In the current study, the application of an additional hydrophobic resin after SBU did not improve retention nor other clinical parameters evaluated in NCCLs after 5 years, regardless of the adhesion strategy applied. Therefore, the results lead us to accept the null hypothesis and corroborate the trend observed in our previous paper at 36 months follow-up .
The incorporation of an extra hydrophobic resin layer in the sequence of simplified adhesives application has been consistently related with a retardation in dentin bond degradation in laboratorial studies [27, 33, 34], also for SBU [20, 24, 27, 28]. The reasons mainly attributed to this beneficial effect are a higher hydrophobicity, a better polymerization efficiency and a thicker film thickness that may protect the adhesive interface against hydrolytic degradation [24, 25, 27, 30, 34, 35]. Therefore, according to our results, we can state that this improvement obtained in vitro studies with SBU does not seem extrapolate to clinical behavior, at least using Scotchbond Multipurpose bonding agent as hydrophobic resin. Two clinical studies also evaluated the application of the same hydrophobic resin coat over one-step self-etch adhesives after 18 months of clinical service, reporting a beneficial effect on the clinical performance, mostly in terms of retention rate in one of them  and no effect in the other . As far as we know, no other studies have evaluated this topic using universal adhesives nor in a long-term period.
In the current clinical trial, Adper Scotchbond Multipurpose was selected as hydrophobic resin layer, based on the two clinical studies cited above [26, 31], instead of Heliobond, a non-solvated bonding resin used in several laboratorial studies [20, 24, 28, 29]. Thus, the composition of the resin used as an extra layer may have influenced the results obtained, as Adper Scotchbond Multipurpose contains HEMA in higher levels than SBU (30–40 wt% vs 15–25% in SBU). Although HEMA promotes the penetration of the resin monomer into demineralized dentin , [22, 36], also leads to water sorption and the formation of poly-HEMA hydrogels that are highly prone to degradation over time . Furthermore, the thicker adhesive layer produced by Scotchbond Multipurpose, might induce greater dimensional alterations due to expansion and contraction from temperature changes, resulting in the deterioration of the bonded interface . In agreement with this, marginal discoloration was not prevented in NCCLs restorations for 3-ER and 2-SE adhesion strategies after five years of clinical service in the present study, despite the fact that the universal adhesive was light-cured separately, as recommended by Ermis et al. .
This absence of clinical benefit on the application of a hydrophobic bonding resin after a universal adhesive should be confirmed using other coatings, such as a non-solvated one or a highly filled resin , instead of a less silica-filled adhesive resin as it was used in the present study.
According to our results, it seems that the adhesion strategy may have been more clinically relevant than an additional hydrophobic layer since a worse performance of NCCL restorations was detected when SBU was applied using the SE strategy. This fact was mainly evidenced by a lower retention rate when the adhesive was applied with both SE strategies compared to ER strategies (75% for 2-SE and 81% for 1-SE strategies vs 100% for both ER strategies), regardless of the application of hydrophobic coating, in accordance with the 36-month recall observations . These results are in agreement with the other 5-year clinical trial testing SBU, in which Matos et al. reported higher retention rate in ER modes (moist and dried dentin) (93%) compared to SE mode (81.4%) . And also, with the results of a meta-analysis comprising clinical studies using universal adhesives in NCCLs with up to 36 months of follow-ups .
Therefore, excellent retention rates were observed when SBU was applied after acid etching, with 3-ER or 2-ER adhesion strategies. In fact, this performance was expected as the application of phosphoric acid of the enamel produces a deeper and more pronounced etch pattern compared to SE mode, resulting in an increased micro-mechanical retention and consequently an optimal bond to enamel . In contrast, the poor etching in enamel margin in SE modes due to less acidic composition of SBU (pH = 2.7), is correlated with the significant differences in marginal discoloration noticed between 2-ER and 1-SE groups at 5-year follow-up period. Again, these results are in line with those reported by Matos et al.,  with SBU. They found less discoloration for the ER strategy compared to the SE strategy after 5 years of clinical function, using both USPHS and FDI criteria. Contrary to these findings, a meta-analysis did not find differences in marginal discoloration in NCCLs restorations according to the strategy applied (SE or ER) in universal adhesives , although only short and medium-term clinical trials were included.
The absence of acid etching of dentin has also been related with a higher retention loss when NCCLs treated with SBU were evaluated after 5 years, even when the enamel was selectively etched . The better long-term performance of this adhesive in dentin using the ER mode contrasts with in vitro reports [11, 40], which may be attributed to the different substrates used in those in vitro studies. Clinical trials are performed in NCCLs, whose main substrate is sclerotic dentin that differs from the healthy (sound) dentin used in laboratory studies. Exposed dentin in such NCCLs contains an hypermineralized layer with denatured collagen and bacteria, and crystalline deposits into the tubules . Acid etching of sclerotic dentin seems to enhance the limited micromechanical interlocking without hampering the chemical interaction of the 10-MDP and carboxyl groups of the polyalkenoic acid with the partially dissolved mineral content as a chemical interaction of SBU in ER mode also has been described in sound dentin as a thin density layer below the hybrid layer, the so-called reaction layer .
There were no significant changes in marginal adaptation among groups in the current study along the time. However, minor marginal defects at the enamel margins were observed in all groups, which could be due to excess adhesive flashes or chipping fractures that were considered as alfa under USPHS criteria, although could have been detected with FDI criteria as they are more sensitive and accurate for the marginal adaptation parameter .
Finally, another limitation of the present long-term study is that the dropout is higher than 20% and that only one universal adhesive was tested, therefore, more clinical studies are warranted to confirm our results.