MedEd WorldDr Khalid · Internal Medicine
Clinical Research Hub

Asking the questions guidelines haven't settled yet

Alongside teaching, Dr Khalid Seidi is building a clinical research programme in internal medicine. The aim is simple: find real gaps in current evidence, study them carefully and honestly, and publish the results in peer-reviewed journals.

Current study Evidence review Study portfolio Research training How we work
In preparation · not yet started

Sodium correction rate in severe hyponatremia

A planned retrospective cohort study of adults admitted with a serum sodium of 120 mEq/L or less. It asks whether the speed of correction in the first 24 hours is linked to survival, once illness severity is properly accounted for.

Why this question

For years, guidelines have advised slow correction to avoid osmotic demyelination syndrome (ODS). Large recent studies and meta-analyses report that faster correction is associated with lower mortality [1, 2]. Other researchers argue that this signal may come from sicker patients correcting more slowly, not from the correction speed itself [3, 4]. No randomised trial has yet compared correction targets with mortality as the outcome.

Study question (PECO)

Population
Adults hospitalised with sodium ≤120 mEq/L
Exposure
Rise of more than 8 mEq/L in the first 24 h
Comparator
Rise of 8 mEq/L or less in the first 24 h
Outcomes
30-day mortality (primary); 90-day and in-hospital mortality, ODS, length of stay

What this study adds

Big database studies can count thousands of patients, but they usually cannot see the details that describe how unwell someone is. A careful chart review can. This study plans to record serum albumin, CRP, frailty, nutritional state and goals of care, alongside the cause of hyponatremia, so that the effect of correction speed can be separated from the severity of illness. The main analysis uses propensity-score weighting, and a pre-specified subgroup looks at patients with SIADH, where correction speed depends more on treatment than on the underlying disease.

Design follows the STROBE and RECORD reporting standards. Ethics committee approval will be obtained before any patient record is accessed.

  1. Step 1Gap identified and literature reviewed
  2. Step 2Protocol drafted and revised (v0.2)
  3. Step 3 · nextEthics committee submission
  4. Step 4Data collection
  5. Step 5Analysis and peer-reviewed publication
Evidence review · updated October 2026

Where the evidence stands

A structured review of the published literature up to October 2026. Every source below was checked in PubMed, and the summaries are written in our own words from each article's abstract.

Meta-analysisSizeMortality with faster correctionODS with faster correction
See et al., 2024 [5]7 cohorts, 6,032 patientsLower (RR 0.51)Higher (RR 3.91)
Ayus et al., 2025 [1]16 cohorts, 11,811 patientsLower (OR 0.67 vs slow)No significant increase
Schneider et al., 2026 [6]43 studiesLower (OR 0.67); certainty very lowHigher (OR 4.23)
Matrisch et al., 2026 [7]11 cohorts, 27,672 patientsLower (OR 0.40 at 8 mEq/L)Higher (OR 3.96)

What the guidelines say

The European guideline (2014) and the US expert panel (2013) still recommend limiting the rise in sodium in the first 24 hours [8, 9]. No major guideline has loosened these limits yet.

The confounding question

Detailed chart-review studies from 2026 found that albumin, CRP, comorbidity and cancer predicted death far better than correction speed, and one found no effect of correction speed on mortality [3, 4].

Who gets ODS

ODS is uncommon, but it clusters in patients with alcohol use disorder, malnutrition, low potassium or phosphate, and liver disease, and can occur even within the usual limits [10, 11].

Bottom line: the observational data consistently link faster correction with lower mortality, but all of it is observational, the certainty is low, and the risk of ODS appears real. Until trials report, careful correction within guideline limits remains standard practice. This page summarises research; it is not treatment advice.

