The correlation between oral and general health-related quality of life in adults: A systematic review and meta-analysis

Abstract

Objective:

The relationship between general health and oral health is critical for understanding the broader implications of oral health on overall well-being and vice versa. The impact of oral and general health on individuals can be comprehensively captured by the concepts oral and general health-related quality of life (OHRQoL and HRQoL), respectively. This systematic review and meta-analysis aimed to synthesize existing evidence on the correlation between OHRQoL and HRQoL across different adult populations.

Methods:

A comprehensive search strategy was executed across six databases (Ovid MEDLINE(R), Embase, CINAHL, APA PsycINFO, Web of Science Core Collection, and Scopus). The search included studies measuring OHRQoL with the Oral Health Impact Profile (OHIP) and HRQoL with a variety of generic patient-reported outcome measures (PROMs). Studies were included if they reported correlations between OHRQoL and HRQoL summary scores in adult populations across dental, medical, or non-patient settings. If a study examined more than one population, each correlation was included for independent analysis. Data extraction and quality assessment were conducted by independent reviewers, with disagreements resolved by a third reviewer. Random effects meta-analysis was used to summarize the OHRQoL-HRQoL correlations.

Results:

From 10 studies, 13 populations (N=6,053 participants) were included in the analysis. The correlation between general health and oral health-related quality of life was of medium size (r=0.41, 95% CI: 0.32–0.50) with high heterogeneity across populations (I²=95%). Results were not unduly influenced by individual populations, study quality, or publication bias.

Conclusions:

The correlation between oral health and general health is of medium size, highlighting the potential for medical-dental integration to enhance patient and community health outcomes.

1. Introduction

Understanding the extent to which general health and oral health are related is crucial for recognizing the significance of oral health as an integral component of overall health. The practical diagnostic and therapeutic implications of this relationship are extensive, as knowledge about one concept can inform the other. While there are countless ways to describe the connection between general and oral health, characterizing this relationship from the patient’s perspective is especially important, as it reflects the lived experience of health impacts.

To effectively summarize the relationship between oral health impact and general health impact within a specific population, expressing this connection through a single numerical value is a necessary step. This number expresses the strength of the oral health-general health relationship. While this provides foundational knowledge for the field of dentistry, the size of the relationship also has practical applications as it helps clinicians, community health workers, and researchers with the implementation of medical-dental integration efforts.

Health-related quality of life (HRQoL) is a multi-dimensional construct. While several definitions for HRQOL have been proposed, there is broad agreement that HRQOL represents the effect of a medical condition and/or its consequent therapy upon a patient.¹ In this study, HRQoL serves as a primary outcome of interest.

Oral health-related quality of life (OHRQoL) can simply be understood as the component of HRQOL that relates to the effects of oral diseases and dental interventions on patients.² It is a four-dimensional construct that measures self-reported oral health impacts across the dimensions of Oral Function, Orofacial Pain, Orofacial Appearance, and Psychosocial Impact.²

Research on the relationship between HRQoL and OHRQoL has been conducted across various groups of individuals. These include dental patients, where the primary focus is on the impact of oral health; medical patients, where overall health concerns take precedence; and the general population, most of whom do not experience significant health or oral health issues that prompt them to seek care.

In dental patients, where oral health impact is the primary concern, Farzadmoghadam et al. (2020) examined how oral health affects HRQoL in dental patients prior to and after dental implant treatment and reported a small correlation.³ Sekulic et al. (2020) studied dental patients with a variety of oral diseases and demonstrated that effective dental interventions improved not only patients’ OHRQoL but also their HRQoL, with a large correlation reported.⁴

For medical patients, where overall health impact is paramount, Aparna et al. (2022) explored the experiences of head and neck cancer and radiation patients, while Rojas-Alcayaga (2022) focused on women with Sjögren’s syndrome. Both studies highlighted that these medical populations experience notable effects on both their HRQoL and OHRQoL, with correlations of medium size.^5,6

The quantity and quality of available research across these populations provide a solid foundation for synthesizing evidence on the relationship between general health and oral health. A systematic review aims to comprehensively identify all relevant literature and reproducibly obtain the same results if conducted again. A meta-analysis can then quantitatively synthesize the evidence across the various studies, providing a more accurate and precise effect estimate while also enabling the investigation of factors influencing the effect estimate.

