Protective effect of extra virgin olive oil (EVOO) consumption on the physical component of health-related quality of life in aging adults

Javier Conde-Pipó, Cristina Molina-Garcia, Julián Arense, José Daniel Jiménez-García, A. Martínez-Amat et al. (+1)•View original

ComprehensiveExpertalloy voice

Welcome back. Today we’re going to take a deep dive into the methodology of a 2026 European Journal of Nutrition paper entitled “Protective effect of extra virgin olive oil (EVOO) consumption on the physical component of health-related quality of life in aging adults,” by Conde-Pipó and colleagues. I’ll walk you through the study design, recruitment and eligibility logic, operationalization of exposure and outcomes, psychometrics, and the statistical analysis plan, and I’ll spend time on the subtle design choices—especially restriction and group definition—that shape what we can and cannot conclude. I’ll also refer explicitly to the tables and figure in the paper, describing them verbally so you can reconstruct how the analysis was organized. Let’s set the stage. This is a cross-sectional, descriptive, and comparative study in nutrition epidemiology, executed in Spain, that asks a very specific question: among physically active middle-aged and older adults who already adhere to a Mediterranean dietary pattern, does higher consumption of extra virgin olive oil appear to attenuate the age-related decline in the physical component of health-related quality of life? The motivation, laid out in the introduction, is that EVOO is a central pillar of the Mediterranean diet and is rich in bioactives—monounsaturated fatty acids like oleic acid, and phenolics like oleocanthal and hydroxytyrosol—that plausibly modulate inflammation, oxidative stress, and pain perception in aging. But most prior work has connected overall Mediterranean pattern adherence with health-related quality of life; very few have isolated EVOO intake specifically. So the authors used design restriction to hold constant two major correlates of quality of life: physical activity and overall Mediterranean dietary adherence, then stratified the sample by EVOO consumption—less than four tablespoons per day versus four or more—and looked at age correlations within those strata. The study population came from an initial pool of 553 individuals recruited “randomly over a 3‑month period” from different regions of Spain, with voluntary participation following informed consent. The phrase “recruited randomly” isn’t unpacked—it likely indicates random outreach rather than probability sampling from a defined frame—but all the telltale features of a volunteer sample are present: informational briefings, the ethical guardrails of the Declaration of Helsinki, and research ethics approval from the University of Granada. The paper then applied four inclusion criteria to that pool: age 41–80 years; an active lifestyle according to WHO guidelines; self-reported good health with no conditions interfering with activities of daily living; and good adherence to the Mediterranean dietary pattern. After exclusions for unmet criteria or incomplete questionnaires, the final analytic sample was 180 participants. That attrition—from 553 to 180—is substantial, and methodologically it’s doing heavy lifting: it’s a restriction strategy to reduce heterogeneity in physical activity and dietary pattern and eliminate individuals whose disease burden might independently depress quality of life. Let’s be concrete about how those inclusion criteria were operationalized. Physical activity was assessed with the Spanish version of the Rapid Assessment of Physical Activity (RAPA-Q), a brief seven-item instrument designed for older adults that uses yes/no items to classify activity levels. The authors aligned classification with the 2020 WHO thresholds—active defined as at least 150 minutes per week of moderate intensity or 75 minutes per week of vigorous intensity. While RAPA-Q is indeed validated in Spanish-speaking populations and is intended to provide a pragmatic activity categorization, it’s worth noting as a methodological nuance that RAPA-Q is not a time-quantified log; it relies on categorical endorsement of activity levels. In this paper, participants were classified as active if their RAPA-Q responses implied they met or exceeded the WHO thresholds. The mapping rules aren’t specified here, but the alignment with WHO thresholds is explicit and serves the purpose of inclusion. Dietary pattern adherence was measured using the 14-item Mediterranean Diet Adherence Screener (MEDAS), itself derived from a longer food frequency questionnaire with a reported correlation of r=0.52, p<0.001 in validation work. MEDAS awards one point for each criterion met across 12 intake frequency items—EVOO, wine, fruits, vegetables, fish, legumes, nuts, meat and derivatives, poultry, butter, pastries, and sweetened beverages—and two items about MedDiet characteristics such as exclusive EVOO use for cooking. Consistent with prior Spanish studies, a score of 10 or higher indicates good adherence to the Mediterranean pattern. Participants who fell below that threshold were excluded. Critically for this study, the EVOO item explicitly queried daily consumption inclusive of all uses—frying, salads, and meals away from home—and defined the tablespoon unit as 13.5 grams. A positive response indicated a daily intake of at least 4 tablespoons, which aligns with approximately 50 mL per day, citing the recommended EVOO intake for MedDiet benefits. This single item then became the stratifying variable: LT4 for those consuming less than 4 tablespoons per day, and MT4 for those consuming 4 or more. Health-related quality of life, the outcome domain of interest, was assessed with the Spanish version of the SF-36. The instrument comprises eight domains—physical function (10 items), role physical (4 items), bodily pain (2 items), general health (5 items), vitality, social function, emotional role, and mental health—each scored from 0 to 100, higher meaning better. The authors also computed the two standard summary scores, the physical component (Comp-P) and the mental component (Comp-M), using the Spain-specific weighting algorithm. They note that cut-off