Low Fasting Insulin Does Not Exclude Elevated Apolipoprotein B: A Real-World Primary Care Study
Low Fasting Insulin Does Not Exclude Elevated Apolipoprotein B: A Real-World Primary Care Study
Stefan Hartmann, PA-C, Muhammed Alo DO, Salaheldin Halasa MD
¹ Iron Direct Primary Care, Melbourne, Florida, USA
Download PDF to view images and the paper.
Watch lecture presented to the Osteopathic Society of Integrative Medicine
Abstract
Background
Insulin resistance and elevated apolipoprotein B (ApoB) frequently coexist in adverse cardiometabolic phenotypes. However, biologic association does not establish that fasting insulin and ApoB provide interchangeable clinical information. Whether individuals with low fasting insulin reliably demonstrate low ApoB in routine primary care remains uncertain.
Objective
To characterize the relationship and discordance between fasting insulin and ApoB among adults undergoing same-day testing in a real-world primary care population, with secondary analyses incorporating hemoglobin A1c and exploratory analyses of female age, LC-MS estradiol, and ApoB.
Methods
We conducted a retrospective observational analysis of routinely collected laboratory data from a community-based primary care practice. The primary cohort included adults aged ≥18 years with numeric ApoB and explicitly fasting insulin measurements obtained on the exact same collection date. Each patient contributed the first chronologically qualifying paired measurement to the primary analysis. The primary outcome was the Spearman correlation between fasting insulin and ApoB. Secondary analyses assessed ApoB prevalence across fasting-insulin quartiles and multivariable associations adjusted for age and sex. Hemoglobin A1c was incorporated as an additional glycemic covariate in participants with same-day measurements. An exploratory female analysis evaluated exact same-day LC-MS estradiol and ApoB measurements.
Results
The primary cohort included 365 adults (62.7% male; mean age 48.0±13.0 years). Mean ApoB was 100.9±28.6 mg/dL, and median fasting insulin was 7.7 [IQR 5.1–12.6]. Fasting insulin and ApoB were positively but weakly correlated (Spearman ρ=0.156; p=0.0029). After adjustment for age and sex, each doubling of fasting insulin was associated with 4.19 mg/dL higher ApoB (95% CI 1.46–6.93; p=0.0028).
Despite this association, substantial discordance was observed. Among participants in the lowest fasting-insulin quartile, 59.8% had ApoB ≥90 mg/dL and 44.6% had ApoB ≥100 mg/dL.
Hemoglobin A1c was available on the same date in 347 participants (95.1%). After additional adjustment for A1c, each doubling of fasting insulin remained associated with 3.47 mg/dL higher ApoB (95% CI 0.66–6.28; p=0.015). Among 79 adults with both A1c <5.7% and fasting insulin in the lowest quartile, 60.8% had ApoB ≥90 mg/dL and 46.8% had ApoB ≥100 mg/dL.
In an exploratory analysis of 133 adult women with exact same-day ApoB and LC-MS estradiol measurements, estradiol was modestly inversely associated with ApoB in unadjusted analysis (Spearman ρ=−0.182; p=0.036), but this association was attenuated after adjustment for age.
Conclusions
Fasting insulin was positively but weakly associated with ApoB in this real-world primary care population. Elevated ApoB remained common among patients with low fasting insulin, including patients who simultaneously had A1c <5.7%. These findings suggest that fasting insulin, chronic glycemia, and ApoB provide overlapping but substantially nonredundant information. Exploratory female analyses demonstrated opposing age-related trajectories of estradiol and ApoB, although a single contemporaneous estradiol measurement did not independently explain ApoB after adjustment for age.
Keywords: apolipoprotein B; ApoB; fasting insulin; hemoglobin A1c; insulin resistance; estradiol; cardiometabolic risk; primary care; discordance
Introduction
Apolipoprotein B is present on circulating atherogenic lipoprotein particles and therefore provides information regarding the number of ApoB-containing particles in circulation. Elevated ApoB reflects increased atherogenic particle burden and represents a clinically relevant dimension of cardiovascular risk.
Insulin resistance is also associated with alterations in lipoprotein metabolism. Human metabolic and epidemiologic studies have demonstrated relationships among insulin resistance, hyperinsulinemia, triglyceride-rich lipoprotein production, and ApoB-containing particles. Consequently, elevated fasting insulin and elevated ApoB frequently coexist.
