Bone is a microenvironment. Anabolic drugs change who lives there.

Bone is not a tissue. It is a microenvironment — a coupled system of osteoblasts, osteoclasts, osteocytes, marrow stromal cells, growth-factor reservoirs, and a calcium-handling apparatus that links to the kidney through PTH and to the immune system through RANKL. A patient’s bone is a hospitable or hostile soil. Drugs that change the bone do not only change its mineral density. They change what can live there.

This matters because bone is the dominant metastatic site for breast, prostate, lung, and renal cancers, and a major site for myeloma. The “vicious cycle” of bone metastasis described by Mundy in the early 2000s — tumour cells secrete PTHrP, PTHrP activates osteoclasts, resorbed bone releases TGF-β and IGF-1, the growth factors feed tumour proliferation, the tumour secretes more PTHrP — runs on the same machinery the bone-remodelling drugs target.

Two of the four major bone-drug classes have been studied head-on in cancer patients with established bone metastases. Two have not. The two that have not are the anabolic ones. The mechanism makes this an important gap.

What the four drug classes do, in microenvironment terms

We composed the bone module from the Lemaire 2004 baseline, the Pivonka 2008 TGF-β extension, and the Graham-Ayati 2013 Wnt / sclerostin / osteocyte extension. The result is a 14-state coupled subsystem capturing osteoblast, osteoclast, osteocyte, RANKL, OPG, PTH, sclerostin, calcium, and the mechanical- load coupling. Run inside the wider body, the bone module talks to the kidney (RAAS, calcium handling), to bone marrow (haematopoiesis, myelosuppression), and to the immune system (RANKL is also a T-cell signal). Each of the four drug classes intersects this picture at a different point.

Bisphosphonates (zoledronic acid, pamidronate) are osteoclast inhibitors. They accumulate in bone and trigger osteoclast apoptosis. Mechanistically, this interrupts the vicious cycle: fewer osteoclasts → less bone resorption → less TGF-β/IGF-1 released into the marrow → less tumour-supportive signal. The Coleman meta-analyses across breast, prostate, and myeloma show real reductions in skeletal-related events. The model reproduces the direction and the order of magnitude.

Denosumab (anti-RANKL antibody) is the same direction as bisphosphonates but acts further upstream. By neutralising RANKL it blocks osteoclast formation in the first place. Same SRE benefit in randomised trials. With one wrinkle: when denosumab is stopped, the RANKL suppression releases rapidly, and the rebound activates a wave of osteoclasts at once. The Cummings 2018 vertebral-fracture observation post-discontinuation is the clinical consequence. The model produces the same rebound under any rapid removal of the RANKL block, with or without the tumour signal present.

Teriparatide (recombinant PTH 1-34) is the anabolic side of the molecule. Pulsed, low-dose PTH paradoxically builds bone — it activates osteoblasts disproportionately more than osteoclasts when delivered in short daily spikes. This is the Neer 2001 pivotal-trial finding and the basis for osteoporosis approval. The mechanism is good for the patient with thin bones.

It is also the mechanism the bone-metastatic tumour cell has been selected to exploit. Many tumour types in the bone microenvironment express the PTH/PTHrP receptor (PTH1R) on their own membrane — PTHrP is part of the vicious cycle’s tumour-side signal. Anabolic PTH 1-34 dosing reaches that receptor with the same affinity it reaches the osteoblast version of it. The carcinogenicity signal in the Vahle 2002 rat study (osteosarcoma in F344 rats at long-term high doses) is a separate but related concern. The pivotal trials excluded patients with active or recent malignancy. The labelled contraindication remains. The mechanistic question — does teriparatide accelerate occult bone metastasis — has not been the subject of a controlled human trial, because no sponsor will run that trial.

Romosozumab (anti-sclerostin antibody) is the newest anabolic. By neutralising sclerostin it removes the brake on canonical Wnt signalling in osteocytes and osteoblasts, which translates into bone formation. FRAME and ARCH demonstrated fracture-rate reduction. ARCH also showed a numerical imbalance in cardiovascular serious adverse events versus alendronate, which led the FDA to add a boxed warning. The cardiovascular signal remains under active investigation. A separate question — whether sclerostin neutralisation affects the metastatic microenvironment — has also not been formally addressed in trials. Wnt signalling is pro-tumour in colorectal, breast, and several other cancers. The model says the direction is plausible. The trial data does not exist.