points for dichotomizing these components into two levels were based on Spanish population median norms by age and gender; in practice, however, their analyses treat Comp-P and its subdomains continuously, using correlations with age. Regarding psychometrics, they report internal consistency in their sample with Cronbach’s alpha exceeding 0.75, and they cite typical Spanish normative alphas for the subscales—physical function around 0.93, role physical 0.95, bodily pain 0.87, general health 0.79—indicating that the measurement noise at the scale level is unlikely to dominate the signal in this population. Sociodemographic and health status variables—age, sex, weight, height, and self-reported diseases in the past 12 months including cardiovascular disease, hypercholesterolemia, diabetes—were captured with an ad hoc questionnaire. That means anthropometry and comorbidities are self-reported; body mass index is thus computed from self-reported height and weight. The reliance on self-report is consistent across the study: diet, activity, and quality of life are all self-reported constructs, though using validated tools where applicable. Now, to the statistical analysis plan. Everything was conducted in R. The authors applied the Kolmogorov–Smirnov test with Lilliefors correction to evaluate distributional normality of continuous variables—appropriate when the mean and variance are estimated from the sample—and Levene’s test to assess homoscedasticity across groups. Given violations of parametric assumptions, they used the Mann–Whitney U test for continuous variable comparisons between the EVOO strata, and the Pearson Chi-square test for categorical comparisons. For the primary association of interest—aging and physical health domains—they computed Spearman’s rank-order correlations between age and the SF-36 physical component and its subdomains, separately within the LT4 and MT4 EVOO groups. Internal reliability was again summarized via Cronbach’s alpha. All p-values were two-tailed, with significance set at 0.05. One subtlety here: the paper reports correlation coefficients accompanied by confidence intervals in Table 2, but it does not state how those intervals were computed—Spearman’s rho CIs are often obtained via bootstrapping or Fisher z-transform approximations. Nonetheless, the presence of CIs helps contextualize the precision of the estimates, especially given the unequal group sizes. Before moving to the figures and tables, let’s pause on the overall design logic. This is classic cross-sectional stratification with design-level restriction. By including only physically active individuals with good Mediterranean diet adherence and no major functional limitations, the authors reduce between-person variability driven by low physical activity, poor diet, or overt disease. In epidemiologic terms, they are attempting to minimize confounding by PA and overall diet by restricting the sample. They then operationalize EVOO dose as a dichotomy at a biologically and culturally meaningful threshold—4 tablespoons per day—and ask: as age increases within each EVOO stratum, does self-perceived physical functioning erode at different rates? That’s not a causal design—temporal ordering for EVOO intake and quality of life is not established—but it is a clean cross-sectional look at effect modification by exposure stratum, with correlations as the summary. The choice of Spearman’s rho is sensible for monotonic but potentially non-linear associations between age and the ordinally bounded SF-36 scales. Let’s interpret the descriptive table first. Table 1, on page 4, presents anthropometric and health status by sex and by EVOO consumption group. The sex distribution is unbalanced: 131 men (72.78%) and 49 women (27.22%), with that difference of course statistically significant. By EVOO group, the distribution is also skewed: 55 in LT4 (30.56%) and 125 in MT4 (69.44%). That asymmetry has analysis implications—precision is higher in MT4—and it may reflect underlying dietary behaviors in this active, MedDiet-adherent Spanish cohort. Among sexes, men were taller and heavier, with higher BMI, and had higher reported prevalence of diabetes and musculoskeletal disease in the past year; those sex-specific differences reached statistical significance. However, across EVOO strata—LT4 versus MT4—there were no statistically significant differences in anthropometry, in the physical component score (Comp-P), or in the listed health conditions. That balance across EVOO groups on observed covariates is helpful for internal validity; it reduces the likelihood that the EVOO strata are simply proxies for different baseline health. But keep in mind there’s no multivariable adjustment; any residual confounding by unmeasured factors—ses, medication, detailed nutrient intake—remains possible. Table 2, also on page 4, is the analytic centerpiece for methodology: it lists the bivariate correlations of each SF-36 physical domain with age, stratified by LT4 and MT4. For LT4, physical functioning correlates with age at r = −0.39 with a 95% confidence interval of approximately −0.58 to −0.17, p=0.003. Role physical is essentially null, r = −0.07, p=0.829, and general health is near zero, r = 0.03, p=0.829. Bodily pain shows a moderate negative association, r = −0.33, CI roughly −0.52 to −0.11, p=0.014. Most importantly, the composite physical component, Comp-P, shows r = −0.35, with the CI spanning −0.58 to −0.09, p=0.009. For MT4, physical functioning also declines with age, r = −0.37, p=0.001, which underscores that physically active, well-nourished people still experience age-related reductions in function. But bodily pain is no longer significantly related to age; the point estimate is modestly positive at 0.10, p=0.234. And the composite Comp-P is essentially flat with respect to age: r = −0.07, CI −0.24 to 0.10, p=0.431. The authors visualize the Comp-P correlation in Figure 1 on page 5, a scatterplot of age on the x-axis and Comp-P on the y-axis, with points colored or separated by EVOO group. You can picture two fitted trend lines: in the LT4 group, there’s a downward slope as age increases—indicating lower perceived physical health with advancing age—while in the MT4 group, the