OBJECTIVES: The objective of this study was to estimate the factors associated with the quality of evidence and its relationship with the recommendations regarding the incorporation of drugs issued by the National Committee for Health Technology Incorporation (Conitec).  METHODS: A paired and independent process of selection, based on Conitec recommendation reports about medicines available in final version to date of 26/6/2015, was performed. Based on data extracted, incremental cost-effectiveness ratio (ICER) and budget impact thresholds were adopted, assuming the 75^th percentile distribution. The relationship between potential quality predictors regarding the evidence presented in the reports was analyzed, together with the relationship to the recommendations issued by Conitec, by running univariate and multivariate (logistic regression) models.  RESULTS: Of 144 available recommendation reports, only 93 (65%) were about medicines. After selection, a total of 67 reports about 60 distinct medicines were included in the study. The medicines were framed in 18 medical specialties, from which stands out oncology, rheumatology, pulmonology and endocrinology. The ICER and budget impact thresholds adopted were R$ 70,000.00 and R$ 85 million, respectively. The bivariate analyses suggest as potential predictors of a low quality of evidence the indication of a rare disease (100% vs 73%; p < 0,05), the external origin of the demand (84% vs 58%; p < 0.05) and the year of the report (no time-dependent pattern). In the model adjusted by the budget impact and the year of recommendation, the low quality of evidence shows a chance approximately four times greater of receiving a non-favorable recommendation from Conitec (OR_adjusted: 4.33; IC95%: 1.26 to 14.93, p < 0.05).  CONCLUSIONS: Our results show the importance and consistency of the quality evaluation of evidence regarding recommendation about the incorporation of drugs into the SUS.