Running the model

We ran the four drug classes against a synthetic patient with bone-resident occult breast-cancer cells (PTHrP-expressing, Wnt-responsive, microscopic at baseline) and tracked tumour burden trajectory over twelve months under each regimen, compared to no bone drug.

Direction-of-effect summary, all reproducing the published mechanism literature:

• Bisphosphonates: tumour burden trajectory is suppressed versus no-treatment — the vicious-cycle interruption is real and measurable in the model.
• Denosumab: suppressed during treatment. If denosumab is then discontinued without a bisphosphonate bridge, the rebound resorption wave releases the stored growth-factor pool back into the marrow; the model predicts a transient acceleration of tumour growth during the rebound window.
• Teriparatide: tumour burden trajectory is higher than no-treatment under any plausible PTH1R-expression parameterisation for the tumour cell. Magnitude depends on receptor density and ligand kinetics. Direction is consistent.
• Romosozumab: uncertain. Direction depends on Wnt pathway parameterisation in the specific tumour type. For Wnt-responsive tumours the model produces an increase; for tumours where Wnt is downregulated the effect is small or zero.

These are mechanism-driven predictions, not anchored to a randomised trial result, because such trials have not been run. They are the kind of hypothesis a screening rule should be written against, not the kind that should drive prescribing without confirmation.

What this implies

The clinical implication is simple. Before prescribing an anabolic bone drug — teriparatide, abaloparatide, romosozumab — in a patient with a known prior malignancy, especially one of the bone-tropic primaries (breast, prostate, lung, renal, myeloma), the mechanistic question should be on the table. Current labels mostly require this. Real-world prescribing under guideline grey areas frequently does not.

The broader implication is a screening direction the field could pursue without waiting for a controlled trial. A baseline PET or whole-body MRI before initiating an anabolic bone agent in a higher-risk patient would detect occult bone disease that would change the prescribing decision. The cost of the imaging is small relative to the cost of accelerating a metastasis that was undetectable at the time the bone drug was started.

The other direction is a registry. The drugs are in widespread use. Linking prescription data to cancer-incidence and bone-metastasis-incidence data prospectively would give the field the answer the trial design did not. The model predicts a measurable signal in the appropriate sub-population. The registry would either confirm or refute it.

Limitations

The tumour-supportive output of the model is a direction, not a magnitude. Receptor density on tumour cells is patient-specific and tumour-specific; we have parameterised it from published expression-microarray and immunohistochemistry literature rather than from prospective measurement. The Wnt pathway question for romosozumab depends heavily on which downstream β-catenin targets are dominant in the specific tumour, which varies by indication. The bone microenvironment also includes immune cells (T-regs, macrophages, MDSCs) that we have abstracted into a smaller compartment than the literature warrants for full immune-coupling work.

We are not oncologists. The mechanism story is, by construction, anchored to the same Lemaire-Pivonka-Graham bone-remodelling tradition that has underpinned the osteoporosis-drug development programmes themselves. What the model adds is the coupling to the tumour compartment that the bone-only trials, by design, did not include.

Mechanism anchors: Lemaire V et al., J Theor Biol 2004;229:293; Pivonka P et al., Bone 2008;43:249; Graham JM, Ayati BP et al., PLoS ONE 2013;8:e63884; Mundy GR, Metastasis to bone — causes, consequences, therapeutic opportunities, Nat Rev Cancer 2002;2:584; Suva LJ et al., Bone metastasis — mechanisms and therapeutic opportunities, Nat Rev Endocrinol 2011. Drug trial anchors: Coleman R et al., Bisphosphonates in early breast cancer (Lancet 2015 meta-analysis); Neer RM et al., NEJM 2001;344:1434 (teriparatide pivotal); Vahle JL et al., Toxicol Pathol 2002;30:312 (osteosarcoma rat study); Cosman F et al., NEJM 2016;375:1532 (romosozumab FRAME); Saag KG et al., NEJM 2017;377:1417 (romosozumab ARCH — cardiovascular signal); Cummings SR et al., JBMR 2018 (denosumab discontinuation vertebral fractures). Model code: unified/replications/bone_met_vicious_cycle/. This note is a mechanism-driven methodology piece, not clinical or prescribing advice.