Comment on: Beneficial effect of Mediterranean diet on disease activity and cardiovascular risk in systemic lupus erythematosus patients: a cross-sectional study

Dear Editor, We read with interest the article by Pocovi-Gerardino et al. [1]. The authors showed that low adherence to the Mediterranean diet (Med diet) among patients with SLE was associated with high levels of disease activity (SLEDAI-2K), organ damage [SLICC/ACR Damage Index (SDI)] and high-sensitivity CRP (hsCRP) as cardiovascular disease (CVD) risk markers. While non-pharmacological interventions such as protection against ultraviolet rays and smoking cessation are important in the lifestyle management of patients with SLE, the lack of data on dietary patterns remains a relevant issue. This study is excellent in that it used validated clinical indicators and provided specific modifiable dietary patterns. This study may help promote management based on shared decision-making between physicians and patients. However, there were some concerns regarding this study.

Our first concern was regarding residual confounding factors in the multivariable regression models. In this study, the same confounding factors were adjusted for via a stepwise algorithm in the heterogeneous outcomes: disease activity measured using the SLEDAI, organ damage measured using the SDI and CVD risk based on the hsCRP level. It is desirable to select clinically important confounders based on previous studies and expert opinions for each outcome. Without evidence- or expert knowledge-based selection, biased estimation due to residual confounding factors will occur. For example, physical activity is an important residual confounding factor, because low physical activity is associated with high levels of hsCRP [2] and low adherence to the Med diet [3]. Another important residual confounding factor for each outcome was socio-economic status (SES). Among SLE patients, low SES is associated with high levels of disease activity [4], organ damage [4] and hsCRP [5] and low adherence to the Med diet [6]. To avoid such misspecification of confounding factors, confounding selection based on evidence and expert knowledge is preferable to a stepwise method, especially in this study setting.

Our second concern was the lack of information regarding outcomes. Although nine variables were selected in each regression model [one exposure variable (Med diet) and eight confounding variables (age, sex, corticosteroid use, immunosuppressant use, antimalarial drug use, smoking status and BMI)], the authors did not provide the number of occurrence events for each outcome, which is essential to evaluate statistical stability. In general, at least 10 events per variable (EPV) should be considered to avoid biased estimation in logistic regression models [7]. In this study, the mean of the SLEDAI was 5.25 (s.d. 2.75) for the low-adherence group of 16, 3.79 (s.d. 2.87) for the medium-adherence group of 121 and 1.44 (s.d. 1.78) for the high-adherence group of 143, suggesting a concern that the number of active SLE cases (SLEDAI ≥5) was low and was not sufficient to meet the required EPV. To prove the results are statistically robust (i.e. preserving sufficient EPV), the authors should have presented the number of outcomes.

Our third concern is the interpretation of conflicting results. This study found an association between hsCRP levels and Med diet adherence scores using linear regression analysis, whereas they did not find an association in logistic regression analysis using a clinically meaningful categorization. Similarly, the Med diet adherence score was associated with SLEDAI scores in the linear regression, whereas good Med diet adherence (defined as ≥10 points) but not medium adherence (defined as 6–9 points) was associated with active SLE in the logistic regression models. Based on these conflicting results, the association between adherence to the Med diet and active SLE or CVD risk remains inconclusive.

In summary, the association between Med diet adherence and disease activity/CVD risk among SLE patients, as shown in this study, remains unclear due to misspecification of confounding factors, lack of outcome data and conflicting results in the linear and logistic regressions. Despite the aforementioned issues, this study is valuable because the authors evaluated a dietary pattern, which is novel and useful for lifestyle management of SLE. Further studies to compensate for these limitations are required.

Rheumatology key message

Funding: No specific funding was received from any bodies in the public, commercial or not-for-profit sectors to carry out the work described in this article.

• Non-statistical confounding selection, outcome data presentation and careful interpretation of the results are essential.

Disclosure statement: The authors declare no conflicts of interest.

Data availability statement

Not applicable.