Thirty-six metal dies were machined out of carbon steel to dimensions shown in Figure 1. During the machining procedure all the surfaces to which the zinc phosphate cement would be attached were finished to roughness values around 6.3 μm. The reason we used the 6.3 μm surface roughness was that a preliminary evaluation of dies with surface roughness values of 3.2, 6.3, 8.0 and 12.5 μm had revealed that a surface roughness value of 6.3 μm was ideal for our study. With such surface texture, the cement did not separate from the cement-model surface. Instead it fractured within the cement or at the crown-cement interface.

To verify the surface roughness values, the finished die surfaces were recorded with a profilometer (Federal Surfanalyzer System 5000, Federal Products Co, Providence, RI). The surface roughness value, R_a, represents the arithmetic average of the absolute values of the measured roughness profile height deviations taken within the scanned length and measured from the mean line. These scanned recordings were made in a cervical to occlusal direction over a length of 3 mm on each metal die. The surface roughness value for that distance was then used to determine the average value for all the dies.

Before impressions were made of the metal dies, a 1.6 mm thick ring was inserted and located to the marginal part of the simulated crown preparation (Figure 2). Impressions were then made in a polyvinylsiloxane impression material (Light Body, President, Coltène AG, Altstätten, Switzerland) supported by an individual tray. The tray had been covered with an adhesive to secure a reliable tray-impression attachment, and the space between the tray and the die was 2 mm. One hour after the impression had been made it was poured with a Type IV gypsum (Silky-Rock, Whip-Mix Corporation, Louisville, KY) and allowed to set during the night. After impression removal and die inspection, the 36 gypsum dies were sent to laboratories making Denzir and Procera copings.

Twenty-four Denzir and twelve Procera copings were ordered from Denzir and Procera certified laboratories. All ceramic copings were made 0.6 mm thick and with a cement space corresponding to 60 μm. That spacing started 0.8 mm from the cervical margin and reached its maximal thickness after 1.2 mm from that margin. When the copings arrived from the laboratories, all 36 copings were checked regarding their fit.

Twelve of the Denzir copings were sandblasted on the inner surfaces with A1₂O₃ (particle size = 50 μm) using an air pressure of 2 bars (200 kPa). The sandblasting process was done with the sandblasting tip located at a distance of 10 mm from the ceramic surface. The centre of the sandblasting stream targeted the transition from the occlusal to the proximal inner surfaces. The entire inner surface was then sandblasted by rotating the coping four times, each time 90 degrees. Each of these locations was sandblasted for 5 s.

The inside of each coping was scanned with the profilometer. The scans were collected within the 0.5 to 1.5 mm interval from the cervical margin. From these scans the R_a values were calculated.

The 1.6 mm ring, located in the cervical region of the preparation when the silicone impression was made was removed and replaced with a 1.55 mm thick washer with an outer diameter of 18 mm (Figure 3). A zinc phosphate cement (Phosphate Cement, Heraeus Kulzer, Dormagen, Germany) was mixed on a room tempered glass plate. For each portion, 1.2 g powder was mixed with 0.5 mL liquid. The powder was divided into six portions (two 1/16^th, one 1/8^th, and three 1/4^th portions). First, one 1/16^th portion was mixed for 10 s, then the second 1/16^th portion for 10 s, followed by the 1/8^th portion for another 10 s. A 1/4^th portion was then added and mixed for 15 s followed by another 1/4^th portion, also mixed for 15 s. The final 1/4^th was then added and mixed for 30 s. Thus, a total mixing time of 1 min and 30 s was used. The mixed cement was then placed inside the ceramic coping, which was rotated 90 degrees as the coping was seated on the metal die. Thirty seconds after completed mixing, a load of 2 N was placed on the crown, and the load acted on the crown for 5 min. Excess material was removed after 7.5 min counted from the time the coping was loaded with the 2 N load.

Fifteen minutes after the initiation of the cementation process the cemented copings with the steel dies and washers were transferred to distilled water or artificial saliva and then stored in an oven at 37°C. The artificial saliva 17 was of the following composition: 0.1 L each of 25 mM K₂HPO₄, 24 mM Na₂HPO₄, 150 mM KHCO₃, 100 mM NaCl, and 1.5 mM MgCl₂. To this were added 0.006 L of 25 mM citric acid and 0.1 L of 15 mM CaCl₂. The pH was then adjusted to 6.7 with NaOH or HCl and the volume made up to 1 L. To avoid bacterial growth, we added 0.05% by weight thymol to the artificial saliva. All chemicals were ACS-grade (American Chemical Society).

