Global complications among patients with mpox: a systematic review and meta-analysis

Summary

Background

Despite increasing global attention, significant knowledge gaps remain regarding the incidence of adverse outcomes associated with mpox. To support the development of World Health Organization (WHO) mpox guidelines, this systematic review and meta-analysis evaluated the incidence of complications among patients with mpox.

Methods

We searched Medline, Embase, CENTRAL, CINAHL, Global Health, medRxiv, bioRxiv, and SSRN from database inception to December 20, 2025, for studies reporting complications in patients with laboratory-confirmed mpox of all ages. Paired reviewers independently screened studies, extracted data, and assessed risk of bias. We conducted proportional meta-analyses using the inverse variance method fixed-effect model with the Freeman-Tukey double arcsine transformation and evaluated the certainty of evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework. We classified mpox as severe or non-severe based on the need for hospitalization or the study's criteria. We registered the protocol with PROSPERO (CRD42024595685).

Findings

Of 4566 records, 170 studies with 127,564 patients (mean age 6.9–43.0 years) proved eligible. Among patients with non-severe mpox, serious complications were rare. Mortality, intensive care unit admission, sepsis, pneumonia, encephalitis, epiglottitis, myocarditis, gastrointestinal bleeding, and pleural effusion all occurred in ≤0.5% of patients (moderate to high certainty evidence). Urinary retention, mechanical ventilation, urethritis, gastroenteritis, and keratitis occurred in 0.7%–1.8% (moderate to high certainty evidence). Severe pain, abscess, tonsillitis, conjunctivitis, and cellulitis occurred in 2.2%–3.7% of patients (moderate certainty evidence) and hospitalization occurred in 4.4% (low certainty evidence). Patients who were vaccinated had a lower hospitalization rate than those who were unvaccinated. Patients in low- and middle-income countries (LMICs) had a higher rate of pneumonia (2.21% versus 0.02%) than those in high-income countries (HICs). In patients with severe mpox, complication rates were substantially higher. All-cause mortality, abscess, sepsis, tonsillitis, keratitis, and mechanical ventilation occurred in 2.5%–5.5% of patients (moderate certainty evidence). Acute kidney injury, pleural effusion, respiratory failure, and conjunctivitis occurred in 1.4%–2.0% (moderate to high certainty evidence). Patients with severe mpox in LMICs had a higher mortality (3.18% versus <0.01%) than those in HICs.

Interpretation

Non-severe mpox is associated with low complication rates and very low mortality, while severe disease carries substantial risks of life-threatening complications and death. These findings provide baseline risk estimates for both patients with severe and non-severe mpox, emphasize the importance of early risk stratification in clinical management, and highlight the need for tailored approaches based on disease severity. The data also reveal health inequities in mpox care between HICs and LMICs, underscoring the urgent need for increased resources in LMICs.

Funding

World Health Organization.

Research in context.

Evidence before this study

A comprehensive understanding of the incidence and pattern of mpox-associated complications is essential for informing prevention strategies, clinical management, and global health policy. Previous systematic reviews and meta-analyses have investigated clinical features, hospitalization, and mortality associated with mpox, but most were limited by small sample sizes, focus on early outbreak data, restricted outcome sets, or a lack of assessment of the certainty of evidence. A comprehensive and up-to-date global synthesis of mpox-associated complication rates, with appropriate evidence grading, has been lacking.

Added value of this study

In this systematic review and meta-analysis including 170 studies and 127,564 patients, we found that, among patients with non-severe mpox, rates of mortality, intensive care unit admission, sepsis, pneumonia, encephalitis, epiglottitis, myocarditis, gastrointestinal bleeding, and pleural effusion proved extremely low (≤0.5%); rates of severe pain, rectal bleeding, abscess, tonsillitis, conjunctivitis, hospitalization, and cellulitis varied between 2.2% and 4.4%. Patients in high-income countries (HICs) had a lower rate of pneumonia than those in low- and middle-income countries (LMICs). Severe mpox cases exhibited substantially elevated risk: all-cause mortality, urinary retention, abscess, sepsis, tonsillitis, pneumonia, keratitis, mechanical ventilation, and cellulitis varied between 2.1% and 5.8%. Patients with severe mpox in HICs had a lower mortality than those in LMICs.

Implications of all the available evidence

To our knowledge, our study provides the most comprehensive global assessment of complications among patients with mpox. These findings underscore the importance of early risk stratification and tailored management for patients based on disease severity, and highlight the urgent need for enhanced resources in LMICs.

