Dyslipidemia negatively impacts the development of nonalcoholic fatty liver disease (NAFLD); yet, the specific relationship between non-high-density lipoprotein cholesterol (non-HDL-C) and NAFLD has not been previously examined. This study aims to explore the association between levels of non-HDL-C and the prevalence of NAFLD in adults residing in Western China. We used data from the Xinjiang physical examination program in 2023, which included a total of 186,965 adults in Kumul District. A variety of statistical techniques were applied, including threshold effect analysis, subgroup analysis, smooth curve fitting, and multivariate logistic regression, to clarify the relationship between non-HDL-C and NAFLD. Of the 79,961 participants, the mean age was 57.21 ± 13.08 years, and the mean non-HDL-C was 3.47 ± 1.10 mmol/L. Smooth curve fitting revealed a non-linear relationship between non-HDL-C and NAFLD, with a significant threshold effect at 4.43 mmol/L (P for likelihood ratio test < .001). Below this inflection point, each unit increase in non-HDL-C was associated with a 35% higher prevalence of NAFLD (OR = 1.35; 95% confidence interval (CI): 1.32-1.37; P < .0001), whereas no significant association was observed above this threshold (OR = 0.98; 95% CI: 0.95-1.02; P = .3830). Conventional quartile-based analyses also showed a positive trend (Q4 OR = 1.90; 95% CI: 1.81-1.98), but these findings need to be interpreted with caution, given the non-linear nature of the underlying relationship. Interaction tests showed that the non-HDL-C-NAFLD association remained significant across most subgroups, with notable interactions observed for body mass index, education, exercise, and hypertension. Our findings suggest a non-linear association between non-HDL-C and NAFLD, with a threshold at around 4.43 mmol/L. Below this level, higher non-HDL-C was linked to increased NAFLD risk, but above it, the association leveled off. This suggests that threshold-based approaches may better capture the true relationship than simple linear models. Further large-scale prospective studies are still needed to confirm these findings.