slope is near-zero, showing stability of the composite physical quality-of-life score across ages 41 to 80 in this cohort. Methodologically, what does this tell us? First, the use of within-stratum correlations avoids the ecological fallacy—associations are estimated at the individual level within exposure strata. Second, the negative physical functioning correlation in both groups indicates that even with high PA and MedDiet adherence—and even with higher EVOO—self-reported physical functioning declines with age. EVOO is not a panacea. But the two domains that aggregate to a broader sense of physical health—bodily pain and the composite score—are differentially associated with age by EVOO intake. That supports the prespecified hypothesis that higher EVOO intake could modulate pain perception and the overall physical health construct in aging, at least cross-sectionally. Turning back to measurements, it’s important to scrutinize the exposure variable. EVOO intake was classified by a single MEDAS item with a daily threshold. It captured total intake across cooking and raw uses and even meals outside the home. The tablespoon was standardized as 13.5 g. A threshold-based dichotomy is pragmatic and maps to dietary guidelines recommending about 50 mL daily intake for the MedDiet’s protective effect. But it also discards granularity—someone at 3.9 tablespoons is grouped with someone at 0.5 tablespoons, and someone at 4 uses sits with someone at 8. If the underlying dose–response is monotone, dichotomization attenuates power and can obscure patterning. On the other hand, because the question is whether those at or above the recommended intake are buffered against age-related decline, a binary classification has clinical interpretability. The outcome measurement—SF-36—was appropriately processed. Using Spain-specific weights for the Comp-P summary ensures cultural adaptation of item contributions. Reporting internal consistency in the analytic sample is good practice. It’s also notable that the authors did not rely solely on the composite but examined domain-level associations—physical functioning, role physical, bodily pain, general health—so we can see that the primary effect separation appears to reside in pain and in the aggregated composite rather than in functional capacity alone. The bodily pain subscale’s breadth is worth emphasizing: as the discussion notes, it reflects perceived pain interference and intensity but does not distinguish etiologies—musculoskeletal, neuropathic, inflammatory. In methodology terms, this is construct validity that is broad rather than specific; it’s a good global measure of bodily pain’s impact on daily life, but it is not mechanistic. For covariates, what did they control? Design restriction did most of the work: only physically active, MedDiet-adherent, generally healthy adults were included. This is a strength for internal validity—major determinants of HRQoL like sedentarism and poor diet are effectively held constant—but it also means the sample is a selective slice of the aging population, and it opens the theoretical door to selection bias. If unmeasured factors influence both EVOO intake and the probability of being physically active and MedDiet-adherent, conditioning on those inclusion criteria could in principle induce collider bias. Given the plausibility that socioeconomic status, health literacy, or regional culinary patterns could influence both, one has to be cautious about causal interpretation. The paper does not draw causal conclusions; it’s appropriately framed as association. Another careful point: While Table 1 suggests balance across EVOO groups for BMI and comorbidities, the sex distribution is skewed in the overall sample, and the analysis does not stratify or adjust by sex when estimating the age–Comp-P associations. If sex modifies the relationship between pain and aging, or if EVOO intake distributions differ by sex within this active cohort, that could confound or modify the observed correlations. The authors do not present sex-stratified correlations or multivariable models that adjust for sex; they rely on ROC-like design restriction and bivariate correlations. Given the large preponderance of men in the sample—about 73%—the precision for women is reduced, and external validity for older women is less certain. Let’s also talk about the statistical testing choices. The authors used Kolmogorov–Smirnov with Lilliefors correction for normality. Many would favor Shapiro–Wilk for small to moderate samples, but K–S with Lilliefors is fine, and given that assumptions for t-tests were not met, the nonparametric Mann–Whitney U for continuous group comparisons is appropriate. For the correlation work, Spearman’s rho is a robust choice for monotonic associations when variable distributions are non-normal or have bounded ranges. They do not report any adjustment for multiple testing across the several correlations in Table 2; in a strictly confirmatory context that would be a concern. But the correlations are focused on a small number of prespecified domains with clear hypotheses—Comp-P and bodily pain—so the family-wise error risk is modest. Still, p-values cluster around the 0.01–0.02 range for the LT4 correlations, which would likely retain significance under modest corrections, and the null in MT4 for Comp-P is comfortably non-significant. Confidence intervals in Table 2 provide more information than raw p-values and are appreciated. From a measurement error standpoint, we have classic non-differential misclassification issues. EVOO intake is self-reported, the threshold is coarse, and recall bias or social desirability could push reporting upward among health-conscious, MedDiet-adherent participants. If misclassification is non-differential with respect to the outcome, it generally biases associations toward the null. That means the observed difference in age–Comp-P correlation across EVOO groups might be conservative. However, if higher pain or lower physical functioning changes how people remember or report their oil intake—for example, people in pain may understate or overstate certain dietary intakes—that could complicate interpretation. The paper acknowledges recall bias potential