Selected references

  1. Ayus JC, Moritz ML, Fuentes NA, et al. Correction rates and clinical outcomes in hospitalized adults with severe hyponatremia: a systematic review and meta-analysis. JAMA Intern Med. 2025;185(1):38-51. doi:10.1001/jamainternmed.2024.5981
  2. Mark DG, Alavi M, Nugent JR, Reed ME, et al. Sodium correction rates and associated outcomes among patients with severe hyponatremia: a retrospective cohort study. Ann Intern Med. 2026;179(3):330-339. doi:10.7326/ANNALS-25-03676
  3. Nagase K, Imaizumi T, Yamamori A, et al. Key predictors of mortality in profound hyponatremia beyond the correction rate. Clin Kidney J. 2026;19(8):sfag219. doi:10.1093/ckj/sfag219
  4. Burst V, Rabii R, Peto-Madew J, Persigehl T, et al. Severe hyponatremia in the emergency department: incidence of cerebral edema and risk of osmotic demyelination syndrome. Acad Emerg Med. 2026;33(1):e70158. doi:10.1111/acem.70158
  5. See XY, Chang YC, Peng CY, et al. Rate of sodium correction and osmotic demyelination syndrome in severe hyponatremia: a meta-analysis. J Crit Care Med (Targu Mures). 2024;10(3):209-212. doi:10.2478/jccm-2024-0030
  6. Schneider AR, Radu LM, Tang KL, et al. Outcomes with sodium overcorrection in chronic hyponatremia: a systematic review and meta-analysis. J Gen Intern Med. 2026;41(8):2277-2288. doi:10.1007/s11606-025-10155-y
  7. Matrisch L, Rau Y, Graßhoff L, Nitschke M, et al. On the optimal sodium correction rate in hyponatraemia and clinical outcome: a meta-analysis. Emerg Med J. 2026;43(10):647-654. doi:10.1136/emermed-2025-215189
  8. Spasovski G, Vanholder R, Allolio B, et al. Clinical practice guideline on diagnosis and treatment of hyponatraemia. Eur J Endocrinol. 2014;170(3):G1-47. doi:10.1530/EJE-13-1020
  9. Verbalis JG, Goldsmith SR, Greenberg A, et al. Diagnosis, evaluation, and treatment of hyponatremia: expert panel recommendations. Am J Med. 2013;126(10 Suppl 1):S1-42. doi:10.1016/j.amjmed.2013.07.006
  10. Seethapathy H, Zhao S, Ouyang T, et al. Severe hyponatremia correction, mortality, and central pontine myelinolysis. NEJM Evid. 2023;2(10):EVIDoa2300107. doi:10.1056/EVIDoa2300107
  11. Tandukar S, Sterns RH, Rondon-Berrios H. Osmotic demyelination syndrome following correction of hyponatremia by ≤10 mEq/L per day. Kidney360. 2021;2(9):1415-1423. doi:10.34067/KID.0004402021
Internal medicine · 4 new studies · October 2026

Study portfolio: the next four questions

Alongside the hyponatremia study, we have prepared four internal medicine studies. Each starts from a question that current guidelines leave open, has a named protocol with primary and secondary objectives, and uses a different study design. Every reference was checked in PubMed or ClinicalTrials.gov in October 2026.

StudyQuestionDesignReporting
MET-B12Is routine vitamin B12 screening in metformin users still justified?Cross-sectionalSTROBE
DOSE-AFHow often are DOACs under-dosed at discharge from internal medicine, and does it matter?Retrospective cohortSTROBE, RECORD
STEROID-CAPHow are corticosteroids used in non-severe pneumonia on medical wards, and at what cost?Retrospective cohort and auditSTROBE, RECORD
STOP-PPIDoes a structured bundle at discharge help patients stop unneeded PPIs?Pragmatic randomised trialSPIRIT, CONSORT 2025
STUDY 1 · MET-B12

Vitamin B12 screening in metformin users

Cross-sectionalOutpatient clinicOne blood sample

The gap. The ADA Standards of Care 2026 still advise checking B12 periodically in people on long-term metformin. A 2026 Danish study that used a strict definition (low B12 and raised methylmalonic acid) found deficiency in only 2.3% of metformin users versus 1.3% of non-users, and questioned whether routine screening is needed. One study from one country cannot settle a worldwide recommendation.

Primary objective. Prevalence of biochemically confirmed B12 deficiency in metformin users and non-users with type 2 diabetes, and the adjusted prevalence ratio.

Secondary objectives and references
  • Dose and duration response
  • Prevalence under older, B12-only definitions
  • Link between confirmed deficiency and peripheral neuropathy
  • How often raised MMA reflects kidney function rather than B12 deficiency

ADA Professional Practice Committee, Diabetes Care 2026, PMID 41358900, doi · Laugesen et al., Endocr Pract 2026, PMID 42567282, doi

STUDY 2 · DOSE-AF

Off-label DOAC dosing in atrial fibrillation

Retrospective cohort12-month follow-upChart review

The gap. US (2023) and European (2024) guidelines ask for DOAC doses that follow the label, yet about one in five patients is under-dosed in practice. Whether this harms patients is unclear: studies disagree, many are underpowered, and it is hard to separate the effect of the low dose from the frailty that prompts it. Most data come from cardiology, not general medical wards.

Primary objective. Proportion of patients with AF discharged from internal medicine on an off-label DOAC dose, using Cockcroft-Gault creatinine clearance and the label criteria.

