In this part, we focus on the influence of the vitrectomy machine parameters on the effective failure rate. As for the clinical findings, first a univariate analysis was performed in order to select the major explanatory variables, then bivariate analysis and finally, a multivariate analysis with a logistic regression, using a step by step method in order to get the model that fits the best the data. The results of the multivariate analysis are presented at the end of this part.

The distribution of the gauge shows that, when a vitrectomy is performed, most surgeon have recourse to a 20 gauge or a 23 gauge. Only 6% use a 25 gauge.

When comparing the efficiency of the three different gauges, it seems that 20 gauge does significantly poorer than 23 gauge. No significant difference appears between 23 gauge and 25 gauge, partly due to the little number of surgeons using a 25 gauge.

However, the PVR distribution differs according to gauge used. The smaller the diameter the less advanced the PVR.

When comparing failure rates for 20 and 23 gauge for a given PVR stage, the trend remains the same : 20 gauge provides both a higher true failure and a higher remaining silicone rate. Although, none of these differences are statistically significant, these variables were considered in the multivariate analysis.

As far as the pump is concerned, a quarter of the surgeons use a flow control machine whereas the other 3 quarters only have access to or prefer a vacuum control machine.

At first sight, when looking at the failure rates depending on the pump used, it appears that a flow control pump is better than a vacuum control pump both in relation to the true failure rate as well as the remaining silicone rate. However, this difference may be explained by the differences in the treated cases with each of these pumps.

If we exclude the aggravating factors and consider the PVR distribution in relation to the pump used we can see that the cases are about the same with a little bit more easy cases stage 0 and difficult cases stage C1 in the venturi pump group. This cannot entirely explain entirely the statistically significant difference.

If we compare the failure rates without choroidal detachment and hypotony for each stage of PVR stage, we can notice that peristaltic pump still gets better results than the venturi pump, however these differences are not statistically different anymore. Yet, the pump was selected to enter the multivariate analysis, the results of which will be presented at the end of this presentation.

Then, different combinations of pump and gauge were compared together. A classification can thus be established and we can determine whether significant differences appear between 2 individual combinations.

The differences become significant when comparing the two types of pumps with 20 gauge.

Those differences are however not significant with 23 gauge.

Finally, concerning the cutting speed, it is noticed that 3 quarters of the surgeons use a high speed whereas the other quarter use a low speed.

No significant difference appears between low and high cutting speed when looking at all the data grouped together.

However, it is still possible to make a comparison using combinations of pump and cutting speed. High cutting speed seems increase the safety of a venturi pump.

Another comparison can be made using combinations of gauge, the pump and the cutting speed. Among the 8 possibilities the peristaltic pump obtains the 4 best results and the worst combination seems to be the Venturi with 20 gauge, especially with low cutting speed.

A multivariate analysis based on a logistic regression was performed. The variables taken into account were the initial clinical findings, the vitrectomy parameters and the maneuvers. A step by step approach was performed.

The result of this logistic regression shows that 2 vitrectomy machine parameters are independently significant. Venturi pump delivers a poorer surgical result whatever the initial findings of the eye : there are three times as much risk of a failure with a venturi pump than with a peristaltic one. Moreover, the multivariate analysis shows that using a high cutting frequency can counterbalance a little bit this tendency. However the combination venturi plus high speed is still worse than peristaltique whatever the cutting frequency.

Vitrectomy machine pump is therefore a major independent explanatory variable of failure rate.