Twenty-eight articles. Twelve etiological pathways. Nine treatment modalities. Five paradoxes. And one consistent finding: no existing guideline tells a clinician which treatment matches which cause.
Every guideline for male hypogonadism — the EAU 2026, the Endocrine Society 2018, the AUA 2024 — organizes treatment by modality: here's what TRT does, here's what SERMs do, here's what hCG does. Choose one.
But the question a clinician actually faces is different: this patient's testosterone is low because of X. What do I do about X?
That question has no algorithmic answer in any published guideline. The ICSM 2024 consensus acknowledged the problem directly: "The dichotomization between functional and organic hypogonadism is not adequate." The Clift and Morgentaler WJMH 2026 study of 9,537 men called for "nuanced algorithms considering age, symptom severity, and testosterone profiles." But neither paper built the algorithm.
This article attempts what those papers requested. It is not a guideline — I am not a clinician. It is a synthesis framework derived from 28 evidence-based deep dives, designed to make visible the decision architecture that currently exists only in the heads of the best-informed specialists.
The Structural Precedent
The AACE 2026 Type 2 Diabetes Algorithm shows how this works in another field. Before you treat the glucose, classify the cause. Their new Diabetes Classification Algorithm asks clinicians to consider whether the patient actually has T2D, LADA, monogenic diabetes, or secondary diabetes — because the treatment changes fundamentally depending on the answer.
No equivalent exists for hypogonadism. The implicit assumption is: testosterone is low → replace testosterone. My 28 articles demonstrate that this assumption fails for the majority of patients, because the cause determines whether replacement is the right layer of intervention.
The Five-Layer Cascade
This framework organizes treatment decisions into five priority layers. Every patient enters at Layer 1 and descends only when upper layers are exhausted or inapplicable.
The narrowing shape is intentional. Most guidelines skip directly to Layer 3. The framework insists on Layers 1 and 2 first — because for many etiologies, they work better, cost less, and preserve fertility.
Five Paths Through the Framework
The cascade is universal. How a patient traverses it is not. Five composite archetypes — drawn from the patterns across 28 articles — show how the same framework produces radically different treatment paths depending on the cause.
Archetype 1: The Metabolic Trap Patient
38 years old · BMI 34 · Total T 280 ng/dL · LH 4.2 · Insulin resistant
Layer 1 dominates. The cause is identifiable and treatable. A GLP-1 receptor agonist breaks the five interlocking feedback loops that lock insulin resistance and low testosterone together. Tirzepatide normalized testosterone in 100% of patients in pilot data. Portillo Canales showed 53% → 77% testosterone normalization in 234 men. The axis is suppressed, not broken — remove the suppressor and it recovers.
Path: Layer 1 (GLP-1 RA, 3–6 months) → monitor T recovery → Layer 2 (SERM) only if T fails to normalize → Layer 3 (TRT) as last resort. Layer 4 alert: monitor lean mass on GLP-1; monitor Pozzi 55% persistence rate — not all recover.
Archetype 2: The Opioid-Induced Patient
45 years old · Chronic pain 5yr · Methadone 80 mg/d · Total T 190 ng/dL · LH 2.1
Layer 1 is optimal but often impossible. Opioids suppress the HPG axis at all three levels — hypothalamic GnRH, pituitary LH/FSH, and direct gonadal toxicity. Buprenorphine switch is the best intervention: Tremonti 2026 showed hypogonadism in 3.3% of opioid-free men vs. 26.7% of continuing users. But switching isn't always possible in chronic pain management.
Path: Layer 1 (buprenorphine switch if feasible) → Layer 2 is an evidence void (no SERM or hCG data in active opioid users) → Layer 3 (TRT if chronic opioids required, with fertility counseling — IM formulations cause 65% azoospermia). Layer 4 alert: TRT improves body composition but NOT pain or quality of life in opioid users (Basaria 2015).
Archetype 3: The Young Fertility-Concerned Patient
29 years old · Total T 250 ng/dL · LH 3.8 · Wants children in 2–3 years
Layer 3 (TRT) is contraindicated as first-line. This patient must never start exogenous testosterone without concurrent hCG — and ideally shouldn't need it at all. Layer 1: identify the root cause. Is it sleep disruption? Metabolic? Fragile axis (Caronia 2011: 12.7% of functional HH patients carry CHH gene variants)?
Path: Layer 1 (identify and address cause) → Layer 2 (enclomiphene preferred — avoids zuclomiphene accumulation, +274 ng/dL average per 2025 meta-analysis, zero adverse events at 1 year per Anno/Kasai 2026). SERM trial also serves as diagnostic: axis response predicts long-term outcome. Layer 3 only with concurrent hCG for fertility preservation.
Archetype 4: The Psychiatric Medication Patient
34 years old · Bipolar on olanzapine · Total T 210 ng/dL · Prolactin 42 ng/mL · LH 2.5
Layer 1 requires collaboration with psychiatry. Olanzapine is a high-prolactin antipsychotic. The triple mechanism — PRL-mediated HPG suppression, direct PI3K-Akt-eNOS impairment, dopaminergic disruption — means treating the testosterone without addressing the drug is a patch on a patch. The management hierarchy: dose reduction → aripiprazole augmentation → switch to a prolactin-sparing agent.
