Your muscles grew.
What happened to your heart?

Training can change cardiac size and geometry. In a prospective study of competitive athletes, endurance and strength training produced different structural and functional patterns. That matters because a nontraining comparison group cannot answer every question about a highly trained person's heart. It also means “athlete's heart” should not become a catch-all explanation for an abnormal scan.^[8]

Skeletal muscle is judged partly by its ability to produce force. The heart must repeatedly fill, eject blood and relax. Increasing its muscle mass does not necessarily improve those coordinated tasks. Wall thickness, chamber size, relaxation and contraction need to be interpreted together, in the context of body size, blood pressure, training and symptoms.

The 2017 Baggish investigation compared experienced male weightlifters with and without substantial AAS exposure. Users had poorer systolic and diastolic measures, while coronary imaging identified greater plaque burden. Using weightlifting controls made the comparison more relevant than comparing lifters with sedentary people, although the study remained observational.^[5]

This article concerns predominantly male cohorts with nonmedical, often mixed AAS exposure. It does not estimate the effects of a particular compound, establish the same risk in women, or substitute for evidence about prescribed testosterone replacement in appropriately selected patients.

Left ventricular ejection fraction, or LVEF, is the fraction of blood in the filled left ventricle ejected during a beat. It is not the percentage of the heart that works. A change from 48% to 54% is six percentage points, not a six-percent relative improvement. The denominator and the measurement conditions matter.

Diastolic function concerns relaxation and filling. Global longitudinal strain, or GLS, tracks deformation during contraction. Because longitudinal shortening is conventionally negative, a less-negative value can indicate worse shortening. LV mass index adjusts estimated muscle mass for body size; it is not a measurement of scar tissue.^[1][5][6]

That separation prevents a common leap: “The echo improved, therefore everything recovered.” A study of pumping function cannot directly establish plaque regression unless plaque was also measured over time. Likewise, a normal calcium score does not itself demonstrate normal microvascular function. The test answers its own question, not every question about the organ.

Abdullah and colleagues published their longitudinal follow-up on October 12, 2026. From an original cohort of 101 AAS users and 71 weightlifting controls, 70 users and 35 controls were analyzed again. The baseline assessments were in 2017–2019, with re-examination in 2025. Their exposure transitions are shown below. Cessation was a real-world choice, not a randomized assignment.^[1]

A transition-based design is more informative about change than a single comparison of current and former users. Each returning person supplies an earlier measurement. However, people who stop may also change other drugs, health care, training or behavior. Those who return may differ from people who cannot or choose not to participate.

Imagine two otherwise unmeasured pathways: a health scare could motivate both cessation and treatment, while a person feeling well might continue using. That is an illustration of selection, not a claim about any specific participant. It shows why a longitudinal association can be persuasive without becoming the equivalent of a randomized experiment.

Those who stopped improved, but the main-text comparison with controls is essential: both former-user groups still had lower LVEF than controls despite normal-range averages, and the discontinued group retained reduced GLS. “Normal range,” “improved” and “indistinguishable from controls” are different statements.^[1]

A group mean also hides the spread. Two groups with the same average can contain very different individual trajectories. Without an appropriate individual-level recovery definition, the table cannot supply a personal probability of recovery or a deadline by which it should occur.

There is another statistical trap here. A significant change in one group and a nonsignificant change in another does not, by itself, prove that the changes differ significantly between groups. The table therefore labels its tests explicitly rather than turning separate P values into a head-to-head treatment claim.

Readers should also ask what “baseline” means. For a previously exposed person, the first study scan is not necessarily a pre-exposure scan. Returning to that measurement can be reassuring without proving a return to the cardiac state that existed before any AAS use.

