Given the age-related changes in the CNS immune system and the pathogenesis of AD described earlier, it seems that neuroinflammation and Th1 skewing of the immune response play a significant role in the pathogenesis of AD. Two dominant phenotypes of CD4+ T cells are the Th1 and Th2 types. Th1 cells are largely responsible for cytotoxic cell-mediated pro-inflammatory responses through coordinating other players such as CD8+ T cells and are essential for clearing intracellular infections (eg. viruses) and malignant cells. Th2 cells are largely responsible for humoral immunity through coordinating antibody generation by B cells (94). As described earlier, there is an increased number of CD8+ T cells, particularly CD8+ T CD28- cells, in the brains of AD patients (80). Memory CD8+ T cells can contribute to the control of CNS infections (95, 96). However, it remains unknown whether and how each subset contributes to CNS immunity. In addition, the presence of these cells positively correlates with the poor antibody response to the influenza vaccine in the elderly (95, 97). These cells also appear to produce high levels of IFNγ and low IL-5 levels, and promote a Th1 cytokine response by CD4+ T cells in response to auto antigenic stimulation (95). It is postulated that the chronic and ubiquitous exposure to autoantigens underpin the prevalence of these CD28-CD8+ clones and the high IFNγ levels in old age (95). Of note as well are the reduced rates of various cancers in AD patients, lending strength to the notion of Th1 skewing in AD pathology (98). In a mouse model of AD, the peripheral administration of antibodies against Aβ has also been shown to induce clearance of Aβ and reduce Aβ burden, suggesting that reduced antibody generation in the elderly might contribute to AD pathogenesis (99). This Th1 skewing of the immune response, namely the production of IFNγ and the reduced antibody production, could result in increased activation of microglia (94). IFNγ working in concert with TNF also triggers Aβ production in neuronal and extra neuronal cells and increased reactive oxygen species (ROS) production by microglia, likely perpetuating AD pathogenesis (80).

There is a recent concept of microglial priming where previous and chronic disturbances in the CNS environment pushes microglia to a phenotype with more marked pro-inflammatory responses than the transient activated phenotype adopted during brief episodes of physiological insults (100). These insults include age-related oxidative stress and DNA damage, traumatic brain injuries (TBI) and CNS infections that ultimately induces a state of neuroinflammation required for microglial priming (100). With respect to microglial phenotypes, they are typically described in vivo as having either a pro-inflammatory activated phenotype associated with increased ROS and pro-inflammatory cytokine production and reduced neurotropic factor production, or a non-activated phenotype associated with increased anti-inflammatory cytokine production (eg. IL-10, TGFβ) and phagocytic activity (100, 101). In the case of primed microglia, these cells undergo morphological and functional changes when exposed to chronic stimulation to resemble a pro-inflammatory activated phenotype (102). However, they are characterized by a higher baseline production of inflammatory markers and mediators, a lower threshold for activations and switching to a pro-inflammatory state, and an exaggerated inflammatory response (102). The state of chronic neuroinflammation that follows can lead to the loss of neurons and the development and deposition of neurofibrillary tangles (NFT) and Aβ (100, 102). This as such might lead to a self-perpetuating cycle of Aβ deposition secondary to the accumulation of primed microglia which in turn stimulate a pro-inflammatory response by microglia. In addition, the reduced clearance of Aβ by mechanisms such as reduced microglial phagocytic capacity, increased IFNγ production and reduced antibody generation in the elderly may further exacerbate this neuroinflammation towards symptomatic AD ( Figure 1 ). Comparisons between the immune system of a healthy individual and that of the elderly and AD patients are summarized in Figure 1 .

There is evidence however that the neuroinflammatory process differs throughout the time course of AD pathogenesis. In a study by Vlad et. al., the risk of developing AD gradually reduced in proportion to the duration of NSAID therapy, reaching to levels almost half that of the general population after a few years (103). Subsequent studies have found similar results (104, 105). However, a Cochrane study has found no significant therapeutic benefit of anti-inflammatory drugs in the treatment of symptomatic AD (106). Supporting this notion of a changing neuroinflammatory process is the finding that oxidative damage is greatest early in AD and decreases with disease progression (80). It has been suggested as such that as AD worsens, Aβ deposition increases alongside compensatory changes to reduce oxidative damage, which might include dysregulated T cell responses (80).