Introduction

Mpox (formerly known as monkeypox), an infectious disease caused by the Mpox virus of the Orthopoxvirus genus,¹ is characterized by two genetic clades—Clade I and Clade II—that are associated with different levels of disease severity and transmissibility.^2,3 The clinical presentation of mpox typically includes a characteristic rash, often accompanied by fever, lymphadenopathy, and other systemic symptoms.^4,5 Although most cases of mpox are self-limiting, specific populations (e.g., immunocompromised individuals and children) may experience severe complications such as secondary infections, sepsis, and even death.5, 6, 7

On August 14, 2024, the World Health Organization (WHO) declared mpox a Public Health Emergency of International Concern (PHEIC),⁸ highlighting the need for a comprehensive and up-to-date understanding of clinical characteristics and complications. Previous systematic reviews addressing the clinical features or complications of mpox were published earlier in the course of the epidemic.9, 10, 11, 12, 13 These reviews focused only on clinical features, hospitalization, and mortality. Further, they did not include data after the recent PHEIC and failed to assess the certainty of evidence. Despite increasing global attention, significant knowledge gaps remain regarding the incidence of complications associated with mpox.

Therefore, in support of the development of WHO mpox guidelines,¹⁴ this systematic review evaluated the incidence of complications among patients with mpox.

Methods

This systematic review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 (PRISMA 2020) statement and Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines.^15,16 We registered this systematic review protocol with PROSPERO (CRD42024595685).

Search strategy

With the aid of an expert librarian, we searched Medline, Embase, Cochrane Central Register of Controlled Trials (CENTRAL), Cumulative Index to Nursing and Allied Health Literature (CINAHL), Global Health, medRxiv, bioRxiv, and SSRN from database inception to September 20, 2024, and updated the searches up to December 20, 2025. Appendix 1 presents the details of the search strategy. We supplemented our search by reviewing the reference lists of eligible studies and relevant systematic reviews.

Eligibility criteria

We included studies that reported on the incidence of complications among patients with mpox virus infection confirmed by polymerase chain reaction (PCR), electron microscopy, or virus isolation. Complications included, but were not limited to, mortality, hospitalization, intensive care unit (ICU) admission, respiratory failure, mechanical ventilation, sepsis, pneumonia, encephalitis, epiglottitis, myocarditis, gastrointestinal bleeding, urinary retention, secondary infection of mpox lesions, complications associated with mucosal lesions, and severe or uncontrolled pain.

We did not apply restrictions on publication language, patient age or sex, clades of monkeypox virus, or comorbidity of patients. We included studies with up to 20% ineligible patients, or those that reported separately on our population of interest. We excluded narrative, scoping or systematic reviews, conference abstracts, and studies with a sample size of less than 20.

Study selection

Using Covidence (https://covidence.org/), pairs of reviewers independently screened titles and abstracts and, subsequently, reviewed the full texts of potentially eligible records. When multiple studies used the same data source and sampling period, we included the study with the largest sample size. When documentation was inadequate to determine whether cases were unique for complications, we minimized double counting by excluding studies using large-scale surveillance data. Reviewers resolved discrepancies through discussion or, if necessary, with the assistance of a third reviewer.

Data extraction

Using a predesigned form (Excel), pairs of reviewers extracted information on study characteristics (first author, study design, publication year, study period, country, and sample size), patient characteristics (age, sex, disease severity, comorbidities, vaccination status, and clade of monkeypox virus), and incidence of complications. Reviewers resolved discrepancies by discussion or, if necessary, with the assistance of a third reviewer.

Risk of bias assessment

Using the Quality In Prognosis Studies (QUIPS) tool, paired reviewers independently assessed the risk of bias of eligible studies.¹⁷ We evaluated four domains: study participation, study attrition, outcome measurement, and statistical analysis and reporting. We rated each domain as either low, moderate, or high risk of bias. We considered studies at overall high risk of bias if we judged one or more domains at high risk; studies at low risk of bias if at least three domains were low risk and the other domain was moderate risk; and the remaining studies at moderate risk of bias. Reviewers resolved discrepancies by discussion and, when necessary, with the assistance of a third reviewer.

Statistical analysis

To estimate the pooled incidence of complications and their associated 95% confidence intervals (CIs), we conducted proportional meta-analyses. We initially applied a random-effects model. However, in the context of proportional meta-analysis with extreme event rates and substantial variation in study size, the random-effects model produced unstable and clinically implausible pooled estimates, likely driven by disproportionate weighting of small studies with extreme proportions. To mitigate this methodological vulnerability and improve estimate stability, we conducted the meta-analysis using the inverse variance method fixed-effect model.¹⁸ To stabilize variances, we applied the Freeman-Tukey double arcsine transformation.¹⁹ We performed pooled analyses for all patients, and separately for patients with severe and non-severe mpox illness, defining severe mpox as an illness that required hospitalization, or according to specific study definitions that typically included hospitalization. If a study included 50% or more participants classified as having severe mpox or those hospitalized for treatment, we categorized patients as predominantly severe, and if fewer than 50% of participants as predominantly non-severe. We assessed the between-study heterogeneity with a visual inspection of forest plots. We followed published guidance suggesting avoiding formal statistical tests in the context of meta-analysis of proportions and therefore did not assess publication bias by Egger's test, Begg's test, or LFK index.¹⁹ To explore possible small-study effects, we visually examined funnel plots for outcomes with more than nine studies. In meta-analyses of proportions, asymmetry in funnel plots can reflect an association between study size and the estimate of proportions, rather than solely as a test for publication bias. We performed analyses using R (version 4.2.1, R Foundation for Statistical Computing).

