Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Contents

Academy · Research Data, Analytics & AI · SD from Interquartile Range

Estimation of the Standard Deviation from the Interquartile Range Alone

When a source paper reports only the first and third quartiles (Q1 and Q3), without the median, minimum or maximum, the standard deviation can still be approximated by dividing the interquartile range by 1.35, the distance between those quartiles in a standard normal distribution.

Pubrica AcademyWritten by a biostatistics reviewer Reviewed bya meta-analysis methodologist Published10 October 2026 Last checked10 October 2026 Reading time7 minutes Sources10 peer-reviewed

The interquartile range (IQR), the distance between the first quartile (Q1) and the third quartile (Q3), is one of the most commonly reported measures of spread in papers that present skewed or ordinal data as medians rather than means. When a meta-analyst requires a standard deviation to pool such a study with others reporting mean and SD directly, and the source paper gives only Q1 and Q3, with no median, minimum, or maximum, the available conversion methods reduce to a single core relationship: under an assumed normal distribution, Q3 minus Q1 corresponds to a fixed, calculable number of standard deviations.

Learning objectives

  • The constant that links the IQR to the SD under a normal-distribution assumption, and its derivation
  • How the estimate's sample-size dependence works, and why Wan and colleagues refined the simple constant for small samples
  • The specific situations in which this estimate is unreliable, regardless of sample size

Common misconceptions

  • “Dividing the IQR by 1.35 is accurate for any dataset, any sample size.” The 1.35 constant is derived from the standard normal distribution and becomes progressively less reliable as sample size decreases or as the underlying data depart from normality, which is why Wan and colleagues published a small-sample correction rather than treating the simple division as sufficient.

The relationship between the interquartile range (IQR) and standard deviation (SD) under normality

For a standard normal distribution, the 75th percentile lies at a z-value of approximately 0.6745 and the 25th percentile at approximately −0.6745, so the distance between them, the interquartile range in standard-deviation units, is approximately 1.34924. This single value is the entire basis for the commonly used approximation:

SD ≈ (Q3 − Q1) / 1.35

The approximation is useful because many papers presenting the median and IQR do not also report a minimum or maximum, which precludes the range-based and five-number-summary methods that require those endpoints. The formula requires only Q1, Q3, and the assumption that the underlying data, although summarised with a median rather than a mean, are approximately normally distributed12.

Small-sample correction of the constant

The 1.349 constant is a population-level property of the normal distribution, but Q1 and Q3 reported from a small sample are only estimates of the true 25th and 75th percentiles, and this estimation error is larger in smaller samples2. Wan and colleagues addressed this directly by deriving sample-size-specific divisors for the case in which only Q1, Q3, and the sample size n are available, refining the fixed 1.35 constant into an adjustment that depends on the number of observations that produced the reported quartiles2. The refined estimator converges to the simple IQR/1.35 formula as n increases but departs from it meaningfully when n is small, which is the range in which clinical subgroup analyses and smaller pilot studies typically fall.

Luo and colleagues extended this line of work further, incorporating the mid-quartile range, the average of Q1 and Q3, as an additional input when it is reported, which measurably improves the accuracy of the derived mean alongside the SD estimate3. For meta-analysts without access to statistical software implementing these refinements, the simple IQR/1.35 approximation remains a reasonable default for moderate-to-large samples, with the explicit caveat that it should be flagged as an approximation in any sensitivity analysis.

Conditions under which the IQR-only estimate fails irrespective of sample size

The normality assumption, rather than sample size, is the principal limitation of this method. If a study reports the median and IQR because the outcome is heavily skewed (hospital length of stay, biomarker concentrations, and time-to-event data are common examples), the reason the authors chose a median-based summary is also the reason the normal-distribution constant will misestimate the true SD15. In right-skewed data, the true SD is typically larger than IQR/1.35 suggests, because a long upper tail extends the actual spread well beyond what the middle 50% of the distribution captures.

A practical warning sign is a Q1 close to zero paired with a large Q3, or a reported median noticeably closer to Q1 than to Q3; either pattern indicates skew strong enough that an IQR-based normal approximation should be treated with considerable caution and, ideally, supplemented by a sensitivity analysis excluding that study, following the same logic that the Cochrane Handbook applies to any imputed statistic entering a pooled estimate6.

