Part V · Immunology and the tumor microenvironment · Chapter 25

Host and tumor interactions

The tumour is not the only variable in the room, and several of the others are measurable.

1 · Endocrine and metabolic context

The host supplies the ligand. Before menopause the ovary is the dominant source of oestrogen. After menopause the dominant source is peripheral aromatisation of androgens in adipose and other tissues. That single fact explains why an aromatase inhibitor is active after menopause and inadequate before it without ovarian suppression. The pharmacology is developed in Endocrine therapy.

The quantitative relationship between circulating hormone and risk is well established. In a reanalysis of nine prospective studies covering 663 women who developed breast cancer and 1765 who did not, risk rose across quintiles of every sex hormone examined. For total oestradiol the relative risk in the highest quintile against the lowest was 2.00, with a 95% confidence interval of 1.47 to 2.71. Free oestradiol gave 2.58, with limits of 1.76 to 3.78, and sex hormone binding globulin was associated with lower risk1.

Those numbers describe circulating concentrations rather than the concentration inside breast tissue, which is a different quantity. Local aromatase activity means tissue availability can diverge from the plasma value, in either direction. A serum oestradiol is therefore a measurement of supply, not of what the receptor in the tumour is seeing. The receptor-level consequences are developed in Estrogen receptor signaling and endocrine resistance.

2 · Obesity, insulin signaling, and metabolic dysfunction

Three mechanisms run under one word and should be separated. Oestrogen supply, insulin and insulin-like growth factor signalling, and adipose inflammation.

Oestrogen supply is the mechanism that reverses direction with menopausal status. In a cohort of 5.24 million United Kingdom adults, each 5 kg/m2 increase in body mass index carried a hazard ratio of 1.05 for postmenopausal breast cancer, with 99% confidence limits of 1.03 to 1.07. The same increase carried a hazard ratio of 0.89 for premenopausal disease, with limits of 0.86 to 0.922. A single exposure with opposite signs either side of menopause is strong evidence that more than one mechanism is operating.

Insulin is a separate axis and it tracks outcome rather than incidence. In 512 women without known diabetes who had early-stage disease, fasting insulin in the highest quartile against the lowest carried an adjusted hazard ratio of 2.1 for distant recurrence, with limits of 1.2 to 3.6. For death the adjusted hazard ratio was 3.3, with limits of 1.5 to 7.03.

Adipose inflammation connects the first two. Crown-like structures, in which macrophages surround necrotic adipocytes, were found in the breast tissue of 14 of 30 women undergoing breast surgery. Their severity correlated with body mass index and with adipocyte size, and inflamed tissue showed elevated aromatase expression and activity4. The adipose interface itself is developed in Adipocytes and the mammary adipose environment.

Caution

An association between a host variable and outcome does not establish that correcting the variable improves outcome. Metformin is the worked example. MA.32 randomised 3649 patients with high-risk non-metastatic breast cancer to metformin or placebo for five years. In the hormone receptor positive population of 2533 patients that carried the primary analysis, the hazard ratio for invasive disease-free survival was 1.01, with a 95% confidence interval of 0.84 to 1.215. The insulin association survives. The inference that lowering insulin pharmacologically helps does not.

3 · Aging, immunosenescence, and clonal hematopoiesis

Age changes the host immune system in ways that matter for immunotherapy. Interferon signalling and antigen presentation in the triple-negative microenvironment decline with age, in mice and in correlative human data, and checkpoint blockade was less effective in aged mice. Priming innate immunity with a STING agonist restored the response in that model6. This is a preclinical result with human correlative support. It is not a clinical finding, and the age structure of checkpoint trials means the question has largely not been asked prospectively.

Clonal haematopoiesis is the better characterised host change. Whole-exome data from 17,182 people unselected for haematologic phenotype showed somatic mutations in leukaemia-associated genes to be rare below age 40. They were present in 9.5% of 2300 people aged 70 to 79, and in 18.4% of 103 aged 90 to 108. Most variants were in DNMT3A, TET2 and ASXL1. Carriage was associated with a hazard ratio of 11.1 for subsequent haematologic cancer and 1.4 for all-cause mortality7.

In people who already have a solid tumour the prevalence is higher. Paired tumour and blood sequencing in 8810 patients with non-haematologic cancers detected clonal haematopoiesis in 25%, with presumptive leukaemia driver mutations in 4.5%. The driver-mutation group had shorter survival and more subsequent haematologic cancer8.