However, association does not imply redundancy. A patient may demonstrate favorable glycemic or insulin-related biomarkers while simultaneously carrying a substantial burden of ApoB-containing particles. From a practical primary-care perspective, an important question is therefore whether fasting insulin can meaningfully predict ApoB status at the level of the individual patient.
Hemoglobin A1c provides an additional opportunity to evaluate this question. If patients with low fasting insulin and relatively low chronic glycemia nevertheless frequently demonstrate elevated ApoB, this would further support the concept that favorable glycemic biomarkers cannot substitute for direct ApoB measurement.
Sex and aging may also contribute to ApoB heterogeneity. In women, reproductive aging is accompanied by substantial changes in circulating estradiol. The availability of LC-MS estradiol measurements within the same clinical population allowed an exploratory assessment of female age, estradiol, and ApoB.
The primary objective of this study was therefore to characterize the relationship and discordance between fasting insulin and ApoB in adults undergoing routine same-day testing. Secondary objectives were to determine whether the insulin–ApoB association persisted after accounting for hemoglobin A1c and to quantify ApoB elevation among patients with both low fasting insulin and A1c <5.7%. An exploratory female analysis evaluated age-related patterns in LC-MS estradiol and ApoB.
Methods
Study Design and Setting
This retrospective observational study used routinely collected laboratory data from a community-based primary care practice in Florida.
The available LabCorp laboratory exports contained sex, age at collection, city, fasting status, numeric laboratory result, test name, patient lab identifiers, and collection date.
ApoB measurements were available from March 2022 through August 2026. Insulin measurements extended from December 2021 through August 2026.
Patient identifiers were used solely to link laboratory measurements belonging to the same individual. No patient names or other direct identifiers are reported in study results, tables, figures, or manuscript outputs.
Data Cleaning
Multiple CSV exports were initially provided for ApoB and insulin. Numeric laboratory results and patient ages were converted to numeric data types, and collection dates were converted to standardized date values.
Repeat testing on different collection dates was considered legitimate longitudinal clinical testing and was preserved.
When multiple qualifying measurements occurred for the same patient on the same date, patient-day values were consolidated before matching. Such occurrences were uncommon and did not materially affect the results.
Insulin measurements were considered fasting only when the laboratory fasting-status field explicitly contained Y. Measurements designated nonfasting, unknown, or with missing fasting status were excluded from primary fasting-insulin analyses.
Primary Study Population
Patients were eligible for the primary analysis if they:
were aged ≥18 years at laboratory collection;
had a numeric ApoB measurement;
had a numeric insulin measurement explicitly designated fasting;
had ApoB and fasting insulin obtained on the exact same collection date; and
could be matched using the available patient identifier fields.
For the primary analysis, each patient contributed only the first chronologically qualifying exact same-day ApoB–fasting-insulin pair.
This produced an independent cross-sectional cohort of 365 adults.
A secondary sensitivity dataset retained all qualifying patient-days, including repeat measurements on different dates.
Laboratory Variables
ApoB was analyzed in mg/dL.
Fasting insulin analyzed in µIU/mL.
Because fasting insulin was right-skewed, it was analyzed both on its original scale and after base-2 logarithmic transformation. A one-unit increase in log₂ fasting insulin therefore represents a doubling of fasting insulin concentration.
ApoB elevation was evaluated using prespecified descriptive thresholds of:
ApoB ≥90 mg/dL
ApoB ≥100 mg/dL
No clinical definition of elevated fasting insulin was imposed. Instead, fasting insulin was divided into empirical quartiles derived from the primary study population.
Hemoglobin A1c Secondary Analysis
Hemoglobin A1c measurements were matched to the primary ApoB–fasting-insulin cohort using patient identifier and exact collection date.
Only laboratory records specifically identified as Hemoglobin A1c and containing numeric results compatible with an A1c percentage were retained. The export contained some historical records in which additional derived glycemic values appeared under related laboratory output; these non-A1c values were excluded.
When more than one plausible Hemoglobin A1c measurement was present for the same patient on the same collection date, the values were consolidated at the patient-day level.
Same-day A1c was available for 347 of 365 primary participants (95.1%).
A1c was analyzed both continuously and using A1c <5.7% as a descriptive subgroup definition.
