Endochondral ossification

The target of the anabolic drive for linear growth of bones, which directly influence height (limb bones, ribs and vertebrae), is endochondral ossification within the growth plate⁽ Reference Kronenberg ^{28 –} Reference Lui, Nilsson and Baron ^{32 )}. This starts with the differentiation cascade initiated by stem cell clonal expansion as proliferative chondrocytes, followed by hypertrophy, cartilage matrix secretion and apoptosis which releases angiogenic factors that stimulate vascular invasion and migration of osteoblasts and osteoclasts, leading to remodelling of calcified cartilage and formation of trabecular bone. Synchronously, the diameter of the long bone diaphysis increases by osteoblastic deposition of cortical bone beneath the periosteum, and the marrow cavity expands as a consequence of osteoclastic bone resorption at the endosteal surface. This ordered process mediates linear growth until adulthood. Progression of endochondral ossification and linear growth is regulated by the combined influence of those systemic endocrine and local paracrine/autocrine anabolic influences⁽ Reference Lui, Nilsson and Baron ^{32 )}, acting together with small molecules which include specific amino acids such as leucine⁽ Reference Millward ^{33 )}, and Zn²⁺, which mediate a signalling cascade via a variety of pathways. Linear growth continues until the growth plates fuse during puberty, but bone mineralisation and consolidation of bone mass continues until peak bone mass is achieved during the third to fourth decade. Considerable progress has been made in unravelling the marked complexity of the regulation of the growth plate⁽ Reference Long and Ornitz ^{29 )}. Here the main features are briefly summarised.

Endocrine regulation

The endocrine signalling of the dietary anabolic drive which acts on the growth plate includes insulin and the growth hormone (GH)–insulin-like growth factor (IGF)-1 axis, a mixed endocrine paracrine–autocrine system, the thyroxine/triiodothyronine (T₄/T₃) axis, androgens, oestrogens, vitamin D, glucocorticoids, and possibly leptin⁽ Reference Nilsson, Marino and De Luca ^{34 ,} Reference Wit and Camacho-Hübner ^{35 )}, making for an extremely complex system which is by no means understood and will only be briefly reviewed here. As far as the insulin–GH–IGF-1 axis is concerned, the Karlberg ICP (infancy, childhood and puberty) model of an insulin/IGF-1-driven fetal infancy phase and a GH/IGF-1-driven childhood phase⁽ Reference Karlberg ^{20 )} may need reassessment, with several authors arguing that GH has an important role in fetal growth⁽ Reference Waters and Kaye ^{36 ,} Reference Harvey and Baudet ^{37 )}. It would appear that: (1) GH can be produced within the fetus both from the pituitary and from other fetal tissues⁽ Reference Waters and Kaye ^{36 ,} Reference Harvey and Baudet ^{37 )} and derives to a limited extent from the placenta⁽ Reference Higgins, Russell and Crossey ^{38 )}, the source of most maternal GH during pregnancy⁽ Reference Higgins, Russell and Crossey ^{38 ,} Reference Zeck, Widberg and Maylin ^{39 )}; (2) there is a positive correlation between placental GH levels in pregnant women and the birth weight of their offspring⁽ Reference Zeck, Widberg and Maylin ^{39 –} Reference McIntyre, Serek and Crane ^{41 )}; (3) a functional GH receptor is widespread in the human fetus at all stages of development⁽ Reference Waters and Kaye ^{36 )}, including on growth plate chondrocytes⁽ Reference Werther, Haynes and Waters ^{42 ,} Reference Fernández-Cancio, Audi and Carrascosa ^{43 )}; and (4) GH increases IGF-1 gene expression in human fetal epiphyseal chondrocytes when conditioned by 1,25-dihydroxyvitamin D₃ (1,25(OH)₂D₃)⁽ Reference Fernández-Cancio, Audi and Carrascosa ^{43 )}.