It seems therefore that the pathogenesis of AD could potentially be divided into three broad stages. In the first instance, there is a period of marked neuroinflammation following physiological insults. This then progresses to the second stage characterized by a self-perpetuating state of chronic stimulation and neuroinflammation that potentially promotes the accumulation of primed microglia and a pro-inflammatory environment. This chronic neuroinflammation and stimulation, caused either by a transient and brief insult like a TBI or by chronic infection by viruses like CMV, pushes the CNS immune system towards immunosenescence, the third stage and final stage, where its ability to adequately maintain the homeostatic neuronal environment is impaired ( Figure 2 ). Cellular senescence is characterized by the inability of cells to proliferate, this process being related to the erosion of telomere (107). Moreover, senescence is characterized by 3 main phases: the induction phase (characterized by telomere shortening, DNA damage, growth factor deprivation); the second phase – DNA damage response (characterized by DNA damage response); and phase 3 – growth arrest (signal transduction molecules p53, p21 trigger growth arrest). However not all primary cells senesce in the same way, for instance fibroblasts present a pre-senescence phase that may be reversible. In contrast T cell senescence does not seem related to shortening of telomere, and they die when they differentiate (107). Whatever the mechanism is, aging (p53 up regulation, mitochondrial dysfunction, DNA damage) and inflammation represent a two-way system in neurodegeneration.

Post-mortem brains of patients without a history of dementia have also been found to have sufficient Aβ and NFT amounts that would have otherwise prompted a diagnosis of AD, supporting the role of a dysfunctional CNS immune response in determining the onset of symptomatic AD (80, 108). One population that has been studied particularly with regards to health in old age are centenarians. Various anti-inflammatory mechanisms such as increased cortisol and IL-10 levels are well documented (80). This might enable centenarians to suppress neuroinflammation in the early stages before chronic neuroinflammation and significant CNS immune dysfunction develops. Interestingly, it has also been observed that administration of IFNβ-1, an antiviral agent, in patients with early AD for twenty-eight weeks resulted in significant improvements in some cognitive functions but not in inhibiting the disease progression (109). This raises the possibility of early antiviral treatment as a potential novel therapeutic option in early AD (109). It is also worth noting that whilst the role of Treg cells in AD pathogenesis remain elusive, centenarians have been noted to have increased numbers of Treg cells and that this positively correlated to improved survival outcomes (58, 110). It is possible that Treg and anti-inflammatory responses may differ significantly between mice models of AD and AD in humans in a manner that is currently poorly understood.

Given the association between Aβ and AD, trials exploring the therapeutic potential Aβ vaccines have been conducted. The first attempt at developing an Aβ vaccine was by Schenk et al. who prepared a vaccine from Aβ_1-42 peptide and Freud’s adjuvant (111). When tested on murine models of AD, the vaccine was found to confer both significant prophylactic and therapeutic benefits in the prevention of Aβ plaque formation and in inhibiting the progression and extent of Aβ deposition respectively (111). This vaccine was then developed for clinical trials but did not progress beyond phase II due to a proportion of patients developing aseptic meningoencephalitis (112). In a follow-up study on patients from this initial trial, post-mortem analysis of the brains of eight of the patients who had received the vaccine and had died with AD showed significant variation in Aβ burden. More notably however was the finding that in seven of the eight vaccinated patients including those with virtually complete Aβ plaque removal there was a severe end stage dementia prior to death (113). Although it is not possible to fully ascertain the therapeutic potential of therapies targeting Aβ from this study alone due to the small sample size analyzed post-mortem in the follow-up and the adverse events associated with the vaccine, it is possible however that future attempts at relieving Aβ burden may eventually reap limited rewards. This lack of association between Aβ burden and the clinical course of AD has been supported by other studies showing similar findings (108, 114). Attempts at using human intravenous immunoglobulins (IVIG) on the basis of preclinical findings showing Aβ clearance by IVIG have also failed to produce significant cognitive improvement in patients. Interestingly, however, was the finding that IVIG significantly reduced Aβ levels and slow cognitive decline in patients with the ApoE4 allele, although cognitive decline was only observed on neuropsychiatric examination with no corresponding improvement in activities of daily living (ADL) (115, 116). As such, it is possible that therapies targeting Aβ clearance may only prove efficacious for a small group of AD patients.

