Periodontitis and Systemic Disease: Association or Causality?

The periodontitis-systemic disease relationship constitutes an important part of clinical periodontal research. Research activity has grown continuously since the late 1980s with about one third of all recent periodontal studies now dealing with the relationship between periodontal disease and systemic disease [1•]. Further to this, a total of 57 different systemic disease conditions are now being researched in relation to periodontal diseases [2].

The periodontitis-systemic disease relationship is commonly described as being ‘two-way’ or ‘bi-directional’. However, the design of many observational epidemiological studies does not allow directionality to be firmly established and so by default, any associations identified will be bi-directional until data emerge that provide clarification. Currently, we do not have a full understanding of the importance of the numerous associations reported and in particular whether they play a part in causation. The conclusion of an expert panel at the joint European Federation of Periodontology/American Association of Periodontology workshop on ‘Periodontitis and Systemic Diseases’ was that ‘reported associations do not imply causality and the establishment of causality would require new studies that fulfil the Bradford Hill or equivalent criteria’ [3].

We have reviewed recent work published since 2014, focusing on prospective studies and systematic reviews where these exist, to examine whether progress has been made. The review is focused on systemic diseases or conditions that could impact the development or progression of periodontitis. Given the constraints of the large number of possible candidates [2], we have limited the focus of this review to diabetes, metabolic syndrome and obesity as they have been suggested to have a causative role in periodontitis. Discussion of the biological mechanisms through which these exemplars could increase the risk of periodontitis is beyond the scope of this review. A further paper in this edition has reviewed studies of periodontitis as a possible risk factor for systemic diseases.

Diabetes mellitus comprises a group of metabolic diseases in which the characteristic phenotype is loss of control of glucose homeostasis, resulting from defects in insulin secretion, insulin action or both [4]. In 2015, there were an estimated 415 million people worldwide with diabetes and this is projected to rise to 642 million by 2040 [5]. In 2015, more than 10% of adults in 25 states of the USA had diabetes [6].

Metabolic syndrome (MetS) relates to a cluster of disorders including excess body fat around the waist and abdominal area, increased blood pressure, elevated plasma glucose, elevated serum triglycerides and reduced serum high-density lipoprotein [16]. The most recent definition from the International Diabetes Federation is the presence of central obesity, defined as a waist circumference exceeding ethnic specific guidelines, plus two from four of the other factors [17]. It has been highlighted that 25% of the world’s adults have MetS and as such, they have a fivefold greater risk of developing T2DM [17].

Obesity has been defined as a systemic disease characterized by excessive body fat accumulation that can lead to adverse impacts on health conditions [22]. The World Health Organization reported about 13% of the world’s adult population (11% of men and 15% of women) were obese in 2014 [23]. The worldwide prevalence of obesity more than doubled between 1980 and 2014 [23]. This increase in global prevalence has been more notable in certain developed countries, such as the UK, where obesity prevalence has increased from 15% in 1993 to 26% in 2014 [24]. In the USA, adult obesity rates now exceed 35% in four states, 30% in 25 states and are above 20% in all states [25].

What overall interpretation can be placed on the studies to date? Many observational studies have reported associations which could reflect causal involvement of an exposure in the aetiology of a disease or condition that subsequently could not be confirmed by large-scale prospective randomized controlled trials (RCTs) [33•]. One classical example is the finding from longitudinal studies that hormone replacement therapy (HRT) led to a reduction in the risk for coronary heart disease (CHD). The obvious conclusion was that HRT had a protective effect; however, when prospective RCTs were completed, it became evident that HRT was actually responsible for an increased risk of CHD [34•]. This was explained by the ‘healthy-user effect’ whereby women who took HRT tended to be better educated, more aware of their health, likely to take other preventive advice, to be less likely to smoke, etc. These factors therefore confounded the association, and this is likely to have affected many observational studies resulting in the identification of spurious associations [35]. A major weakness of observational studies is that there is no protection against confounding. Most studies use statistical techniques to adjust for obvious confounders; however, there may be residual confounding reflecting the difficulty in controlling for all dimensions of relevant factors [33•]. This is a particular problem for studies investigating exposures such as obesity and diseases such as periodontitis which can be affected by factors which are difficult to measure such as behaviour in social networks [36], socioeconomic status and health service utilization [33•]. In addition, there is no protection against unknown or unrecognized confounders.

