Previous studies suggest that the mechanism of action of thymosin is reconstitution of immune defects rather than augmentation of relatively normal levels of cellular immunity. To test this hypothesis, we correlated in vitro assays with clinical course in 55 patients with bronchogenic carcinoma who were randomized to receive subcutaneous inoculations of thymosin 60 mg/m2 (T60), thymosin 20 mg/m2 (T20), or no thymosin (P) twice weekly during the first six weeks of an intensive chemotherapy program. Overall tumor response did not differ significantly among the three treatment groups (p >.12); however, survival of the T60 group was significantly greater than that of the P group (p =.017). Duration of response for patients achieving a complete remission was significantly longer in the T60 group than in the P group (p =.02). In an attempt to define an association between the increased survival of the T60 group and in vitro assays that show changes with tumor presence and extent, comparisons were made between survival and pretreatment peripheral blood levels of carcinoembryonic antigen (CEA), α2HS-glycoprotein (α2HS), and total thymus-dependent lymphocytes (T cells). Independent of therapy, pretreatment CEA levels correlated directly with tumor extent (p=.004) and inversely with degree of tumor response (p=.006) and survival (p=.006). The effect of T60 compared with the P group on survival was similar in patients with pretreatment CEA levels >4 Ng/ml and those with levels >4 Ng/ml. However, among those with pretreatment total T-cell and α2HS levels below the median for all patients, survival of the T60 group was significantly greater than that of the P group (p=.006 and p=.036, respectively); for patients with higher pretreatment total T-cell and α2HS levels, survival did not differ significantly between the T60 and P groups (p=.28 and p=.17, respectively). Analysis by tests of interaction shows that the effect of T60 on survival depended on pretreatment levels of α2HS (p=.046) and, to some extent, total T-cells (p=.17). Thus the association between survival and type of treatment may be related to pretreatment levels of these two parameters of cellular immunity. These results support the hypothesis that patients with relatively low levels of immunity benefit most from administration of thymosin. Additional studies with larger patient populations and other tumors are needed to confirm these findings. Cancer 43:863–870, 1979.