Risk Factors for the Occurrence of Benign Paroxysmal Positional Vertigo: A Systematic Review and Meta-Analysis

Background and Purpose: The lifetime prevalence of benign paroxysmal positional vertigo (BPPV) is high, especially in the elderly. Patients with BPPV are more susceptible to ischemic stroke, dementia, and fractures, severely reducing quality of life of patients. Many studies have analyzed risk factors for the occurrence of BPPV. However, the results of these studies are not identical. We performed this meta-analysis to determine potential risk factors associated with the occurrence of BPPV. Methods: PubMed, EMBASE, and the Cochrane Library (January 2000 through March 2020) were systematically searched for eligible studies analyzing risk factors for the occurrence of BPPV. Reference lists of eligible studies were also reviewed. We selected observational studies in English with a control group and sufficient data. Pooled odds ratios (ORs) or the mean differences (MDs) and 95% confidence intervals (CIs) were calculated to measure the impacts of all potential risk factors. Heterogeneity among studies was evaluated using the Q-test and I2 statistics. We used the random-effect model or the fixed-effect model according to the heterogeneity among the included studies. Results: We eventually included 19 studies published between 2006 and 2019, including 2,618 patients with BPPV and 11,668 participants without BPPV in total. In this meta-analysis, the occurrence of BPPV was significantly associated with female gender (OR = 1.18; 95% CI, 1.05–1.32; P = 0.004), serum vitamin D level (MD = −2.12; 95% CI, −3.85 to −0.38; P = 0.02), osteoporosis (OR = 2.49; 95% CI, 1.39–4.46; P = 0.002), migraine (OR = 4.40; 95% CI, 2.67–7.25; P < 0.00001), head trauma (OR = 3.42; 95% CI, 1.21–9.70; P = 0.02), and total cholesterol level (MD = 0.32; 95% CI, 0.02–0.62; P = 0.03). Conclusion: Female gender, vitamin D deficiency, osteoporosis, migraine, head trauma, and high TC level were risk factors for the occurrence of BPPV. However, the effects of other risk factors on BPPV occurrence need further investigations.


INTRODUCTION
Benign paroxysmal positional vertigo (BPPV) is one of the most common types of vestibular vertigo, accounting for ∼17-42% of patients with vertigo (1,2). Patients suffering from BPPV are characterized by transient episodes of vertigo provoked by head position changes (3). The lifetime prevalence of BPPV is estimated at 2.4%, and the 1-year prevalence of BPPV in the elderly is much higher than that in other age groups (4). In addition, some studies have suggested that patients with BPPV were more susceptible to future ischemic strokes, dementia, and fractures, which severely reduces quality of life of patients, especially in the elderly (5)(6)(7). Thus, identifying potential risk factors for the occurrence of BPPV can help prevent this disease. Furthermore, some serum indicators may also help improve the clinical misdiagnosis of some atypical BPPV.
Although canalith repositioning maneuver is an effective treatment for BPPV, nearly 50% of patients experienced at least one recurrence in 2 years after treatment (8). Many of the risk factors investigated in this meta-analysis, such as hypertension and migraine, may also be risk factors for BPPV recurrence, which may help improve the treatment and prognosis of this disease (9).

Literature Search Strategy
The electronic databases PubMed, EMBASE, and the Cochrane Library (January 2000 through March 2020) were systematically searched by two researchers (JB Chen and WS Zhao) for eligible observational studies analyzing risk factors for the occurrence of BPPV. The MeSH terms "Risk Factors, " "Benign Paroxysmal positional vertigo, " and all related free words were combined to search relevant literature as comprehensively as possible. Reference lists of all eligible studies were also reviewed to identify other potentially relevant studies.

Selection Criteria
Articles included in this meta-analysis must meet the following criteria: (1) clearly define the experimental group (patients diagnosed with BPPV) and the control group (participants or patients without any history of vertigo); (2) all BPPV patients included in studies were diagnosed by a characteristic history of recurrent positional vertigo or a typical nystagmus during Dix-Hallpike tests or Roll test; (3) reported sufficient data on risk factors investigated in our meta-analysis; (4) the outcome was BPPV; (5) case-control studies, cohort studies, or other observational English studies analyzing relevant risk factors for occurrence of BPPV. The following studies were excluded from this meta-analysis: (1) sufficient information could not be obtained; (2) the outcome was the recurrence of BPPV, not the occurrence of BPPV.