Treatment selects which clones expand. Radiation, platinum and topoisomerase II inhibitors preferentially select mutations in DNA damage response genes, including TP53, PPM1D and CHEK2. Serial sampling showed those clones outcompeting others under exposure9. This is the selection argument of Temporal heterogeneity and clonal evolution applied to the marrow instead of the tumour.

In practice

The consequence for plasma testing is immediate. In 124 patients with metastatic cancer and 47 controls without cancer, 53.2% of cell-free DNA mutations in patients and 81.6% in controls had features consistent with clonal haematopoiesis10.

A circulating tumour DNA report showing TP53, PPM1D or CHEK2 at low allele fraction may be reporting the marrow rather than the tumour. Ask whether the assay sequenced matched white blood cells. If it did not, do not treat such a variant as tumour-derived. The blind spots of plasma assays generally are set out in Liquid biopsy and longitudinal monitoring.

4 · Pregnancy, postpartum involution, and tissue remodeling

Parity is protective in the long run and hazardous in the short run, and the crossover is slow enough that both statements describe most patients at some point.

A pooled analysis of 15 prospective cohorts, covering 18,826 incident breast cancers in women under 55, quantified the shape. Compared with nulliparous women, the hazard ratio for parous women peaked around five years after birth at 1.80, with a 95% confidence interval of 1.63 to 1.99. It crossed below one about 24 years after birth and reached 0.77 after 34 years, with limits of 0.67 to 0.88. The pattern was driven by oestrogen receptor positive disease, and breastfeeding did not modify it11.

Prognosis follows the same clock. Among 701 women aged 45 or younger with stage I to III disease, diagnosis within ten years of childbirth was associated with higher risk of distant metastasis. The effect was most marked in stage I and II disease, and it was present in both receptor-positive and receptor-negative cancers12.

The mechanism proposed for this is postpartum involution. Weaning triggers a programmed regression of the gland with the tissue signature of wound healing, including fibrillar collagen deposition and cyclooxygenase-2 expression. Human breast cancer cells placed in an involuting mouse mammary microenvironment formed large tumours with abundant fibrillar collagen and an invasive phenotype. Inhibiting cyclooxygenase-2 reduced the involution-associated collagen fibrillogenesis, tumour growth and lung infiltration13. Matrix and mechanics generally are developed in Extracellular matrix composition and tissue mechanics.

Caution

The term pregnancy-associated breast cancer runs together two different exposures. Diagnosis during pregnancy and diagnosis in the years after delivery are not the same situation, and the postpartum window carries the excess risk11,12. Studies that pool them dilute the finding, so check which definition was used before comparing two estimates from this literature. Management in young women is covered in Young adults, fertility, and pregnancy.

5 · Gut microbiome, intratumoral bacteria, and the mycobiome

Three separate claims travel under one heading: gut bacteria altering systemic oestrogen exposure, bacteria living inside the tumour, and fungi doing the same. They have different evidence and different failure modes.

The gut claim has a specific mechanism. Bacterial beta-glucuronidases deconjugate oestrogens excreted in bile, returning free hormone to the circulation, so the composition of the gut community can alter systemic oestrogen exposure. The aggregate of enteric genes capable of metabolising oestrogens has been named the estrobolome14. The route to oestrogen receptor positive disease is plausible. No intervention has tested it.

The intratumoral claim is observational and it singles breast cancer out. A survey of 1526 tumours and adjacent normal tissue across seven cancer types found tumour-type-specific bacterial communities, with breast cancer carrying a particularly rich and diverse one. The bacteria were mostly intracellular, in both cancer cells and immune cells15.

Function has been tested only in mice. Depleting intratumoral bacteria in a spontaneous murine breast tumour model reduced lung metastasis without changing primary tumour growth. Bacteria carried inside circulating tumour cells promoted survival under fluid shear stress by reorganising the actin cytoskeleton16. That is a specific, testable mechanism for a step in dissemination, and it is currently a mouse result.

Fungi have been characterised in the same way, across 17,401 samples and 35 cancer types, at low abundance and with cancer-type-specific composition17.

Caution

Low-biomass sequencing is exquisitely vulnerable to contamination and to misassignment of reads to the wrong organism. A re-analysis of a widely cited pan-cancer microbiome dataset found errors that overturned the original findings, and concluded that most of the microbes reported as cancer-associated were not present in the samples at all18. The underlying biology may still be real, and the directly imaged intracellular bacteria in breast tissue are harder to explain away15. Quantitative microbiome claims should not be repeated without knowing which dataset and which pipeline produced them.