Subgroup analysis

If data were available, we performed the following prespecified subgroup analyses:

• 1.Countries: low- and middle-income countries (LMICs) versus high-income countries (HICs) (hypothesis: higher incidence of complications in LMICs).
• 2.Clades of monkeypox virus: clade I versus clade II (hypothesis: higher incidence of complications in patients with clade I monkeypox virus infection). If the source documents did not specify the clade, we based inferences regarding clade on the period and country of data collection, according to published epidemiological descriptions.
• 3.Age: infants and toddlers (<5 years) versus children (5–12 years) versus adolescents (>12–18 years) versus adults (>18–64 years) versus elderly (≥65 years) (hypothesis: higher incidence of complications in infants, toddlers, and the elderly).
• 4.Vaccination status (e.g., previous smallpox vaccine, MVA-BN (LC-16), ACAM2000): patients who were vaccinated versus patients who were unvaccinated (hypothesis: higher incidence of complications in patients who were unvaccinated).

We prioritized the use of within-study analyses whenever such data (e.g., reported stratified estimates within individual studies) were available from at least two studies for a subgroup. When within-study information was insufficient (e.g., reported by fewer than two studies), we conducted between-study subgroup analyses using study-level characteristics. Due to limited availability of within-study information for age and vaccination, we also performed meta-regression analyses for the proportion of patients <18 years (hypothesizing a higher incidence of complications in studies with a higher proportion of patients <18 years) and proportion of patients who were vaccinated (hypothesizing a lower incidence of complications in studies with a higher proportion of patients who were vaccinated). To assess the credibility of the apparent subgroup effect, we used a modified version of the Instrument for assessing the Credibility of Effect Modification Analyses (ICEMAN) tool tailored specifically for prognostic studies.²⁰ ICEMAN evaluates this credibility based on factors including prespecification including direction of effects, within-study versus between-study comparisons, number of hypotheses tested, and the risk of chance findings. We adapted only the wording and interpretation of individual items to suit prognostic subgroup analyses in observational studies, rather than intervention effect modification. These adaptations preserved the original intent of the ICEMAN framework while improving its applicability to prognosis research. One of our team (GG) who led the development of the ICEMAN instrument oversaw this modification.

Certainty of evidence

To assess the certainty of the evidence, we used GRADE (Grading of Recommendations Assessment, Development and Evaluation) guidance for overall prognosis in broad populations, in which observational studies estimating event rates in unselected populations begin as high certainty evidence and may be rated down based on predefined domains.²¹ By assessing the domains of risk of bias, imprecision, inconsistency, and indirectness, we rated the overall certainty of the evidence for each outcome as ‘‘high’’, ‘‘moderate’’, ‘‘low’’, or ‘‘very low’’. We assessed imprecision based on the width of the 95% CIs. We did not rate down for imprecision when the CI width was <3%; we rated down one level for widths of 3–6%, two levels for 6–12%, and three levels for >12%.

Ethics statement

This study is a systematic review and meta-analysis of previously published literature and did not involve primary data collection from human participants. Ethical approval and informed consent were not required.

Role of the funding source

The funder (World Health Organization) had no role in study design, data collection, analysis, and interpretation, or writing of the manuscript and the decision to submit.

Results

The electronic search identified 4566 records. After screening 3152 unduplicated titles and abstracts and 613 full texts, 170 studies proved eligible (Fig. 1). Appendix 2 lists the eligible studies.

Fig. 1.

Study selection.

Characteristics of included studies

Appendix 3 summarizes the characteristics of the 170 eligible studies, including 71 single-center studies and 99 multicenter studies. The number of patients ranged from 22 to 20,690 patients, including a total of 127,564 patients. Across included studies, reported mean age varied from 6.9 to 43 years, with a median study-level mean age of 35 years. The proportion of male patients ranged from 0% to 100%, with most studies reporting over 90% males. Among 151 studies reporting HIV status, the median proportion of patients living with HIV was 43.0% (range 0.5%–100%). Among 149 studies reporting detailed data of severity or hospitalization status, the median proportion of severe or hospitalized patients was 8.8% (range 0%–100%). Twenty-four studies primarily focused on patients with severe mpox and 146 included primarily patients with non-severe mpox.

Risk of bias

Among the 170 included studies, 153 (90.0%) proved at low risk of bias across all reported outcomes; 8 (4.7%) at moderate risk of bias and 4 (2.4%) at high risk of bias for all reported outcomes; 2 (1.2%) had a high risk for specific outcomes (e.g., secondary bacterial infection) and 3 (1.8%) had a moderate risk for specific outcomes. The primary limitations were substantial attrition rates and insufficient description of outcome measurement (Appendix 4).