Appropriate use of the estimate in a systematic review

Because this statistic is imputed rather than reported, it should be labelled as such in the review’s data-extraction table and, where the statistical plan of the review allows, tested in a sensitivity analysis comparing pooled results with and without the studies that rely on an IQR-derived SD67. Reporting the method used for each imputed SD in the methods section, rather than presenting every included study’s SD as though it had been reported directly, is consistent with the PRISMA 2020 expectation that methodological choices affecting the synthesis be disclosed explicitly9.

Where a study reports both an IQR and a usable sample size but the underlying variable is known or suspected to be skewed, converting the median and IQR into a mean and SD and pooling on the natural scale is sometimes less appropriate than analysing the available medians directly through a methods-without-meta-analysis synthesis, an option that the Cochrane Handbook explicitly discusses for this situation6. The IQR/1.35 approximation is a useful tool, but it is intended for approximately normal data, not as a universal converter for any reported measure of spread.

A worked example and a counterexample with skewed data

Consider a study of n=200 that reports Q1=48 and Q3=72 for a continuous outcome, with no stated median, minimum, or maximum. The simple approximation gives SD ≈ (72−48)/1.35 = 24/1.35 ≈ 17.82. Because n is reasonably large in this case, the sample-size-specific divisor of Wan and colleagues for this scenario converges closely to the fixed 1.35 constant, so the simple approximation and the refined formula would give very similar results, within a fraction of a unit of each other2.

By contrast, consider a study of n=15 that reports Q1=2 and Q3=40 for a length-of-stay outcome, with a reported median of 8. The simple approximation gives SD ≈ 38/1.35 ≈ 28.1, but the position of the median, far closer to Q1 than to Q3, is a clear indication of skew: in a normal distribution the median would fall almost exactly midway between the two quartiles, at approximately 21, not at 815. In this second case, both the small sample size and the evident skew support treating the IQR/1.35 estimate as a rough approximation at best, flagging the study for a sensitivity analysis, and considering a narrative or methods-without-meta-analysis treatment of this outcome alongside, rather than instead of, the quantitative estimate6.

These two examples also illustrate a useful general practice when extracting data from a set of IQR-only studies: calculating the implied symmetric midpoint, (Q1+Q3)/2, and comparing it with the reported median is quick, and a large discrepancy is often apparent before any SD calculation is attempted. Incorporating this check into a standard extraction template, alongside the sample size, ensures that skew is flagged consistently across all included studies rather than identified only incidentally on a case-by-case basis18.

It should also be recognised that this method answers a narrower question than it might initially appear to: it estimates the SD that would make the reported IQR and sample size consistent with a normal distribution, not the SD of the actual, possibly non-normal, underlying data. For data that are close to normal, these two quantities are nearly identical, which is why the method performs well in that case. For data that are not, the method still produces a value, but that value describes the best-fitting normal approximation rather than the true spread of the original observations, a distinction that should be borne in mind whenever an IQR-derived SD is reported alongside directly measured SDs from other included studies in the same table24. Making this distinction explicit in the reporting of a review is a modest but meaningful act of transparency that helps readers calibrate their confidence in the pooled result, which is the purpose of disclosing any derivation method.

Key points

Key points

With only Q1 and Q3 reported, SD can be approximated as (Q3-Q1)/1.35 under a normality assumption, refined for small samples by Wan and colleagues' sample-size-specific divisors; the approximation degrades, independent of sample size, whenever the underlying data is meaningfully skewed.

Practical application

Practical application

  1. The use of the median and IQR in the paper should be examined as a possible indication of skew. A median distant from the midpoint of Q1 and Q3 directly indicates that the normal-distribution constant will misestimate the true SD.
  2. For small studies, the sample-size-specific Wan formula should be used rather than the constant of 1.35. The constant is a large-sample approximation and loses accuracy precisely in the smaller studies common in subgroup or pilot data.
  3. Every IQR-derived SD should be labelled as imputed in the extraction table. The distinction between directly reported and derived statistics should remain visible through to the final synthesis.
  4. Where feasible, a sensitivity analysis excluding IQR-imputed studies should be performed. Such an analysis shows whether the pooled conclusion depends on the imputation method rather than on directly reported data.