6 · Circadian disruption and the timing of biology

Two claims are usually presented together, and the evidence for them runs in opposite directions: circadian disruption causes breast cancer, and circadian phase governs how an existing breast cancer behaves.

The first claim is not supported by the prospective data. Three United Kingdom cohorts were analysed together and pooled with seven previously published prospective studies, giving 4660 breast cancers among women reporting night shift work. The combined relative risk for any night shift work was 0.99, with a 95% confidence interval of 0.95 to 1.03. For 20 or more years of night shift work it was 1.01, with limits of 0.93 to 1.1019. Those limits are narrow, so this is a null result rather than an inconclusive one, and it should be reported as such.

The second claim is supported and it is striking. Most spontaneous intravasation of circulating tumour cells, in patients with breast cancer and in mouse models, occurred during the rest phase. Rest-phase circulating tumour cells were highly metastasis-competent, while cells generated during the active phase were largely not. Single-cell profiling showed upregulation of mitotic genes confined to the rest phase20.

The measurement consequence is direct and under-appreciated, because a circulating tumour cell count depends on the time of day the blood was drawn. Any study comparing counts across cohorts without fixing sampling time has an uncontrolled variable. Dissemination biology is developed in Metastatic dissemination and organ tropism, and plasma-based monitoring in Liquid biopsy and longitudinal monitoring.

The two findings are compatible. Disrupting the clock need not cause cancer for the clock to govern how a cancer that already exists spreads.

7 · Chronic stress and adrenergic signaling

The pathway is sympathetic. Noradrenaline acts on beta-adrenergic receptors expressed by tumour cells and by stromal and immune cells in the microenvironment. The innervation itself is described in Nerves and neuro-immune interaction.

The animal result is large. Stress-induced neuroendocrine activation had negligible effect on primary tumour growth in an orthotopic mouse model of breast cancer. It produced a 30-fold increase in metastasis to lymph nodes and lung. The effect was mediated by beta-adrenergic signalling through recruitment of macrophages into the primary tumour, and propranolol reversed it21. An effect of that magnitude in mice should be read as a demonstration that the pathway exists, not as a prediction of effect size in people.

The human evidence is a window trial, and its endpoint should be named accurately. Sixty patients were randomised to oral propranolol or placebo for seven days before surgery. Propranolol downregulated mesenchymal gene expression in the primary tumour and altered intratumoral recruitment of neutrophils, natural killer cells and dendritic cells22. That is target engagement in human tumour tissue. It is not a recurrence endpoint, and no randomised trial has tested whether beta-blockade changes breast cancer outcomes.

The honest position has three parts held separately. The mechanism is established. The pathway is drugable in humans and engagement has been shown in tissue. Clinical benefit is unproven. Observational studies of beta-blocker users are confounded by the indication for the drug, and they should not be used to close the gap.

8 · Exercise, diet, and modifiable host biology

The observational evidence in breast cancer is consistent. The Nurses' Health Study followed 2987 women with stage I to III disease. Women reporting 9 to 14.9 metabolic equivalent task hours of activity per week had a relative risk of breast cancer death of 0.50 against women reporting under 3 hours. The 95% confidence interval was 0.31 to 0.82, and the benefit was concentrated in women with hormone-responsive tumours. Nine metabolic equivalent task hours is roughly three hours of walking at an average pace23.

Consistency across observational cohorts supports a hypothesis without establishing causation. Physical activity tracks fitness, comorbidity, treatment tolerance and occult disease burden, and none of those is fully adjustable.

The randomised evidence exists and it is not in breast cancer. CHALLENGE randomised 889 patients with resected colon cancer who had completed adjuvant chemotherapy to a three-year structured exercise programme or to health education materials. At a median follow-up of 7.9 years the hazard ratio for disease-free survival was 0.72, with a 95% confidence interval of 0.55 to 0.94. Five-year disease-free survival was 80.3% against 73.9%24. This is the first level-one demonstration that a behavioural intervention changes cancer outcome, and it is in a different disease.

Diet has a randomised result in prevention. The Women's Health Initiative dietary modification trial randomised 48,835 postmenopausal women to a low-fat dietary pattern with increased vegetable, fruit and grain intake, or to usual diet. At a median follow-up of 19.6 years, deaths attributed to breast cancer numbered 132 in the intervention group against 251 in the comparison group, giving a hazard ratio of 0.79 with limits of 0.64 to 0.9725. The intervention changed several things at once, so the active component is not identified.