Path: Layer 1 (psychiatric medication management, not unilateral) → Layer 5 immediately (no quantified T recovery timeline after antipsychotic switch — total evidence void) → Layer 2 (if axis remains suppressed after drug change) → Layer 3 as last resort. Layer 4 alert: 57% reduced BMD in this population — screen bone density early. Non-adherence risk OR 1.96 if sexual side effects untreated.
Archetype 5: The Post-AAS Patient
32 years old · Former AAS user (3 cycles, 2 years, stopped 8 months ago) · Total T 150 ng/dL · LH 1.2 · FSH 0.8 · INSL3 low
Layer 1 reveals hidden damage. This patient's testosterone is low, but that's not the full story. INSL3 — the biomarker testosterone misses — reveals Leydig cell damage that persists even if T recovers. The HAARLEM study dismantled the recovery myth. Bulut JCEM 2025 confirmed testicular damage via hCG stimulation test.
Path: Layer 1 (INSL3 + hCG stimulation test for damage assessment) → Layer 2 (SERM trial to test axis responsiveness — if poor response per CloTASH data, this predicts chronic need) → Layer 3 (TRT long-term if axis unresponsive) → Fertility counseling critical at every layer (İbis BJU Int 2026: short-term hCG accelerates fertility recovery, but long-term users have poor SERM response).
Five patients. Five different primary layers. One framework. The modality-organized guideline would tell all five to consider TRT, SERMs, or hCG. The etiology-organized framework tells each where to start — and what to monitor for along the way.
The Evidence Void Map
This is the most important visual in 28 articles. It makes visible what clinicians are forced to keep invisible: the vast gaps between what we prescribe and what we've actually tested.
RCT evidence (n≥100 or meta-analysis) Observational or small RCTs only No data
| Treatment | Efficacy RCT | CV Safety | Long-term (>3yr) | Fertility | Bone | Cognition | Depression | Combo Data |
|---|---|---|---|---|---|---|---|---|
| TRT Injectable | ● | ● | ● | ● | ● | ● | ● | — |
| TRT Topical | ● | ● | ● | ● | ● | ● | ● | — |
| hCG Monotherapy | ● | ● | ● | ● | ● | ● | ● | ● |
| Enclomiphene | ● | ● | ● | ● | ● | ● | ● | ● |
| Clomiphene Citrate | ● | ● | ● | ● | ● | ● | ● | ● |
| T + hCG Combo ★ | ● | ● | ● | ● | ● | ● | ● | ● |
| GLP-1 RA (indirect) | ● | ● | ● | ● | ● | ● | ● | ● |
| Pulsatile GnRH | ● | ● | ● | ● | ● | ● | ● | ● |
★ T + hCG is the treatment 58.8% of patients receive in real-world practice (Clift/Morgentaler 2026, n=9,537). It has been studied as a combination in zero RCTs.
Read the table by row and you see a treatment's evidence profile. Read it by column and you see a domain's coverage. But the critical reading is the color pattern: the sea of red. TRT is the only modality with green in cardiovascular safety — because TRAVERSE cost ~$100 million and enrolled 5,246 men. No equivalent investment exists for any alternative. We recommend alternatives as "safer" while having zero safety data for them. That is not evidence-based medicine — it is assumption-based medicine.
The most uncomfortable row is T + hCG. Three out of five patients in the Clift/Morgentaler cohort received this combination. Every clinical trial that generated the evidence for TRT studied monotherapy. The combination patients — the majority — are living in an evidence void. It is as if we studied aspirin and ibuprofen separately while 60% of patients take both, and no one studies what happens when they do.
Monitoring by Cause, Not by Drug
Current monitoring protocols are drug-centric: hematocrit, PSA, liver function, lipids. The framework adds etiology-specific monitoring that no single guideline captures because each was compiled from a different subset of the literature.
| Patient Etiology | Additional Monitoring (beyond standard) | Source Article |
|---|---|---|
| Metabolic / Obesity | HOMA-IR, HbA1c, body composition (DXA), lean mass if on GLP-1, T recovery trajectory without replacement (3–6 mo before Layer 2/3) | #13, #28 |
| Opioid-Induced | Opioid dose/type/duration, buprenorphine switch response, cortisol (10% multi-axis suppression), T recovery post-dose-reduction | #4 |
| Antipsychotic-Induced | Prolactin (baseline + every dose change), BMD early screening, psychiatric non-adherence risk, switch response timeline (no data exists) | #17, #8 |
| SSRI-Induced | Erectile function longitudinal (PSSD risk), T before/after dose changes, bupropion switch consideration | #18 |
| Post-AAS (PPAAH) | INSL3 (Leydig damage marker), hCG stimulation test, semen analysis longitudinal, SERM response as axis test | #10 |
| Veteran / TBI | Pituitary MRI, multi-axis assessment (GH, cortisol, TSH), blast exposure history, PTSD screening | #11 |
| Sleep-Disruption | Sleep architecture (not just duration), AHI, nocturia frequency, T pre/post sleep intervention | #14 |
| Hyperprolactinemia | PRL trend, MRI, visual fields, recovery predictors (T <7.4 nmol/L at 6 mo = 91% sens for non-recovery), cabergoline 5-year withdrawal rule | #8 |
| Age-Related / LOH | Free T (SHBG rises with age — total T misleads), AR CAG repeats if available, critical window consideration for cognitive protection | #7, #22 |
Every row in this table was synthesized from a different article. No single guideline contains all of them. That is the gap the framework fills — not by generating new evidence, but by connecting evidence that already exists across 28 separate domains.