As described earlier, current evidence suggests a role for anti-inflammatory therapies in AD particularly as prophylaxis and treatment during the prodromal stage. At present, there is increasing interest in the inflammasome, particularly NLRP3, as a novel target for anti-inflammatory therapy research in AD. The NLRP3 inflammasome inhibitor dapansutrile (OLT1177) has been investigated for its therapeutic potential in other inflammatory conditions, having recently completed a phase II trial in treating osteoarthritis and acute gout flare (117). Given the association between NLRP3 and Aβ and pro-inflammatory responses in AD, dapansutrile has recently been investigated pre-clinically for the treatment of AD (118). In this study, dapansutrile was shown to reverse cognitive impairment, reduce microglial activation and cortical plaques, and normalize plasma metabolic markers of AD in murine models (118).

Since the success of the anti-TNF therapy infliximab in the treatment of Crohn’s disease, there has been a considerable interest in the following years in developing therapies targeting specific cytokines for various inflammatory conditions (119). One such drug currently of interest in the treatment of AD is etanercept. Similar to infliximab, etanercept is a TNF inhibitor. Various trials conducted have corroborated the positive therapeutic potential of etanercept in treating AD (120). However, in a recent phase II trial, although etanercept was found to be generally well tolerated by patients there was no statistically significant changes in cognition, behavior or global function (121). Interestingly, one mechanism by which etanercept might improve AD symptoms and neuropathology if any is that it modulates TNF regulatory function of synaptic transmission (120).

Naturally occurring anti-inflammatories and the role of certain diets have also been explored for their therapeutic potential in AD. One particular diet well-known for its anti-inflammatory properties is a Mediterranean diet (MD). This diet is characterized by large intake of fruit, vegetables, wholegrains, nuts and legumes; moderate consumption of fish, poultry and alcohol, and restricted intake of red and processed meats alongside olive oil as the main source of fat (122). Observational studies have found that greater adherence to the MD is associated with improved cognitive performance, slower cognitive decline and reduced risk of cognitive impairment and AD (122). The MD has also been associated with protective brain structures and function such as increased cortical thickness, greater brain volumes, reduced hippocampal atrophy rate and reduced Aβ accumulation (122). However, it seems that adherence to the MD for at least a few years is required to reap the cognitive benefits. In the PREDIMED study, adherence to the MD over 4-6 years was particularly associated with improved global cognition, memory and executive function (122). However, no cognitive benefits were seen in populations on the MD for up to one year, although improved global cognition and episodic memory were seen in those with greater adherence to the diet (123–125). These findings are not surprising as they reflect the findings of trials investigating the use of NSAIDs for treating AD as described earlier (103–106). It would be interesting however to explore in greater detail the role of a MD in managing symptomatic AD as it may be possible that there may be other beneficial compounds that may influence the disease course either through regulating neuroinflammation or by other pathways.

The role of Aβ in AD pathogenesis was first proposed as part of the ‘amyloid cascade’ hypothesis following the observation that virtually all people with Down’s Syndrome (DS) will exhibit AD pathology by 40 years of age, and that there was concomitant overexpression of Aβ from the cloning of the APP gene with trisomy 21 (126). APP cloning and subsequent overexpression of Aβ also causes familial EOAD (127). However, as described earlier, post-mortem studies showing a lack of correlation between Aβ levels and clinical AD, and the current lack of therapeutic success in therapies targeting Aβ suggest that a dysfunctional immune response might play a more significant role in AD pathogenesis. In fact, people with DS typically have immune systems that are not dissimilar to that seen in AD and old age. For example, poor vaccine responses are frequently observed in DS patients alongside decreased immunoglobulin levels and B cell populations (128). Increased Th1/Th2 ratio with corresponding decreased risk of various cancers, decreased T cell proliferation and reduced naïve T cell populations are also seen (128, 129). Therefore, it is possible that a dysfunctional immune system might be the main player in the pathogenesis of AD, with Aβ likely exacerbating the process but not being a causative agent. For example, the observation that Aβ deposition is initially diffuse with gradual progression to neuritic plaques but accumulating at an exponential rate after 40 years of age in DS patients might be due to an earlier initiation and acceleration of the process driving the CNS immune system towards a state of immunosenescence resembling that seen in AD patients (126). Similarly, it is possible that overexpression of amyloid such as is seen with DS patients might be a key player driving the progression towards immunosenescence. Interestingly as well is the finding that DS individuals are more sensitive to IFN stimulation due to the overexpression of its receptor located on the extra chromosome 21 (130). It is possible that this might lead to increased activation of microglia and ROS production that in turn accelerates aging and depletion of immune reserves via chronic inflammation (131)

