For the first time in Italy, eight different researches for the treatment of reading of dyslexic subjects, are compared for efficacy and efficiency. Efficacy was evaluated considering the differences in fluency and accuracy of reading obtained by the different treatments, against the expected "natural" development of reading without intensive treatment of dyslexic subjects. Efficiency was measured dividing the measure of efficacy with the length and intensity of treatment, that is months and hours per month, respectively. Treatments with the best values of efficacy and efficiency are those aimed at automatizing the relationship between syllables and their phonological correspondences using ad hoc software and the treatment based on the Balance Model that exploit the neurophysiological bases of reading. We also discuss the reasons of the less efficacy and efficiency results obtained by other treatments and the future perspectives of treatments for dyslexia in Italy.