explicitly. A final methodological note on sample size and power: there’s no mention of a priori sample size calculation or detectable effect size estimates. The LT4 group has n≈55; the MT4 group n≈125. For Spearman’s correlations, power to detect r around 0.3 at alpha 0.05 is good in MT4 and marginal in LT4; nonetheless, they observe r = −0.35 in LT4 for Comp-P with p=0.009, which is stable enough. The null in MT4, r = −0.07, is not a function of low power; the sample is larger and the point estimate is near-zero. Let me describe Figure 1 more concretely, because it’s a good illustration of how the analysis is framed. On page 5, the figure is titled “Association between age and Comp-P by EVOO consumption groups.” Imagine two clouds of points plotted against age from early forties up to eighty. In the LT4 group, as age increases, the points tend to decline on the y-axis—a visible, downward scatter with perhaps a fitted line sloping negative; in contrast, the MT4 points look more horizontally spread, the line fairly flat. The authors are essentially using this as a visual credibility check on the Spearman correlations in Table 2. The figure’s impression matches the tabulated statistics. What would I suggest as improvements or extensions from a methods standpoint? Several things that the authors themselves acknowledge. First, the cross-sectional design precludes causal inference; longitudinal cohort follow-up or randomized EVOO supplementation within a MedDiet-adherent active population would be the gold standard to test whether increasing EVOO attenuates the slope of physical HRQoL decline. Second, a more granular exposure variable—either continuous tablespoons per day, or better yet, biomarker triangulation such as urinary hydroxytyrosol or plasma oleic acid profiles—would mitigate self-report bias and capture dose–response. Third, multivariable modeling—linear models or generalized additive models—could adjust for sex, BMI, comorbidity count, and perhaps region, and test interaction terms to examine whether the age–Comp-P relationship differs by EVOO group after covariate adjustment. Given the bounded and potentially non-linear nature of SF-36 composites, spline terms could model curvilinear age effects. Fourth, the RAPA-Q mapping to WHO thresholds could be more transparent; alternatively, accelerometry in a subsample would provide an objective check on PA inclusion. Finally, acknowledging and testing for multiple comparisons explicitly would tighten the inferential discipline, though it likely wouldn’t change the key inferences here. To round out the methodology discussion, let’s connect briefly to the consistency checks within the paper. In the Results, the authors note that there were no significant differences between LT4 and MT4 in anthropometry, Comp-P scores, or health conditions at baseline, which reduces concerns about confounding by those measured variables. In the Discussion and Strengths and Limitations, they explicitly call out the key limitations we just reviewed: cross-sectional design; self-report dietary data; uncontrolled variation in total dietary intake beyond the MEDAS threshold; and the restricted, physically active sample limiting generalizability to less active or metabolically compromised older adults. They also recognize the breadth of the bodily pain construct in SF-36 and refrain from overinterpreting mechanistic specificity. From a methodological transparency perspective, these acknowledgments matter. It’s also worth emphasizing the thoughtful choice to restrict on two axes—activity and MedDiet adherence. Many observational nutrition studies attempt to statistically adjust for these in regression models; restriction is a complementary tool that strengthens internal validity by design. The trade-off is representativeness. If your interest is understanding the marginal effect of EVOO intake in the general aging population, you’d want a broader sample and modeling adjustments. But if your aim is to ask, among people who plausibly should be doing well—active, MedDiet-adherent—does EVOO make an additional difference?—this design is fit for purpose. Two last, fine-grained details demonstrate methodological care. First, they used Spain-specific weights to compute SF-36 summary components, which is crucial because the factor structure and item loadings can vary by culture and language; using non-local norms would misestimate Comp-P. Second, they report Cronbach’s alpha in their sample exceeding 0.75, which reassures us that measurement error at the scale level isn’t unusually high in this cohort. Let me close by synthesizing the methodological through-line. This study asks a focused question within a carefully restricted cross-sectional sample, using validated Spanish instruments for physical activity, diet adherence, and quality of life, and a guideline-based dichotomy for EVOO intake. The primary analysis is bivariate, stratified by exposure, with nonparametric tests chosen based on diagnostic checks for distributional assumptions. The descriptive balancing across strata and the consistency of the visual and tabular correlation evidence boost confidence in the internal signal. At the same time, the reliance on self-report, the dichotomization of exposure, the limited covariate adjustment, and the selective sample shape both the precision and the generalizability of inferences. The authors are appropriately cautious: they present their findings as suggestive, not causal, and they call for longitudinal and randomized work to confirm whether higher EVOO consumption truly mitigates age-related decline in self-perceived physical health and pain. If you scan page 4’s Table 2 in your mind one more time, what you see is a methodological fingerprint: clean, prespecified correlations by strata, reporting estimates, confidence intervals, and p-values. And if you look at Figure 1 on page 5, you can almost feel how a single design choice—stratifying at four tablespoons—transforms a general observation about aging and function into a nuanced story about how a key Mediterranean oil might buffer the way we experience our bodies as we grow older. That, in a nutshell, is the value of the methodological architecture in this paper.