Secondary objectives and references
  • Patient and prescriber factors linked to under-dosing
  • 12-month stroke or systemic embolism, major bleeding and death
  • How many dosing decisions change when eGFR replaces Cockcroft-Gault

Joglar et al., Circulation 2024, PMID 38033089, doi · Van Gelder et al., Eur Heart J 2024, PMID 39210723, doi · Mongkhon et al., Future Sci OA 2026, PMID 42544439, doi · Cho et al., Heart 2026, PMID 40930594, doi · Pereira et al., Int J Cardiol 2022, PMID 35513121, doi

STUDY 3 · STEROID-CAP

Corticosteroids in non-severe pneumonia

Retrospective cohortGuideline auditMedical wards

The gap. A Cochrane review found that corticosteroids lowered mortality in severe pneumonia but not in non-severe pneumonia, and they raise blood glucose. Recent ICU trials in severe pneumonia point in different directions. On general medical wards, where most pneumonia is treated, how often steroids are given without another reason (such as COPD or wheeze), and what that costs patients, is poorly described.

Primary objective. Proportion of patients with non-severe community-acquired pneumonia who receive systemic corticosteroids within 24 hours, overall, without an alternative indication, and by ward.

Secondary objectives and references
  • Effect of early steroids on hyperglycaemia within 7 days, overall and by diabetes status
  • Time to clinical stability, 30-day readmission, secondary infection and 30-day death
  • Whether the association with death differs between severe and non-severe pneumonia

Stern et al., Cochrane 2017, PMID 29236286, doi · Jones et al. (ATS guideline), Am J Respir Crit Care Med 2026, PMID 40679934, doi · Dequin et al., N Engl J Med 2023, PMID 36942789, doi · REMAP-CAP Investigators, Intensive Care Med 2025, PMID 40261382, doi · Edland & Waterer, Curr Opin Pulm Med 2026, PMID 41603097, doi · Michels et al., Lancet Respir Med 2026, PMID 41856148, doi

STUDY 4 · STOP-PPI

Stopping unneeded PPIs at hospital discharge

Randomised trialOpen-label, pragmatic12-week outcome

The gap. Many people take a proton pump inhibitor without a lasting reason. Deprescribing trials so far were mostly in outpatients or primary care. In older multimorbid inpatients, only about one in six had their PPI stopped at discharge with usual care. No randomised trial has tested a structured deprescribing bundle at the point of discharge from internal medicine.

Primary objective. Whether a deprescribing bundle (taper, leaflet, letter to the family doctor, pharmacist call at 2 weeks) increases the proportion of patients off regular PPI at 12 weeks, compared with usual care.

Secondary objectives and references
  • Reflux symptoms at 12 weeks
  • PPI restarted by 6 months
  • Safety: upper GI bleeding and readmission by 6 months

Boghossian et al., Cochrane 2017, PMID 28301676, doi · Farrell et al., Can Fam Physician 2017, PMID 28500192 · Aubert et al., CMAJ Open 2023, PMID 36854455, doi · Koggel et al., United European Gastroenterol J 2026, PMID 42894466, doi · Ongoing: NCT06129474, NCT05629143

Status and practice data. These are planned studies; none has started recruiting, and each needs research ethics approval first (and trial registration for STOP-PPI). For training, each study also comes with a fully synthetic practice dataset in the Virtual Hospital below. Those practice numbers are invented and are never presented as findings.
Training material · synthetic data

The Virtual Hospital: learning research by doing it

Before touching real patient records, every step of the study was rehearsed on a computer-generated hospital. The same package is a practical way for junior doctors to learn how a clinical study is built, from approvals to a finished paper.

Important: this is not real research data. The Virtual Hospital uses invented, computer-generated patients. Its numbers, figures and practice paper exist only for training. They are not study findings and must never be presented, submitted or cited as real results.

1 · Approvals

Which bodies must approve a chart-review study, a worked ethics application, a consent waiver and the questions a committee is likely to ask.

2 · Data

A synthetic patient dataset with a data dictionary, so you can build Table 1 and run the main analysis yourself in Excel, SPSS or R.

3 · Analysis

A reference analysis with propensity-score weighting and spline plots to check your own numbers against.

4 · Writing

A full practice paper in STROBE format, including bias, ethics and AI-disclosure sections, with the STROBE checklist.

5 · The hidden answer

The data were built with a known "true" effect, so learners can see how confounding can mislead an analysis and how better adjustment brings the estimate back towards the truth.

6 · Real literature

The PubMed-verified evidence review above, which is real and reusable, unlike the synthetic data.

How we work

  • Ethics committee approval before any patient data are accessed; all data de-identified.
  • Reporting to recognised standards (STROBE, RECORD) and authorship by ICMJE criteria.
  • Every reference checked in PubMed or ClinicalTrials.gov, with a DOI or PMID.
  • Our own words: no copied or paraphrased text from other authors.
  • Open disclosure of any AI assistance, as journals now require.
  • Synthetic or training data always labelled and kept apart from real results.

Interested in collaborating?

Clinicians, statisticians and students who would like to join a study or learn with the training package are welcome to get in touch.

Message on WhatsApp