In the postnatal growth plate, insulin, GH and IGF may have overlapping functions. There can be cross-talk between IGF-1 and insulin at the receptor level⁽ Reference Zhang, He and Tsang ^{44 )} and post-receptor signalling is identical⁽ Reference Siddle ^{45 ,} Reference Guntur and Rosen ^{46 )}, involving the Ras/mitogen-activated protein kinase pathway which signals cell proliferation and the PKB/Akt kinase and mammalian target of rapamycin complex 1 (mTORC1) pathway which stimulates cell growth. Also, although the GH receptor is distinctive, some of the GH Janus activated kinase-2 (JAK2)-initiated signalling is similar to insulin/IGF-1 post-receptor signalling⁽ Reference Carter-Su, Schwartz and Argetsinger ^{47 ,} Reference Brooks, Wooh and Tunny ^{48 )}. Because the GH receptor is found on chondrocytes of all zones of the growth plate⁽ Reference Parker, Hegde and Buckley ^{49 )}, GH action is more complex than simply activating expression of IGF-1, bone morphogenetic protein and other genes in resting-zone chondrocytes through the JAK/STAT pathway⁽ Reference Darvin, Joung and Yang ^{50 )}. Also IGF-1, like GH, can stimulate proliferation of resting-zone chondrocytes and chondrocyte hypertrophy⁽ Reference Nilsson, Marino and De Luca ^{34 )}. For IGF-1, both its endocrine action through its mainly hepatic-derived circulating form and its paracrine–autocrine action contributes to postnatal growth⁽ Reference Stratikopoulosa, Szabolcsb and Dragatsisc ^{51 ,} Reference Elis, Courtland and Wu ^{52 )}. However, the distinctive feature of IGF-1 action is its regulation through the IGF-binding proteins (IGFBP) 1–6⁽ Reference Clemmons ^{53 )}. These have a variety of functions, with IGFBP-1, -2, -4 and -6 mainly inhibitory by preventing receptor binding. Also IGFBP-1, which is increased in plasma of protein-deficient rats⁽ Reference Jousse, Bruhat and Ferrara ^{54 )}, increases IGF-1 clearance from the plasma and subsequent catabolism⁽ Reference Thissen, Davenport and Pucilowska ^{55 )}. IGFBP-3 can protect IGF-1 by maintaining a circulating ternary complex with IGF-1 and the acid-labile subunit (ALS), a circulating hepatic-derived glycoprotein. ALS functions to prolong the half-life of the IGF-1, IGFBP-3 (or IGFBP-5) binary complexes⁽ Reference Boisclair, Rhoads and Ueki ^{56 )}. IGFBP-5 can enhance IGF-1 activity, including induction of chondrocyte proliferation⁽ Reference Kiepe, Ciarmatori and Hoeflich ^{57 )}. Both IGFBP-3 and -5 can be made by chondrocytes under IGF-1 control⁽ Reference Kiepe, Ciarmatori and Hoeflich ^{57 )}. Interestingly, in the context of a role for Zn in the growth plate, IGFBP-3 possesses a metal-binding domain⁽ Reference Forbes, McCarthy and Norton ^{58 )}, and Zn²⁺ influences binding of IGF to cell surface-associated IGFBP-3 and -5, possibly affecting IGF bioavailability⁽ Reference McCusker ^{59 )}.

As for thyroid hormones, the thyroid-stimulating hormone (TSH) receptor is expressed in growth plate cartilage and cultured chondrocytes, consistent with evidence of its action to inhibit chondrocyte proliferation⁽ Reference Bassett and Williams ^{31 )}. T₃ is widely involved⁽ Reference Bassett and Williams ^{31 )}. Its indirect effects involve enhancing the direct effects of IGF-1 on cartilage and influencing the levels of GH-binding protein, IGF-1, IGF-2, IGFBP-3 and IGF bioactivity⁽ Reference Williams, Robson and Shalet ^{60 )}. It also acts through thyroid hormone receptor (TR)α1, one of three T₃ nuclear receptors (the other two TRα2 and TRα3 acting as antagonists) to inhibit proliferation and stimulate prehypertrophic and hypertrophic chondrocyte differentiation. In addition, it has a number of non-genomic actions involving various signal transduction pathways⁽ Reference Bassett and Williams ^{31 )}. Thus, overall thyroid hormone is essential for coordinated progression of endochondral ossification, acting to stimulate genes that control chondrocyte maturation and cartilage matrix synthesis, mineralisation and degradation⁽ Reference Bassett and Williams ^{31 )}.

Both androgens and oestrogen contribute to the pubertal growth spurt⁽ Reference Nilsson, Marino and De Luca ^{34 )}. Androgens can stimulate longitudinal bone growth acting directly on growth plate chondrocytes and indirectly by increased local IGF-1 expression⁽ Reference Nilsson, Marino and De Luca ^{34 )}. Androgens also act in boys after aromatisation to oestrogens which signal through the ER-α receptor to mediate the growth spurt directly and, like androgens, through stimulation of the GH–IGF-1 axis. Eventually with skeletal maturation and epiphyseal fusion, the proliferative capacity of the growth plate chondrocytes is exhausted⁽ Reference Nilsson, Marino and De Luca ^{34 ,} Reference Wit and Camacho-Hübner ^{35 )}.

Finally, vitamin D is involved in endochondral ossification, although this is poorly understood in terms of its direct as opposed to indirect influence, and its influence may be subtle. Although chondrocytes contain both the vitamin D receptor (at least in hypertrophic but not proliferating chondrocytes⁽ Reference Wang, Zhu and DeLuca ^{61 )}) and can produce 1,25(OH)₂D, the major skeletal manifestations of vitamin D deficiency, rickets and osteomalacia, can be corrected by increasing the intestinal absorption of Ca and phosphate, indicating the importance of indirect effects⁽ Reference Bikle ^{62 )}. However, direct effects are observed in the human fetal epiphyseal growth plate with 1,25(OH)₂D, like T₃, a potent inhibitor of chondrocyte proliferation but enhancing chondrocyte differentiation in terms of stimulating expression of several GH–IGF axis genes such as IGF-1, IGFBP-3 and GHR⁽ Reference Fernández-Cancio, Audi and Carrascosa ^{43 )}. Vitamin D also participates with parathyroid hormone (PTH) in a functional paracrine feedback loop in the growth plate between 1,25(OH)₂D and PTH-related protein (PTHrP). Thus 1,25(OH)₂D decreases PTHrP production, while PTHrP increases chondrocyte sensitivity to 1,25(OH)₂D by increasing vitamin D receptor production⁽ Reference Bach, Rutten and Hendriks ^{63 )}.