Data Extraction and Quality Assessment
Two reviewers (JB Chen and WS Zhao) independently assessed the quality of each study included in this meta-analysis using the Newcastle Ottawa Scale (27). Studies were evaluated according to three dimensions including selection, comparability, and outcome (cohort studies) or exposure (case-control studies). Any discrepancies between the two reviewers were resolved through discussion with another author (XJ Yue). The total NOS scores of all included articles are shown in Table 1. Studies with NOS scores ≥7 were considered high quality.
A standardized pre-extraction form was used to extract available data, including study characteristics, sample demographic information, medical comorbidities, and serum indicators. For each risk factor, we performed a detailed analysis and compared their definitions in the original literature. Data extraction was independently completed by the same two reviewers according to the revised extraction form from January 2020 to February 2020. All disagreements between the two reviewers were fully discussed, and furthermore a third reviewer (XJ Yue) was consulted for unresolved discrepancies to reach a consensus. The following data were extracted for each included study: (1) Study characteristics: fist author, study region, sample size, publication year, and study design (case-control or cross-sectional study); (2) sample demographic information: gender, age (mean ± SD), body mass index (BMI), smoking, drinking, and regular exercise; (3) medical comorbidities of participants: osteoporosis, osteopenia, migraine, stroke, head trauma, hypertension (HTN), diabetes mellitus (DM), and hyperlipidemia; (4) serum indicators: total cholesterol level (TC) (mmol/L) and serum vitamin D level (ng/ml).

Statistical Analysis
The impacts of all potential risk factors on the occurrence of BPPV were measured by calculating odds ratios (ORs) or mean differences (MDs) and 95% confidence intervals (CIs). ORs were calculated for categorical variables including female gender, osteoporosis, osteopenia, migraine, stroke, head trauma, hypertension, DM, hyperlipidemia, smoking, drinking, and regular exercise. MDs were calculated for continuous variables including age, serum vitamin D level, and TC level. Heterogeneity among studies was tested and quantified using the Cochrane Q-test and I 2 statistics. A fixed-effect model was used when heterogeneity was not significant (I 2 < 50%) and a randomeffect model was used when heterogeneity was significant (I 2 > 50%) (30). In addition, funnel plots of some risk factors were used to assess the publication bias in included studies. All statistical analyses were performed using the Review Manager 5.3 software.

Study Selection and Characteristics
The literature search produced a total of 256 records. Six additional records were identified through screening the reference lists of each study included in this meta-analysis. After 49 duplicates were removed, we further excluded 158 records through screening the titles/abstracts. The remaining 55 studies were assessed by reviewing the full text in detail. Finally, 19 studies published between 2006 and 2019 were included in our meta-analysis. A flow diagram of the literature selection was present in Supplemental Figure 1. A total of 14,286 participants were included in this meta-analysis, including 2,618 patients with BPPV and 11,668 controls without BPPV. Most studies were conducted in Asia. Furthermore, 5 studies were prospective (17-19, 23, 26), 12 were retrospective (10-16, 20, 22, 24, 28, 29), and 2 were cross-sectional (4,25). In addition, the NOS scores of each study ranged from 5 to 9, indicating a medium and high quality of all included studies. Baseline characteristics of each study and pooled results for each risk factor were summarized in Tables 1, 2, respectively. Funnel plots of some risk factors showed that no significant publication bias was found in the included studies (Supplemental Figures 2-5)

Female Gender
Fifteen studies involving 13,819 participants analyzed the relationship between female gender and the occurrence of BPPV.
Four studies were not included in this risk factor analysis, because the participants in these studies were all male or female. The pooled results showed that female had a slightly higher risk of BPPV compared with male (OR = 1.18; 95% CI, 1.05-1.32; P = 0.004) (Figure 1). We used a fixed-effect model, because the statistical heterogeneity between these studies was not significant (I 2 = 49%; P = 0.02).

Age
Thirteen studies including 7,056 participants reported sufficient data between age and the occurrence of BPPV. The pooled results showed that age was not associated with BPPV occurrence (MD = 0.56; 95% CI, −0.17-1.29; P = 0.13) (Figure 1). These results may be partly due to the fact that many included studies controlled the age between the experimental and control groups. We used a fixed-effect model, because the statistical heterogeneity between these studies was not significant (I 2 = 20%; P = 0.24).

Serum Vitamin D Level
Seven studies including 1,254 participants measured serum vitamin D level to investigate the relationship between serum vitamin D level and BPPV occurrence. Significant relationship was found between serum vitamin D level and BPPV in our analysis. The vitamin D level was lower in patients with BPPV than in controls (MD = −2.12; 95% CI, −3.85 to −0.38; P = 0.02) (Figure 1). Statistical heterogeneity was significant (I 2 = 75%; P = 0.0006). As shown in Supplemental Figure 6, the results of sensitivity analysis were consistent with previous analysis (MD = −3.09; 95% CI, −3.95 to −2.23; P < 0.00001; I 2 = 22%; P = 0.27).