Let’s set the stage. This is a cross-sectional, descriptive, and comparative study in nutrition epidemiology, executed in Spain, that asks a very specific question: among physically active middle-aged and older adults who already adhere to a Mediterranean dietary pattern, does higher consumption of extra virgin olive oil appear to attenuate the age-related decline in the physical component of health-related quality of life? The motivation, laid out in the introduction, is that EVOO is a central pillar of the Mediterranean diet and is rich in bioactives—monounsaturated fatty acids like oleic acid, and phenolics like oleocanthal and hydroxytyrosol—that plausibly modulate inflammation, oxidative stress, and pain perception in aging. But most prior work has connected overall Mediterranean pattern adherence with health-related quality of life; very few have isolated EVOO intake specifically. So the authors used design restriction to hold constant two major correlates of quality of life: physical activity and overall Mediterranean dietary adherence, then stratified the sample by EVOO consumption—less than four tablespoons per day versus four or more—and looked at age correlations within those strata.

The study population came from an initial pool of 553 individuals recruited “randomly over a 3‑month period” from different regions of Spain, with voluntary participation following informed consent. The phrase “recruited randomly” isn’t unpacked—it likely indicates random outreach rather than probability sampling from a defined frame—but all the telltale features of a volunteer sample are present: informational briefings, the ethical guardrails of the Declaration of Helsinki, and research ethics approval from the University of Granada. The paper then applied four inclusion criteria to that pool: age 41–80 years; an active lifestyle according to WHO guidelines; self-reported good health with no conditions interfering with activities of daily living; and good adherence to the Mediterranean dietary pattern. After exclusions for unmet criteria or incomplete questionnaires, the final analytic sample was 180 participants. That attrition—from 553 to 180—is substantial, and methodologically it’s doing heavy lifting: it’s a restriction strategy to reduce heterogeneity in physical activity and dietary pattern and eliminate individuals whose disease burden might independently depress quality of life.

Let’s be concrete about how those inclusion criteria were operationalized. Physical activity was assessed with the Spanish version of the Rapid Assessment of Physical Activity (RAPA-Q), a brief seven-item instrument designed for older adults that uses yes/no items to classify activity levels. The authors aligned classification with the 2020 WHO thresholds—active defined as at least 150 minutes per week of moderate intensity or 75 minutes per week of vigorous intensity. While RAPA-Q is indeed validated in Spanish-speaking populations and is intended to provide a pragmatic activity categorization, it’s worth noting as a methodological nuance that RAPA-Q is not a time-quantified log; it relies on categorical endorsement of activity levels. In this paper, participants were classified as active if their RAPA-Q responses implied they met or exceeded the WHO thresholds. The mapping rules aren’t specified here, but the alignment with WHO thresholds is explicit and serves the purpose of inclusion.

Dietary pattern adherence was measured using the 14-item Mediterranean Diet Adherence Screener (MEDAS), itself derived from a longer food frequency questionnaire with a reported correlation of r=0.52, p<0.001 in validation work. MEDAS awards one point for each criterion met across 12 intake frequency items—EVOO, wine, fruits, vegetables, fish, legumes, nuts, meat and derivatives, poultry, butter, pastries, and sweetened beverages—and two items about MedDiet characteristics such as exclusive EVOO use for cooking. Consistent with prior Spanish studies, a score of 10 or higher indicates good adherence to the Mediterranean pattern. Participants who fell below that threshold were excluded. Critically for this study, the EVOO item explicitly queried daily consumption inclusive of all uses—frying, salads, and meals away from home—and defined the tablespoon unit as 13.5 grams. A positive response indicated a daily intake of at least 4 tablespoons, which aligns with approximately 50 mL per day, citing the recommended EVOO intake for MedDiet benefits. This single item then became the stratifying variable: LT4 for those consuming less than 4 tablespoons per day, and MT4 for those consuming 4 or more.

Health-related quality of life, the outcome domain of interest, was assessed with the Spanish version of the SF-36. The instrument comprises eight domains—physical function (10 items), role physical (4 items), bodily pain (2 items), general health (5 items), vitality, social function, emotional role, and mental health—each scored from 0 to 100, higher meaning better. The authors also computed the two standard summary scores, the physical component (Comp-P) and the mental component (Comp-M), using the Spain-specific weighting algorithm. They note that cut-off points for dichotomizing these components into two levels were based on Spanish population median norms by age and gender; in practice, however, their analyses treat Comp-P and its subdomains continuously, using correlations with age. Regarding psychometrics, they report internal consistency in their sample with Cronbach’s alpha exceeding 0.75, and they cite typical Spanish normative alphas for the subscales—physical function around 0.93, role physical 0.95, bodily pain 0.87, general health 0.79—indicating that the measurement noise at the scale level is unlikely to dominate the signal in this population.