The main inhibitory endocrine influence is cortisol, which is increased in energy deficiency and inflammation and its action is discussed below (see the Inflammation and endochondral ossification section).

Paracrine signalling within the growth plate

The linkage between nutritionally mediated endocrine changes in insulin, GH/IGF-1, T₃ androgens and oestrogens, as described above, and endochondral ossification is complex and not entirely understood. A brief summary of those key factors involved is reported here⁽ Reference Kronenberg ^{28 ,} Reference Long and Ornitz ^{29 ,} Reference Bassett and Williams ^{31 ,} Reference Xie, Zhou and Chen ^{64 )}. Chondrogenesis is initiated by one of the bone morphogenetic protein (BMP) family of signalling molecules, together with the SOX series of transcription factors. Chondrocyte proliferation is directly stimulated by Indian hedgehog, which ensures that sufficient chondrocyte proliferation occurs by increasing PTHrP signalling, which in turn inhibits chondrocyte hypertrophic differentiation. Wingless/integrated protein signalling promotes hypertrophic differentiation, whereas fibroblast growth factors (FGF) FGF18 and 19, acting through FGFR3, inhibit chondrocyte proliferation and differentiation and stimulate their apoptosis. Epidermal growth factor, vascular endothelial growth factor and transforming growth factor (TGF)-β also play a part, most probably with additive effects as with the additive effects of IGF-1 and TGF-β1 observed on human chondrocytes in culture⁽ Reference Seifarth, Csaki and Shakibaei ^{65 )}.

One of the principal apparent functions of the endocrine system is to allow rapid growth only when the organism is able to consume abundant and appropriate nutrients. Thus any GH-driven linear growth can only occur when GH can induce IGF-1 and the various other paracrine growth factors as a result of the appropriate dietary signalling from key nutrients, such as amino acids and Zn, to increase growth factor levels. This anabolic drive is apparent in the growing rat in which growth and protein synthesis rates in muscle and bone reflect circulating levels of insulin, IGF-1 and free T₃, in turn a function of dietary protein intakes⁽ Reference Jepson, Bates and Millward ^{66 –} Reference Yahya, Tirapegui and Bates ^{71 )}, and especially leucine which enhances insulin secretion from islet β-cells as well as having direct tissue effects (see Millward⁽ Reference Millward ^{33 )}). However, in children, length growth is very much slower and saltatory (discontinuous) with 90–95 % of infant development growth-free⁽ Reference Lampl, Veldhuis and Johnson ^{72 )}, and the physiology and endocrinology of this saltatory growth is by no means understood. What is known is that GH secretion, which is pulsatile during the day and night and extremely variable on a day-to-day basis, does appear to relate to height growth velocity, in that synchrony between successive changes in height and GH output is associated with increased stature in healthy children⁽ Reference Gill, Tillmann and Veldhuis ^{73 )}. Furthermore, in infants each saltation appears to follow periods of increased sleep⁽ Reference Lampl and Johnson ^{74 )}, and because from early childhood the onset of sleep is a robust stimulus for GH secretion⁽ Reference Van Cauter and Plat ^{75 )}, this may link sleep quality to length growth. In this case low-quality sleep of children in a poor environment could be part of the aetiology of stunting.

Direct anabolic influences of amino acids and zinc

The extent and nature of direct anabolic influences of amino acids or Zn on endochondral ossification are poorly understood. What is known is that chondrocyte development in cell culture is sensitive to amino acid levels⁽ Reference Ishikawa, Chin and Schalk ^{76 )} and that in a wide variety of cell types, the major pathway through which amino acids influence cell growth and protein synthesis is as obligatory upstream regulators of the mTORC1 signalling pathway, with leucine as the main effector⁽ Reference Millward ^{33 ,} Reference Laplante and Sabatini ^{77 ,} Reference Kimball and Jefferson ^{78 )}. Given that mTORC1 signalling has been reported to promote osteoblast differentiation from pre-osteoblasts in mouse primary calvarial cells⁽ Reference Chen and Long ^{79 )}, it can be expected that similar signal-transduction pathways operate in chondrocytes. Dietary leucine is not likely to be limited by poor protein quality since it occurs in high concentrations in most dietary proteins including cereals⁽ Reference Millward ^{33 )} and its signalling role may simply indicate intake of protein in general. In this context it is the case, as discussed below, that plasma leucine levels are particularly low in stunted children in Malawi⁽ Reference Semba, Shardell and Ashour ^{80 )}.