Results including all patients

Certainty proved moderate to high for most outcomes (Table 1). Rates of mortality, ICU admission, sepsis, encephalitis, epiglottitis, myocarditis, and gastrointestinal bleeding proved extremely low (all at or below 5 in 1000) (Table 1). Rates of pneumonia, urinary retention, pleural effusion, urethritis, and acute kidney injury ranged between 0.6 and 1.4%. Rates of gastroenteritis, abscess, mechanical ventilation, tonsillitis, conjunctivitis, respiratory failure, keratitis, hospitalization, cellulitis, and severe pain ranged between 2.0% and 5.2% (Table 1). Rates of proctitis (7.6%), rectal bleeding (7.4%), pharyngitis (9.0%), secondary bacterial infection (9.4%), and severe or serious rash (31.3%) proved more common (Table 1, Appendices 5 and 6).

Table 1.

GRADE assessment for complication rates in overall patients.

Outcomes	No of studies	Effect			Certainty of the evidence
		No of events	No of individuals	Rate (95% CI)
All-cause mortality	103	341	73,672	event rate 0 per 1000 (0–<0.01)	⨁⨁⨁⨁ High
ICU admission	27	100	31,846	event rate 0.4 per 1000 (0.1–0.8)	⨁⨁⨁⨁ High
Sepsis	15	180	11,657	event rate 5 per 1000 (3–6)	⨁⨁⨁◯ Moderatea
Pneumonia	14	72	3821	event rate 6 per 1000 (3–10)	⨁⨁⨁◯ Moderatea
Encephalitis	14	21	7739	event rate 0.2 per 1000 (0–1)	⨁⨁⨁⨁ High
Epiglottitis	3	3	701	event rate 1 per 1000 (0–7)	⨁⨁⨁⨁ High
Myocarditis	9	10	5617	event rate 0 per 1000 (0–0.07)	⨁⨁⨁⨁ High
Gastrointestinal bleeding	2	10	2184	event rate 3 per 1000 (1–7)	⨁⨁⨁◯ Moderatea
Urinary retention	5	6	514	event rate 8 per 1000 (0.6–19)	⨁⨁⨁⨁ High
Pleural effusion	2	7	777	event rate 8 per 1000 (2–16)	⨁⨁⨁◯ Moderatea
Urethritis	10	141	5438	event rate 12 per 1000 (8–15)	⨁⨁⨁◯ Moderatea
Acute kidney injury	3	5	282	event rate 14 per 1000 (2–34)	⨁⨁⨁◯ Moderateb
Gastroenteritis	3	8	243	event rate 23 per 1000 (6–48)	⨁⨁◯◯ Lowab
Abscess	13	85	2008	event rate 31 per 1000 (23–40)	⨁⨁⨁◯ Moderatea
Mechanical ventilation	2	24	777	event rate 26 per 1000 (16–39)	⨁⨁⨁◯ Moderatea
Tonsillitis	14	129	3314	event rate 30 per 1000 (24–36)	⨁⨁⨁◯ Moderatea
Conjunctivitis	13	412	8191	event rate 33 per 1000 (29–37)	⨁⨁⨁◯ Moderatea
Respiratory failure	3	19	749	event rate 20 per 1000 (11–32)	⨁⨁⨁⨁ High
Keratitis	6	29	880	event rate 25 per 1000 (15–37)	⨁⨁⨁⨁ High
Cellulitis	12	191	4669	event rate 25 per 1000 (20–30)	⨁⨁⨁◯ Moderatea
Severe pain	15	375	4706	event rate 51 per 1000 (45–58)	⨁⨁⨁◯ Moderatea
Hospitalization	123	6366	96,456	event rate 52 per 1000 (51–54)	⨁⨁⨁◯ Moderatea
Pharyngitis	8	57	536	event rate 90 per 1000 (66–117)	⨁⨁◯◯ Lowab
Proctitis	40	1392	13,013	event rate 76 per 1000 (72–81)	⨁⨁⨁◯ Moderatea
Rectal bleeding	4	42	435	event rate 74 per 1000 (50–101)	⨁⨁◯◯ Lowab
Secondary bacterial infection	48	1194	10,359	event rate 94 per 1000 (89–100)	⨁⨁⨁◯ Moderatea
Severe or serious rash	3	350	1117	event rate 313 per 1000 (286–340)	⨁⨁◯◯ Lowab

^aRated down one level for inconsistency due to the high heterogeneity.

^bRated down one level for imprecision due to the wide 95% confidence interval.

Patients with non-severe mpox

Certainty proved moderate to high for most outcomes (Table 2). For patients with non-severe mpox, rates of mortality, ICU admission, sepsis, pneumonia, encephalitis, epiglottitis, myocarditis, gastrointestinal bleeding, and pleural effusion proved extremely low (all at or below 5 in 1000) (Table 2). Rates of urinary retention, mechanical ventilation, urethritis, gastroenteritis, and keratitis ranged between 0.7 and 1.8%. Rates of cellulitis, severe pain, rectal bleeding, abscess, tonsillitis, conjunctivitis, and hospitalization ranged between 2.2% and 4.4% (Table 2). Rates of pharyngitis (8.6%), proctitis (7.4%), secondary bacterial infection (7.4%), and severe or serious rash (31.3%) proved more common (Table 2 and Appendix 5). Substantial statistical heterogeneity was observed for several outcomes (I² range 0%–98.7%).

