Contents
Academy · Research Data, Analytics & AI · Mean vs Mean Difference
Mean and Mean Difference in Meta-Analysis: Definitions, Pooling and Measurement Scales
A mean describes one group on its own measurement scale; a mean difference is the effect-size statistic defined by the Cochrane Handbook for pooling the same comparison across multiple trials. Conflation of the two is a common and avoidable cause of meta-analyses that cannot be computed.
A mean is a descriptive statistic belonging to a single group: it is the arithmetic average of a continuous outcome, such as blood pressure or a pain score, measured within one arm of one study. A mean difference (MD) is an effect-size statistic belonging to a comparison: it is the difference between two group means within the same study, intervention minus control, calculated on the original measurement scale that both groups share. The practical distinction matters because only the second of these, the mean difference, is a quantity a meta-analysis can pool across studies using the Cochrane Handbook's inverse-variance method, and only when every contributing study measured the outcome on the same scale and in the same units. A raw group mean reported in a results table is not, by itself, an effect size at all; it is one half of the information a mean difference requires.
Learning objectives
- The precise definitional difference between a mean and a mean difference as the Cochrane Handbook uses the terms
- Why mean difference can only be pooled across studies that share the same measurement scale
- When a mean difference is the wrong effect-size choice, and standardized mean difference is the Handbook's specified alternative
Common misconceptions
- “If two studies both report means for the same outcome, those means can be pooled directly.” The Cochrane Handbook's chapter on analysing data specifies that mean difference pooling is only valid when the included studies measured the outcome using the same scale and the same units; where studies used different scales for a conceptually similar outcome, the Handbook directs review authors to standardized mean difference instead, precisely because raw means from different scales are not directly comparable.
The fundamental distinction between the two terms
A mean is reported for a single group: for example, the average systolic blood pressure in the treatment arm of a single trial, expressed in millimetres of mercury. A mean difference is reported for a comparison within a single study: the mean of the treatment arm minus the mean of the control arm, again expressed in millimetres of mercury, because it is a difference between two values on the same original scale1. The Cochrane Handbook's chapter on analysing data and undertaking meta-analyses, written by Deeks, Higgins, and Altman, defines the mean difference explicitly as an effect measure for continuous outcomes calculated on the original units of measurement, and distinguishes it from the standardized mean difference, which rescales that same kind of comparison into standard-deviation units specifically so that studies using different original scales can be pooled together1. A reviewer who writes “the mean was 12.4” and a reviewer who writes “the mean difference was 3.1” are describing two different kinds of quantity: the first is a within-group descriptive average, whereas the second is a between-group comparison and is itself the input that a meta-analysis pools.
The requirement for a shared scale when pooling mean differences
Borenstein, Hedges, Higgins, and Rothstein’s methods text on meta-analysis explains the mechanical reason why a mean difference can be pooled only across studies sharing the same measurement scale and units: the pooling step takes a weighted average of each study’s mean difference, weighted by the inverse of its variance, and this averaging is mathematically and conceptually meaningful only if every mean difference being averaged is expressed in the same units2. Averaging a 3.1 mmHg difference from one trial with a 0.4-point difference on a 10-point visual analogue scale from another produces a number with no coherent interpretation, because the two values do not measure change on comparable scales, even if both describe conceptually similar outcomes. The Cochrane Handbook’s guidance is explicit on this point: the mean difference is the appropriate choice only when all contributing studies used the same instrument, or instruments known to report on a directly comparable scale with the same units1. For the same reason, a rigorous review protocol specifies, before data extraction, which measurement scale or instrument the included studies are expected to use, since that decision determines whether pooling of mean differences will be possible once the data have been collected.
Circumstances in which the mean difference is inappropriate
When included studies measure a conceptually similar outcome with genuinely different instruments — for example, one trial using the Hamilton Depression Rating Scale and another using the Beck Depression Inventory — pooling of mean differences is not possible, because the two scales are not expressed in the same units. For this situation the Cochrane Handbook recommends the standardized mean difference, which divides each study’s mean difference by that study’s pooled standard deviation, converting each result into a common, scale-free metric of standard-deviation units before pooling1. Hedges and Olkin’s foundational statistical methods text, and the later practical tutorial by Cuijpers and colleagues, both describe this standardization as the direct solution to the specific problem that the mean difference cannot address: combining results that share a construct but not a measurement scale34. Determining which situation applies — the same scale across studies, or different scales measuring a similar construct — is the single decision that establishes whether the mean difference or the standardized mean difference is the correct effect measure for a given meta-analysis; an incorrect decision produces a pooled estimate that appears plausible on a forest plot but does not mean what it appears to mean.