It has also recently been proposed that Aβ might be an innate effector molecule (132). For example, Aβ has been shown to have antimicrobial properties against viruses, bacteria and fungi and enact this by forming oligomers (133). In addition, overexpression of Aβ provides increased resistance to bacterial and viral infections (134). Aβ has also been proposed to functionally resemble that of cytokines and should be considered as such (132). This theory might account for the close spatial and temporal association between microglia and Aβ frequently observed in AD. It also strengthens the argument that AD might be brought on by CNS immunosenescence secondary to chronic stimulation like persistent CMV infection as the CNS immune system attempts to clear the infection (80).

Human umbilical cord blood cells (HUCBC) have been garnering increasing interest as a potential strategy for remodulating the CNS immune response amid observations that HUCBC inhibits Th1 whilst promoting Th2 responses in murine models of stroke and neurodegenerative diseases (80, 135, 136). Hippocampal administration of HUCBC in murine models of AD have also been demonstrated to reduce neuropathological changes and improve cognitive impairment (137). Notably, there was a reduction in Aβ deposits and tau hyperphosphorylation, improved spatial learning and memory, a switching in cytokine profile exhibited by microglia from a pro-inflammatory phenotype to an anti-inflammatory one, and that these alternatively activated microglia were associated with Aβ (137). Such findings have led to the undertaking of a phase I clinical trial involving HUCBC mesenchymal stem cells (MSC) administration into the hippocampus of nine patients with mild to moderate AD (138). This method was found to be well tolerated and safe, although the efficacy was not able to be properly assessed in this study (138). Subsequent clinical trials have corroborated the safety profile of HUCBC-MSC administration into the brain. However, clinical outcomes have been less encouraging, with no significant improvement in cognition being reported (139). These clinical trials and their findings are summarized by Liu and colleagues (139). In addition to various technical issues such as the specific cell stage to be transplanted that remain to be resolved, it has been posited that the lack of clinical success with stem cell therapy research for AD could be due to inappropriate timing of stem cell transplantation (139). Although CSF biomarkers have strong predictive value in diagnosing or excluding AD in patients with mild dementia, individual clinical courses tend to be more variable which may reflect different stages and severity of neuropathology that may render subsequent stem cell transplantation ineffective (139).

Another strategy that might be useful to explore would be the reconstitution of Treg cells in AD patients. It has recently been shown that Treg anti-inflammatory function is impaired in patients with AD, but had increased suppressive activity on effector T cell proliferation and macrophage activation following ex vivo clonal expansion (140). This suggests that Treg function and immunophenotype are impaired in AD, but that administration of ex vivo clonally expanded of Tregs might be a viable therapeutic option (140). More recently, in vivo expansion of Tregs has been increasingly explored as a preferred method to ex vivo expansion, one main reason being that it bypasses the major obstacle of low Treg frequencies in lymphoid tissues that as such requires in vitro expansion prior to adoptive transfer (140). In vivo expansion of Tregs can be achieved via administration of low-dose IL-2 (141). IL-2 therapy has been shown to reduce Aβ load and improve neuroplasticity that was associated with recovery of memory deficits (60, 141). However, due to the short t_1/2 of IL-2, one method that may help to overcome this is the co-administration of IL-2 with an IL-2 specific monoclonal antibody, such as JES6-1A12, which increases its t_1/2 (141).

Findings such as the functional recovery of Tregs following ex vivo expansion in AD patients suggest that strategies tailored toward alteration of the AD microenvironment may help restore immune balance. One novel strategy could be in the form of transfusions of plasma or specific plasma constituents from young donors to elderly or AD recipients. Murine models of plasma transfusion including parabiosis studies have shown improvement in both neuropathological changes with reduction of Aβ and tau burden and improvement in cognition (142, 143). Various clinical trials exploring different types of therapeutic plasma exchange (TPE) in patients with established AD have generally yielded promising results, the findings of which are summarized by Imbimbo and colleagues (144). However, concrete conclusions are difficult to draw at this stage due to the inherent limitations of the study designs such as small sample sizes and strict exclusion criteria (144). It is also worth noting that one reason for the particular interest in TPE over other blood products such as whole blood might be due to previous associations found between blood transfusions and an increased risk of dementia or AD (145).

