The relationship between systemic lupus erythematosus (SLE) and thyroid diseases is still controversial. Due to confounders and reverse causation, previous studies were not convincing. We aimed to investigate the relationship between SLE and hyperthyroidism or hypothyroidism by Mendelian randomization (MR) analysis. We performed a two-step analysis using bidirectional two-sample univariable and multivariable MR (MVMR) to explore the causality of SLE and hyperthyroidism or hypothyroidism in three genome-wide association studies (GWAS) datasets, including 402,195 samples and 39,831,813 single-nucleotide polymorphisms (SNPs). In the first step analysis, with SLE as exposure and thyroid diseases as outcomes, 38 and 37 independent SNPs strongly (P < 5*10-8) associated with SLE on hyperthyroidism or SLE on hypothyroidism were extracted as valid instrumental variables (IVs). In the second step analysis, with thyroid diseases as exposures and SLE as outcome, 5 and 37 independent SNPs strongly associated with hyperthyroidism on SLE or hypothyroidism on SLE were extracted as valid IVs. In addition, MVMR analysis was performed in the second step analysis to eliminate the interference of SNPs that were strongly associated with both hyperthyroidism and hypothyroidism. 2 and 35 valid IVs for hyperthyroidism on SLE and hypothyroidism on SLE were obtained in MVMR analysis. MR results of two steps analysis were estimated respectively by multiplicative random effects-inverse variance weighted (MRE-IVW), simple mode (SM), weighted median (WME) and MR-Egger regression methods. Sensitivity analysis and visualization of MR results were performed by heterogeneity test, pleiotropy test, leave-one-out test, scatter plots, forest plots and funnel plots. The MRE-IVW method in the first step of MR analysis revealed that SLE was causally associated with hypothyroidism (OR = 1.049, 95% CI = 1.020-1.079, P < 0.001), but not causally associated with hyperthyroidism (OR = 1.045, 95% CI = 0.987-1.107, P = 0.130). In the inverse MR analysis, the MRE-IVW method revealed that both hyperthyroidism (OR = 1.920, 95% CI = 1.310-2.814, P < 0.001) and hypothyroidism (OR = 1.630, 95% CI = 1.125-2.362, P = 0.010) were causally associated with SLE. Results from other MR methods were consistent with MRE-IVW. However, when MVMR analysis was performed, there was no longer a causal relationship of hyperthyroidism on SLE (OR = 1.395, 95% CI = 0.984-1.978, P = 0.061), nor was there a causal relationship of hypothyroidism on SLE (OR = 1.290, 95% CI = 0.823-2.022, P = 0.266). The stability and reliability of the results were confirmed by sensitivity analysis and visualization. Our univariable and multivariable MR analysis revealed that systemic lupus erythematosus was causally associated with hypothyroidism, but did not provided evidence to support a causal relationship of hypothyroidism on SLE or between SLE and hyperthyroidism.