The HAARLEM echocardiography study followed 31 men through a self-selected AAS cycle lasting a median 16 weeks. LVEF fell by 4.9 percentage points (95% CI −7.2 to −2.5), while LV mass rose 28.3 g (95% CI 14.2 to 42.4). Measures returned toward baseline after a median eight-month recovery interval.^[2]

Only 25 underwent the recovery scan: four restarted AAS before it and two missed it for other reasons. That qualification matters. The reassuring recovery finding describes the observed follow-up group, not a guarantee for continuous exposure or for everyone who began the study.^[2]

In Fyksen's long follow-up, median within-person LVEF change was +3 percentage points (IQR 1 to 8) among those who stopped, versus −2 (IQR −6 to 2) among continued users. The between-group comparison was significant. However, the sample was small, exposure and treatment changed, and not everyone from the original cohort returned.^[3]

Rasmussen's earlier study found altered strain in former users without detecting focal fibrosis on its MRI assessment. That combination cautions against both extremes: an abnormal functional marker does not prove visible scar, and a negative scar assessment does not erase an abnormal functional marker.^[6]

These studies should not be stitched into a single recovery curve. Different participants, exposure histories, methods and starting points prevent that calculation. Their contribution is triangulation: related questions examined in different ways, with results that support improvement while leaving its completeness and timing unresolved.

Bulut's 2024 PET study examined 90 strength-training men. Mean myocardial flow reserve did not differ significantly between groups (P=.15). More current and former users fell below prespecified flow-reserve thresholds, but adjusted pairwise comparisons were not significant. Both findings belong in the story.^[4]

This is a signal warranting follow-up, not proof of permanent microvascular injury. It also cannot establish that every former user has impaired flow. The primary null finding limits certainty even when secondary ways of examining the data look concerning.^[4]

Clinical events are a separate layer. A Danish registry study followed 1,189 men sanctioned for AAS use and 59,450 matched population controls for an average eleven years. The abstract reported higher adjusted hazards of myocardial infarction, cardiomyopathy and heart failure, including a heart-failure hazard ratio of 3.63 (95% CI 2.01 to 6.55).^[7]

That is a relative association in a particular cohort, not a 3.63% chance of heart failure. The general-population comparator, selection into the sanctioned cohort and unmeasured differences affect interpretation. This design does not isolate a specific steroid or tell us how event risk changes after confirmed cessation. Those questions require different data.

The new longitudinal paper also found that antihypertensive-treated continuing users had more favorable LVEF changes than untreated continuing users. Medication was not randomly allocated, and treatment details were collected at follow-up. This cannot establish that adding a particular drug neutralizes continued AAS exposure.^[1]

Consider what would be needed for that stronger claim: comparable starting risk, defined medication exposure, adherence, repeated blood-pressure measurements, control of other drugs and sufficient clinical follow-up. A favorable imaging association is useful for generating a question. It cannot answer every part of that question by itself.

Nor should adjusting statistically for blood pressure be interpreted as proving that pressure is irrelevant. A variable can be measured imprecisely, change over time, or form part of the pathway linking an exposure to an outcome. Removing its statistical contribution is not equivalent to conducting an experiment that eliminates it.

Mixed exposure creates another boundary. When participants have used multiple AAS and other performance-enhancing drugs, a cohort result cannot be assigned to trenbolone, testosterone or any other single agent. A memorable compound name should not substitute for a measured compound-specific comparison.

The practical reading skill is to keep the outcome, time interval and comparison attached to every claim. Ask whether a paper measured symptoms, imaging, hospitalization or death; whether it followed the same people; and whether exposure was verified or reconstructed from memory. “Recovery” becomes much clearer once those details are specified.

For an individual, a clinician can interpret a scan alongside exposure history, symptoms, blood pressure, other substances and prior measurements. A website cannot decide whether a particular result represents athletic adaptation, measurement variability or disease. It also cannot infer a safe return to exposure from an improving number.

There is a useful distinction between uncertainty and indifference. Uncertainty about the exact size or reversibility of harm does not mean that the observed functional changes should be ignored. Equally, evidence of harm does not mean that improvement is impossible. The longitudinal studies make room for an evidence-based discussion of both.