Literature search

We searched the PubMed, Embase, and Cochrane Library databases for studies that investigated the association between PD and HNC published up to March 22, 2013. The following search terms were used: (1) "head and neck cancer" OR "oral cancer" OR "oropharyngeal cancer" OR "pharyngeal cancer" OR "laryngeal cancer" and (2) "periodontal disease" OR "periodontitis" OR "periodontal attachment loss" OR "periodontal pocket" OR "alveolar bone loss". We also reviewed the reference lists of pertinent articles and recent reviews.

Eligibility criteria

Cohort studies, case-control studies, and cross-sectional studies evaluating the risk of HNC events in relation to PD were considered eligible for inclusion if the following criteria were met: (1) full-text could obtain; (2) clear diagnostic criteria for PD and HNC were reported; and (3) the adjusted and/or unadjusted hazard ratios (HRs), odds ratios (ORs) or relative risks (RRs), and associated 95% confidence intervals (CIs) or the numbers of events that can calculate them were reported. If more than one study covered the same population, only the report containing the most comprehensive information on that population was included. Two authors independently evaluated the eligibility of all retrieved studies, and disagreements were resolved by discussion or consultation with a third author.

Data extraction

Two authors independently extracted and tabulated the following basic information on each study: first author’s surname, publication year, design, country of origin, sample size, number of events, age range, assessment of PD and HNC, tumor site and pathologic type of HNC, crude and/or adjusted point estimates and 95% CIs, and covariate features included in the multivariable model.

Data analysis

ORs were used to measure association across studies. HRs were directly considered as RRs, and RRs were transformed into ORs with this formula [21]:RR=OR/[(1−P₀)+(P₀×OR)], where P₀ is the incidence of the outcome of interest in the non-exposed group.

The standard error (SE) of the resulting converted OR was subsequently determined using the formula: SElog(RR)=SElog(OR)×log(RR)/log(OR). Given that these transformations can overestimate the variance of ORs derived from RRs [22], we performed a sensitivity analysis by omitting the study in which this transformation had been applied.

We computed pooled ORs and relevant 95% CIs using Review Manager (RevMan, version 5.2, Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2012). We transformed ORs by taking their natural logarithms as well as calculating SEs and the corresponding CIs [23]. We then pooled log (ORs) and their SEs using the inverse variance method. Heterogeneity was assessed with the I² statistic [24] with low, moderate, and high I² values of 25%, 50%, and 75%, respectively [25]. When I² ≤ 25% (indicating no evidence of heterogeneity), we used the fixed effects model; otherwise, the random-effects model was used. We generated forest plots sorted by the publication year.

When heterogeneity was present, we performed a sensitivity analysis excluding studies reporting only HRs or RRs. Planned subgroup analyses were conducted based on the geographical region, study design, PD assessment method, tumor sites, and crude and adjusted ORs. Finally, we assessed evidence of publication bias by visually inspecting of funnel plot.

Study selection and characteristics

Among the 625 records initially searched, seven eligible studies involving two cohort studies [12,15] and six case-control studies [12–18] were included in the meta-analysis. A detailed flowchart of the selection process was shown in Figure 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Flow chart from identification of eligible studies to final inclusion.

PD, periodontal disease; HNC, head and neck cancer.

https://doi.org/10.1371/journal.pone.0079017.g001

Table 1 shows the major characteristics of the seven studies. The included studies were all published in English and preformed in the following countries (n= 8): USA, Brazil, Sweden, Russia, Romania, Poland, Argentina, and Cuba. Among them, six were single -center studies [12,13,15–18], and one contained two multicenter studies [14]. All cases were histologically, pathologically, or cytologically confirmed as HNC. One study was oral tumor (non-specific) [12], one was oropharyngeal cancer [15], two were oral and oropharyngeal SCCs [13,16], and four were HNSCCs (including oral cavity, pharynx, and larynx SCCs) [14,17,18]. One study reported the total number and events of case and control groups [16], three studies only reported adjusted ORs and 95% CIs [12,14,18], one only reported adjusted HR and 95% CI [15], and two reported both crude and adjusted data [13,17]. Table 2 shows the adjusted covariates. PD was defined by alveolar bone loss (ABL) with radiography in 3 studies [13,15,17], by clinical attachment loss (CAL) in one study [12], by community Periodontal Index of Treatment Needs (CPITN) score in one study [16], by tooth mobility in one study [18], and by poor mouth conditions in one study [14].