Therefore, we conducted a systematic review with meta-analysis to study the correlation between general health and oral health, offering a comprehensive of this relationship.

2. Methods

2.1. Search Strategy

This systematic review and meta-analysis is registered in the PROSPERO registry and has been assigned study ID CRD42024555921.⁷ This project was developed, executed, and documented in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.⁸

A search strategy was designed by a health sciences librarian (NTM) and utilized controlled vocabulary and natural language keywords to encompass the concepts of oral health-related quality of life (OHRQoL) and general health-related quality of life (HRQoL). The search was developed in Ovid MEDLINE(R) ALL and then translated across five databases: MEDLINE(R) ALL (via Ovid), Embase via Ovid (Embase+Embase Classic), CINAHL, APA PsycINFO (via Ovid), Web of Science Core Collection, and Scopus from their inception to February 3, 2023. Search results were limited to the English language, humans, and adults, but no other limitations were applied. MEDLINE(R) ALL is included in Appendix I. To ensure no relevant records were omitted, we hand-searched reference lists of relevant records, including those from the included studies. The search was rerun on June 10, 2024, prior to the final analysis to ensure the most recent records have been captured.

Search results were then uploaded to Covidence⁹ for deduplication and screening. Two calibrated expert reviewers (DRP and AI) conducted independent title/abstract and full-text screening and quality assessment, while a third reviewer (MTJ) resolved any conflicts that arose during the process.

2.2. General Health and Oral Health Outcome Measurement

General health was assessed by patient-reported outcome measures (PROMs) to evaluate HRQoL. Commonly used HRQoL instruments include the Medical Outcomes Study 36-item¹⁰ and 12-item Short Form¹¹ (SF-36 and SF-12, respectively), the World Health Organization Quality of Life Assessment (WHOQoL)¹² and the European Quality of Life 5-Dimension (EQ-5D)¹³ questionnaires.

Oral health was assessed by a dental patient-reported outcome measure (dPROM), the Oral Health Impact Profile (OHIP).¹⁴ The OHIP is the most widely used multi-item dPROM, which has many lengths and language versions and has been shown to be a valid and reliable tool for OHRQoL measurement.^15,16 The tool is flexible in that it can be used in a variety of settings, including community, dental, and medical populations.

2.3. Inclusion and Exclusion Criteria

This study included populations investigating the unadjusted correlation, i.e., a correlation that did not take other factors into account, between OHRQOL and HRQoL summary scores. in adult populations (≥18 years) in community, dental, or medical settings. Included populations must be adults (≥18 years) in community, dental, or medical settings. OHRQoL measured using at least one version of the Oral Health Impact Profile (OHIP), including OHIP-5,^15–16 OHIP-14,¹⁹ OHIP-19,²⁰ OHIP-20,²¹ or OHIP-49,¹⁴ with the original item response format (ordinal scale ranging from 0 = “never” to 4 = “very often”) was included. Additionally, we only included studies that measured HRQoL using a patient-reported outcome measure (PROM) listed in the systematic reviews by Pequeno et al. (2020) or Hernández-Segura et al. (2022) (Table 1).^22,23 To be eligible, studies must be published in English, present primary data, and be in the form of journal articles. Eligible study designs include randomized trials and observational studies, such as cross-sectional, cohort, and case-control studies.

Table 1.

Generic HRQoL PROMs Eligible for Inclusion in this Systematic Review and Meta-analysis

Abbreviated generic HRQoL instrument	Included in the systematic review by Pequeno et al. (2020)22	Included in the systematic review by Hernández-Segura et al. (2022)23
AQoL		X
AQoL-4D	X
AQoL-6D		X
AQoL-8		X
CASP-16	X
CAT-Health		X
CDC-HRQoL-4	X
CDC-HRQoL-14	X
EQ-5D	X	X
EQ-5D-5L		X
EuroQol-VAS	X
EUROHIS-QoL 8-item	X
FSQ		X
HALex		X
HINT-20		X
HUI-2		X
HUI-3		X
MQLI		X
NHP		X
PAT-5D-QOL		X
PROMIS	X
QoL scale	X
QWB-SA		X
SF-6D		X
SF-8	X
SF-12	X	X
SF-36	X	X
Stark QoL		X
SWED-QUAL		X
WHOQOL-BREF		X