Table 2.

GRADE assessment for complication rates in non-severe patients.

Outcomes	No of studies	Effect			Certainty of the evidence
		No of events	No of individuals	Rate (95% CI)
All-cause mortality	85	217	70,674	event rate 0 per 1000 (0–<0.001)	⨁⨁⨁⨁ High
ICU admission	23	64	31,181	event rate 0.4 per 1000 (0.1–0.8)	⨁⨁⨁⨁ High
Sepsis	8	138	10,658	event rate 5 per 1000 (4–7)	⨁⨁⨁◯ Moderatea
Pneumonia	9	44	3275	event rate 3 per 1000 (1–6)	⨁⨁⨁◯ Moderatea
Encephalitis	6	8	5869	event rate 0 per 1000 (0–0.5)	⨁⨁⨁⨁ High
Epiglottitis	3	3	701	event rate 1 per 1000 (0–7)	⨁⨁⨁⨁ High
Myocarditis	8	9	5331	event rate 0 per 1000 (0–0.01)	⨁⨁⨁⨁ High
Gastrointestinal bleeding	2	10	2184	event rate 3 per 1000 (1–7)	⨁⨁⨁◯ Moderatea
Pleural effusion	1	1	395	event rate 3 per 1000 (0–11)	⨁⨁⨁⨁ High
Urinary retention	3	4	423	event rate 7 per 1000 (0.2–19)	⨁⨁⨁⨁ High
Mechanical ventilation	1	3	395	event rate 8 per 1000 (1–19)	⨁⨁⨁⨁ High
Urethritis	9	129	5366	event rate 11 per 1000 (8–14)	⨁⨁⨁◯ Moderatea
Gastroenteritis	2	3	203	event rate 12 per 1000 (0.2–35)	⨁⨁⨁◯ Moderateb
Keratitis	2	11	420	event rate 18 per 1000 (7–34)	⨁⨁⨁⨁ High
Cellulitis	9	155	4204	event rate 22 per 1000 (18–27)	⨁⨁⨁◯ Moderatea
Severe pain	8	116	3823	event rate 24 per 1000 (19–29)	⨁⨁⨁◯ Moderatea
Rectal bleeding	2	9	239	event rate 26 per 1000 (8–51)	⨁⨁◯◯ Lowab
Abscess	8	58	1321	event rate 34 per 1000 (24–46)	⨁⨁⨁◯ Moderatea
Tonsillitis	11	109	2791	event rate 30 per 1000 (23–37)	⨁⨁⨁◯ Moderatea
Conjunctivitis	9	382	6665	event rate 37 per 1000 (33–42)	⨁⨁⨁◯ Moderatea
Hospitalization	114	4931	94,504	event rate 44 per 1000 (42–45)	⨁⨁◯◯ Lowac
Pharyngitis	6	44	422	event rate 86 per 1000 (60–115)	⨁⨁◯◯ Lowab
Proctitis	36	1331	12,773	event rate 74 per 1000 (70–79)	⨁⨁⨁◯ Moderatea
Secondary bacterial infection	32	720	7952	event rate 74 per 1000 (68–80)	⨁⨁⨁◯ Moderatea
Severe or serious rash	3	350	1117	event rate 313 per 1000 (286–340)	⨁⨁◯◯ Lowab

^aRated down one level for inconsistency due to the high heterogeneity.

^bRated down one level for imprecision due to the wide 95% confidence interval.

^cRated down one level for indirectness due to the definition of severe cases also taking hospitalization into account.

Within-study subgroup analyses revealed no significant subgroup effects of age on hospitalization rates (Appendix 7.1). Patients who were vaccinated had a lower hospitalization rate than patients who were unvaccinated (pooled ratio of rates 0.39, 95% CI 0.22–0.71; P for interaction = 0.002; Appendix 7.2). We assessed the credibility of this subgroup effect as moderate (Appendix 8.1). Between-study subgroup analyses (Fig. 2A) revealed significant differences in certain outcomes (e.g., all-cause mortality, secondary bacterial infection) by clades (Appendix 7.3) and by countries (Appendix 7.4). Patients from high-income countries (HICs) had a lower rate of pneumonia (0.02% versus 2.21%, P for interaction <0.001) than those from low- and middle-income countries (LMICs) (moderate credibility) (Fig. 2A and Appendices 8 and 9). However, due to the between-study comparison and test of many effect modifiers, we rated the credibility of other subgroup effects as low or very low (Appendix 8). Meta-regression analyses did not identify credible subgroup effects for outcomes by proportion of patients <18 years or proportion of vaccinated patients (Appendices 10 and 11).

Fig. 2.