A role for Zn within the growth plate would be expected given that up to 10 % of human genes code for proteins with Zn-binding domains⁽ Reference Cousins, Liuzzi and Lichten ^{81 )}, with >100 Zn²⁺-dependent enzymes and >2000 Zn²⁺-dependent known transcription factors⁽ Reference Andreini, Banci and Bertini ^{82 )}. Labile Zn²⁺ in extra- and intracellular compartments exerts a wide range of influences on cellular functions with an imbalance in Zn²⁺ homeostasis linked to dysfunction⁽ Reference Cousins, Liuzzi and Lichten ^{81 ,} Reference Prasad ^{83 )}, and a role in growth-related signalling is emerging involving those transporters which mediate intracellular Zn trafficking.

Of the two families of Zn transporters which either reduce (ZnT 1–10) or increase (ZIP 1–14) intracellular Zn, transporter-mediated Zn signalling roles appear to involve the ZIP family⁽ Reference Cousins, Aydemir and Lichten ^{84 –} Reference Hojyo, Fukada and Shimoda ^{87 )}. Thus an up-regulation of ZIP8 in human osteoarthritis cartilage increases chondrocyte intracellular Zn²⁺ levels, activating the metal-regulatory transcription factor 1 (MTF1) which increases matrix-degrading enzymes⁽ Reference Kim, Jeon and Shin ^{85 )}. ZIP13 may be important for bone BMP and TGF-β signalling during chondrocyte maturation on the basis of ZIP13 knockout mice studies and a ZIP13 loss of function mutation in human Ehlers–Danlos syndrome which is associated with short stature⁽ Reference Fukada, Civic and Furuichi ^{86 )}. ZIP14 controls G-protein-coupled receptor (GPCR)-mediated signalling and ZIP14 knock-out mice exhibit growth retardation, attributable to disrupted GPCR signalling in the growth plate and elsewhere⁽ Reference Hojyo, Fukada and Shimoda ^{87 )}. As indicated above, the metal-binding domain of IGFBP-3⁽ Reference Forbes, McCarthy and Norton ^{58 )} allows Zn²⁺ to influence binding of IGF to cell surface-associated IGFBP-3 and consequent IGF bioavailability⁽ Reference McCusker ^{59 )}. Taken together, these examples show that the availability of Zn²⁺ could influence the regulation of growth in diverse ways.

Inflammation and endochondral ossification

Inflammation, which occurs as part of many chronic disease conditions⁽ Reference Wong, Dobie and Altowati ^{88 )}, and to a greater or lesser extent in response to infective insults by bacterial pathogens or immunogenic macromolecules either ingested or translocated across a compromised gut mucosa, is associated with increased and sustained production of pro-inflammatory cytokines, chemokines, adhesion molecules, eicosanoids, NO, oxygen radicals, FGF21⁽ Reference Hanada and Yoshimura ^{89 ,} Reference Turner, Nedjai and Hurst ^{90 )} as well as the activin A–follistatin system⁽ Reference Phillips, de Kretser and Hedger ^{91 ,} Reference Sozzani and Musso ^{92 )}. This is usually coupled to both GH and insulin resistance and a reduction in circulating IGF-1 and IGFBP-3⁽ Reference Wong, Dobie and Altowati ^{88 ,} Reference Sederquist, Fernandez-Vojvodich and Zaman ^{93 )}. Clinically, low-grade inflammation often results in increased circulating hepatic acute-phase reactants, C-reactive protein (CRP) and α-1 acid glycoprotein, elevations in inflammatory markers in stool such as myeloperoxidase, α-1 antitrypsin, and neopterin, and reduced circulating IGF-1 and IGFBP-3 (for example, Prendergast et al. ⁽ Reference Prendergast, Rukobo and Chasekwa ^{94 )} and DeBoer et al. ⁽ Reference DeBoer, Scharf and Leite ^{95 )}). In rat models these responses associated with insults such as endotoxaemia⁽ Reference Jepson, Pell and Bates ^{96 ,} Reference Hasegawa-Ishii, Inaba and Umegaki ^{97 )}, a parasitic infection⁽ Reference Omwega, Bates and Millward ^{98 )} or elevated levels of glucocorticoid⁽ Reference Odedra and Millward ^{99 )} induce a profound insulin resistance⁽ Reference Jepson, Pell and Bates ^{96 )}, and markedly inhibit tissue protein synthesis and growth. Children suffering from specific inflammatory diseases such as inflammatory bowel disease, Crohn’s disease, ulcerative colitis, and juvenile idiopathic arthritis, who exhibit physiologically elevated levels of glucocorticoids and proinflammatory cytokines, usually display abnormal growth patterns as well as delayed puberty⁽ Reference Wong, Dobie and Altowati ^{88 ,} Reference Sederquist, Fernandez-Vojvodich and Zaman ^{93 )}. Furthermore, linear growth can be increased in these patients by the inhibition of these cytokines⁽ Reference Wong, Dobie and Altowati ^{88 ,} Reference Sederquist, Fernandez-Vojvodich and Zaman ^{93 )}.

