**3. Discussion**

This systematic review supports that PD is associated with PAOD, which may lead to the hypothesis that PD may be a risk factor for PAOD. To date, many studies describe associations between periodontitis or oral disease and atherosclerosis in general. Therefore, we wanted to specifically focus on studies that concern peripheral vascular disease (PAOD) as a potential consequence of periodontitis in this systematic review. However, it has to be stated that the published literature is not absolutely certain about the term PAOD for "peripheral artery occlusive disease". In publications, it is frequently used for lower extremity artery disease (LEAD). Indeed, other peripheral localizations, including the carotid and vertebral, upper extremities, mesenteric and renal arteries, are also frequently a ffected, mainly by atherosclerosis, and complete the family of peripheral arterial diseases [6]. In addition, there is sometimes no di fferentiation between extracranial and cerebral atherosclerotic pathologies. Hence, for this systematic review, we excluded studies that were linked to carotid/cerebral sclerosis, coronary sclerosis as well as vascular sclerosis in general. We also excluded animal studies as we wanted to focus on clinical evidence for an association between the two pathologies.

All studies that were included in this systematic review could detect an association between PD or tooth loss and PAOD irrespective of study design, outcome measure, and study population.

This supports that the consistency of the association is high. It must be noted, however, that there is inherent bias, since risk factors for PAOD also can cause PD. Standardized e ffect sizes, which provide a measure of the strength of the association, have mostly been determined by logistic regression analyses and reported as ORs together with 95% CIs. After making adjustments for age, gender, and other cardio vascular disease (CVD) risk factors, the reported ORs ranged from somewhat over two in NHANES [18] to over eight in the small case-control study published by Soto-Barreras et al. [21]. In general, the smaller case-control studies yielded higher ORs than the larger cohort studies. Measures of hazard ratios or relative risk estimates are available from only a few studies. Data from the Nurses' Health Study demonstrated a significant association between incident tooth loss and PAOD and reported a hazard ratio of 1.3 for PAOD in women with PD vs. women without PD. The meta-analysis published by Yang et al. 2018 [20] reported a statistically significant relative risk of 1.7 for PAOD in people with PD vs. those without PD.

Taken together, these results sugges<sup>t</sup> that severe PD increases the risk for PAOD to a similar extent as PD increases the risk for cardiovascular events, which with respect to the latter was shown to be ≈1.20-fold in adjusted models from meta-analyses of prospective cohort studies [33,34]. Smoking, a profound risk factor for PAOD, is associated with PAOD with odds ratios ranging between 1.7 and 7.4 [3]. With respect to diabetes, a twofold increased rate of macroalbuminuria and a threefold increased rate of end-stage renal disease were found in diabetics who also had severe periodontitis compared to diabetics without severe periodontitis [35]. Moreover, cardiorenal mortality resulting from ischaemic heart disease and diabetic nephropathy was three times higher in diabetics with severe PD compared to periodontally healthy diabetics [36]. The risk of PD for preterm delivery ranged between 4.45 and 7.07, depending on the gestational age [37]. Severe maternal PD was also shown to be associated with preterm low birth weight with an odds ratio of 7.5 [38]. All referenced studies considered a wide range of suspected confounders and included corresponding adjustments. Thus, PD may be a risk factor for multiple, widespread diseases. On the other hand, the possibility that some of the weak associations may be due to residual confounding by unrecognized confounders should not be neglected.

If PD is a causal or, at the least, an important contributor involved in the pathogenesis of PAOD, one would expect that PD precedes the onset of PAOD. However, only very limited information exists with respect to the temporality of both diseases. According to results from the prospective Health Professionals Follow-up Study, tooth loss seemed to precede PAOD, since the incidence of PAOD was most strongly associated with tooth loss in a period of 2 to 6 years prior to the occurrence of PAOD [17]. The fact that tooth loss in the previous 2 to 6 years was more strongly associated with PAOD than tooth loss in the previous 2 years or 6 to 8 years suggests that 6 years may be too distant and 2 years may be too recent for tooth loss to have an impact on PAOD. However, these reported time-dependent di fferences in the strength of the association were based on only the disparity of only a few PAOD incidences and may, thus, have been chance findings.