Results of subgroup analysis by clades and countries in patients with non-severe mpox (A) and severe mpox (B). Footnotes: The horizontal axis represents the outcome; the vertical axis represents the rate of the outcome; The bubble size represents the number of eligible studies for a specific outcome; the green color indicates moderate credibility of the subgroup effect between clade I and II or between high-income countries (HICs) and low- and middle-income countries (LMICs); the orange color indicates low credibility of the subgroup effect; the red color indicates very low credibility of the subgroup effect; the gray color indicates no subgroup effects (P for interaction ≥0.05) between clade I and II or between HICs and LMICs.

Patients with severe mpox

Certainty varied widely, from very low to high (Table 3). For patients with severe mpox, the rates of encephalitis and myocarditis proved extremely low (under 5 in 1000; high certainty). Rates of acute kidney injury, pleural effusion, respiratory failure, and conjunctivitis ranged between 1.4% and 2.0% (moderate to high certainty) (Table 3). Rates of urinary retention, abscess, sepsis, all-cause mortality, tonsillitis, pneumonia, keratitis, mechanical ventilation, and cellulitis ranged between 2.1% and 5.8% (low to moderate certainty). Very low to low certainty evidence suggested that complications with an incidence rate exceeding 10% included pharyngitis (10.9%), gastroenteritis (12.5%), rectal bleeding (16.0%), urethritis (16.7%), secondary bacterial infections (17.9%), proctitis (23.4%), and severe pain (27.3%) (Table 3 and Appendix 5). Substantial statistical heterogeneity was observed for some outcomes (I² range 0%–96%).

Table 3.

GRADE assessment for complication rates in severe patients.

Outcomes	No of studies	Effect			Certainty of the evidence
		No of events	No of individuals	Rate (95% CI)
Encephalitis	8	13	1870	event rate 4 per 1000 (1–9)	⨁⨁⨁⨁ High
Myocarditis	1	1	286	event rate 3 per 1000 (0–10)	⨁⨁⨁⨁ High
Acute kidney injury	3	5	282	event rate 14 per 1000 (2–34)	⨁⨁⨁◯ Moderateb
Pleural effusion	1	6	382	event rate 16 per 1000 (5–31)	⨁⨁⨁⨁ High
Conjunctivitis	4	30	1526	event rate 18 per 1000 (11–26)	⨁⨁⨁⨁ High
Respiratory failure	3	19	749	event rate 20 per 1000 (11–32)	⨁⨁⨁⨁ High
Urinary retention	2	2	91	event rate 21 per 1000 (0–66)	⨁⨁◯◯ Lowc
Abscess	5	27	687	event rate 25 per 1000 (13–39)	⨁⨁⨁◯ Moderatea
Sepsis	7	42	999	event rate 25 per 1000 (15–37)	⨁⨁⨁◯ Moderatea
All-cause mortality	18	124	2998	event rate 29 per 1000 (23–36)	⨁⨁⨁◯ Moderatea
Tonsillitis	3	20	523	event rate 31 per 1000 (17–48)	⨁⨁⨁◯ Moderateb
Keratitis	4	18	460	event rate 34 per 1000 (18–54)	⨁⨁⨁◯ Moderateb
Pneumonia	5	28	546	event rate 46 per 1000 (29–67)	⨁⨁◯◯ Lowab
Mechanical ventilation	1	21	382	event rate 55 per 1000 (34–80)	⨁⨁⨁◯ Moderateb
Cellulitis	3	36	465	event rate 58 per 1000 (37–83)	⨁⨁◯◯ Lowab
Pharyngitis	2	13	114	event rate 109 per 1000 (56–174)	⨁◯◯◯ Very lowac
Gastroenteritis	1	5	40	event rate 125 per 1000 (37–248)	⨁◯◯◯ Very lowd
Rectal bleeding	2	33	196	event rate 160 per 1000 (111–215)	⨁◯◯◯ Very lowac
Urethritis	1	12	72	event rate 167 per 1000 (88–263)	⨁◯◯◯ Very lowd
Secondary bacterial infection	16	474	2407	event rate 179 per 1000 (163–195)	⨁⨁◯◯ Lowab
Proctitis	4	61	240	event rate 234 per 1000 (181–290)	⨁◯◯◯ Very lowac
Severe pain	7	259	883	event rate 273 per 1000 (243–303)	⨁⨁◯◯ Lowab

^aRated down one level for inconsistency due to the high heterogeneity.

^bRated down one level for imprecision due to the wide 95% confidence interval.

^cRated down two levels for imprecision due to the wide 95% confidence interval.

^dRated down three levels for imprecision due to the wide 95% confidence interval.

Between-study subgroup analyses revealed that patients from HICs had a lower all-cause mortality rate (<0.01%) than LMICs (3.18%) (P for interaction <0.001, Fig. 2B and Appendix 12). We assessed the credibility of this subgroup effect as moderate (Appendix 13). Between-study subgroup analyses (Fig. 2B and Appendix 9) also found significant differences in all-cause mortality and secondary bacterial infection rates between patients with clade I and clade II (Appendix 12.1) and in rates of secondary bacterial infection, sepsis, cellulitis, abscess and proctitis between patients from HICs and LMICs (Appendix 12.2). However, due to the between-study comparison and test of many effect modifiers, we rated the credibility of these subgroup effects as low or very low (Appendix 13). Meta-regression analyses did not identify subgroup effects for all-cause mortality and secondary bacterial infection by proportion of vaccinated patients (Appendix 14).