The Combination Therapy Paradox
The Clift/Morgentaler study is a landmark because it is the first large dataset of what clinicians actually prescribe. The numbers are striking: 58.81% received T + hCG, 8.06% received T + hCG + tadalafil, 7.08% received clomiphene (81.93% combined with T). Roughly three out of four patients received combination therapy.
Now look at the evidence base. Every RCT that generated our confidence in TRT — including TRAVERSE — studied monotherapy. The hCG studies (Coviello 2005, n=29; Hsieh 2013, n=26) showed that adding hCG to TRT preserves intratesticular testosterone and spermatogenesis. But zero RCTs have studied T + hCG as a combined treatment for hypogonadism symptoms, optimal dosing, long-term safety, or interaction effects.
This is not a minor gap. The standard of care in practice has never been studied as the standard of care in research. When a clinician says "the evidence supports TRT," they mean the evidence supports TRT monotherapy — which is not what they are prescribing to 75% of their patients.
What the Paradoxes Tell Us About Monitoring
Five consequence articles — bone, cognitive, cardiovascular, depression, immune — revealed the same pattern: the relationship between testosterone and outcomes is not linear. Every domain has a paradox, and every paradox has an etiology-specific dimension that current monitoring misses.
The bone paradox: TRT improves density but TRAVERSE showed HR 1.43 for fractures. The resolution is compartment-specific — trabecular gains, cortical losses — and the E2 threshold for cortical protection (~25–30 pg/mL) means patients with low aromatase activity need bone drugs on top of testosterone, not instead of it.
The cognitive paradox: low T predicts dementia, but it may be the elevated FSH — not the low T — driving neurodegeneration. TRT suppresses gonadotropins; SERMs raise them. No one is studying whether the treatment modality matters for brain outcomes.
The cardiovascular paradox: Mendelian randomization shows higher genetic T → higher CAD risk, mediated through blood pressure. But TRAVERSE found TRT safe. The resolution: lifetime genetic exposure ≠ therapeutic replacement in deficient men. Both can be true.
Layer 4 of the framework exists because these paradoxes are etiology-dependent. The metabolic trap patient on GLP-1 faces a different set of paradoxes (lean mass loss, T durability after discontinuation) than the post-AAS patient on TRT (INSL3 damage masking, Leydig cell depletion). Monitoring must follow the cause, not just the drug.
What Would Need to Change
The framework reveals five research priorities that would convert red cells to green in the evidence void map:
1. A T + hCG combination RCT. Not a fertility substudy — a primary efficacy and safety trial of the combination that 58.8% of patients actually receive. This is the single largest evidence gap in the field.
2. Cardiovascular outcome trials for alternatives. Enclomiphene, hCG monotherapy, and clomiphene have zero MACE data. TRAVERSE cost $100 million. The investment asymmetry means we cannot claim alternatives are safer — only that they haven't been tested.
3. Etiology-specific treatment trials. No RCT has randomized by cause. A trial comparing Layer 1 → Layer 2 → Layer 3 stepwise approach vs. direct TRT in metabolic HH would test the framework itself.
4. Drug-induced HH recovery timelines. After antipsychotic switch, after SSRI discontinuation, after opioid dose reduction — the recovery trajectories that would guide Layer 1 waiting periods are almost entirely unknown.
5. Treatment modality and brain outcomes. The gonadotropin hypothesis — that elevated FSH/LH drive neurodegeneration — means treatment choice may matter for cognitive outcomes. TRT suppresses gonadotropins; SERMs and hCG raise them. The differential brain effect has never been studied.
A Framework, Not an Answer
Twenty-eight articles taught me that secondary hypogonadism is not one condition. It is twelve conditions that share a common endpoint — low testosterone — while diverging in mechanism, optimal treatment, paradox profile, and evidence density. The modality-organized guideline treats them as one. The etiology-organized framework treats them as twelve.
This framework is not a clinical guideline. It is a map of the decision architecture that already exists in the literature but has never been assembled in one place. It shows where the evidence supports confident decisions (green cells), where we're relying on weaker data (yellow cells), and where we're guessing (red cells). The sea of red is not an indictment of the field — it is a reflection of a condition that has been studied modality-by-modality when it needed to be studied cause-by-cause.
The ICSM 2024 consensus said the binary classification is inadequate. The Clift/Morgentaler study called for nuanced algorithms. The AACE showed that etiology-first works in metabolic disease. This framework is an attempt to answer those calls.
The five layers. The five paths. The evidence void map. Together, they form the treatment decision architecture that 28 deep dives reveal — and that no single guideline has yet built.