In practice

Recommend structured physical activity. The basis is a randomised disease-free survival benefit in another cancer24 plus consistent observational data in this one23. Say which is which when a patient asks.

Name the harm honestly. Musculoskeletal adverse events were more frequent in the exercise arm of the randomised trial, occurring in 18.5% against 11.5%24.

Do not promise a survival benefit from exercise in breast cancer. It has not been shown in a randomised trial in this disease.

Do not prescribe metformin for the cancer5. Treat the diabetes.

9 · Host factors that change efficacy and toxicity

Four host variables change what a given dose does, and each rests on a different kind of evidence.

Body composition changes exposure. Among 55 women receiving capecitabine for metastatic breast cancer, about a quarter were sarcopenic by a computed tomography measure of skeletal muscle at the third lumbar vertebra, including women of normal weight and obese women. Toxicity occurred in 50% of sarcopenic patients against 20% of the others, and time to progression was shorter26. Dosing by body surface area does not measure lean mass, and lean mass is where much of the metabolism happens.

Adiposity may change the efficacy of an endocrine agent. In an exploratory analysis of the ATAC trial, women with body mass index above 35 kg/m2 had more recurrences than women below 23, with an adjusted hazard ratio of 1.39 and limits of 1.06 to 1.82. The relative advantage of anastrozole over tamoxifen was non-significantly greater in thin women27. The hypothesis is incomplete aromatase suppression at higher fat mass. The analysis was exploratory, and it is not a basis for changing a dose.

Germline pharmacogenetics gives two worked examples that point in opposite directions. Prospective DPYD genotyping in 1103 evaluable patients identified 85 heterozygous variant carriers who received reduced starting doses of fluoropyrimidine. Severe toxicity in carriers fell relative to a historical cohort of carriers treated at full dose28. By contrast, 1203 patients from the ATAC trial were genotyped for CYP2D6, of whom 588 had received tamoxifen. After a median follow-up of ten years there was no significant association between metaboliser status and recurrence in the tamoxifen-treated group29.

Putting those two side by side gives the lesson. Both hypotheses had a clean mechanism, and one survived prospective testing while the other did not. Mechanistic plausibility is not evidence, and much of the CYP2D6 controversy turned on which tissue was genotyped rather than on the biology.

Comorbidity, organ function and polypharmacy are the fourth category, and they are developed in Older adults and complex comorbidity and Treatment toxicity and supportive care.

10 · Social adversity as biology, allostatic load and the bridge to Part XIII

Social adversity is usually written about as a determinant of access, and it is one. It is also a chronic physiological exposure with measurable mediators. This section makes the second claim, because the first is already well made elsewhere and the second is what belongs in a chapter about host biology.

Allostatic load is the accumulated physiological cost of repeated adaptation to stressors. It is indexed across neuroendocrine, metabolic, inflammatory and cardiovascular systems30. Say how the index is built, because it is a scoring convention and not a measurement of a substance. Each component biomarker is dichotomised, usually at the worst sample quartile, and the flags are summed. The threshold is a convention chosen for reproducibility, exactly like the counting rules discussed in HER2 heterogeneity.

The breast cancer evidence is direct. Among 4459 patients with stage I to III disease, high allostatic load carried a hazard ratio of 1.46 for all-cause mortality. The 95% confidence interval was 1.11 to 1.93, after adjustment for sociodemographic, clinical and treatment factors. The gradient was dose-dependent across quartiles, reaching 1.79 in the highest quartile against the lowest. The association persisted after adjustment for the Charlson Comorbidity Index. Mean allostatic load was higher in Black patients, in unmarried patients, and in patients with government-supplied insurance31.

That last pair of findings is the hinge. Allostatic load is patterned by social position, and it predicts mortality after treatment and comorbidity have been accounted for. Those two facts together describe a biological pathway running from social exposure to outcome. They do not describe a proxy for undertreatment, which would have been absorbed by the treatment adjustment.

A second and independent line points the same way. Among 2835 women in the Nurses' Health Study, social isolation before diagnosis was compared with social integration. Isolated women had a hazard ratio of 1.66 for all-cause mortality, with limits of 1.04 to 2.65. For breast cancer mortality specifically the hazard ratio was 2.14, with limits of 1.11 to 4.1232.