Sociodemographic and health status variables—age, sex, weight, height, and self-reported diseases in the past 12 months including cardiovascular disease, hypercholesterolemia, diabetes—were captured with an ad hoc questionnaire. That means anthropometry and comorbidities are self-reported; body mass index is thus computed from self-reported height and weight. The reliance on self-report is consistent across the study: diet, activity, and quality of life are all self-reported constructs, though using validated tools where applicable.

Now, to the statistical analysis plan. Everything was conducted in R. The authors applied the Kolmogorov–Smirnov test with Lilliefors correction to evaluate distributional normality of continuous variables—appropriate when the mean and variance are estimated from the sample—and Levene’s test to assess homoscedasticity across groups. Given violations of parametric assumptions, they used the Mann–Whitney U test for continuous variable comparisons between the EVOO strata, and the Pearson Chi-square test for categorical comparisons. For the primary association of interest—aging and physical health domains—they computed Spearman’s rank-order correlations between age and the SF-36 physical component and its subdomains, separately within the LT4 and MT4 EVOO groups. Internal reliability was again summarized via Cronbach’s alpha. All p-values were two-tailed, with significance set at 0.05. One subtlety here: the paper reports correlation coefficients accompanied by confidence intervals in Table 2, but it does not state how those intervals were computed—Spearman’s rho CIs are often obtained via bootstrapping or Fisher z-transform approximations. Nonetheless, the presence of CIs helps contextualize the precision of the estimates, especially given the unequal group sizes.

Before moving to the figures and tables, let’s pause on the overall design logic. This is classic cross-sectional stratification with design-level restriction. By including only physically active individuals with good Mediterranean diet adherence and no major functional limitations, the authors reduce between-person variability driven by low physical activity, poor diet, or overt disease. In epidemiologic terms, they are attempting to minimize confounding by PA and overall diet by restricting the sample. They then operationalize EVOO dose as a dichotomy at a biologically and culturally meaningful threshold—4 tablespoons per day—and ask: as age increases within each EVOO stratum, does self-perceived physical functioning erode at different rates? That’s not a causal design—temporal ordering for EVOO intake and quality of life is not established—but it is a clean cross-sectional look at effect modification by exposure stratum, with correlations as the summary. The choice of Spearman’s rho is sensible for monotonic but potentially non-linear associations between age and the ordinally bounded SF-36 scales.

Let’s interpret the descriptive table first. Table 1, on page 4, presents anthropometric and health status by sex and by EVOO consumption group. The sex distribution is unbalanced: 131 men (72.78%) and 49 women (27.22%), with that difference of course statistically significant. By EVOO group, the distribution is also skewed: 55 in LT4 (30.56%) and 125 in MT4 (69.44%). That asymmetry has analysis implications—precision is higher in MT4—and it may reflect underlying dietary behaviors in this active, MedDiet-adherent Spanish cohort. Among sexes, men were taller and heavier, with higher BMI, and had higher reported prevalence of diabetes and musculoskeletal disease in the past year; those sex-specific differences reached statistical significance. However, across EVOO strata—LT4 versus MT4—there were no statistically significant differences in anthropometry, in the physical component score (Comp-P), or in the listed health conditions. That balance across EVOO groups on observed covariates is helpful for internal validity; it reduces the likelihood that the EVOO strata are simply proxies for different baseline health. But keep in mind there’s no multivariable adjustment; any residual confounding by unmeasured factors—ses, medication, detailed nutrient intake—remains possible.

Table 2, also on page 4, is the analytic centerpiece for methodology: it lists the bivariate correlations of each SF-36 physical domain with age, stratified by LT4 and MT4. For LT4, physical functioning correlates with age at r = −0.39 with a 95% confidence interval of approximately −0.58 to −0.17, p=0.003. Role physical is essentially null, r = −0.07, p=0.829, and general health is near zero, r = 0.03, p=0.829. Bodily pain shows a moderate negative association, r = −0.33, CI roughly −0.52 to −0.11, p=0.014. Most importantly, the composite physical component, Comp-P, shows r = −0.35, with the CI spanning −0.58 to −0.09, p=0.009. For MT4, physical functioning also declines with age, r = −0.37, p=0.001, which underscores that physically active, well-nourished people still experience age-related reductions in function. But bodily pain is no longer significantly related to age; the point estimate is modestly positive at 0.10, p=0.234. And the composite Comp-P is essentially flat with respect to age: r = −0.07, CI −0.24 to 0.10, p=0.431. The authors visualize the Comp-P correlation in Figure 1 on page 5, a scatterplot of age on the x-axis and Comp-P on the y-axis, with points colored or separated by EVOO group. You can picture two fitted trend lines: in the LT4 group, there’s a downward slope as age increases—indicating lower perceived physical health with advancing age—while in the MT4 group, the slope is near-zero, showing stability of the composite physical quality-of-life score across ages 41 to 80 in this cohort.