A worked illustration of the distinction
Consider three hypothetical trials of the same blood-pressure medication, all measuring systolic blood pressure in mmHg with a standard sphygmomanometer. Trial A reports a treatment-arm mean of 128 mmHg and a control-arm mean of 134 mmHg; the mean difference for Trial A is therefore -6 mmHg (treatment mean minus control mean), still expressed in mmHg. Trial B reports a treatment-arm mean of 131 and a control-arm mean of 137, also giving a mean difference of -6 mmHg, although its raw group means differ from those of Trial A. This illustrates why a meta-analysis pools the mean differences (-6 and -6) rather than the raw group means (128, 134, 131, 137): the raw means reflect the baseline blood pressure of each trial’s patient population, which can differ for reasons unrelated to the treatment, whereas the mean difference isolates the within-trial treatment effect that is comparable across trials1. If Trial C instead reported its outcome using a differently calibrated home blood-pressure monitoring device, or measured a related but distinct construct such as a composite cardiovascular risk score, its mean difference would not be directly combinable with the mmHg-based mean differences of Trials A and B; this is precisely the scale-incompatibility problem that the standardized mean difference is designed to address13.
Reporting the distinction in a systematic review
A review’s methods section should state explicitly, before results are known, which effect measure will be used for each outcome and why, rather than leaving this as an implicit choice made during analysis. The Cochrane Handbook’s chapter on choosing effect measures recommends that this be decided at the protocol stage, precisely because the decision is a substantive methodological commitment rather than a detail that can be adjusted once the pooled estimate has been calculated5. In the results, a mean difference should always be reported with its original units stated explicitly, for example “mean difference -6 mmHg (95% CI -9 to -3)”, so that the reader need not infer whether a bare number refers to a within-group mean, a between-group mean difference or a standardized, unit-free SMD. Lipsey and Wilson’s practical meta-analysis text notes that such unit ambiguity in a results table is a common source of reader confusion in published reviews, and recommends labeling every reported number with both its statistical identity (mean, MD or SMD) and its units or scale wherever applicable, precisely to prevent one quantity from being mistaken for another6. The same labeling discipline applies to a review’s abstract and plain-language summary, where space constraints often lead authors to omit the unit or the statistical label; this is precisely where a reader unfamiliar with the methodology is most likely to mistake a mean difference for a simple group mean, so the units and the label should be retained even in a tightly word-limited summary rather than being the first detail removed. A short parenthetical such as “(mean difference, mmHg)” immediately after a number requires only a few characters and removes this ambiguity entirely, a disproportionately small cost relative to the confusion it prevents in a document that many readers will only skim. Review teams that draft a results table template at the protocol stage, before any data are extracted, can incorporate this labeling convention into the column headers from the outset, which removes the need to apply it manually once the data arrive and the writing team is focused on the findings rather than on formatting.
Key points
Key points
A mean describes one group; a mean difference describes a within-study comparison expressed on the original measurement scale, and it can only be pooled across studies that share that same scale and units; otherwise, the Cochrane Handbook specifies the standardized mean difference.
Practical application
Practical application
- Every included study should be confirmed to use the same scale before the mean difference is selected. The instrument and units reported by each study should be checked before it is decided whether pooling of mean differences is mathematically valid for the outcome.
- The expected measurement scale should be specified in the protocol before data extraction. Specifying the scale in advance avoids the discovery during analysis that the included studies cannot be pooled with the planned effect measure.
- The standardized mean difference should be used when scales differ. Where studies measure a similar construct with genuinely different instruments, the standardized mean difference (SMD) should be used rather than forcing the pooling of mean differences across incompatible scales.
Frequently asked questions
No. A meta-analysis pools a comparison (an effect size, such as a mean difference or standardized mean difference) between groups within each study, not a single group's raw mean on its own.
The resulting pooled estimate is not interpretable, because it averages numbers expressed in incompatible units; the Cochrane Handbook specifies the standardized mean difference for precisely this situation.
No. When all included studies share the same scale and units, the mean difference is preferred because it keeps the result in clinically interpretable original units; the standardized mean difference is the alternative specifically for scale incompatibility across studies.
A mean difference is always a within-study comparison between two arms (intervention minus control), not a single arm's mean on its own.