A prespecified sensitivity model evaluated:
ApoB = log₂(fasting insulin) + age + sex + A1c
A further discordance analysis examined participants who had both:
fasting insulin within the lowest study quartile, and
A1c <5.7%.
Exploratory Female LC-MS Estradiol Analysis
An exploratory secondary analysis evaluated relationships among female age, serum estradiol, and ApoB.
Estradiol measurements were restricted to laboratory records identified as:
Estradiol, LCMS, Endo Sci
Adult women aged ≥18 years with numeric LC-MS estradiol measurements were included in descriptive age analyses.
For direct estradiol–ApoB analyses, measurements were matched by patient identifier and exact collection date.
To maintain independence, each woman contributed only her first chronologically qualifying exact same-day ApoB–estradiol pair.
Estradiol was strongly right-skewed and was therefore assessed using Spearman rank correlation and log₂ transformation for regression analyses.
A multivariable model assessed whether estradiol remained associated with ApoB after accounting for chronological age.
Menopausal status, menstrual-cycle phase, hormone-replacement therapy exposure, dose, formulation, route of administration, and treatment initiation dates were not available in the laboratory export. Therefore, no patient was classified as premenopausal, postmenopausal, treated, or untreated based solely on age or serum estradiol concentration.
Statistical Analysis
Continuous variables were summarized as mean±standard deviation and median [interquartile range]. Categorical variables were reported as number and percentage.
The prespecified primary analysis was the Spearman rank correlation between fasting insulin and ApoB.
A 95% confidence interval for Spearman ρ was estimated by bootstrap resampling.
Secondary analyses included Pearson correlation between log₂-transformed fasting insulin and ApoB.
ApoB concentration and prevalence of ApoB ≥90 and ≥100 mg/dL were examined across empirical fasting-insulin quartiles.
Multivariable linear regression evaluated:
ApoB = β₀ + β₁(log₂ fasting insulin) + β₂(age) + β₃(sex)
HC3 heteroscedasticity-robust standard errors were used.
Separate logistic regression models evaluated ApoB ≥90 and ApoB ≥100 mg/dL.
The A1c sensitivity model additionally included hemoglobin A1c as a continuous covariate.
Age-stratified insulin–ApoB analyses were performed for:
18–39 years
40–59 years
≥60 years
These subgroup analyses were considered exploratory.
For repeated-measures sensitivity analysis, all qualifying patient-days were analyzed using patient-clustered standard errors.
All tests were two-sided with α=0.05.
Analyses were performed using Python, pandas, SciPy, and statsmodels.
Artificial Intelligence–Assisted Data Analysis and Manuscript Preparation
OpenAI ChatGPT (GPT-5.6 Sol; OpenAI, San Francisco, California, USA) was used as an artificial intelligence–assisted analytical and writing tool during this study.
The AI system provided substantial assistance with inspection and cleaning of laboratory CSV datasets; identification of duplicate files and observations; construction of analytic cohorts; patient- and date-level matching of ApoB, fasting insulin, hemoglobin A1c, and LC-MS estradiol measurements; selection and execution of statistical analyses; sensitivity analyses; organization and interpretation of statistical outputs; preparation of figures; organization of relevant literature; and preparation and revision of manuscript text.
The research question, clinical hypotheses, source data, selection of clinically relevant variables, interpretation of findings in the clinical context, and decisions regarding the final manuscript and conclusions were directed by the human authors.
AI-generated statistical outputs, interpretations, references, figures, and manuscript text were subject to human review. The human authors retain responsibility for verification of the final statistical results, accuracy of citations, integrity of the analysis, originality of the manuscript, and all conclusions presented.
The AI system was not considered an author and bears no responsibility for the final manuscript.
Ethics and Privacy
This study represents a retrospective analysis of routinely collected clinical laboratory data.
[INSERT FORMAL IRB / EXEMPT / QUALITY-IMPROVEMENT DETERMINATION, INSTITUTION, PROTOCOL NUMBER IF APPLICABLE, AND WAIVER-OF-CONSENT LANGUAGE.]
Direct patient identifiers contained in the original clinical exports were used only to accurately link serial and same-day laboratory measurements. Identifiers were not included in aggregate statistical outputs, manuscript tables, figures, or reported results.