Small-study effects

Visual inspection of funnel plots suggested a possible association between study size and the estimate for some outcomes, with smaller studies tending to report wider variations in effect estimates (Appendix 15).

Discussion

Among patients with non-severe mpox, moderate to high certainty evidence demonstrates that rates of mortality, ICU admission, sepsis, pneumonia, encephalitis, epiglottitis, myocarditis, gastrointestinal bleeding, and pleural effusion were all extremely low (all at or below 5 in 1000); rates of urinary retention, mechanical ventilation, urethritis, gastroenteritis, and keratitis proved between 0.7 and 1.8%; with low to moderate-certainty, rates of cellulitis, severe pain, rectal bleeding, abscess, tonsillitis, conjunctivitis, and hospitalization ranged between 2.2% and 4.4%; with low to moderate-certainty, rates of pharyngitis (8.6%), proctitis (7.4%), secondary bacterial infection (7.4%), and severe or serious rash (31.3%) proved more common (Table 2). Patients from HICs had a lower rate of pneumonia (0.02% versus 2.21%) than those from LMICs.

For patients with severe mpox, moderate to high certainty evidence demonstrated that rates of encephalitis and myocarditis were extremely low (under 5 in 1000); rates of acute kidney injury, pleural effusion, respiratory failure, and conjunctivitis ranged between 1.4% and 2.0%; with low to moderate-certainty; rates of mortality, urinary retention, abscess, sepsis, tonsillitis, pneumonia, keratitis, mechanical ventilation, and cellulitis ranged between 2.1% and 5.8%. Pharyngitis (10.9%), gastroenteritis (12.5%), rectal bleeding (16.0%), urethritis (16.7%), secondary bacterial infections (17.9%), proctitis (23.4%), and severe pain (27.3%) proved more frequent, though all supported by low or very low-certainty evidence. Patients from HICs had a lower all-cause mortality rate (<0.01% versus 3.18%) than those from LMICs.

Strengths of this systematic review include a comprehensive search across multiple databases without language restrictions that identified 170 studies with a cumulative sample size of 127,564 patients. The review process incorporated duplicate independent screening, data extraction, and risk of bias assessments, and the application of GRADE approach to evaluate the certainty of evidence for each outcome. Based on the WHO guideline development panel's discussion and suggestions, we summarized complication rates separately for patients with non-severe and severe mpox. To enhance clinical applicability, we conducted comprehensive subgroup analyses investigating variations in complication rates across key demographic and clinical parameters, including vaccination status, age groups, viral clades, and country income levels, evaluating the credibility of apparent subgroup effects using the ICEMAN instrument modified for assessment of prognosis.²⁰

Our review has limitations. Despite our comprehensive search, the available data for some outcomes were limited. The paucity of evidence precluded within-study subgroup comparisons for most prespecified subgroup analyses (e.g., clade). Due to multiple comparisons and the nature of between-study comparisons, between-study subgroup analyses yielded, for most subgroup analyses, findings with low or very low credibility. Although age is a potential determinant of risk, limited and inconsistently reported age-stratified data precluded robust within-study subgroup analyses.

When multiple studies utilized identical datasets and overlapping study periods, we retained those with larger sample sizes. To detect potential population overlaps, we cross-verified author institutional affiliations, geographic settings, and enrollment timeframes. Despite these methodological precautions, the potential for residual population overlap persists.

In most included studies, severity classification was based on clinical status at presentation or early during admission, although temporal overlap between severity assessment and reported complications cannot be entirely excluded. We acknowledge that this severity definition inherently reflects greater clinical burden at presentation. Accordingly, stratified analyses of severe versus non-severe cases should be interpreted as descriptive comparisons across clinically defined severity strata rather than only evidence that severity independently causes subsequent complications.

Because the random-effects model yielded unstable and clinically implausible estimates, we used an inverse-variance fixed-effect model to improve stability. However, this approach assumes a common underlying effect, a questionable assumption.

Although we excluded studies with fewer than 20 participants a priori, small studies with extreme event rates may still have influenced the pooled estimates. Visual inspection of funnel plots identified some asymmetry, which may suggest a possible association between study size and the estimate for some outcomes. In meta-analyses of proportions, funnel plots asymmetry can arise for reasons other than publication bias or missing studies, including variance instability, transformation effects, and heterogeneity. These findings should therefore be interpreted cautiously. Although we visually examined funnel plots, we could not fully exclude the publication bias.