Abbreviations: HRQOL, health-related quality of life; PROM, patient reported outcome measure; QOL, Quality of Life; AQoL, Assessment of Quality of Life; AQoL-4D and AQoL-6D, Assessment of Quality of Life, 4- and 6-dimensions respectively; AQoL-8, Assessment of Quality of Life 8-item; CASP-16, Control, Autonomy, Self-realization and Pleasure; CAT-Health, computer-adaptive test-health; CDC HRQoL-4, Healthy Days core questions; CDC HRQoL–14, Healthy Days measures; EQ-5D, European Quality of Life Instrument (5L, 5 level); EuroQol-VAS, European Quality of Life, Visual Analogue Scale; EUROHIS-QoL 8-item, European Health Interview Survey Quality of Life 8-item questionnaire; FSQ, functional status questionnaire; HALex, health and activity limitation index; HINT-20, 20-item Health-related Quality of Life Instrument; HUI, Health Utility Index (2- and 3-item); MQLI, multicultural quality of life index; NHP, Nottingham Health Profile; PAT-5D-QOL, Paper-and-pencil semi-adaptive questionnaire for 5 domains of health-related quality of life; PROMIS, Patient-Reported Outcomes Measurement Information System-Global Health Scale; QoL scale, Quality of life scale; QWB-SA, Quality of Well-being Self-Administered; SF-6D, Medical Outcomes Study Short Form, 6 Dimensions; SF-8, −12, and −36, Short-Form Health Survey with 8-, 12-, and 36-Items respectively; Stark QoL, Stark Quality of Life Questionnaire; SWED-QUAL, Swedish Health-related Quality of Life Survey; WHOQOL-BREF, World Health Organization Quality of life.

Studies were excluded if they only correlated OHRQoL summary scores with HRQoL component scores, such as the mental and physical health components of the SF-36.¹⁰

2.4. Data Extraction

Data extraction methods were developed to reflect those of project Mapping Oral Disease Impact with a Common Metric (MOM) to support the goal of standardization, precision, and quality.²⁴ Complete data extraction was performed independently by two reviewers (DRP and MTJ) for all studies that included a correlation between OHIP summary scores and HRQoL instrument summary scores. Disagreements were discussed between the two reviewers, and a final decision was reached. In situations where studies reported multiple correlations, such as in Reissmann’s work²⁵, where the correlation in two distinct populations was determined, each correlation was included independently in the analysis.

If a study presented both a summary score correlation and an adjusted correlation, only the unadjusted correlation was extracted. Adjusted correlations were not considered as the relationship between OHRQoL and HRQoL was the target without taking into account additional factors.

2.5. Quality Assessment

To assess the quality and risk of bias in the included studies, we adapted a quality assessment tool originally designed for prevalence studies to suit the objectives of this systematic review.²⁶ The adapted tool, detailed in Appendix II, included criteria tailored to evaluate studies reporting correlations between OHRQoL and HRQoL. After a calibration session, two reviewers (DRP and AI) independently assessed study quality, resolving discrepancies through consensus, with input from a third reviewer (MTJ) when needed.

2.6. Statistical Analysis

In the meta-analyses of correlation coefficients, Fisher’s z transformation was applied to achieve normal approximation and stabilize the variability of the correlation coefficients. After the meta-analyses, the pooled Fisher’s z values were back-transformed to the original correlation scale for interpretation.²⁷ The analyses were conducted using the restricted maximum likelihood (REML) method to account for between-study variance.²⁸ After synthesizing the correlation coefficients, we conducted a subgroup analysis to explore variations in the correlations. We hypothesized that the source of the health impact within the populations classified as “dental”, “medical,” and “non-patient” populations, could explain the study heterogeneity. “Non-patients” refers to individuals who are not specifically seeking care for medical or dental issues. Additionally, non-patient populations may exhibit different health impact experiences and correlational outcomes due to varying levels of health awareness and access to care. We anticipated that the correlations would differ depending on the population, as dental patients (e.g., patients who are partially edentulous, or have other dental impacts) may experience less impact on their HRQOL compared to medical patients (e.g., individuals with cancer), whose OHRQOL might be more severely affected.