Bone Mineral Density
Bone mineral density measurements were expressed as T scores and we specifically analyzed the effects of osteoporosis and osteopenia on BPPV. Osteopenia was defined as −2.5 <T score <-1.0, and osteoporosis was defined as T score ≤-2.5. Eight studies including 3,944 participants investigated the effects of osteoporosis on the occurrence of BPPV. Our analysis indicated that osteoporosis was a risk factor for BPPV occurrence (OR = 2.49; 95% CI, 1.39-4.46; P = 0.002) (Figure 2). The I 2value was 79%, suggesting significant heterogeneity among these studies. Six studies involving 1,484 participants were included in osteopenia analysis. No significant relationship was found between osteopenia and BPPV (OR = 1.11; 95% CI, 0.76-1.62; P = 0.59) (Figure 2). The I 2 -value was 63%.
The same two studies also analyzed the correlation between stroke and the occurrence of BPPV. The pooled results showed no significant correlation between BPPV and stroke (OR = 3.58; 95% CI, 0.43-29.93; P = 0.24) (Figure 2), with significant heterogeneity between the two studies (I 2 = 93%; P = 0.0002).

Head Trauma
Two studies including 4,523 participants investigated the relationship between head trauma and BPPV occurrence. Our analysis indicated that head trauma was a risk factor for BPPV occurrence (OR = 3.42; 95% CI, 1.21-9.70; P = 0.02) (Figure 2). The I 2 -value was 67%, indicating significant heterogeneity between the two studies.

Hypertension
Twelve studies including 10,869 participants evaluated the effects of hypertension on the onset of BPPV. The pooled results suggested no significant association between BPPV and hypertension (OR = 1.26; 95% CI, 0.97-1.62; P = 0.08) (Figure 3). This risk factor was analyzed by a random-effect model (I 2 = 65%; P = 0.001). Significant heterogeneity between studies limited the accuracy of the results.

Diabetes Mellitus
Twelve studies including 10,869 participants reported the relationship between DM and BPPV occurrence. The pooled evidence showed that DM was not associated with BPPV occurrence (OR = 1.04; 95% CI, 0.86-1.25; P = 0.71) (Figure 3). No significant heterogeneity was detected among these studies, and a fixed-effect model was used (I 2 = 18%; P = 0.27).

Hyperlipidemia and TC Level
Four studies including 9,426 participants investigated the influence of hyperlipidemia on the occurrence of BPPV. Our analysis showed no significant association between hyperlipidemia and BPPV occurrence (OR = 1.50; 95% CI, 0.88-2.53; P = 0.13) (Figure 3). The I 2 -value was 86%, so a random-effect model was used.
Three studies involving 582 participants measured total cholesterol level to assess their influence on BPPV occurrence. The pooled evidence showed that patients with BPPV have a higher TC level than controls (MD = 0.32; 95% CI, 0.02-0.62; P = 0.03) (Figure 4). The I 2 -value was 66%, indicating significant heterogeneity between these studies.
Correlations between drinking and BPPV occurrence were performed in six studies involving 1,830 participants. No significant association was found between drinking and BPPV (OR = 0.64; 95% CI, 0.29-1.43; P = 0.28) (Figure 4). The I 2 -value was 89%, suggesting significant heterogeneity among included studies.