Overall estimates

The overall OR estimates for each study were pooled to give a total estimate of risk (Figure 2). Obviously heterogeneity was observed (p < 0.001, I² = 89%), so the result based on the random-effects model showed that exposure to PD can increase the HNC risk 2.63 times (OR = 2.63, 95%CI = 1.68 - 4.14, p < 0.001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Forest plot of periodontal disease and risk of head and neck cancer, studies are pooled with random-effects model.

Guha E 2007, the study conducted in Europe, included Russia, Romania, and Poland; Guha LA 2007, the study conducted in Latin-America, included Argentina, Cuba, and Brazil.

https://doi.org/10.1371/journal.pone.0079017.g002

When the study that reported only HR was omitted [15], the result showed that exposure to PD can increase the HNC risk 3.01 times (OR = 3.01, 95%CI = 1.84 - 4.93, p < 0.001), and heterogeneity was also present (p < 0.001, I² = 89%). Thus, the overall result was not changed in the sensitivity analysis.

Subgroup analysis

Table 3 shows the results of subgroup analysis by adjustment for covariates, study design, PD assessment, tumor site, and ethnicity. All these analyses indicated that PD was a significant risk factor of HNC, except for the cohort design (OR = 2.23, 95% = 0.58 - 8.61, p = 0.24), oral and oropharyngeal SCCs (OR = 4.82, 95%CI = 0.67 - 34.48, p = 0.12), Latin American population (OR = 4.68, 95%CI = 0.75 - 29.33, p = 0.10), and PD assessment by ABL (OR = 1.11, 95%CI = 0.74 - 5.98, p = 0.16). However, when the random-effects model was switched to the fixed effect model for these exceptions, the results were all significant, i.e., for cohort design (OR = 1.72, 95% = 1.17 - 2.52, p = 0.005), oral and oropharyngeal SCCs (OR = 5.95, 95%CI = 2.81 - 12.61, p < 0.001), Latin American population (OR = 2.22, 95%CI = 1.72 - 2.86, p < 0.001), and ABL (OR = 2.73, 95%CI = 2.11 - 3.53, p < 0.001).

Publication bias

Figure 3 shows that the funnel plot was not asymmetrical, thereby indicating no significant publication bias.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Funnel plot based on overall results of included 2 cohort and 6 case-control studies.

Guha E 2007, the study conducted in Europe, included Russia, Romania, and Poland; Guha LA 2007, the study conducted in Latin-America, included Argentina, Cuba, and Brazil.

https://doi.org/10.1371/journal.pone.0079017.g003

Main findings

HNC is known for its rapid clinical progression and poor prognosis, and the survival rates barely improved over the past 40 years. Identifying and preventing the risk factors of HNC are important tasks. Smoking and alcohol are the classical and major risk factors of HNC [3], and successful tobacco and alcohol control programs for reducing the risk of HNC have been developed [1,26]. However, HNC cannot be completely prevented. PD is also considered as an important risk factor of HNC. Accordingly, this study was aimed to estimate the pooled size of PD in HNC risk by a meta-analysis. The pooled effect estimates from seven included papers involving eight studies demonstrated a 2.63 fold (95%CI = 1.68 - 4.14) increased risk of HNC in PD patients compared with PD-free individuals. Sensitivity and subgroup analyses revealed an association between the two diseases when analysis was restricted to adjustment for covariates, study design, PD assessment, tumor site, and ethnicity.