Key pro-inflammatory cytokines known to inhibit endochondral ossification include TNFα, IL-1 (particularly IL-1β) and IL-6 (in some⁽ Reference Nakajima, Naruto and Miyamae ^{100 )} but not all studies⁽ Reference Wong, Dobie and Altowati ^{88 )}). Each of these signals via its specific type I cytokine receptor which are structurally divergent from other cytokine receptor types. High concentrations of these cytokines suppress growth in a dose-dependent manner by decreasing chondrocyte proliferation and hypertrophy while increasing apoptosis⁽ Reference Sederquist, Fernandez-Vojvodich and Zaman ^{93 )}. TNFα, a potent mediator of the inflammatory action of the innate immune system, is secreted mainly from activated macrophages, and by many other cell types in response to endotoxins⁽ Reference Wajant, Pfizenmaier and Scheurich ^{101 )}. It induces cytokine production, activation and/or expression of adhesion molecules, with additional functions linked with lipid metabolism, coagulation, insulin resistance and endothelial function, and is one of the most important and pleiotropic cytokines, mediating inflammatory and immune responses. TNFα is produced endogenously throughout the growth plate and can inhibit chondrocyte proliferation, especially in combination with IL-1β⁽ Reference Mårtensson, Chrysis and Savendahl ^{102 )}. IL-1β expression is induced mainly in response to microbial molecules and it signals via both mitogen-activated protein kinases and the transcription factor NF-κB, thereby resulting in further pro-inflammatory cytokine expression. IL-1β induces rapid dedifferentiation of chondrocytes in culture⁽ Reference Seifarth, Csaki and Shakibaei ^{65 )} and acts in synergy with TNFα to inhibit rat longitudinal bone growth in culture⁽ Reference Mårtensson, Chrysis and Savendahl ^{102 )}. IL-6 is a pleiotropic cytokine with a wide range of biological activities⁽ Reference Kishimoto ^{103 )} involving haematopoiesis, B-cell maturation and T cell activation, differentiation and regulation, the secretion of acute-phase proteins by the liver, which it does in cooperation with IL-1 while inhibiting hepatic IGF-1 production. Most importantly, IL-6 has direct inhibitory effects on growth-plate chondrocytes⁽ Reference Nakajima, Naruto and Miyamae ^{100 ,} Reference Fernandez-Vojvodich, Zaman and Savendahl ^{104 )}.

Activin A is released rapidly into the circulation during inflammation, in advance of TNFα or IL-6⁽ Reference Phillips, de Kretser and Hedger ^{91 )}. Activin A and its antagonist follistatin are members of the TGF-β superfamily, deriving from a wide range of resting and activated immune and other cell types including bone cells, and modulating several aspects of the inflammatory response, including release of pro-inflammatory cytokines, NO production and immune cell activity⁽ Reference Phillips, de Kretser and Hedger ^{91 )}. Increased circulating levels of activin A occur in inflammatory conditions such as inflammatory bowel disease and rheumatoid arthritis and during bacterial septicaemia, hepatitis C infection, and trauma⁽ Reference Sozzani and Musso ^{92 )}. Less is known about the role of this system in the growth plate, but both activin A and follistatin are produced by human osteoblasts, and are involved in controlling the extent of mineralisation and the prevention of over-mineralisation of bone tissue⁽ Reference Eijken, Swagemakers and Koedam ^{105 ,} Reference Bowser, Herberg and Arounleut ^{106 )}.

FGF21, one of the endocrine-like FGF which is also expressed by chondrocytes, is increased by inflammatory stimuli⁽ Reference Feingold, Grunfeld and Heuer ^{107 )} and by undernutrition in mice⁽ Reference Kubicky, Wu and Kharitonenkov ^{108 )}, and in humans in late-stage fasting⁽ Reference Fazeli, Lun and Kim ^{109 )} and sepsis⁽ Reference Gariani, Drifte and Dunn-Siegrist ^{110 )}, and directly inhibits chondrocyte proliferation and differentiation as well as preventing GH-mediated stimulation of chondrocyte proliferation and differentiation⁽ Reference Wong, Dobie and Altowati ^{88 ,} Reference Kubicky, Wu and Kharitonenkov ^{108 ,} Reference Wu, Levenson and Kharitonenkov ^{111 )}.

Inflammation is also accompanied by elevated levels of cortisol, which inhibits length growth directly in terms of chondrocyte proliferation, hypertrophy and cartilage matrix production⁽ Reference Wong, Dobie and Altowati ^{88 ,} Reference Klein ^{112 )} and by the blockade of anabolic growth factor signalling⁽ Reference Nilsson, Marino and De Luca ^{34 )}. Children treated with corticosteroids for various reasons exhibit impaired growth⁽ Reference Mushtaq and Ahmed ^{113 )}. In the rat, exogenous corticosterone at physiologically elevated plasma levels inhibits longitudinal bone growth within 24 h with total arrest after 4 d, at which time epiphyseal cartilage width had shrunk, proteoglycan synthesis was markedly suppressed but overall protein synthesis was unchanged⁽ Reference Tirapegui, Yayha and Bates ^{70 ,} Reference Yahya, Tirapegui and Bates ^{71 )}. One target appears to be apoptosis of proliferative growth-plate chondrocytes⁽ Reference Chrysis, Zaman and Chagin ^{114 )}, particularly activation of the caspase cascade triggering Bax-mediated mitochondrial apoptosis⁽ Reference Zaman, Chrysis and Huntjens ^{115 )}, as well as suppression of the PKB/Akt kinase IGF-1 signalling pathway⁽ Reference Chrysis, Zaman and Chagin ^{114 )}.