In addition, whether the study quality influenced results, or individual populations had an undue influence on the results (leave-one-out method by excluding one population at each analysis) was investigated. Funnel plots were generated and visually inspected to assess the potential for publication bias.

Correlation coefficients were classified according to Cohen’s guidelines²⁹ as small (0.1), medium (0.3), and large (0.5), based on their absolute magnitudes. We considered correlations <0.1 as “trivial.” All analyses were performed using R (version 4.3.3) with the “metafor” package (4.6–0)³⁰

3. Results

3.1. Literature search

We found 944 records in our search. After removing 611 duplicates, we had 333 unique records to screen based on their titles and abstracts. We excluded 237 of these, leaving 96 reports for a detailed review. After reviewing these, 10 studies met our criteria and were included in the final analysis. Figure 1 shows this selection process in detail. The meta-analysis included 13 populations derived from 10 different studies.

Figure 1:

PRISMA flowchart of report selection process

3.2. Study Characteristics

The studies span from 2013 to 2024 (Table 2). The ten studies covered a wide range of populations, settings, and methodologies across nine countries; Brazil, Chile, Germany, India, Iran, Ireland, Serbia, Taiwan, and USA. Convenience sampling was most common, and sample sizes ranged from 30 up to 2076 participants. The populations featured participants with mean ages ranging from 42 to 79 years old. Two populations included Sjögren’s syndrome patients (a disease that is significantly more common in female populations), while others included a more balanced mix of genders. The populations studied included other medical groups like head and neck cancer patients, dental patient populations, as well as non-patient population settings. Most studies used the OHIP-14 to assess OHRQoL, with two studies using different variations of the tool (OHIP-49 and OHIP-5). The most commonly used HRQoL instrument was the SF-12 and SF-36.^10,11 Recall periods were often unspecified, and while most studies used bivariate analysis, some employed Structural Equation Modeling (SEM). Zucoloto et al (2016) was the only included study that provide an adjusted correlation, as they presented a correlation adjusted for age, pain, and presence of chronic disease.³¹ This range of study and population characteristics highlighted the diverse approaches to examining the relationship between oral health and general health.

Table 2:

Population Characteristics

First Author	Year	Country	Sample Size	Age of Population (years)	Women (%)	Population	Population Setting	OHIP Version	HRQoL Instrument	Recall Period (OHIP)	Recall Period (HRQoL)	Method	Adjustment Performed
Aparna, KS5	2022	India	150	Mean ± SD: 49.75 ± 13.44	34	head and neck cancer/radiation patients	medical population	OHIP-14	SF-12	missing	missing	Bivariate (Pearson’s correlation coefficient)	none
da Mata C32	2019	Ireland	93	Mean: 79.2	53	frail, older patients	medical population	OHIP-14	EQ-5D	missing	missing	Bivariate (Pearson’s correlation coefficient)	none
da Mata C32	2019	Ireland	234	Mean: 71.8	56	non-frail, older people	non-patient population	OHIP-14	EQ-5D	missing	missing	Bivariate (Pearson’s correlation coefficient)	none
Farzadmoghadam M3	2020	Iran	93	Mean: 42 Range: 18 – 81	61	partially edentulous dental patients	dental population	OHIP-14	EQ-5D	3 months	missing	Bivariate (Spearman’s correlation coefficient)	none
Farzadmoghadam M3	2020	Iran	93	Mean: 42 Range: 18 – 81	61	implant patients	dental population	OHIP-14	EQ-5D	3 months	missing	Bivariate (Spearman’s correlation coefficient)	none
Kuo HC33	2018	Taiwan	517	age 60 – 74: 288 subjects age 75+: 229 subjects	52	older people	non-patient population	OHIP-14	WHOQoL-Bref	12 months	missing	Bivariate (Pearson’s correlation coefficient)	none
Paulson DR7	2024	USA	607	Mean ± SD: 43.7 ± 17.6	68	general population	non-patient population	OHIP-5	PROMIS v.1.2 Global Health	12 months	12 months	SEM	none
Reissmann DR25	2013	Germany	311	Mean ± SD: 61.7 ± 13.7	57	dental patients	dental population	OHIP-14	SF-36	1 month	missing	SEM	none
Reissmann DR25	2013	Germany	811	Mean ± SD: 50.1 ± 16.3	53	general population	non-patient population	OHIP-14	SF-36	1 month	missing	SEM	none
Rojas-Alcayaga G6	2022	Chile	31	Mean ± SD: 43.5 ± 13.4 Range: 29 – 68	100	women with Sjogren’s syndrome	medical population	OHIP-14	EuroQol-VAS	missing	missing	Bivariate (Spearman’s correlation coefficient)	none
Sekulic S4	2020	USA	2076	Mean ± SD: 54.7 ± 16.2	60	dental patients	dental population	OHIP-49	PROMIS v.1.1 Global Health	1 month	missing	SEM (with HRQoL factor)	none
Vujovic S34	2023	Serbia	30	Mean ± SD: 63.93 ± 10.5	97	patients with primary Sjogren’s syndrome	medical population	OHIP-14	EQ-5D	missing	missing	Bivariate (Spearman’s correlation coefficient)	none
Zucoloto ML31	2016	Brazil	1007	Mean ± SD: 45.7 ± 12.5	72	dental patients	dental population	OHIP-14	SF-36	missing	4 weeks	SEM	age, pain, and presence of chronic disease

3.3. Quantitative data analysis, including subgroup analysis

The size of HRQOL-OHRQoL correlations varied. A “large” correlation (r=0.50) was observed five times, a “medium” correlation (r=0.30) four times, and a “small” correlation (r=0.10) also four times. About two-thirds of the correlation coefficients fell between 0.25 (“small”) and 0.57 (“large”), centering around a summary estimate of r=0.41 (“medium”). The 95% confidence interval (CI) further supported this classification, with values ranging from 0.32 to 0.50, indicating “medium” correlations (where an upper limit of 0.5 is considered “medium” in this context).

The 95% prediction interval, i.e., the range of possible correlations found in a new study that would be selected at random from the same population as the studies in the meta-analysis, was considerably wider than the 95% confidence interval reported above. The prediction interval ranged from r=0.07 (“negligible” correlation) to 0.67 (“large” correlation).

Study population heterogeneity (the proportion of total variance between populations that is attributed to heterogeneity rather than sampling error) was 93%, a level considered “high” according to Higgins et al.³⁵ and statistically significant (Q = chi2(12) = 109.92, p<0.001).

The major factor that was hypothesized to contribute to heterogeneity, the origin of the health impact, categorized as “dental patients,” “medical patients,” and “non-patient” subjects, failed to explain differences between populations. The three subgroup estimates were 0.41, 0.42, and 0.43, which were nearly identical to the overall synthesized correlation coefficient of 0.41 (Figure 2). The subgroup test indicated non-significant differences between subgroups (p=0.97). Due to the small number of populations in the subgroups, uncertainty for these estimates, as described by the 95% CIs, was considerable.

Figure 2.

Subgroup Analysis of HRQOL-OHRQoL Correlations by Population Settings

3.4. Quality Assessment

The majority of studies exhibited no to low bias (Figure 3). For bias assessment questions 5–7, 100% of the studies showed no bias (green). Only studies that employed appropriate statistical methods, such as Structural Equation Modeling (SEM) and bivariable analyses, and used reliable instruments like OHIP and generic HRQoL measures from notable systematic reviews were included in this project. Our inclusion criteria ensured that the selected studies maintained a high methodological standard, resulting in no observed bias in these areas.