Results
Primary Cohort
The primary cohort included 365 unique adults, comprising 229 men (62.7%) and 136 women (37.3%).
Mean age was 48.0±13.0 years, with an age range of 18–85 years.
Mean ApoB was:
100.9±28.6 mg/dL
Median ApoB was:
96 [IQR 81–117] mg/dL
Mean fasting insulin was:
10.0±7.5
Median fasting insulin was:
7.7 [IQR 5.1–12.6]
Overall:
227 participants (62.2%) had ApoB ≥90 mg/dL.
175 participants (47.9%) had ApoB ≥100 mg/dL.
Table 1. Characteristics of the primary same-day paired cohort
Characteristic | Overall | Women | Men |
n | 365 | 136 | 229 |
Age, years | 48.0±13.0 | 48.3±12.3 | 47.8±13.4 |
Age 18–39 | 98 (26.8%) | 34 (25.0%) | 64 (27.9%) |
Age 40–59 | 184 (50.4%) | 69 (50.7%) | 115 (50.2%) |
Age ≥60 | 83 (22.7%) | 33 (24.3%) | 50 (21.8%) |
ApoB, mg/dL | 100.9±28.6 | 95.8±25.8 | 104.0±29.8 |
ApoB, median [IQR] | 96 [81–117] | 92 [78–109] | 102 [82–121] |
Fasting insulin | 10.0±7.5 | 9.5±7.4 | 10.3±7.5 |
Fasting insulin, median [IQR] | 7.7 [5.1–12.6] | 7.3 [4.7–11.4] | 7.7 [5.5–13.1] |
ApoB ≥90 mg/dL | 227 (62.2%) | 76 (55.9%) | 151 (65.9%) |
ApoB ≥100 mg/dL | 175 (47.9%) | 52 (38.2%) | 123 (53.7%) |
Primary Relationship Between Fasting Insulin and ApoB
Fasting insulin demonstrated a statistically significant but weak positive relationship with ApoB:
Spearman ρ = 0.156p = 0.0029
The 95% bootstrap confidence interval for Spearman ρ was approximately:
0.056–0.253
Using log₂-transformed fasting insulin, Pearson correlation similarly demonstrated only a modest association:
r = 0.136; p = 0.0095
Thus, fasting insulin was associated with ApoB at the population level but accounted for relatively little interindividual variation in ApoB.
Multivariable Analysis
After adjustment for age and sex, each doubling of fasting insulin was associated with:
+4.19 mg/dL ApoB
95% CI:
+1.46 to +6.93 mg/dL
p = 0.0028
Each additional decade of age was associated with approximately:
+5.64 mg/dL ApoB
Male sex was independently associated with approximately:
+8.0 mg/dL ApoB
The overall model explained approximately:
10.1% of ApoB variance (R²=0.101).
Table 2. Multivariable linear model for ApoB
Predictor | Adjusted ApoB difference | 95% CI | p |
Doubling fasting insulin | +4.19 mg/dL | +1.46 to +6.93 | 0.0028 |
Age, per 10 years | +5.64 mg/dL | +3.28 to +7.99 | <0.00001 |
Male sex | +8.00 mg/dL | +2.46 to +13.54 | 0.0048 |
Discordance Across Fasting-Insulin Quartiles
Fasting insulin was divided into four empirical quartiles.
Table 3. ApoB according to fasting-insulin quartile
Quartile | Insulin range | n | Median insulin | Mean ApoB | Median ApoB | ApoB ≥90 | ApoB ≥100 |
Q1 | 1.3–5.1 | 92 | 3.9 | 97.6 | 95 | 59.8% | 44.6% |
Q2 | 5.2–7.7 | 93 | 6.3 | 98.1 | 94 | 54.8% | 47.3% |
Q3 | 7.8–12.6 | 90 | 9.9 | 99.1 | 93 | 55.6% | 38.9% |
Q4 | 12.7–48.4 | 90 | 17.7 | 109.0 | 106 | 78.9% | 61.1% |
Higher fasting insulin identified a population with greater prevalence of elevated ApoB. However, substantial discordance was present.