Sources of heterogeneity

Substantial heterogeneity was observed among the studies because of the differences in population characteristics, study design, PD ascertainment, HNC sites, and adjustment for confounding factors. Our sensitivity analysis suggests that the study of Michaud et al [15] did not contribute to the heterogeneity. Subgroups analysis showed increased heterogeneity in ethnicity and study design groups, that heterogeneity was not caused by ethnicity or study design. However, we cannot exclude them, because the sample sizes were smaller than the overall population. Heterogeneity was decreased in PD ascertainment, HNC sites, and adjustment for confounding factors indicating that they were the factors of heterogeneity. Similar to ethnicity and study design, we cannot exclude the influence of sample sizes. The small number of cases and healthy controls can also increase the possibility of heterogeneity.

Implications for practice and research

The oral cavity can become a reservoir of respiratory pathogens [27], mainly because of dental plaque and thus contribute to PD [28]. Patients with oral cancer often have poor oral health in general, i.e., they have carious teeth and PD. However, oral health -related variables are also associated with the use of tobacco and alcohol. Thus whether PD is an independent risk factor or only a maker of HNC is unknown. Interestingly, all the six studies [12–15,17,18] with adjustment for covariates including smoking and alcohol showed significantly higher risk (OR = 2.23, 95% = 1.46 - 3.43). This result indicated that PD cannot act as a confounder or an effect modifier for tobacco smoking and alcohol. Therefore, PD was an independent risk factor of HNC. Accordingly, we could deduce that an effective oral hygiene regimen may effectively prevent HNC progression, and an effective PD intervention treatment should be able to improve the prognosis of HNC patients. However, sufficient relevant randomized controlled trials must be conducted to substantiate these findings. If these findings are confirmed, the government medical departments should urgently to increase their budgets for health education and health intervention programs.

Specific oral bacteria have been suggested to be involved in carcinogenesis [29]. Significant differences have been observed when microbial populations in mouth mucosa are compared between healthy and malignant sites. PD is due to oral microorganisms [27]; however, the carcinogenic mechanisms that can explain how PD causes HNC is unclear. Whether PD is only a biomarker for microbe-associated risk of HNC is also undertermined. Therefore, animal experiments should be performed to investigate the true mechanism between PD and HNC, particularly to determine the mechanism and identify the pathogens involved both in PD and HNC. The results can give implications for clinical practice and support the link between the two diseases. If such a link is proven, the related organizations such as the International Agency for Research on Cancer and the American Dental Association should list PD as a risk factor of HNC and formulate policies to promote oral health care.

The included studies were performed in non-Asian countries, and most Asian countries are developing countries. Given that access to dental care is limited in these countries, we predict that the prevalence/incidence of HNC in these countries will higher than the developed countries (where access to dental care is better). Therefore, we suggest that the developing countries carry out relevant studies to provide evidence and verify this prediction. In addition, the tools for PD assessment varied in the included studies. In future studies, researchers should choose a validated and optimal measurement (e.g., CPITN [30]) of PD, to decrease heterogeneity and increase accuracy.

Limitations of study

This study is the first meta-analysis investigating the association between PD and HNC and has some limitations. First, our meta-analysis was based on observational studies, and all included studies with adjustment groups have controlled the classical and major risk factors (smoking and alcohol) even though this confounding factor is difficult to control in epidemiological studies [31].Thus, whether PD is a risk factor or only a significantly biomarker for oral microorganisms. Second, we performed a thorough literature review and examination of reference lists in published articles to identify all published studies, and the funnel plot indicated that no obviously publication bias existed; however, the plot was not very symmetrical and the number of studies for our meta-analysis is not large. Third, heterogeneity was detected in our meta-analysis. Heterogeneity between studies should not be ignored even if it is highly common in the meta-analysis of genetic association studies. Moreover, we performed subgroup analyses to verify the data, but heterogeneity was also observed. Fourth, the sample sizes in many studies were relatively small. Compared with the studies having a large sample size, studies with a small sample size may overestimate the true association. A large sample study with either finding may better reflect a true association because of its sufficient statistical power. Fifth, the assessment tools for identifying PD were different, and relevant data of developing countries based on the currently included studies were not obtained. Overall, these aforementioned limitations may affect our final conclusions.