The relative value of predictive factors of cochlear implant performance depends on follow-up time

Objective: Although cochlear implantation is widely applied, its outcomes are not very predictable on an individual basis. Several factors, such as the duration of deafness and the pre-operative speech understanding, have been identified as potential predictors. However, the relative value of these predictors varies between studies. This study evaluated the effect of the duration of post-operative follow-up on the value of pre-operative predictors. Study Design: 91 subjects implanted between 2000 and 2005 with a HiRes90K or CII implant with a HiFocus I electrode array were included in this study. The group was divided into subpopulations of positioner users (n=28) and nonpositioner users (n=63), as well as into young implanted (< 65 yr, n=66) and late implanted (> 65 yr, n=25). Speech understanding was evaluated using a CD to present CVC words in free field at 65 dB SPL. Testing was conducted at increasing intervals from 1 wk to 2 yr following hook-up. Correlations with the duration of deafness, age at implantation and pre-operative CVC scores were calculated directly. Multiple regression analysis were applied using the Iowa predictive model directly, or with this model extended to consider age at implantation or the presence of an electrode positioner. All analysis were performed on the entire population and for the subgroups described above. Results: While age at implantation showed no correlation with speech perception in the first year after implantation, after 2 yrs a negative correlation was found (r=-0.25;p < 0.05). Inversly age at onset of deafness showed a strong correlation in the initial phase (r=0.4; p < 0.0001 at 2wk) but this correlation diminished with longer implant use (r < 0.01; p > 0.5 at 2yr ). Duration of deafness showed a similar decrease in correlation with longer follow up time, remaining only barely significant after 2 yr (r=0.25; p < 0.05). No significant correlation was found between preand post-operative speech understanding. In line with these findings, the coefficients of the Iowa predictive model turned out to be heavily dependent upon the implant experience accrued before an evaluation was conducted. Conclusion: The relative importance of predictive factors is highly dependent on the post-operative experience. Therefore, predictive models should be based on follow-up times of at least 2 years to allow poorer performers approach their ultimate performance level. This is especially so for the duration of deafness: recipients with a longer duration of deafness have a much shallower learning curve, yet finally obtain similar results to those with shorter durations of deafness. Implantation at an older age limits the long term speech understanding improvement, resulting in a significantly poorer score after 2 yr