Frequently asked questions

It is the distance, in standard deviations, between the 25th and 75th percentiles of a standard normal distribution, which is approximately 1.349; dividing a reported IQR by this constant reverses that relationship to estimate the SD that would produce the observed IQR under normality.

No. When min and max are also available, Wan and colleagues' and Luo and colleagues' more information-rich estimators combine the range-based and IQR-based signals into a single, more accurate estimate, which is a distinct calculation from the IQR-only case covered here.

It can be applied mechanically, but the normal-distribution assumption underlying the 1.35 constant is weaker for ordinal and count outcomes than for continuous measurements, so any resulting SD should be treated as a rougher approximation in those cases.

No. Sample size affects how precisely Q1 and Q3 estimate the true population quartiles, not whether the underlying data distribution is actually normal; a very large sample with genuinely skewed data will still produce a biased SD estimate from this method.

References

  1. Wan X, Wang W, Liu J, Tong T. Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range. BMC Medical Research Methodology. 2014;14:135. read 10 October 2026. View on BMC Medical Research Methodology
  2. Luo D, Wan X, Liu J, Tong T. Optimally estimating the sample mean from the sample size, median, mid-range, mid-quartile range, and/or range. Statistical Methods in Medical Research. 2018;27(6):1785-1805. read 10 October 2026. View on Statistical Methods in Medical Research
  3. Hozo SP, Djulbegovic B, Hozo I. Estimating the mean and variance from the median, range, and the size of a sample. BMC Medical Research Methodology. 2005;5:13. read 10 October 2026. View on BMC Medical Research Methodology
  4. Higgins JPT, Li T, Deeks JJ (editors). Chapter 6: Choosing effect measures and computing estimates of effect. In: Cochrane Handbook for Systematic Reviews of Interventions, version 6.4. Cochrane, 2023. read 10 October 2026. View the Cochrane Handbook, Chapter 6
  5. Walter SD, Yao X. Effect sizes can be calculated for studies reporting ranges for outcome variables in systematic reviews. Journal of Clinical Epidemiology. 2007;60(8):849-852. read 10 October 2026. View on Journal of Clinical Epidemiology
  6. Deeks JJ, Higgins JPT, Altman DG (editors). Chapter 10: Analysing data and undertaking meta-analyses. In: Cochrane Handbook for Systematic Reviews of Interventions, version 6.4. Cochrane, 2023. read 10 October 2026. View the Cochrane Handbook, Chapter 10
  7. Altman DG, Bland JM. Standard deviations and standard errors. BMJ. 2005;331(7521):903. read 10 October 2026. View on The BMJ
  8. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327(7414):557-560. read 10 October 2026. View on The BMJ
  9. Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. read 10 October 2026. View on The BMJ
  10. Mulrow CD. Rationale for systematic reviews. BMJ. 1994;309(6954):597-599. read 10 October 2026. View on The BMJ

Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.

How to cite this article

Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.

Pubrica Academy. (2026, October 10). Estimation of the Standard Deviation from the Interquartile Range Alone. Pubrica. https://pubrica.com/academy/estimation-of-standard-deviation-with-only-interquartile-range/

This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.

IQR-to-SD Conversion Worksheet

This method applies when a source study reports only Q1 and Q3 (plus n) and an estimated SD is required.

[ ] Confirmed only Q1, Q3, and n are reported (no median, min, or max)
[ ] Checked whether the median sits roughly midway between Q1 and Q3 (skew warning sign)
[ ] Applied sample-size-specific divisor (Wan et al.) for small n, or IQR/1.35 for larger n
[ ] Labelled the resulting SD as imputed in the extraction table
[ ] Flagged the study for a sensitivity analysis excluding IQR-imputed SDs

Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.

How Pubrica can help

Pubrica’s data extraction team applies the sample-size-specific Wan and Luo estimators, rather than a fixed constant, when converting IQR-only studies into usable SDs for a meta-analysis, and identifies indications of skew before any figure enters a pooled estimate.

Find something else

Search the Pubrica Academy

Access is free and requires no account. Enter a topic, or open the full library to filter by category, sub-topic or resource type.

The normality assumption underlying an IQR-derived standard deviation

A scoping assessment is free and takes two business days. It provides a written assessment of what the work requires and which specialist would undertake it.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.