Whilst it is currently unclear how TPE with plasma from young donors might exert their therapeutic effects, several theories have been proposed. Vascular cell adhesion molecule 1 (VCAM1), an important molecule in mediating attachment of immune cells to the endothelial surface, has recently been found to be overexpressed in the brains and plasma of both aged mice and humans (146). In addition, systemic administration of anti-VCAM1 immunoglobulins and deletion of the Vcam1 gene in brain endothelial cells (BECs) was shown to reverse microglial reactivity and cognitive defects in mice (146). Therefore, it has been posited that increased inflammatory cytokine levels in old plasma increase VCAM1 expression on BECs, resulting in peripheral leukocyte recruitment and BEC inflammation, in turn activating microglia (146). Another theory for the mechanism of TPE in AD is that old and young plasma may contain youthful and pro-aging factors that influence that ultimately influence the CNS environment and immune response (142). Just as old plasma may contain higher levels of pro-aging pro-inflammatory cytokines, young plasma may contain factors that promote healthy aging of the CNS (142). For example, systemic administration of growth differentiation factor 11 (GDF11) was shown to increase neural stem cell numbers and vascular remodeling in the hippocampi of aged mice (142). However, more research is required to identify these factors in old and young plasma before specific supplementation with these factors can be considered as a viable therapeutic approach for AD.

Given the qualitatively and quantitatively poor production of immune cells seen in the elderly and in AD patients, bone marrow transplantation might be a potential therapy for AD. It has been found, for example, that transplantation of a young bone marrow into old recipient mice preserved cognitive functions, reduced microglial activation in the hippocampus and preserved synaptic connections (147). However, even if bone marrow transplantation was demonstrated to be highly effective in treating AD, the practical limitations in implementation are likely to exclude it as a treatment offered by healthcare systems. These may include availability of compatible bone marrows for transplantation and ethical considerations over which AD patients should receive it and if they should receive them at all in preference over other patient groups like childhood hematological cancers to name a few. Perhaps a more feasible approach might be strategies aimed at preserving and restoring thymic function. Various strategies to rejuvenate thymic function currently being explored are broad and diverse, and are discussed in detail in a review by Thomas and colleagues (148). It should also be noted, however, that in addition to these novel therapies simple strategies such as regular physical exercise have also been shown to have prophylactic and therapeutic benefits in reducing the risk of AD and improving cognitive decline in AD patients (149). The means through which they exert their beneficial effects include the slowing of age-related deterioration in thymic output and decreasing pro-inflammatory cytokine levels associated with neurodegeneration or “inflammaging” (150).

Identifying suitable biomarkers for monitoring different stages of the AD pathogenic course would be important for future research into therapeutic options for AD. Various biomarkers directly relating to the changing immune response in the AD continuum have been explored, one example being the blood compound neopterin. Neopterin is produced by IFNγ-stimulated macrophages and its association with cell-mediated responses has made it a common biomarker for monitoring such processes (80, 151–153). Notably, neopterin levels are increased in AD patients compared to age-matched controls and have been suggested to reflect serum immunity to CMV (80). In addition, neopterin levels were significantly associated with cognitive decline in AD patients (151). However, whilst neopterin levels were significantly higher in AD relative to MCI patients, there has been no significant difference in neopterin levels between MCI patients and cognitively normal individuals (154). Significantly increased neopterin levels were also only observed in patients with moderate dementia as assessed by their mini mental state score (MMSE) (153). This suggests that the usefulness of neopterin as a biomarker for AD is restricted to patients with established AD and is unlikely to be useful for this purpose during the preclinical phase for screening.