Even within the lowest fasting-insulin quartile, approximately:
6 in 10 patients had ApoB ≥90 mg/dL
nearly 1 in 2 had ApoB ≥100 mg/dL
Compared with Q1, participants in Q4 had:
2.80-fold higher adjusted odds of ApoB ≥90 mg/dL95% CI 1.42–5.50p=0.0028
For ApoB ≥100 mg/dL, the adjusted Q4-versus-Q1 odds ratio was:
2.0795% CI 1.12–3.81p=0.0197
These results demonstrate both a population-level association and substantial patient-level discordance.
Hemoglobin A1c Secondary Analysis
Same-day hemoglobin A1c was available for 347 of 365 participants (95.1%).
Mean A1c was:
5.45±0.57%
Median A1c was:
5.4 [IQR 5.2–5.6]%
The observed range was:
4.5–11.0%
A1c demonstrated a modest positive association with fasting insulin:
Spearman ρ=0.207; p=0.00010
A1c was also modestly associated with ApoB:
Spearman ρ=0.192; p=0.00032
Within this 347-patient subset, adjustment for age and sex yielded an ApoB increase of:
+4.01 mg/dL per doubling of fasting insulin95% CI 1.28–6.74p=0.0040
After A1c was added to the model, the fasting-insulin association remained statistically significant:
+3.47 mg/dL ApoB per doubling of fasting insulin95% CI 0.66–6.28p=0.015
Thus, adjustment for chronic glycemia attenuated the insulin coefficient modestly but did not eliminate the association.
Low Insulin Plus A1c <5.7%
A total of 281 participants in the matched A1c cohort had A1c <5.7%.
Among the 79 participants who simultaneously had:
A1c <5.7%, and
fasting insulin in the lowest study quartile (≤5.1),
mean ApoB was:
97.5 mg/dL
and median ApoB was:
95 mg/dL
Despite the combination of low fasting insulin and A1c <5.7%:
60.8% had ApoB ≥90 mg/dL
46.8% had ApoB ≥100 mg/dL
Table 4. Glycemic-context sensitivity analysis
Analysis | Result |
Same-day ApoB + insulin + A1c cohort | n=347 |
Mean A1c | 5.45±0.57% |
A1c–insulin Spearman ρ | 0.207; p=0.00010 |
A1c–ApoB Spearman ρ | 0.192; p=0.00032 |
Insulin doubling → ApoB, adjusted for age/sex | +4.01 mg/dL; p=0.0040 |
Insulin doubling → ApoB, additionally adjusted for A1c | +3.47 mg/dL; p=0.015 |
A1c <5.7% + insulin Q1 | n=79 |
ApoB ≥90 in low-A1c/low-insulin subgroup | 60.8% |
ApoB ≥100 in low-A1c/low-insulin subgroup | 46.8% |
The persistence of elevated ApoB among participants with both low fasting insulin and A1c <5.7% further demonstrates that favorable glycemic biomarkers do not reliably identify individuals with low ApoB.
Exploratory Age-Stratified Insulin–ApoB Analysis
The relationship between fasting insulin and ApoB varied according to age.
Age | n | Spearman ρ | p |
18–39 years | 98 | 0.349 | 0.00042 |
40–59 years | 184 | 0.183 | 0.0127 |
≥60 years | 83 | −0.045 | 0.685 |
The association was strongest among younger adults and was essentially absent among participants aged ≥60 years.
A formal interaction between age group and log₂ fasting insulin was statistically significant:
p=0.016
Because these analyses were exploratory, this finding should be interpreted as hypothesis-generating.
Exploratory Female Age, LC-MS Estradiol, and ApoB Analysis
The LC-MS estradiol dataset contained 741 adult female measurements from 445 unique women.
Using one chronologically first measurement per woman, estradiol demonstrated a strong inverse relationship with age:
Spearman ρ=−0.577p<10⁻⁴⁰
Median LC-MS estradiol declined markedly across age:
Female age | n | Median LC-MS estradiol |
18–29 | 54 | 104 |
30–39 | 107 | 111 |
40–49 | 114 | 85 |
50–59 | 102 | 8.95 |
≥60 | 68 | 5.5 |
Female ApoB demonstrated the opposite age-related pattern.
Using the first available adult female ApoB measurement per patient:
Female age | n | Mean ApoB |
18–29 | 12 | 73.1 mg/dL |
30–39 | 28 | 91.1 mg/dL |
40–49 | 47 | 92.4 mg/dL |
50–59 | 40 | 99.3 mg/dL |
≥60 | 38 | 109.5 mg/dL |
A total of 133 unique adult women had qualifying exact same-day ApoB and LC-MS estradiol measurements.