DISCUSSION
This systematic review and meta-analysis indicated that female gender, vitamin D deficiency, osteoporosis, migraine, head trauma, and high TC level were risk factors for the occurrence of BPPV. There was no sufficient evidence to suggest that age, osteopenia, stroke, HTN, DM, hyperlipidemia, smoking, drinking, and physical inactivity were associated with BPPV occurrence. The accuracy of some of our results may be limited to significant heterogeneity or the limited number of included studies, so further research was needed to confirm some of our results.
Although many included studies controlled the sex ratio between the experimental and control groups, our analysis showed that women were more likely to develop BPPV than men. Previous studies have also suggested that women had a higher incidence of BPPV than in men, especially in the elderly women (4). This relationship may be related to estrogen deficiency in postmenopausal women, as estrogen may promote the development of osteoporosis and even BPPV (18). In addition, women BPPV patients have a higher risk of recurrence than men (9,31). Therefore, further research between estrogen levels and BPPV may help early diagnosis and prevention of BPPV.
Our analysis of serum vitamin D level suggested that vitamin D deficiency appeared to be a risk factor for the occurrence of BPPV. This result was consistent with a previous meta-analysis (32). BPPV significantly increased the risk of fractures and osteoporosis, which may be related to vitamin D deficiency in BPPV patients (33,34). Moreover, serum vitamin D level can be affected by estrogen deficiency (35), which may help explain why BPPV was more common in postmenopausal women. Thus, serum vitamin D level may be used for the auxiliary diagnosis of atypical BPPV as a serum predictor. In addition, some studies showed that vitamin D supplements can effectively improve symptoms of patients with BPPV (36) and have preventive effects on BPPV recurrence (37). Hence, vitamin D supplements may have important effects on improving the diagnosis and prognosis of patients with BPPV.
Our analysis results indicated that osteoporosis was a risk factor for BPPV occurrence, but osteopenia was not.
A previous systematic review also showed that BPPV may be associated with osteoporosis or osteopenia (38). Many studies suggested that bone mineral density values in BPPV patients were lower than those in controls (39). In addition, osteoporosis and osteopenia may also be associated with BPPV recurrence (40,41). Thus, treatment of osteoporosis may help prevent the occurrence of BPPV and improve the prognosis of BPPV patients (42). Further studies were needed to determine the effects of BMD on BPPV occurrence and recurrence.
The pooled results showed that BPPV has no significant relationship with hyperlipidemia, but BPPV patients have a higher TC level. An increased TC level was a risk factor for BPPV occurrence. A higher TC level or hyperlipidemia can cause vascular damage in the inner ear, which may lead to BPPV occurrence (4). In addition, a recent study found that the three rs2074880 genotypes in the CACNA1A (Calcium Voltage-Gated Channel Subunit Alpha1 A) gene were associated with increased levels of cholesterol in BPPV patients (28). The relationship between TC level and BPPV has not been adequately studied. Further studies were required to confirm these results.
BPPV was frequently induced by secondary factors such as head trauma, migraine, or other inner ear diseases. Recent studies showed that migraine (43) and head trauma (29) were significantly associated with an increased incidence of BPPV. Most included studies excluded patients with any history of vestibular or neurological diseases, including head trauma and migraine. Our analysis still showed that migraine and head trauma were risk factors for BPPV occurrence. However, the limited number of studies included or significant heterogeneity may limit the accuracy of these results.
Some studies have investigated associations between vascular risk factors and BPPV, such as hypertension, DM, and hyperlipidemia, but the results were controversial (4,19). In addition, vascular comorbidities may also be risk factors for BPPV recurrence (8,9). However, our analysis showed that migraine and high TC level were risk factors for BPPV occurrence, while HTN, DM, hyperlipidemia, and stroke were not. The limited number of eligible studies or significant heterogeneity among studies may limit the accuracy of these results. Large-scale studies of these risk factors were needed to confirm the reliability of these results.
Previous studies suggested that smoking has adverse effects on middle ear diseases and hearing loss (44) and even makes the treatment of vertigo ineffective (45). However, some studies have shown that smoking can reduce the incidence of BPPV, prevent the recurrence of BPPV, and shorten the recovery time of BPPV (16). The relationship between smoking and BPPV was quite controversial and had not been adequately investigated. We expected that smoking was a potential risk factor for BPPV occurrence, but pooled results showed no significant relationship between smoking and BPPV occurrence. Significant heterogeneity among studies may limit the accuracy of this results. Further investigations were needed to establish the effects of smoking on BPPV.
Our analysis showed no significant association between BPPV and physical inactivity. However, previous studies showed that moderate physical exercise can prevent the occurrence of BPPV and decrease the risk of falls and fractures, especially in the elderly (46). Intense physical activity may trigger posttraumatic BPPV without head trauma (47), but a study showed that BPPV caused by intense physical activity was a rare condition (48). Some included studies did not give specific definition, which may limit the accuracy of this result. The role of regular exercise and moderate exercise in BPPV needed further investigations.

LIMITATIONS
Inevitably, there were several limitations in this meta-analysis. First, searches were restricted to English literature, which means that potentially high-quality literature may not be included in our analysis. Second, some potential risk factors were not analyzed in our analysis, because too few published studies were available, such as coronary heart disease, serum uric acid level, and albumin level. Third, subgroup analysis of each risk factor was not performed due to insufficient data. Furthermore, many included studies were retrospectively conducted in Asia and BPPV had many levels of its severity, which may limit the reliability of our results. In addition, for some risk factors, the limited number of included studies, significant heterogeneity, or ambiguous definition may limit the accuracy of these results. Large-scale randomized controlled trial (RCT) studies were necessary to confirm the reliability of our results.

CONCLUSION
This meta-analysis was based on 19 studies involving a total of 14,286 participants, which provided strong evidence that female gender, vitamin D deficiency, osteoporosis, high TC level, migraine, and head trauma were risk factors for the occurrence of BPPV. However, the effects of other risk factors on BPPV occurrence needed further investigations. Further investigations should focus on exploring potential mechanisms, how to effectively intervene in high-risk populations, and preventing these risk factors as much as possible.

DATA AVAILABILITY STATEMENT
All datasets presented in this study are included in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
JC and WZ contributed to literature search, data analysis, and drafting and revision of the manuscript. JC and XY contributed to data collection and crafting and revision of the tables and figures. PZ given constructive suggestions for the revision of this manuscript. All authors contributed to the article and approved the submitted version.