It is well acknowledged that immune changes are seen in the peripheral blood of AD patients. There is growing evidence showing peripheral immune changes in AD such as reduced CD4+ T cells and regulatory T cells (155). However, although T cell and B cell numbers and function have been shown to be altered in AD, there is a great amount of discordance between the results which are summarized in the review by Martorana et al. (155). The cytokine profile in the peripheral blood has also been documented reflecting the pro-inflammatory state associated with AD and are discussed in a review by Park and colleagues (156). Whilst attempts to accurately profile peripheral immune changes in AD may eventually lead to the discovery of new biomarkers, current understanding of these changes does not support their usefulness as biomarkers for monitoring the pre-symptomatic stage of AD. This is mainly due to virtually all current research into this being focused on patients with established AD. In addition, as discussed earlier, whilst a general state of neuroinflammation might predominate earlier in the AD continuum, the skewing of the immune response might be occurring much later in the course. It has also been found that the time spent in each phase of the continuum varies between individuals influenced by various factors (157). Essentially, what then remains to be more adequately understood is what the threshold is for pushing the immune system onto the self-perpetuating chronic inflammatory state towards immunosensence and what are the events that can precipitate this process. As discussed previously, events like CMV infection can lead to a state of chronic immune stimulation that depletes the limited immune reserves. This process of inflammaging is thought to play a key role in determining healthy and pathological aging (158). Better understanding and recognition of this threshold could help to identify patients for early intervention and produce better outcomes in old age.

It should be noted however that one main obstacle to research into peripheral immune changes in individuals with AD is the likely unwillingness of AD patients and their relatives to participate in the study given the substantial challenges already being experienced by them from the disease. This might be circumvented to some extent by gaining consent for post-mortem analysis of immune changes in the tissues, which might also have an added advantage of any findings directly reflecting the pathogenic process compared to peripheral blood immune changes which may differ. However, delays in acquiring and analyzing the samples and processing of the samples may alter the samples and potentially produce misleading findings. Alternatively, it might prove useful to further investigate the skull bone marrow especially for its therapeutic potential. The skull bone marrow provides a rapid source of immune cells to the brain parenchyma via direct vascular networks, which have been found in both mice and humans. Additionally, given the lack of a BBB in the dura mater, immune processes and cells may be more easily targeted by therapeutics and findings in mice models may be more easily translatable to humans (159).

Current attempts at producing a dementia risk assessment tool have begun making its way into clinical trials. For example, the Cardiovascular Risk Factors, Aging and Incidence of Dementia (CAIDE) score that takes into account risk factors like age, hypertension, hypercholesterolaemia, physical inactivity, obesity and educational attainment has demonstrated cognitive benefits in at-risk individuals identified by their CAIDE score following a multimodal intervention in the FINGER trial (160, 161). However, it may be possible that such risk scores may be limited in their effectiveness given the significant proportion of the population who may present with such a constellation of risk factors (162). This might make the large-scale implementation of multimodal interventions such as that used in the FINGER trial impracticable. It might prove more useful in future to design a risk assessment tool that factors in known precipitants and biomarkers of inflammaging such as CMV seropositivity and previous TBI to better target interventions to individuals that may already be progressing towards immunosenescence and AD.

There is good evidence to suggest that AD might be a consequence of CNS immunosenescence secondary to chronic inflammaging. The lack of therapeutic success of Aβ-directed and anti-inflammatory therapies alongside the recent discovery of novel properties of Aβ also prompts for a reconsideration of the role and function of Aβ, from being an etiological factor in AD pathogenesis to an innate immune response mediator. It is also important to note that whilst animal models have aided in our understanding of AD, there is still an unmet need for more humanized ex vivo models of the disease. This is due to the limitations intrinsic to current models such as 5XFADs that present an accelerated form of aging that does not occur in humans, in addition to differing brain areas and gene signatures between humans and the animal model. It is possible that single RNA sequencing may help in defining better understanding the disease. Additionally, characteristic AD features such as tau pathology may prove more useful as biomarkers rather than as potential therapeutic targets. However, future research into identifying new biomarkers especially in the preclinical stage of AD may aid development of early therapeutic interventions that may alter its clinical course.

SEJC performed the scientific literature search, design the figures and wrote the review. ES supervised, completed and finalized the review. All authors contributed to the article and approved the submitted version.

This work was supported by Alzheimer’s Research UK (Pilot Grant ARUKPPG2013B-2 to ES), Fondazione Italiana Sclerosi Multipla-FISM to ES (2014/R/21). And research contract grant number TMTL1DR to ES.

The authors declare the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.