Within this paired cohort, higher estradiol was modestly associated with lower ApoB in unadjusted analysis:
Spearman ρ=−0.182p=0.036
However, the association did not persist after adjustment for age.
In an age-adjusted model, each doubling of estradiol was associated with approximately:
+1.08 mg/dL ApoB
95% CI:
−1.50 to +3.66 mg/dL
p=0.41
Age remained strongly associated with ApoB.
These findings indicate that the apparent inverse estradiol–ApoB relationship was largely attributable to their opposing associations with female age rather than an independent association between a single serum estradiol measurement and ApoB.
Repeated-Measures Sensitivity Analysis
When repeat testing on different dates was retained, 442 qualifying adult patient-days were available.
Using patient-clustered standard errors, each doubling of fasting insulin was associated with:
+4.86 mg/dL ApoB
95% CI:
+1.79 to +7.92
p=0.0019
The close agreement between this estimate and the primary independent-patient analysis supports the robustness of the primary result.
Discussion
Principal Findings
In this real-world primary care population, fasting insulin and ApoB were significantly associated, but the relationship was modest.
The most important finding was therefore not the absence of an insulin–ApoB relationship. Rather, it was the presence of substantial discordance between the two biomarkers.
Participants with the highest fasting insulin had substantially greater prevalence of elevated ApoB. Nevertheless, 59.8% of adults in the lowest fasting-insulin quartile still had ApoB ≥90 mg/dL, and 44.6% had ApoB ≥100 mg/dL.
These findings distinguish association from clinical interchangeability.
Fasting insulin contains information related to ApoB burden, but it does not contain enough information to reliably infer an individual patient's ApoB concentration.
Glycemic Context Strengthens the Discordance Finding
The A1c analysis provides an important extension of the primary finding.
Fasting insulin remained independently associated with ApoB after adjustment for A1c, suggesting that the observed relationship was not merely a reflection of chronic glycemia.
More importantly, substantial ApoB elevation persisted among individuals with both low fasting insulin and A1c <5.7%.
Among these patients, approximately 61% had ApoB ≥90 mg/dL, while nearly 47% had ApoB ≥100 mg/dL.
Thus, the combination of relatively low chronic glycemia and low fasting insulin did not reliably identify patients with low ApoB.
This finding is clinically relevant because favorable glycemic markers may create an impression of broadly favorable cardiometabolic status. The present data demonstrate that such an inference cannot reliably be extended to ApoB-containing lipoprotein burden.
Fasting Insulin and ApoB Provide Nonredundant Information
The modest correlation observed between fasting insulin and ApoB is biologically plausible.
Insulin signaling, hepatic lipid metabolism, adiposity, and triglyceride-rich lipoprotein production are interconnected. Accordingly, a population with higher insulin concentrations would reasonably be expected to demonstrate greater ApoB burden on average.
However, ApoB is also affected by numerous determinants that are incompletely represented by fasting insulin.
The overall multivariable model containing age, sex, and fasting insulin explained only approximately 10% of ApoB variance.
Therefore, substantial ApoB heterogeneity remained unexplained even after accounting for these readily available clinical characteristics.
The findings should not be interpreted as evidence that fasting insulin or insulin resistance is unimportant. Rather, they suggest that fasting insulin and ApoB characterize overlapping but distinct aspects of cardiometabolic physiology.
Age-Related Differences
The exploratory age analysis revealed a potentially important pattern.
Fasting insulin and ApoB demonstrated their strongest association among adults aged 18–39 years, a weaker relationship among adults aged 40–59 years, and essentially no correlation among adults aged ≥60 years.
If replicated, this may suggest that ApoB concentration becomes increasingly influenced by determinants other than fasting insulin as patients age.
Potential contributors could include genetic factors, age-related metabolic changes, medication exposure, hormonal changes, or other unmeasured clinical variables.
Because these analyses were exploratory and involved subgroup comparisons, they require confirmation in larger populations.
Female Age, Estradiol, and ApoB
The exploratory female analysis demonstrated strikingly opposing age-related patterns.
LC-MS estradiol declined markedly with advancing female age, whereas ApoB progressively increased.