Figure 3:

Quality Assessment Summary

The greatest risk of bias among the studies was attributable to question 1, sample representativeness, where 50% of the included studies exhibited bias. This was attributable to half of the studies using convenience sampling compared to consecutive or random sampling. Overall, while studies used objective and standardized criteria to measure QoL reliably, along with appropriate statistical analysis methods, variations in sample representativeness and some unclear or insufficient data analyses across the identified sample highlight areas where bias could potentially influence the results.

Study quality affected the results somewhat, with better studies having higher HRQoL-OHRQoL correlations than lower-quality studies. Deleting the study with the worst study quality (the only study that did not describe the study setting informatively) in the leave-one-out analysis (section 3.5) increased the summary estimate from 0.41 to 0.46 (still a “medium” effect), and the 95% CI slightly narrowed from 0.32–0.50 to 0.39–0.54 (still mainly “medium” effects).

3.5. Leave-one-out analysis

The leave-one-out analysis demonstrated the robustness of the results, with summary estimates ranging from 0.39 to 0.44, all of which were very close to the overall summary estimate of 0.41. This consistency suggests that the main findings were not unduly influenced by any individual population, reinforcing the stability of the OHRQoL-HRQoL correlation across diverse populations.

3.6. Small-study effects/publication bias

Publication bias was observed, with a small but statistically significant effect (Egger’s test: P=0.04). The funnel plot and trim-and-fill analysis indicated that a small/medium-sized study was missing, leading to a conservative estimate of the HRQoL-OHRQoL correlation of 0.41. When including the imputed study (14 populations), the correlation slightly increased to 0.43 (95% CI: 0.34 to 0.52). However, the difference between the two estimates (0.43 vs. 0.41) was minimal, suggesting that publication bias had a minimal effect on the results. (Figure 4).

Figure 4:

Funnel plot after imputation of one additional study (N=14 studies)

4. Discussion

Our systematic review with meta-analysis found a correlation between oral health and health-related quality of life of medium size (r=0.41), indicating that oral health and general health, from the patient’s perspective, shared about 17% of their information.

This result seemed to be robust against methodological influences such as study quality, influence from individual populations, and small-study effects (e.g., publication bias). While the size of the OHRQoL correlation demonstrated high heterogeneity among populations, i.e., the true correlation between the two concepts varied substantially, the source of this variation could not be identified in this study. We could exclude the population type, categorized as dental patients, medical patients, non-patients, as an influential factor, but additional sources of heterogeneity were not explored in the absence of strong hypotheses.

The relationship between oral health and general health can be approached from several angles. One approach might involve cataloging systemic diseases and comparing them with all oral diseases, illustrating how the structure and function of the entire body’s orofacial system interact with overall health. This systematic review with meta-analysis approached the relationship from the patient’s perspective. The concept HRQoL describes the impact of health status on a person’s quality of life. In this study, seven HRQOL instruments were included, measuring the target construct based on somewhat different underlying models. This variation is evident in correlations between these instruments; for example, the EQ-D5 correlated with SF-12 scores with r=0.41 (medium effect size) and r=0.68 (large effect size).³⁶ The concept OHRQoL describes the impact of oral health status on a person’s quality of life. This study used a globally used³⁷ and psychometrically well investigated² instrument – the Oral Health Impact Profile.

Our project has limitations which can mainly be attributed to a sample size that was not large. A sample size of 13 sample populations limited subgroup and sensitivity analyses. For example, we only investigated the influence of population type on the heterogeneity of the correlations. Other factors, such as the language version of the HRQOL/OHRQoL instruments or the place of the studies could be relevant but could not be studied.

Overall, our sample size was sufficient to provide precise results for the main result, as the 95% confidence interval indicated that the summary measure will likely stay in the medium effect size category when the study is repeated.

Some factors that influence the magnitude of the oral health-general health relationship using HRQOL/OHRQoL instruments are known, but we have not formally investigated them:

1. Using longer instruments will provide higher correlations as the construct includes more indicators. Using a 49-item OHIP should provide more precise results than using a 5-item OHIP. However, both instrument scores correlated with r=0.93 in international general population subjects and dental patients,^38,39 and consequently, we deemed the influence of this factor small.