Methodologically, what does this tell us? First, the use of within-stratum correlations avoids the ecological fallacy—associations are estimated at the individual level within exposure strata. Second, the negative physical functioning correlation in both groups indicates that even with high PA and MedDiet adherence—and even with higher EVOO—self-reported physical functioning declines with age. EVOO is not a panacea. But the two domains that aggregate to a broader sense of physical health—bodily pain and the composite score—are differentially associated with age by EVOO intake. That supports the prespecified hypothesis that higher EVOO intake could modulate pain perception and the overall physical health construct in aging, at least cross-sectionally.

Turning back to measurements, it’s important to scrutinize the exposure variable. EVOO intake was classified by a single MEDAS item with a daily threshold. It captured total intake across cooking and raw uses and even meals outside the home. The tablespoon was standardized as 13.5 g. A threshold-based dichotomy is pragmatic and maps to dietary guidelines recommending about 50 mL daily intake for the MedDiet’s protective effect. But it also discards granularity—someone at 3.9 tablespoons is grouped with someone at 0.5 tablespoons, and someone at 4 uses sits with someone at 8. If the underlying dose–response is monotone, dichotomization attenuates power and can obscure patterning. On the other hand, because the question is whether those at or above the recommended intake are buffered against age-related decline, a binary classification has clinical interpretability.

The outcome measurement—SF-36—was appropriately processed. Using Spain-specific weights for the Comp-P summary ensures cultural adaptation of item contributions. Reporting internal consistency in the analytic sample is good practice. It’s also notable that the authors did not rely solely on the composite but examined domain-level associations—physical functioning, role physical, bodily pain, general health—so we can see that the primary effect separation appears to reside in pain and in the aggregated composite rather than in functional capacity alone. The bodily pain subscale’s breadth is worth emphasizing: as the discussion notes, it reflects perceived pain interference and intensity but does not distinguish etiologies—musculoskeletal, neuropathic, inflammatory. In methodology terms, this is construct validity that is broad rather than specific; it’s a good global measure of bodily pain’s impact on daily life, but it is not mechanistic.

For covariates, what did they control? Design restriction did most of the work: only physically active, MedDiet-adherent, generally healthy adults were included. This is a strength for internal validity—major determinants of HRQoL like sedentarism and poor diet are effectively held constant—but it also means the sample is a selective slice of the aging population, and it opens the theoretical door to selection bias. If unmeasured factors influence both EVOO intake and the probability of being physically active and MedDiet-adherent, conditioning on those inclusion criteria could in principle induce collider bias. Given the plausibility that socioeconomic status, health literacy, or regional culinary patterns could influence both, one has to be cautious about causal interpretation. The paper does not draw causal conclusions; it’s appropriately framed as association.

Another careful point: While Table 1 suggests balance across EVOO groups for BMI and comorbidities, the sex distribution is skewed in the overall sample, and the analysis does not stratify or adjust by sex when estimating the age–Comp-P associations. If sex modifies the relationship between pain and aging, or if EVOO intake distributions differ by sex within this active cohort, that could confound or modify the observed correlations. The authors do not present sex-stratified correlations or multivariable models that adjust for sex; they rely on ROC-like design restriction and bivariate correlations. Given the large preponderance of men in the sample—about 73%—the precision for women is reduced, and external validity for older women is less certain.

Let’s also talk about the statistical testing choices. The authors used Kolmogorov–Smirnov with Lilliefors correction for normality. Many would favor Shapiro–Wilk for small to moderate samples, but K–S with Lilliefors is fine, and given that assumptions for t-tests were not met, the nonparametric Mann–Whitney U for continuous group comparisons is appropriate. For the correlation work, Spearman’s rho is a robust choice for monotonic associations when variable distributions are non-normal or have bounded ranges. They do not report any adjustment for multiple testing across the several correlations in Table 2; in a strictly confirmatory context that would be a concern. But the correlations are focused on a small number of prespecified domains with clear hypotheses—Comp-P and bodily pain—so the family-wise error risk is modest. Still, p-values cluster around the 0.01–0.02 range for the LT4 correlations, which would likely retain significance under modest corrections, and the null in MT4 for Comp-P is comfortably non-significant. Confidence intervals in Table 2 provide more information than raw p-values and are appreciated.

From a measurement error standpoint, we have classic non-differential misclassification issues. EVOO intake is self-reported, the threshold is coarse, and recall bias or social desirability could push reporting upward among health-conscious, MedDiet-adherent participants. If misclassification is non-differential with respect to the outcome, it generally biases associations toward the null. That means the observed difference in age–Comp-P correlation across EVOO groups might be conservative. However, if higher pain or lower physical functioning changes how people remember or report their oil intake—for example, people in pain may understate or overstate certain dietary intakes—that could complicate interpretation. The paper acknowledges recall bias potential explicitly.