Among women with exact same-day measurements, higher estradiol was modestly associated with lower ApoB in unadjusted analysis. However, this relationship disappeared after chronological age was incorporated into the model.
This finding is important because it prevents overinterpretation of the cross-sectional estradiol association.
A single serum estradiol concentration did not independently account for the age-related ApoB increase observed among women.
Female reproductive aging involves multiple physiologic changes that cannot be represented by a single estradiol measurement. In addition, estradiol can exhibit substantial intraindividual variability and may be affected by exogenous hormone administration.
The clinical practice represented in this study commonly uses estradiol therapy in postmenopausal women. Such treatment could plausibly alter the expected relationship between chronological age and circulating estradiol. However, medication exposure, treatment initiation dates, doses, formulations, and routes were not available in the laboratory dataset.
Consequently, no causal conclusion regarding estradiol therapy and ApoB can be drawn from the present analysis.
The opposing age trajectories of estradiol and ApoB are compatible with a potential influence of reproductive aging on ApoB-containing lipoprotein metabolism. However, PCSK9 activity or concentration was not measured, and the present study cannot determine whether PCSK9 or another pathway mediates this association.
Clinical Implications
The principal clinical implication of this study is non-substitution.
Low fasting insulin should not be interpreted as evidence that ApoB is low.
Similarly, the presence of A1c <5.7% in a patient with low fasting insulin does not reliably identify low ApoB.
Conversely, patients with high fasting insulin were substantially more likely to demonstrate elevated ApoB, supporting the concept that adverse metabolic phenotypes often cluster.
These findings suggest that fasting insulin, hemoglobin A1c, and ApoB should be interpreted according to the distinct physiologic information each provides rather than assuming that one can reliably substitute for another.
Strengths
Several features strengthen this analysis.
First, the primary comparison used exact same-day laboratory measurements, eliminating substantial temporal mismatch between biomarkers.
Second, fasting status was explicitly documented for included insulin measurements.
Third, the primary cohort used only one qualifying observation per patient, preventing frequently tested individuals from disproportionately influencing inferential statistics.
Fourth, 95% of the primary cohort also had same-day A1c, allowing unusually complete glycemic-context sensitivity analysis.
Fifth, the principal insulin–ApoB finding remained robust when repeat patient-days were incorporated with patient-clustered standard errors.
Sixth, the availability of LC-MS estradiol allowed a separate exploratory analysis of female aging and hormone-related patterns using measurements collected within the same practice environment.
Finally, these data arose from routine clinical practice rather than an experimental recruitment protocol, thereby characterizing biomarker relationships encountered during real-world primary care.
Limitations
This study has several important limitations.
The retrospective design prevents causal inference.
Testing was performed according to clinical practice rather than population-based random sampling. Patients undergoing ApoB, insulin, A1c, or estradiol testing may therefore differ systematically from the general community population.
Important clinical covariates were unavailable, including:
body mass index;
waist circumference;
fasting glucose;
triglycerides;
diabetes diagnosis;
lipid-lowering medication use;
statin therapy;
ezetimibe therapy;
PCSK9-directed therapy;
diet;
alcohol exposure;
smoking;
physical activity;
renal function; and
other relevant medications.
Medication exposure is particularly important because lipid-lowering therapy could lower ApoB independently of fasting insulin and thereby contribute to apparent biomarker discordance.
Fasting insulin is not a definitive measurement of insulin resistance. HOMA-IR could not be systematically calculated because matched fasting glucose was not included in the present analysis, and clamp-derived insulin sensitivity was not available.
Hemoglobin A1c is also influenced by factors unrelated to glycemia and should not be interpreted as a perfect measure of average glucose exposure.
The female estradiol analysis lacked menstrual-cycle information, menopausal status, hysterectomy/oophorectomy history, hormone-therapy exposure, treatment dose, formulation, route, and initiation date. Accordingly, age served only as an imperfect proxy for reproductive aging.
PCSK9 was not measured, preventing direct evaluation of proposed mechanistic pathways linking reproductive aging, estradiol, and ApoB.
The analysis was conducted within a single primary care practice with a moderate sample size and may not generalize to other geographic, demographic, or clinical populations.
Finally, several secondary and exploratory subgroup analyses were performed and should therefore be considered hypothesis-generating pending external replication.