If we accept that confounding is a challenging issue, how can we test whether a disease or condition such as obesity is a causative factor for periodontitis? One way to control for confounding is to apply a randomized design. RCTs are used to test the effectiveness of interventions and are the gold standard for validating a particular method of treatment. From ethical and practical viewpoints, it is difficult to envisage an RCT with a test group who are provided with a mechanism to become obese and a control group who maintain normal weight. There are few conclusions on the natural history of disease that are supported by prospective RCTs [37]. The best that might be achieved are prospective cohort studies looking at incident cases; however, these are also liable to bias and confounding.

Mendelian randomization (MR), a method which exploits the random assignment of genes from parents to their children at meiosis, could provide a solution [33•]. In MR, a genetic variant associated with a phenotype is used as an instrumental variable (IV) to evaluate the causal relationship between that phenotype and the outcome of interest. The fundamental consideration is that the IV is regarded as independent of confounders [38•]. Therefore, if we can identify a gene variant which is closely linked to a putative risk factor, then we may be able to use MR to assess whether it is a true causal aetiological factor. In this context, variants in a number of genes have been shown to be associated with obesity [39, 40]. Subsequently, MR has been used to show that greater adiposity leads to higher CRP levels with no evidence of any reverse pathway [41], and that higher body mass index (BMI) has a causal relationship with cardiometabolic traits including heart failure, metabolic syndrome and type 2 diabetes [38•].

The Gene Lifestyle Interactions and Dental Endpoint (GLIDE) consortium applied MR to test whether there was a causal relationship between obesity and periodontitis [42•]. GLIDE had access to BMI, periodontal status and relevant genotype data from 13 studies of up to 68,761 individuals. Genotypes at FTO (rs1121980), MC4R (rs17782313) and TMEM18 (rs6548238) which represent the largest known effect sizes for BMI in European populations were used to provide a genetic risk score (GRS). Definitions of periodontitis were those applied in each of the participating cohorts. Higher BMI was significantly associated to periodontitis after adjustment for known confounders in subjects (n = 29,459) who had been clinically assessed for periodontitis (odds ratio 1.13 (95% CI 1.10–1.17). In obese compared with normal weight participants, the relationship was stronger OR = 1.33 (95% CI 1.18–1.50). The genotypes used were associated with BMI with a statistically significant trend in mean BMI per GRS unit. However, when BMI-associated genotypes were studied in relation to periodontitis, there were no significant associations in those with clinically assessed periodontitis (10,972 cases, 9681 controls) with the pooled OR = 1.00 (95% CI 0.97–1.03) per risk allele. The authors concluded that their MR analysis did not support a role for total adiposity as a causal risk factor for periodontitis [42•].

We completed a further MR analysis on 1271 dentate men in Northern Ireland who were participating in the PRIME study (for details of participants, see [43]). The adiposity-associated variant rs17782313 in melanocortin-4 receptor (MC4R) was used as an instrumental variable for waist circumference (WC) in an MR design. There was a significant association between variants in rs17782313 and WC (p = 0.01). Increased abdominal adiposity was significantly associated (p = 0.007) with increased periodontal attachment loss (PAL) which was used to measure periodontitis. However, when the instrumental variable was applied to investigate the effect of WC on PAL, the association was not significant. This further independent study using an MR approach did not support a causal relationship between abdominal adiposity and periodontitis. The results of these recent MR studies suggest that the epidemiological association between obesity and periodontitis may be confounded. The possibility of reverse causality, i.e. in the direction from periodontitis to obesity cannot be excluded by the MR studies described. Further speculation on periodontitis as a possible causative factor for obesity is beyond the scope of this review.

Evidence from many observational studies, including an increasing number of prospective studies, supports associations between the conditions described (diabetes, MetS, obesity) with periodontitis. There is a need for balanced professional judgement of how meaningful these associations are. In this context, one recurrent problem is the substantial variations in the definitions used to identify periodontitis with few studies that meet stringent criteria for periodontitis [32]. Confounding remains the most challenging issue in the interpretation of the associations found between various health states such as those discussed in this review (diabetes, MetS, obesity) and periodontal disease. However, where it has been possible to interrogate one possible association, in the absence of confounding, recent studies using an MR approach do not provide evidence that supports obesity as a causative factor for periodontitis in contradiction to the outcomes of observational studies [42•]. One limitation to this method is that it can only be used where there are suitable functional polymorphisms, or markers linked to functional polymorphisms, that are relevant to the exposure of interest. However, it is possible that these may be identified; for example, there is increasing evidence that genetic variants are associated with T2DM [44]. Until further progress is made, we should accept that the associations reported may be important and periodontal diagnostic procedures should be routinely carried out in patients who present with diabetes, MetS and obesity.