2. Correlations derived from latent variable methods such as a structural equation model (SEM) should be higher because they take measurement error into account as compared to Pearson and Spearman correlations. In our previous study, Pearson and Spearman correlations of 0.39 and 0.36, respectively, were considerably lower than the SEM-derived correlations 0.51 and 0.52.⁴⁰ For studies reporting latent variable methods-derived correlations, we recommend simple, bivariable correlations should also be reported.

3. Analyses adjusting for other variables’ influence on the HRQOL/OHRQoL correlation should provide lower correlations. Correlations adjusted for age, gender, and tooth count were indeed lower than the unadjusted correlation, but the change was small.⁴⁰ For studies reporting adjusted correlations, we recommend that unadjusted correlations should also be reported.

As mentioned above for subgroup and sensitivity analyses, the sample size prevented informative analyses for these three methodological factors that influence correlation magnitude.

5. Conclusion

Oral health and general health are bidirectionally connected. Oral health is a component of overall health, and therefore, it influences general health. In return, general health also affects oral health. However, the magnitude of this relationship, expressed as a correlation, was not known, but it is critical to understanding its practical relevance. Our findings provide, for the first time, a summary of the size of the oral health-general health relationship, quantified in a single number (r=0.41). This has significant implications for clinical and public health contexts, as the strength of this correlation is a key factor in determining its applicability for diagnostic, therapeutic, and preventive interventions in dental, medical, and community settings. The medium-sized correlation identified in this systematic review and meta-analysis suggests that the connection between oral health and general health is substantial enough to support practical applications, such as medical-dental integration in patient care and public health initiatives. The shared information between the two domains provides a foundation for improving the health of individuals and communities through coordinated care strategies across settings.

Appendix I: Medline(R) ALL search strategy

Ovid MEDLINE(R) ALL <1946 to June 10, 2024>
1	HRQoL.tw,kf.
2	(general adj3 “health related quality of life”).tw,kf.
3	PROMIS.tw,kf.
4	systemic health.tw,kf.
5	1 or 2 or 3 or 4
6	exp Oral Health/ and exp “Quality of Life”/
7	OHRQoL.tw,kf.
8	(“oral health” adj3 “related quality of life”).tw,kf.
9	OHIP.tw,kf.
10	“oral health impact profile”.tw,kf.
11	6 or 7 or 8 or 9 or 10
12	5 and 11
13	exp animals/ not exp humans/
14	12 not 13
15	(exp child/ or exp infant/ or exp adolescent/) not exp adult/
16	14 not 15
17	limit 16 to the English language

Appendix II: Adapted Quality Assessment Tool for Studies Reporting Correlations between OHRQoL and HRQoL

Appraisal tool adapted from Munn et al, 2014.^* Not applicable items from Munn et al, 2014, for the assessment of correlations were excluded from the assessment tool (highlighted in gray) and reasons for exclusion are depicted.