Conclusion
In adults undergoing exact same-day laboratory testing in routine primary care, fasting insulin was positively but weakly associated with ApoB.
Higher fasting insulin identified a population with greater prevalence of elevated ApoB; however, substantial discordance remained.
Nearly 60% of adults in the lowest fasting-insulin quartile had ApoB ≥90 mg/dL, and nearly 45% had ApoB ≥100 mg/dL.
This discordance persisted when chronic glycemia was considered. Among adults with both A1c <5.7% and fasting insulin in the lowest quartile, approximately 61% had ApoB ≥90 mg/dL and 47% had ApoB ≥100 mg/dL.
These findings suggest that fasting insulin, hemoglobin A1c, and ApoB provide overlapping but substantially nonredundant information.
Exploratory female analyses additionally demonstrated opposing age-related trajectories of LC-MS estradiol and ApoB, although a single contemporaneous estradiol concentration did not independently predict ApoB after age adjustment.
Direct measurement of ApoB therefore provides information that cannot be reliably inferred from favorable fasting-insulin or glycemic biomarkers alone.
Prospective studies incorporating adiposity, glucose-insulin dynamics, medication exposure, hormone therapy, genetic factors, and cardiovascular outcomes are warranted to further characterize the clinical implications of ApoB discordance.
Figures
Figure 1. Fasting insulin versus ApoB
Figure 2. ApoB Prevalence Across Insulin Quartiles
Figure 3. Prevalence of elevated ApoB across fasting-insulin quartiles
Figure 4. Age-related distributions of LC-MS estradiol and ApoB among women
Author Contributions
Stefan Hartmann, PA-C: Conceptualization; clinical hypothesis generation; data acquisition; data curation; study design; methodology; clinical interpretation; project administration; manuscript preparation; writing—original draft; writing—review and editing; final manuscript responsibility.
Muhammed Alo DO : Cardiology perspective; clinical interpretation; scientific insight; critical review and revision of the manuscript.
Salaheldin Halasa MD: Scientific insight and challenged me to perform this analysis.
All listed human authors should review and approve the final submitted manuscript and agree to be accountable for their respective contributions.
Artificial Intelligence Disclosure
OpenAI ChatGPT (GPT-5.6 Sol) provided substantial assistance with data cleaning, cohort construction, statistical analysis, sensitivity analyses, figure development, organization of results, literature organization, and manuscript drafting and revision.
The AI system was not considered an author. The human authors directed the study questions and clinical hypotheses, retained access to and responsibility for the source data, reviewed the analyses and manuscript content, and assume responsibility for the accuracy, integrity, originality, and conclusions of the final publication.
Acknowledgments
The authors acknowledge the clinical staff and patients whose routine care generated the laboratory data used in this retrospective analysis.
References
Marston NA, Giugliano RP, Melloni GEM, et al. Association of Apolipoprotein B-Containing Lipoproteins and Risk of Myocardial Infarction in Individuals With and Without Atherosclerosis: Distinguishing Between Particle Concentration, Type, and Content. JAMA Cardiol. 2022;7(3):250-256. doi:10.1001/jamacardio.2021.5083.
Lamarche B, Tchernof A, Mauriège P, et al. Fasting Insulin and Apolipoprotein B Levels and Low-Density Lipoprotein Particle Size as Risk Factors for Ischemic Heart Disease. JAMA. 1998;279(24):1955-1961. doi:10.1001/jama.279.24.1955.
Sung KC, Hwang ST. Association between insulin resistance and apolipoprotein B in normoglycemic Koreans. Atherosclerosis. 2005;180(1):161-169. doi:10.1016/j.atherosclerosis.2004.11.009.
Pont F, Duvillard L, Florentin E, Gambert P, Vergès B. Early kinetic abnormalities of ApoB-containing lipoproteins in insulin-resistant women with abdominal obesity. Arterioscler Thromb Vasc Biol. 2002;22:1726-1732. doi:10.1161/01.ATV.0000032134.92180.41.
Agoons DD, Agoons BB. Apolipoprotein B and Glycemic Indices in Normoglycemic Adults: Analysis of the National Health and Nutrition Examination Survey, 2007–2016. Cureus. 2025;17(6):e85656. doi:10.7759/cureus.85656.
Comments