A final methodological note on sample size and power: there’s no mention of a priori sample size calculation or detectable effect size estimates. The LT4 group has n≈55; the MT4 group n≈125. For Spearman’s correlations, power to detect r around 0.3 at alpha 0.05 is good in MT4 and marginal in LT4; nonetheless, they observe r = −0.35 in LT4 for Comp-P with p=0.009, which is stable enough. The null in MT4, r = −0.07, is not a function of low power; the sample is larger and the point estimate is near-zero.

Let me describe Figure 1 more concretely, because it’s a good illustration of how the analysis is framed. On page 5, the figure is titled “Association between age and Comp-P by EVOO consumption groups.” Imagine two clouds of points plotted against age from early forties up to eighty. In the LT4 group, as age increases, the points tend to decline on the y-axis—a visible, downward scatter with perhaps a fitted line sloping negative; in contrast, the MT4 points look more horizontally spread, the line fairly flat. The authors are essentially using this as a visual credibility check on the Spearman correlations in Table 2. The figure’s impression matches the tabulated statistics.

What would I suggest as improvements or extensions from a methods standpoint? Several things that the authors themselves acknowledge. First, the cross-sectional design precludes causal inference; longitudinal cohort follow-up or randomized EVOO supplementation within a MedDiet-adherent active population would be the gold standard to test whether increasing EVOO attenuates the slope of physical HRQoL decline. Second, a more granular exposure variable—either continuous tablespoons per day, or better yet, biomarker triangulation such as urinary hydroxytyrosol or plasma oleic acid profiles—would mitigate self-report bias and capture dose–response. Third, multivariable modeling—linear models or generalized additive models—could adjust for sex, BMI, comorbidity count, and perhaps region, and test interaction terms to examine whether the age–Comp-P relationship differs by EVOO group after covariate adjustment. Given the bounded and potentially non-linear nature of SF-36 composites, spline terms could model curvilinear age effects. Fourth, the RAPA-Q mapping to WHO thresholds could be more transparent; alternatively, accelerometry in a subsample would provide an objective check on PA inclusion. Finally, acknowledging and testing for multiple comparisons explicitly would tighten the inferential discipline, though it likely wouldn’t change the key inferences here.

To round out the methodology discussion, let’s connect briefly to the consistency checks within the paper. In the Results, the authors note that there were no significant differences between LT4 and MT4 in anthropometry, Comp-P scores, or health conditions at baseline, which reduces concerns about confounding by those measured variables. In the Discussion and Strengths and Limitations, they explicitly call out the key limitations we just reviewed: cross-sectional design; self-report dietary data; uncontrolled variation in total dietary intake beyond the MEDAS threshold; and the restricted, physically active sample limiting generalizability to less active or metabolically compromised older adults. They also recognize the breadth of the bodily pain construct in SF-36 and refrain from overinterpreting mechanistic specificity. From a methodological transparency perspective, these acknowledgments matter.

It’s also worth emphasizing the thoughtful choice to restrict on two axes—activity and MedDiet adherence. Many observational nutrition studies attempt to statistically adjust for these in regression models; restriction is a complementary tool that strengthens internal validity by design. The trade-off is representativeness. If your interest is understanding the marginal effect of EVOO intake in the general aging population, you’d want a broader sample and modeling adjustments. But if your aim is to ask, among people who plausibly should be doing well—active, MedDiet-adherent—does EVOO make an additional difference?—this design is fit for purpose.

Two last, fine-grained details demonstrate methodological care. First, they used Spain-specific weights to compute SF-36 summary components, which is crucial because the factor structure and item loadings can vary by culture and language; using non-local norms would misestimate Comp-P. Second, they report Cronbach’s alpha in their sample exceeding 0.75, which reassures us that measurement error at the scale level isn’t unusually high in this cohort.

Let me close by synthesizing the methodological through-line. This study asks a focused question within a carefully restricted cross-sectional sample, using validated Spanish instruments for physical activity, diet adherence, and quality of life, and a guideline-based dichotomy for EVOO intake. The primary analysis is bivariate, stratified by exposure, with nonparametric tests chosen based on diagnostic checks for distributional assumptions. The descriptive balancing across strata and the consistency of the visual and tabular correlation evidence boost confidence in the internal signal. At the same time, the reliance on self-report, the dichotomization of exposure, the limited covariate adjustment, and the selective sample shape both the precision and the generalizability of inferences. The authors are appropriately cautious: they present their findings as suggestive, not causal, and they call for longitudinal and randomized work to confirm whether higher EVOO consumption truly mitigates age-related decline in self-perceived physical health and pain.

If you scan page 4’s Table 2 in your mind one more time, what you see is a methodological fingerprint: clean, prespecified correlations by strata, reporting estimates, confidence intervals, and p-values. And if you look at Figure 1 on page 5, you can almost feel how a single design choice—stratifying at four tablespoons—transforms a general observation about aging and function into a nuanced story about how a key Mediterranean oil might buffer the way we experience our bodies as we grow older. That, in a nutshell, is the value of the methodological architecture in this paper.