, a checklist designed specifically to help researchers evaluate journal legitimacy before submitting.

The consequences of publishing in a predatory journal go beyond wasted fees. It can undermine your credibility, make your work invisible to legitimate citation networks, and in some cases affect funding or promotion decisions if the publication is later flagged. Before you submit anywhere, ask: can I independently verify this journal’s editorial board, indexing, and review process? If the answer is no, that uncertainty is reason enough to look elsewhere.

Want to Contribute? 

We welcome original contributions and invite you to submit your work for consideration.

Basic requirements
  • File format: Submit your manuscript as an editable .docx file.
  • Font: 12-point font.
  • Line spacing: Double-spaced.
  • Word count: Manuscripts should be fewer than 1,000 words.
  • References: Use Vancouver referencing style and ensure all references are accurate and complete.
  • Figures/Images: If your submission includes an image, please provide:
    • The image in a suitable high-resolution format.
    • The corresponding PDF/source file, if applicable.
    • The image separately from the main manuscript.
  • Originality: Submissions should be original and should not contain plagiarized or previously published material.
  • Citations: All statements requiring references should be appropriately cited, with references numbered consecutively in the order of appearance.
  • Permissions: Authors are responsible for obtaining permission for any copyrighted material, images, or figures that are not their own.
  • Language: Blogs should be written in clear, concise English and carefully proofread before submission.
How to submit
  • Please send your completed blog and any accompanying figures or files to: archofficial.info@gmail.com​
  • We look forward to reading your work and welcoming your contributions.
Scroll to Top