Given that some or all of the proinflammatory cytokines and activin A–follistatin are produced within the growth plate⁽ Reference Sederquist, Fernandez-Vojvodich and Zaman ^{93 ,} Reference Eijken, Swagemakers and Koedam ^{105 ,} Reference Bowser, Herberg and Arounleut ^{106 )}, the implication is that individually they may be involved in normal bone growth regulation with their inhibitory effect resulting from their combined effects and/or at high concentrations. The key question which remains is the profile and thresholds in the changes from normal to increased levels of these various mediators induced by low-grade inflammation, which does inhibit linear growth.

Type 1 nutrients

Of the type 1 nutrients, iodine deficiency does influence linear growth, as discussed below. For Fe and vitamin A, intervention studies suggest that linear growth is not affected by their individual deficiency⁽ Reference Ramakrishnan, Nguyen and Martorell ^{118 )}. For the bone minerals, it is usually assumed that low Ca intakes are not responsible for stunting. In animal models, while Ca deficiency results in low bone mineralisation and reduced bone strength, it does not result in reduced bone length⁽ Reference Chen, Hayakawa and Emura ^{119 )}. It has been suggested that slow growth rates of children in many developing countries may represent an adaptation to limited Ca supply and poor bioavailability even after adaptation of losses to match low intakes. However, where these low intakes occur, case reports indicate biochemical signs of hyperparathyroidism, low bone mineral contents, and rickets even in the presence of adequate vitamin D status⁽ Reference Prentice and Bates ^{120 )}. Ca supplementation studies in developing countries mostly found no significant effects on linear growth retardation and studies of bone mineral acquisition in younger children reported no height effects⁽ Reference Prentice, Ginty and Stear ^{121 )}. Interestingly, Ca supplementation for 13 months in adolescents which did increase bone mineral content, increased height in boys but not girls⁽ Reference Stear, Prentice and Jones ^{122 )}, possibly through an interaction between the sex hormone systems and GH/IGF⁽ Reference Prentice, Ginty and Stear ^{121 )}.

Type 2 nutrients

Type 2 nutrients, of which linear growth inhibition is an immediate response to their deficiency, include protein (specifically N and indispensable amino acids), Zn, P and the main electrolytes K, Na and Mg. Of these, there is evidence that deficiencies of protein and Zn can occur in the human diet, especially for populations consuming diets based on starchy roots with little or no animal-source foods (ASF); their potential role in the nutritional aetiology of stunting is discussed below. There is some limited evidence for phosphate deficiency occurring through a dietary lack⁽ Reference Waterlow ^{23 )}, although phosphate deficiency is likely to be very rare. Indeed, average intakes of P and Mg from most diets are usually substantially greater than their biological requirements and, if not, increased absorption and conservation of their endogenous losses may prevent an inadequate dietary supply contributing to the poor linear growth of Third World children⁽ Reference Prentice and Bates ^{120 )}.

Multiple nutrient deficiencies and growth

Limitations in the type 1–type 2 classification need to be recognised. First, some type 1 nutrients can influence growth in situations where the symptoms of their deficiency may be missed, one example being mild iodine deficiency⁽ Reference Zimmermann ^{123 –} Reference Zimmermann, Jooste and Mabapa ^{128 )}. Second, endochondral ossification is a complex system involving multiple hormonal and small molecule signalling systems responding to multiple dietary factors. This means that multiple type 1 and type 2 nutrient deficiencies may induce growth effects through interactions of their deficiencies which would only be identifiable by multiple nutrient supplementation studies. Finally the type 1/type 2 concept was developed in relation to postnatal growth and is probably less relevant to fetal growth where maternal deficiencies of many nutrients including those involved in cell division such as folate and vitamin B₁₂ can influence fetal development. In this context it is not surprising, as discussed below, that multiple micronutrient supplementations of pregnant women have been shown to be beneficial for birth outcomes including birth weight.

Protein and energy deficiency

Because animal growth is usually rapid, it is very sensitive to dietary protein intake and quality, reflecting a metabolic demand for protein almost entirely driven by deposition in lean tissue. The protein efficiency ratio (PER) rat growth assay became the industry-standard method of assessing dietary protein quality⁽ Reference Sarwar, Peace and Botting ^{129 )}, until it was recognised that the assay markedly overestimated the importance of protein quality in the human diet⁽ Reference Millward ^{130 )}. This means that while rat growth studies enable the mechanisms of the dietary activation of growth to be evaluated, they are much less likely to indicate how variation in dietary protein quality influences linear growth in children with a quite different metabolic demand⁽ Reference Millward ^{131 )}.