References
- Deeks JJ, Higgins JPT, Altman DG. Chapter 10: Analysing data and undertaking meta-analyses. In: Higgins JPT, Thomas J, Chandler J, et al, eds. Cochrane Handbook for Systematic Reviews of Interventions. read 10 October 2026. View on training.cochrane.org
- Borenstein M, Hedges LV, Higgins JPT, Rothstein HR. Introduction to Meta-Analysis. John Wiley & Sons; 2009. read 10 October 2026. View on wiley.com
- Hedges LV, Olkin I. Statistical Methods for Meta-Analysis. Academic Press; 1985. read 10 October 2026. View on elsevier.com
- Cuijpers P, Weitz E, Cristea IA, Twisk J. Pre-post effect sizes should be avoided in meta-analyses. Epidemiology and Psychiatric Sciences. 2017;26(4):364-368. read 10 October 2026. View on Cambridge Core
- Higgins JPT, Li T, Deeks JJ. Chapter 6: Choosing effect measures and computing estimates of effect. Cochrane Handbook for Systematic Reviews of Interventions. read 10 October 2026. View on training.cochrane.org
- Lipsey MW, Wilson DB. Practical Meta-Analysis. Sage Publications; 2001. read 10 October 2026. View on sagepub.com
- Deeks JJ, Higgins JPT, Altman DG. Statistical algorithms in Review Manager 5. Cochrane Handbook Archive. read 10 October 2026. View on training.cochrane.org
- DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled Clinical Trials. 1986;7(3):177-188. read 10 October 2026. View on Controlled Clinical Trials
- Cuijpers P, Weitz E, Cristea IA, Twisk J. Pre-post effect sizes should be avoided in meta-analyses. Epidemiology and Psychiatric Sciences. 2017;26(4):364-368. (publisher record) read 10 October 2026. View on Cambridge Core
- Schünemann HJ, Higgins JPT, Vist GE, et al. Chapter 15: Interpreting results and drawing conclusions. Cochrane Handbook for Systematic Reviews of Interventions. read 10 October 2026. View on training.cochrane.org
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). Mean and Mean Difference in Meta-Analysis: Definitions, Pooling and Measurement Scales. Pubrica. https://pubrica.com/academy/difference-between-mean-and-mean-difference/
Pubrica Academy. “Mean and Mean Difference in Meta-Analysis: Definitions, Pooling and Measurement Scales.” Pubrica, 10 Oct. 2026, pubrica.com/academy/difference-between-mean-and-mean-difference/.
Pubrica Academy (2026) Mean and Mean Difference in Meta-Analysis: Definitions, Pooling and Measurement Scales. Pubrica. Available at: https://pubrica.com/academy/difference-between-mean-and-mean-difference/ (Accessed: 15 October 2026).
Pubrica Academy. “Mean and Mean Difference in Meta-Analysis: Definitions, Pooling and Measurement Scales.” Pubrica. October 10, 2026. https://pubrica.com/academy/difference-between-mean-and-mean-difference/
Pubrica Academy, ‘Mean and Mean Difference in Meta-Analysis: Definitions, Pooling and Measurement Scales’ (Pubrica, 10 October 2026) <https://pubrica.com/academy/difference-between-mean-and-mean-difference/> accessed 15 October 2026.
Pubrica Academy. Mean and Mean Difference in Meta-Analysis: Definitions, Pooling and Measurement Scales. [Internet]. Pubrica; 2026 Oct 10 [cited 2026 Oct 15]. Available from: https://pubrica.com/academy/difference-between-mean-and-mean-difference/
Pubrica Academy, “Mean and Mean Difference in Meta-Analysis: Definitions, Pooling and Measurement Scales,” Pubrica, Oct. 10, 2026. [Online]. Available: https://pubrica.com/academy/difference-between-mean-and-mean-difference/
@misc{/la_mean_vs_mean_difference_meta_analysis_2026,
author = {{Pubrica Academy}},
title = {Mean and Mean Difference in Meta-Analysis: Definitions, Pooling and Measurement Scales},
howpublished = {Pubrica},
year = {2026},
month = {oct},
url = {https://pubrica.com/academy/difference-between-mean-and-mean-difference/},
note = {Accessed: <span data-cite-acc="iso">2026-10-15</span>}
}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.
MD vs SMD Decision Note
This checklist should be applied before an effect measure is chosen for a continuous outcome.
[ ] Listed the exact instrument/scale and units each included study used for this outcome [ ] Confirmed whether all studies share the same scale and units [ ] If yes: proceed with mean difference (MD), keeping results on the original scale [ ] If no: switch to standardized mean difference (SMD) and note this decision in the protocol [ ] Documented this decision in the review's statistical methods section
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 biostatistics team checks the measurement scales across the included studies before analysis and confirms whether the mean difference or the standardized mean difference is the statistically valid effect measure for the meta-analysis.