OBJECTIVES: To perform a probabilistic sensitivity analysis to analyze previously reported results about the cost-effectiveness of oral hypoglycemic agents (OHA's) in the initial oral drug treatment of patients diagnosed with type 2 diabetes mellitus in public primary attention in Mexico. METHODS: A probabilistic sensitivity analysis was made in order to analyze results previously reported in which a deterministic sensitivity analysis was performed to study the cost-effectiveness of three OHA's: metformin, glibenclamide and acarbose.  We used TreeAge-Pro® software for programming and simulating a Markov model of two health states (HbA1C≤7% or HbA1C>7%) and twelve cycles of 1 month for a time horizon of 1 year.  The parameters of monthly success probability as beta distributions and monthly costs as lognormal distributions of therapeutic alternatives were computed through a parametrization of data.  Monte Carlo´s simulations were computed for cohorts of 10,000 patients for each treatment option. RESULTS: The results of the Monte Carlo´s simulations showed very close iterations clouds for metformin and glibenclamide showing evident dominance of both over acarbose.  In the acceptability curve generated, for a willingness to pay (WTP) = 0  the probabilities to be cost-effective were 49.46 %, 43.04 % and 7.50 % for glibenclamide, metformine and acarbose, respectively,  whereas for a WTP = 1 mexican GDP per capita  (US $ 7876.00 in 2009) were 66.26 %, 26.98 % and 6.76%.  The glibenclamide versus metformin incremental cost-effectiveness analysis showed similar results as mentioned before, showing 59.72% of iterations below the WTP = 1 mexican GDP per capita line. CONCLUSIONS: The probabilistic sensitivity analysis showed which the initial drug therapy with glibenclamide or metformin have advantage over acarbose. There is not sufficient evidence to say glibenclamide has advantage over metformin for WTP near to cero, as in low to middle income countries where containment of expenditures is important.