The concept of the anabolic drive⁽ Reference Millward and Rivers ^{18 )} and the protein-stat model of dietary protein-mediated growth control⁽ Reference Millward ^{27 )} was developed from a detailed study of the nutritional and endocrinological regulation of muscle and bone length growth in the rat⁽ Reference Jepson, Bates and Millward ^{66 ,} Reference Yahya, Bates and Millward ^{67 ,} Reference Yahya and Millward ^{69 –} Reference Yahya, Tirapegui and Bates ^{71 )}. In response to protein deficiency, slowed tibial length growth was apparent by 3 d⁽ Reference Yahya and Millward ^{69 )}, and this was graded according to the extent of dietary protein deficiency, ceasing with a near-protein-free diet. Tibial length growth was restored in a graded way with increasing dietary protein concentration over a wide range of protein intakes.

As for energy deficiency, while food restriction has long been known to inhibit linear growth⁽ Reference Dickerson and McCance ^{132 )}, the specific effects of energy deficiency, independently from protein deficiency, are more difficult to evaluate. Energy restriction in young rats to 25 % of normal intakes or total starvation induced immediate weight loss but maintained tibial length growth over 4 d, albeit at reduced rates⁽ Reference Tirapegui, Yayha and Bates ^{70 )}, although corticosterone treatment inhibited tibial growth immediately. Progressive food restriction (75, 50 and 25 % ad libitum) of diets containing increasing concentrations of protein, to maintain constant protein intakes, arrested body-weight growth in all groups, with weight loss in the severely restricted rats, but allowed some tibial length growth to continue at all levels of energy restriction with only a slow development of inhibition. Taken together, these experiments strongly support a dietary control mechanism for linear growth in which protein intake is the most powerful macronutrient influence, acting to promote high circulating levels of insulin, IGF-1 and free T₃, which together with amino acids act on the growth plate to optimise endochondral ossification at the level of protein and proteoglycan synthesis maintaining both ribosomal capacity (RNA:protein ratio) and activity (protein synthesis:RNA)⁽ Reference Jepson, Bates and Millward ^{66 ,} Reference Yahya, Bates and Millward ^{67 ,} Reference Yahya and Millward ^{69 ,} Reference Yahya, Tirapegui and Bates ^{71 )}. The inhibition of length growth by energy deficiency or corticosteroids was dissociated from and preceded inhibition of protein synthesis and ³⁵S uptake, possibly reflecting an elevation of rates of proteolysis and proteoglycan degradation associated with elevated corticosterone levels.

Zinc deficiency

Notwithstanding the importance of Zn for pre- and postnatal development⁽ Reference Hambidge and Krebs ^{133 )}, evaluating the mechanism of its actions in growth control has proved remarkably difficult. Weight and length growth inhibition is the immediate response to Zn deficiency in the rat and its influence on catch-up growth in infants on a high-energy soya-based formula, which slowed unless additional Zn was added, resulted in the analogy of Zn with an essential amino acid⁽ Reference Golden and Golden ^{134 )}, given the large number of Zn²⁺-dependent enzymes and transcription factors in tissues. Although some very limited initial muscle growth occurs in the severely Zn-deficient (ZD) rat, a normal Zn concentration is maintained, expanding the muscle Zn pool, by drawing on Zn mobilised from bone⁽ Reference Giugliano and Millward ^{135 ,} Reference Rossi, Migliaccio and Corsi ^{136 )}. However, the key problem of animal studies of ZD is to allow for the characteristic inhibition of appetite, i.e. cyclic changes in food intake and associated body weight⁽ Reference Giugliano and Millward ^{135 ,} Reference Williams and Mills ^{137 )}. Careful pair-feeding studies indicated that ZD specifically inhibited body weight and muscle growth⁽ Reference Giugliano and Millward ^{135 )}, through both a corticosteroid-induced blockade due to anorexia and reduced food intake, and an impaired insulin response to food intake when appetite returns, with insufficient insulin to fully activate the translational phase of protein synthesis⁽ Reference Giugliano and Millward ^{138 )}. This indicates a Zn deficiency-induced impairment of the anabolic drive at a key initial step of insulin production. However, in terms of linear growth, while 28 d of a very-low-Zn (ZD) diet⁽ Reference Rossi, Migliaccio and Corsi ^{136 )} induced marked differences between ZD and pair-fed animals in bone Zn and bone histology, the reductions in femur weight and length, in circulating IGF-1, and in thicknesses of both the overall growth plate and of the hypertrophic cartilage, differed little between ZD and pair-fed animals. When food intake and serum IGF-1 levels were maintained in ZD rats given megestrol acetate, a synthetic progestin used clinically to correct anorexia, growth inhibition persisted, a relatively unambiguous demonstration of the inhibition of the anabolic signalling by IGF-1 with Zn deficiency⁽ Reference Browning, MacDonald and Thornton ^{139 )}. Subsequent studies with ZD rats⁽ Reference MacDonald, Wollard-Biddle and Browning ^{140 )} demonstrated the need for Zn in the IGF-1 mediation of cell division in cultured 3T3 cells, possibly at the level of a specific step in Ca-dependent mitogenic signal transduction through the IGF-1 receptor protein kinase C (PKC) pathway⁽ Reference MacDonald ^{141 )}, PKC being a Zn metalloenzyme. It remains to be seen whether similar Zn effects explain Zn’s role in regulating endochondral ossification, and/or whether low growth-plate Zn²⁺ levels decrease chondrocyte IGF bioavailability through impaired surface-associated IGFBP-3 function given its metal-binding domain⁽ Reference Forbes, McCarthy and Norton ^{58 ,} Reference McCusker ^{59 )}.