Item No.	Items for the assessment of risk of bias	Yes /No/Unclear/ Not applicable
1.	Was the sample representative of the target population?	Evaluate the type of sampling from source population to eligible patients (random sample better than consecutive sample, consective sample better than convenience sample):
		Yes: Random sample (general population) or consecutive sample (patient population)
		No: Convenience sample
		Unclear: Sampling type not mentioned
		Note: Is random sampling possible? Only feasible when the population can be enumerated, i.e., all the units of the population are known. For most clinical populations this is difficult because patients enter the treatment center on an ongoing basis.
2.	Were study participants recruited in an appropriate way?	Is source population appropriate?
		Yes: Population registers or similar instruments that cover general population well. Are treatment center’s patients representative for patients in general?
		No: Evidence exists that source population is not appropriate.
		Unclear: No information provided
3.	Was the sample size adequate?	Not applicable: Sample size determines the precision of the estimate. Studies’ sample size is always adequate for the purpose of this study.
4.	Were the study subjects and the setting described in detail?	Yes: Sufficient information so that reader knows what subjects were studied in what setting. For example, it should be clear to the reader what disease or oral health condition the individuals had. For example, if patients/individuals with dental caries were studied, caries should be diagnosed with an accepted diagnostic system.
		No: Sufficient information not provided.
		Unclear: Should avoid using this category as it is not clear whether this category is different from the “no.”
5.	Was the data analysis conducted with sufficient coverage of the identified sample?	Evaluate the subjects and item nonresponse as follows: Subject nonresponse - Compare eligible patients to actual patients (how many subjects/patients participated from the targeted sample? General population example: 1000 subjects were identified and invited to the study and 600 subjects participated. Patients: A consecutive sample of 100 patients was targeted and 95 patients participated). Item nonresponse – Compare actual subjects to actual subjects with data: 600 general population subjects participated, but only 580 had sufficient OHIP and HRQoL data to compute summary scores. Add the subject and item response rates together and evaluate the result against the thresholds mentioned below to determine the quality assessment category.
		General population subjects:	Patients:
		Yes: 50% or larger	Yes: 80% and larger
		No: < 50%	No: <80%
		Unclear: Information to assess non-response not provided
		Note: The bias from non-response does not depend on the actual magnitude of the response rate. A response rate of 20% in a general population random sample could be perfectly fine if no distortion would be present and this sample represents the target population well. Typically, the distortion factors are not known. Therefore, with a “high” response rate less opportunity exist that the sample is biased. Response rates in patients (as opposed to general population) are usually not a problem.
6.	Were objective and standard criteria used for the measurement of the condition?	Yes: Both, OHIP language-version and the specific HRQoL instrument language-version, are referenced; or authors provide evidence for OHIP and HRQoL instruments’ score validity in the paper.
		No: OHIP language-version and the specific HRQoL instrument language-version are not referenced and no evidence of score validity is provided.
		Note: Since this criterion was part of our study’s inclusion criteria, this question should technically always have a “Yes” response; so please double check to make sure it’s a “Yes.” If it’s a ‘No’ then study must be excluded from review.
7.	Was the condition measured reliably?	Yes: Both, OHIP language-version and the specific HRQoL instrument language-version, are referenced; or authors provide evidence for OHIP and HRQoL instruments’ score reliability in the paper.
		No: OHIP language-version and the specific HRQoL instrument language-version are not referenced and no evidence of score reliability is provided.
		Note: Since this was part of our study inclusion criteria, this question should technically always have a “Yes” response; so please double check to make sure it’s a “Yes.” If it’s a ‘No’ then study must be excluded from review.
8.	Was there appropriate statistical analysis?	Yes: The correlation measure was sufficiently described in the methods section so that it is clear what measure was used [look for the following: Bivariable correlations (Pearson’s or Spearman’s correlation coefficients, or standardized beta coefficients from simple linear regression models, or SEM correlations].
		No: The correlation measure was NOT sufficiently described in the methods section so that it is clear what measure was used. For example, authors just mentioned in the results section that variables “correlated” without describing in the methods section what correlation was computed.
		Unclear: This category should be avoided as it is not clear whether this category is different from the “No” response.
9.	Are all important confounding factors/subgroups/differences identified and accounted for?	Not applicable: Relationship between OHRQoL and HRQoL were assessed purely from a descriptive perspective without any intent to investigate causality. Hence, the concept of confounding does not apply to the purpose of our study.
10.	Were subpopulations identified using objective criteria?	Not applicable: For each correlation coefficient provided in the manuscript, regardless of whether the correlation coefficient was for the total sample or for subgroups, e.g., separate correlation coefficients for men and women, subjects younger than 60 years and subjects 60 years and older, patients with disease A and disease B, a risk of bias assessment is performed.
		For each correlation coefficient, selection bias (questions #1, 2, 4, and 5), measurement bias (questions #6 and 7), and the analysis (question #8) will be evaluated. While measurement bias is likely similar in subgroups, selection bias can differ. For example, patients with disease A could be well characterized, but patients with disease B could not be well characterized, resulting in a “Yes” for disease A and a “No” for disease B for question #4.
		When bias assessments for all correlation coefficients are performed, the worst category will be selected to represent the quality of the entire study.

^*Original tool: Munn Z, Moola S, Riitano D, Lisy K. The development of a critical appraisal tool for use in systematic reviews addressing questions of prevalence. International journal of health policy and management. Aug 2014;3(3):123–8. doi:10.15171/ijhpm.2014.71