The plausibility of a causal or, at least, an important involvement of PD in the development of PAOD mostly relates to experimental data showing that inflammation is involved in the pathophysiology of atherosclerosis, which in turn is involved in the development of PAOD. This inflammation could

be caused by a direct involvement of periodontal pathogenic bacteria, which enter the vascular wall via the bloodstream. The study by Figuero et al. used nested polymerase chain reaction (PCR) to detect three periodontal pathogens in subgingival, vascular, and blood samples. Although positive test results were obtained in high fractions of the subgingival samples (>70%) and the vascular and blood samples (7 to 11.4%), patients with and without PD did not di ffer with respect to the levels of the targeted bacteria. Therefore, a direct involvement of the bacteria seems inconclusive at this stage.

The studies by Nishida et al. 2016 [30] and Armingohar et al. 2015 [29] support that inflammatory mediators, such as serum amyloid A and anti-inflammatory mediators, such as interleukin-10, which are released from the oral sites a ffected by PD into the bloodstream, thereby modulating systemic inflammation may be involved in the pathomechanism of PD-induced PAOD. In addition, autoimmunity induced by PD via the immune response of the host to the bacterial HSP60 homolog GroEL produced by *P. gingivalis* (the main oral pathogen involved in PD) could play a role [31]. Support for this mechanism comes from the study by Choi et al., who successfully established *P. gingivalis*–specific T-cell lines from atheroma lesions isolated from PD patients. However, the study included only two patients, and the origin of the lesions remained unclear.

It is evident from the results section that our search-strategy yielded only a few publications that dealt with the pathomechanism of the association between PD and PAOD. A large fraction of the published mechanistic studies concerned animal, in vitro and ex vivo studies describing the link between PD and vascular sclerosis in general rather than that between PD and PAOD specifically. These studies were, however, not eligible for this review based on the pre-defined exclusion criteria shown in Figure 3. Roles of oral infections in the pathomechanism of atherosclerosis, in general, were discussed in grea<sup>t</sup> detail in a recent review published by Aarabi et al. [15]. Briefly, there is a wealth of support for at least four plausible pathogenic mechanisms: (1) low-level bacteremia by which oral bacteria enter the bloodstream and invade and damage the arterial wall; (2) systemic inflammation induced by inflammatory mediators, which are released from the sites of the oral inflammation into the bloodstream; (3) autoimmunity to host proteins which results from the host immune response to specific components of oral pathogens; (4) pro-atherogenic e ffects resulting from specific bacterial toxins that are produced by oral pathogenic bacteria. In addition, recent genome-wide association studies supported that PD and PAOD share at least one important predisposing genetic risk haplotype that is located at chromosome 9p21.3 in a locus known as *ANRIL*/*CDKN2B-AS1* [39,40]. The risk haplotype a ffects the structure and expression of ANRIL, which is a long non-coding RNA (lncRNA) that, such as other lncRNAs, regulates genome methylation, thereby a ffecting the expression of multiple genes by *cis* and *trans* mechanisms. How precisely ANRIL contributes to the risk of PD and PAOD on the molecular level is currently unclear.

So far, many, but not all, studies demonstrated the presence of bacterial DNA in a large number of atheromas, but only very few could demonstrate the successful isolation of viable bacteria from an atherosclerotic plaque. In fact, to the best of our knowledge, there is not a single study available that could demonstrate isolation and cultivation of viable *P. gingivalis* from atherosclerotic tissue. In addition, it should be noted that long-term treatment with antibiotics, such as roxithromycin and rifalazil, showed no benefit in patients with an established diagnosis of PAOD [41,42]. Nevertheless, it seems prudent at this stage to recommend that patients with PAOD should be routinely referred to a dentist, and periodontitis should be appropriately treated if present.