Fourteen days after cementation, half of the specimens (3 Denzir as received, 3 Denzir being sandblasted, and 3 Procera, all stored in distilled water; and 3 Denzir as received, 3 Denzir being sandblasted, and 3 Procera, all stored in artificial saliva) were tested in tension (Figure 4) until failure using a specially designed testing device in an Instron Universal Testing machine at a load rate of 0.5 mm/min. After one year (3 Denzir as received, 3 Denzir being sandblasted, and 3 Procera, all stored in distilled water; and 3 Denzir as received, 3 Denzir being sandblasted, and 3 Procera, all stored in artificial saliva), the remaining 18 specimens were also tested as described earlier.

The force values, needed to dislodge the copings, were used for the statistical evaluation. One-way and two-way ANOVA:s were used to determine significant differences between materials, storage medium and storage time as well as interactions of these (ANOVA, SAS Institute, Cary, NC, USA). Comparisons between the individual groups were also conducted using Duncan's multiple range tests. All test were conducted on the 95% significance level.

The profilometer readings of the metal die surfaces gave an average surface roughness value (R_a) of 5.49 ± 0.98 μm. These roughness values were primarily based on wave shaped surface where the distance between the peaks was around 200 μm and where the main peak-to-main valley distance was around 20 μm (Figure 5).

The surface roughness values of the insides of the different coping groups are shown in Table 1. No significant difference existed between the three combinations (p = 0.2239).

The statistical analysis revealed that the most important factor affecting the retention force was storage time (Tables 2 and 5). There was no difference between the two main materials or whether the Denzir copings had been sandblasted or not (Table 3).

Comparing the storage media could not prove whether such a difference existed (p = 0.082) (Table 4). In this comparison, no consideration was taken for the different material groups and storage times. When storage time only was compared there was a significant increase in retention force with time (Table 5).

As seen in the Tables 6, 7 and 8, there are large differences among the different test groups (standard deviation ~30% of the mean value). From Table 2, we can also see that there are no significant interactions, although the material/storage and storage/time interactions are pretty close.

Of the Procera copings, two copings fractured during testing after 14 days of storage and one coping fractured after one year. Of all tested Denzir copings, not a single coping fractured. However, careful inspection with transilluminating light revealed that one of the Denzir copings tested after 1 year had a crack that extended from the cervical region to the occlusal region.

Table 1 shows that there was no significant difference in surface roughness between the three evaluated ceramic coping groups. The low value of the sandblasted Denzir copings suggest that the machining process generated a surface roughness that was at least as rough as a machined and sandblasted Denzir surface. Because of these findings, sandblasting conducted under the conditions evaluated in this study is not recommended for Denzir copings.

The statistical evaluation of variables such as material, storage and time as well as interactions of these variables revealed that storage time was significant regarding retention force (Table 2). Storage medium and interaction between time and storage medium were almost significant on the 95% significance level.

There was no difference between the three material groups regarding retentive force (Table 3). That finding most likely relates to the similarities in surface roughness values among the three groups (Table 1). The similarities in retentiveness among the three material groups are important. Because of the published success rate of Procera after 10 years in clinical service 13, our in vitro results suggest that Denzir copings, sandblasted or not, and cemented with zinc phosphate cement are likely to perform equally well as Procera crowns, at least regarding retention.

Based on the lower fracture frequency identified among the Denzir copings, our findings suggest that Denzir copings might perform even better than Procera crowns. Of the twelve tested Procera copings, three complete fractures occurred in the Procera copings, while of the twenty-four tested Denzir copings only one had a detectable crack that did not even result in a clear fracture during testing. At the present time, though, one cannot exclude that these differences are coincidental. However, the higher fracture toughness of Denzir, almost twice as high as that of Procera, probably explains the lower fracture tendency of Denzir identified in this study.