We identified substantial statistical heterogeneity for several outcomes. Contributors to the heterogeneity likely include variation in study design, healthcare setting, and disease severity distribution. Although we used a fixed-effect model to improve estimate stability, pooled estimates, rather than being applicable to all contexts, reflect average incidence. For outcomes with important differences in incidence rates between studies, following GRADE guidance we rated down certainty of evidence for inconsistency. Thus, when important differences in incidence occurred, we addressed heterogeneity in our certainty assessments.

The heterogeneity in outcome definitions and reporting across studies may have influenced our pooled estimates. Data on certain subgroups (especially children, pregnant women, and elderly patients) were limited, limiting robust estimates and subgroup-specific conclusions for these vulnerable populations. Most studies included predominantly adult male patients (often >90%) and focused on the Clade IIb outbreak, potentially limiting generalizability of our findings to female populations and other epidemiological contexts.

We categorized studies as predominantly severe if they included 50% or more participants classified as having severe mpox or those hospitalized for treatment, and categorized the remaining studies as non-severe. This method may have resulted in overestimation of complication rates among patients with non-severe mpox, while potentially underestimating complication rates in the patients with severe mpox. The rationale for the chosen threshold is as follows: If we set the threshold for studies of patients with severe mpox at greater than 50% (i.e. somewhere between 51% and 99%), there will be studies classified as non-severe (i.e. less than the threshold) in which the majority of patients are actually severe. If we set the threshold for studies of patients with severe mpox at less than 50% (i.e. somewhere between 1% and 49%), there will be studies classified as severe (i.e. greater than the threshold) in which the majority of patients are actually non-severe. The choice of 50% minimizes this problem.

Previous reviews9, 10, 11 reported hospitalization rates ranging from 14.3% to 35% and mortality rates ranging from 0.03% to 4%. Our study provides lower estimates with hospitalization rates of 5.2% for all cases and 4.4% for non-severe cases, with mortality below 0.001% in non-severe and 2.9% in severe cases. This discrepancy, likely stems from previous reviews primarily incorporating data from the early phase of the 2022–2023 outbreak. Improved healthcare resource preparedness (e.g., ICU bed availability, optimized treatment protocols, and enhanced staffing), and strengthened population immunity acquired through vaccination and natural infection likely explain our lower rates of complications.

Previous systematic reviews9, 10, 11^,22,23 primarily focused on limited sets of outcomes, such as neurological complications, oral manifestations, or general clinical features, without assessing the full spectrum of complications or their incidence rates. Our review is the first to comprehensively quantify the risk of over 20 different complications with corresponding certainty of evidence ratings using the GRADE approach. Our analysis stratifies outcomes by disease severity; incorporates more recent data; includes a substantially larger cumulative sample size of nearly 130,000 patients; and includes subgroup analyses examining the influence of vaccination status, age, virus clade, and country income level on outcomes with rigorous assessment of subgroup credibility.

Our findings have important implications for clinical and health policy practice. Policy makers can use the baseline estimates provided in this review to estimate the absolute impact of interventions. The substantial differences in complication rates between non-severe and severe mpox cases underscore the importance of early clinical assessment and close monitoring of patients at risk of deterioration. Accordingly, healthcare systems should develop standardized protocols for monitoring patients with clinically severe mpox or those requiring hospitalization. Our review also identified several common complications for patients with mpox. Decision-makers can prioritize strategies that address these conditions to optimize management for these patients.²⁴ Our data also reveal health inequities in mpox care between HICs and LMICs, underscoring the urgent need for increased resources in LMICs.^25,26

For research priorities, identifying reliable predictors of disease progression from non-severe to severe mpox will require large prospective studies. The low or very low-certainty evidence for several outcomes highlights knowledge gaps requiring further investigation. Future research should focus on providing within-study data addressing important subgroup comparisons, such as age, virus clade, and vaccination status. Therapeutic clinical trials should stratify participants based on disease severity to better evaluate intervention efficacy across the spectrum of mpox presentations.

In conclusion, non-severe mpox is associated with low complication rates and very low mortality, whereas severe disease carries substantial risks of life-threatening complications and death. These findings highlight the critical need for early risk stratification and tailored management based on disease severity. The data also reveal health inequities in mpox care between HICs and LMICs, underscoring the urgent need for increased resources in LMICs.

Contributors

GG, QH, and YG conceived and designed the study. YG and QH designed and performed the search strategy. YG, WSZ, RL, ML, YZ, PL, WYZ, KN, WT screened and selected the articles. YG, WSZ, RL, DR, ML, YZ, ZL, PL, WYZ, KN, WT, FY, and JZ extracted the data and assessed the risk of bias. YG and QH analyzed the data. GG and JT supervised the data analyses. YG and QH rated the certainty of evidence. GG provided methodological advice on GRADE assessment. YG, GG, and QH interpreted the data. YG and QH drafted the manuscript. YG, FX, JT, GG, and QH revised the manuscript. All authors approved the final version of the manuscript. YG and QH accessed and verified the underlying data. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Data sharing statement

Data in this systematic review with meta-analysis are extracted from published studies available on the internet. All processed data are presented in this article and the appendix.

Declaration of interests

We declare no competing interests.