We studied the X-ray and optical absorption properties of 13 Gamma Ray Burst afterglows observed by BeppoSAX . We found that X-ray absorption in addition to the Galactic one along the line of sight is highly statistically significant in the two cases with the best statistics ( probability > 99.9 \% ) . In three other cases the presence of X-ray absorption is marginally significant ( probability \sim 97 \% ) . Measured rest frame absorbing column densities of hydrogen , N _ { H } , range from 0.1 to 10.0 \times 10 ^ { 22 } cm ^ { -2 } ( at 90 % confidence level ) assuming a solar metal abundance . X-ray absorption may be common , although the quality of present data does not allow us to reach a firm conclusion . We found that the rest frame column densities derived from XMM and Chrandra data as quoted in the literature are in good agreement with the BeppoSAX estimated rest frame N _ { H } range , supporting our result . For the same GRB afterglow sample we evaluated the rest frame visual extinction A _ { Vr } . We fitted the optical-NIR afterglow photometry with a power law model corrected at short wavelengths by four different extinction curves . By comparing X-ray absorptions and optical extinction , we found that if a Galactic-like dust grain size distribution is assumed , a dust to gas ratio lower than the one observed in the Galaxy is required by the data . A dust to gas ratio \sim 1/10 than the Galactic one , as in the Small Magellanic Cloud ( SMC ) environment , has been tested using the SMC extinction curve , which produces good agreement between the best fit N _ { H } and A _ { Vr } . We note , however , that the best fit N _ { H } values have been obtained by assuming solar metal abundances , while the metallicity of the SMC ISM is \sim 1/8 the solar one ( Pei 1992 ) . If such low metallicity were assumed , the best fit N _ { H } values would be higher by a factor of \sim 7 , providing a significant increase of the \chi ^ { 2 } . Alternative scenarios to explain simultaneously the optical and X-ray data involve dust with grain size distributions biased toward large grains . Possible mechanisms that can bring to such a grain size distribution are discussed .