The candidate mediators are already in this chapter rather than hypothetical. Chronic sympathetic activation has a demonstrated tumour-level effect and a drugable receptor21,22. Glucocorticoid and melatonin rhythms govern the generation of metastasis-competent circulating tumour cells20. Perceived isolation has characterised neuroendocrine consequences, including altered hypothalamic pituitary adrenal activity33. What is missing is not a pathway but quantification of how much of any observed outcome gap each pathway carries.

Caution

None of this licenses treating social adversity as a fixed biological property of a person or a group. Allostatic load is a consequence of exposure and is in principle reversible. Reading a measured biological difference as an inherent one is the specific failure mode set out in How disparities research can itself cause harm, and it has caused harm before in this field.

This section sits in Part V rather than Part XIII on purpose. Quarantining equity into a part of its own carries an implicit claim, which is that it is a delivery problem bolted onto the end of a biological argument. The measurements above contradict that claim. Social exposure has neuroendocrine, metabolic and immune mediators, the same mediators this chapter has already used to explain obesity, circadian phase and adrenergic signalling. The decomposition of an outcome gap into biology, access and delivery is taken up in Decomposing an outcome gap into biology, access, and delivery, the gene and environment interaction literature in Gene and environment interaction, social genomics, and biological embedding, and the full argument in Disparities across the breast cancer continuum.

Interplay

Host variation is a source of heterogeneity that no amount of tumour sequencing will find. Two patients with the same tumour genotype can differ in circulating insulin, in marrow clonality, in sympathetic tone, in gut community composition and in the timing of their disease relative to a pregnancy. Each of those changes either what the tumour experiences or what the treatment does. The heterogeneity framework in Part VI is built around variation within and between tumours. This chapter argues that variation between hosts belongs on the same axis, and that some of it is measurable and some of it is modifiable.