Future studies regarding CADCAM technologies need to focus on flaw formation that might be induced during manufacturing. One may suspect that a milling process like the one used to make the Denzir copings, induce more flaws than a pressing and sintering technique as the one used to make Procera copings. However, there is no proof available supporting that assumption at the present time. In the case of Procera, one cannot exclude the possibility that flaws are induced when copings are pressed and that these flaws may not heal completely during sintering. Besides, during sintering and cooling, thermal stresses may be induced that trigger crack formation in the future.

From the above argumentation, flaws may very well be induced during manufacturing of both Denzir and Procera copings. Thus, differences in either fracture toughness or flaw sizes/densities, or a combination of the two, would explain why the Procera copings had higher fracture tendency. The higher fracture toughness of zirconia favours Denzir and would explain the lower fracture frequency seen in these copings. However, whether the flaws introduced in Denzir copings are smaller or bigger than those present in Procera copings is not known and needs to be investigated further. Flaw formation during manufacturing becomes very important when we compare different zirconia crowns that now are available on the market. Some of them are made by milling industrially sintered and processed zirconia, while other are made by milling presintered zirconia that is then sintered.

During our evaluation, we used the force levels generated by the copings that fractured during testing. One could argue that such values should be excluded because the samples fractured. However, we did not exclude those samples of the following reasons: First. the force levels on the copings that fractured were not lower than those of those that did not fracture. Second, we were not able to determine whether the fracture occurred before or after debonding had occurred because of the speed of the dislodgement/fracturing process.

Storage of the copings in artificial saliva resulted in force values almost significantly higher than those stored in water (Table 4). A possible explanation is that some of the ions, for example phosphate ions, diffused into the cement and pushed the setting reaction toward an increased precipitation reaction. Such an explanation can be related to the setting reaction of zinc phosphate cements. As the storage time increased, the required force needed to dislodge the copings also increased (Table 5). A likely explanation is that as time passed the setting reaction became more complete. There is also a possibility that corrosion of the steel dies and release of iron ions from the dies affected the setting reaction of the zinc phosphate cement. Such a corrosion process might also have increased the surface roughness at the cement-dye interface and thereby also increased the mechanical retention.

Even though time improved the retention of the cemented copings, one should not extrapolate that value to the clinical situation. Clinically, the coping would be exposed to different loads during the entire observation time. In our study, no such forces acted on the cemented coping from time of cementation to time of testing. However, the improved results with time shows that storage media such as water and artificial saliva by themselves do not decrease the retention force. This finding is important, because it implies that other factors are more important when we try to explain why the retention of zinc phosphate cemented crowns sometimes fail over time. Loading conditions and fracture toughness of the luting agent are such factors.

Our results suggest that a clinical evaluation of Denzir crowns cemented with zinc phosphate cement are likely to perform as well as Procera crowns cemented with zinc phosphate cement. However, based on Burke et al.'s 5 review, which supported the use of resin cements, one can question the rational of even considering using zinc phosphate cement as a luting agent in a clinical study. There are at least two reasons justifying such a clinical study. First, by assuming that the high success rate of zinc phosphate cemented Procera crowns is likely to be equally high with Denzir copings, justifies such a study ethically. Second, the simplicity of using zinc phosphate cements, their ease of removal from marginal regions after setting, and the ease with which a zinc phosphate cemented crown can be removed if remake is needed, are beneficial clinical advantages that cannot be neglected.

Having justified the use of zinc phosphate cement in a clinical study, it is also important to emphasize that such an evaluation should consider retention and solubility of the luting agent too. In the Procera study conducted by Ödman and Andersson 13, retention failures requiring recementation were not included in their impressive success rate. Present evidences suggest that resin bonding improves the results with ceramic restorations 5, even though these claims are not conclusive 1819. There is no doubt that retention is an important factor to consider, but one must also accept that strong bonding can also be a drawback if the crown needs to be removed. In the latter case, a well-bonded ceramic restoration can be a bigger clinical challenge than the need for recementing a less well-bonded restoration.

One often hears the claim that resin cements decrease the fracture frequency of ceramics. Such a claim is justified for some ceramic systems, but is may not be valid when we are dealing with high strength ceramic copings like the ones used in both Procera and Decim. Instead, some clinical studies dealing zinc phosphate cemented alumina copings are so good that one can question whether resin bonded copings will outperform these results. It is first when comparative studies take all these pros and cons into consideration, as we know whether resin bonded alumina or zirconia copings outperform zinc phosphate cemented alumina or zirconia copings.