See Integrative biological interplay

References

  1. Key T, Appleby P, Barnes I, Reeves G. Endogenous sex hormones and breast cancer in postmenopausal women: reanalysis of nine prospective studies. J Natl Cancer Inst 2002 94:606-616. PMID 11959894
  2. Bhaskaran K, Douglas I, Forbes H, et al. Body-mass index and risk of 22 specific cancers: a population-based cohort study of 5.24 million UK adults. Lancet 2014 384:755-765. PMID 25129328
  3. Goodwin PJ, Ennis M, Pritchard KI, et al. Fasting insulin and outcome in early-stage breast cancer: results of a prospective cohort study. J Clin Oncol 2002 20:42-51. PMID 11773152
  4. Morris PG, Hudis CA, Giri D, et al. Inflammation and increased aromatase expression occur in the breast tissue of obese women with breast cancer. Cancer Prev Res (Phila) 2011 4:1021-1029. PMID 21622727
  5. Goodwin PJ, Chen BE, Gelmon KA, et al. Effect of metformin vs placebo on invasive disease-free survival in patients with breast cancer: the MA.32 randomized clinical trial. JAMA 2022 327:1963-1973. PMID 35608580
  6. Sceneay J, Goreczny GJ, Wilson K, et al. Interferon signaling is diminished with age and is associated with immune checkpoint blockade efficacy in triple-negative breast cancer. Cancer Discov 2019 9:1208-1227. PMID 31217296
  7. Jaiswal S, Fontanillas P, Flannick J, et al. Age-related clonal hematopoiesis associated with adverse outcomes. N Engl J Med 2014 371:2488-2498. PMID 25426837
  8. Coombs CC, Zehir A, Devlin SM, et al. Therapy-related clonal hematopoiesis in patients with non-hematologic cancers is common and associated with adverse clinical outcomes. Cell Stem Cell 2017 21:374-382. PMID 28803919
  9. Bolton KL, Ptashkin RN, Gao T, et al. Cancer therapy shapes the fitness landscape of clonal hematopoiesis. Nat Genet 2020 52:1219-1226. PMID 33106634
  10. Razavi P, Li BT, Brown DN, et al. High-intensity sequencing reveals the sources of plasma circulating cell-free DNA variants. Nat Med 2019 25:1928-1937. PMID 31768066
  11. Nichols HB, Schoemaker MJ, Cai J, et al. Breast cancer risk after recent childbirth: a pooled analysis of 15 prospective studies. Ann Intern Med 2019 170:22-30. PMID 30534999
  12. Goddard ET, Bassale S, Schedin T, et al. Association between postpartum breast cancer diagnosis and metastasis and the clinical features underlying risk. JAMA Netw Open 2019 2:e186997. PMID 30646210
  13. Lyons TR, O'Brien J, Borges VF, et al. Postpartum mammary gland involution drives progression of ductal carcinoma in situ through collagen and COX-2. Nat Med 2011 17:1109-1115. PMID 21822285
  14. Kwa M, Plottel CS, Blaser MJ, Adams S. The intestinal microbiome and estrogen receptor-positive female breast cancer. J Natl Cancer Inst 2016 108:djw029. PMID 27107051
  15. Nejman D, Livyatan I, Fuks G, et al. The human tumor microbiome is composed of tumor type-specific intracellular bacteria. Science 2020 368:973-980. PMID 32467386
  16. Fu A, Yao B, Dong T, et al. Tumor-resident intracellular microbiota promotes metastatic colonization in breast cancer. Cell 2022 185:1356-1372. PMID 35395179
  17. Narunsky-Haziza L, Sepich-Poore GD, Livyatan I, et al. Pan-cancer analyses reveal cancer-type-specific fungal ecologies and bacteriome interactions. Cell 2022 185:3789-3806. PMID 36179670
  18. Gihawi A, Ge Y, Lu J, et al. Major data analysis errors invalidate cancer microbiome findings. mBio 2023 14:e0160723. PMID 37811944
  19. Travis RC, Balkwill A, Fensom GK, et al. Night shift work and breast cancer incidence: three prospective studies and meta-analysis of published studies. J Natl Cancer Inst 2016 108:djw169. PMID 27758828
  20. Diamantopoulou Z, Castro-Giner F, Schwab FD, et al. The metastatic spread of breast cancer accelerates during sleep. Nature 2022 607:156-162. PMID 35732738
  21. Sloan EK, Priceman SJ, Cox BF, et al. The sympathetic nervous system induces a metastatic switch in primary breast cancer. Cancer Res 2010 70:7042-7052. PMID 20823155
  22. Hiller JG, Cole SW, Crone EM, et al. Preoperative beta-blockade with propranolol reduces biomarkers of metastasis in breast cancer: a phase II randomized trial. Clin Cancer Res 2020 26:1803-1811. PMID 31754048
  23. Holmes MD, Chen WY, Feskanich D, Kroenke CH, Colditz GA. Physical activity and survival after breast cancer diagnosis. JAMA 2005 293:2479-2486. PMID 15914748
  24. Courneya KS, Vardy JL, O'Callaghan CJ, et al. Structured exercise after adjuvant chemotherapy for colon cancer. N Engl J Med 2025 393:13-25. PMID 40450658
  25. Chlebowski RT, Aragaki AK, Anderson GL, et al. Dietary modification and breast cancer mortality: long-term follow-up of the Women's Health Initiative randomized trial. J Clin Oncol 2020 38:1419-1428. PMID 32031879
  26. Prado CMM, Baracos VE, McCargar LJ, et al. Sarcopenia as a determinant of chemotherapy toxicity and time to tumor progression in metastatic breast cancer patients receiving capecitabine treatment. Clin Cancer Res 2009 15:2920-2926. PMID 19351764
  27. Sestak I, Distler W, Forbes JF, Dowsett M, Howell A, Cuzick J. Effect of body mass index on recurrences in tamoxifen and anastrozole treated women: an exploratory analysis from the ATAC trial. J Clin Oncol 2010 28:3411-3415. PMID 20547990
  28. Henricks LM, Lunenburg CATC, de Man FM, et al. DPYD genotype-guided dose individualisation of fluoropyrimidine therapy in patients with cancer: a prospective safety analysis. Lancet Oncol 2018 19:1459-1467. PMID 30348537
  29. Rae JM, Drury S, Hayes DF, et al. CYP2D6 and UGT2B7 genotype and risk of recurrence in tamoxifen-treated breast cancer patients. J Natl Cancer Inst 2012 104:452-460. PMID 22395643
  30. McEwen BS. Protective and damaging effects of stress mediators. N Engl J Med 1998 338:171-179. PMID 9428819
  31. Obeng-Gyasi S, Elsaid MI, Lu Y, et al. Association of allostatic load with all-cause mortality in patients with breast cancer. JAMA Netw Open 2023 6:e2313989. PMID 37200034
  32. Kroenke CH, Kubzansky LD, Schernhammer ES, Holmes MD, Kawachi I. Social networks, social support, and survival after breast cancer diagnosis. J Clin Oncol 2006 24:1105-1111. PMID 16505430
  33. Cacioppo JT, Cacioppo S, Capitanio JP, Cole SW. The neuroendocrinology of social isolation. Annu Rev Psychol 2015 66:733-767. PMID 25148851