Iodine deficiency

Although dietary iodine is an essential nutrient for maintenance of the T₄/T₃ system, animal models of iodine deficiency are problematic. This is because of a well-documented conservation of iodine at the level of T₃. Thus rats fed an iodine-deficient diet for long periods⁽ Reference Riesco, Taurog and Larsen ^{142 )}, or for two generations⁽ Reference Moreno-Reyes, Egrise and Boelaert ^{143 ,} Reference Ren, Guo and Zhang ^{144 )}, or pigs fed a goitrogenic diet⁽ Reference Spiegel, Bestetti and Rossi ^{145 )}, maintain plasma T₃ and some growth although symptoms of hypothyroidism can be observed⁽ Reference Riesco, Taurog and Larsen ^{142 )}. T₃ conservation is thought to involve increased iodine uptake into the thyroid, increased T₃ secretion from the thyroid and continued extrathyroidal T₄ to T₃ conversion⁽ Reference Zimmermann and Boelaert ^{124 )}. Because of this, early rat studies involved thyroidectomy to lower T₃ levels and such studies showed the slow growth was associated with a reduction in DNA-dependent RNA polymerase activity⁽ Reference Tata and Widnell ^{146 )}, and a loss of the capacity for tissue protein synthesis in terms of ribosomal RNA⁽ Reference Brown, Bates and Holliday ^{147 )}. T₃ administration by osmotic minipump revealed a dose–response of growth, muscle protein synthesis and ribosomal RNA to circulating T₃ levels⁽ Reference Brown and Millward ^{148 )}, similar to the relationship between plasma T₃ levels and muscle ribosomal RNA observed when growth was manipulated by protein deficiency⁽ Reference Jepson, Bates and Millward ^{66 )}. Since these early animal studies, studies in GM mice have been the main source of information on the complex physiological relationship between centrally regulated thyroid status and the peripheral anabolic actions of T₃ on the growth plate outlined above⁽ Reference Bassett and Williams ^{31 )}.

Energy intake

The introduction of the term ‘protein–calorie malnutrition’ in the 1960s resulted in a debate into the extent of a protein, as opposed to a food or energy gap⁽ Reference Hegsted ^{157 )}, and this prompted intervention studies. Undernourished Indian children, consuming a cereal-based diet with some buffalo milk, judged to provide sufficient protein but inadequate energy were given an energy-rich low-protein supplement (310 kcal (1300 kJ) and 3 g protein/d)⁽ Reference Gopalan, Swaminathan and Kumari ^{158 )}. The intervention increased both height and weight gain over 14 months, enabling growth on the 50th percentile of American children. Furthermore, supplementation protected the linear growth of the children from an outbreak of measles during the study, whereas the measles outbreak depressed height gain and induced some weight loss of the non-supplemented children.

A similar conclusion that energy intake was the limiting factor for linear growth was reached following the INCAP longitudinal intervention study of Guatemalan mothers, infants and severely stunted children in 1969–1977⁽ Reference Habicht, Martorell and Rivera ^{159 )}. The provision of either a good-quality high-protein (protein:energy ratio 28 %) supplement (Atole), or a protein-free low-energy supplement (Fresco) as control, each with some micronutrients, resulted in a greater increase in height with Atole than Fresco at 3 years. However, analysis of variation in the actual intakes of the two supplements consumed by both pregnant women and children showed that energy, not protein, intake predicted both birth weight and height at 3 years of age⁽ Reference Habicht, Martorell and Rivera ^{159 )}. This was consistent with the background diets of this community (mainly maize and beans, with tomatoes, sorghum and cassava)⁽ Reference Martorell, Habicht and Rivera ^{160 )} which provided sufficient protein but inadequate energy intakes. Allen has argued⁽ Reference Allen ^{161 )} that their very low energy intake explains why Guatemalan children were more stunted that those from Egypt, Mexico and Kenya studied in the Nutrition Collaborative Research Support Program (CRSP). Overall, the Atole-supplemented children showed only small benefits of supplementation and remained severely stunted even though many of them consumed more energy and much more protein than their requirements⁽ Reference Martorell and Klein ^{154 ,} Reference Habicht, Martorell and Rivera ^{159 )}. Given their impoverished environment, it is possible that enteric dysfunction as discussed below was an additional important factor limiting growth.

Krebs et al. ⁽ Reference Krebs, Mazariegos and Chomba ^{162 )} suggested the possibility of energy insufficiency in their meat intervention study in stunted infants and toddlers in which stunting rates increased during the trial (see below under Animal-source foods and linear growth). They gave 293–439 kJ/d (70–105 kcal/d), a very modest amount compared with their energy requirements of up to 3350 kJ/d (800 kcal/d).

Malcolm’s studies of the stunted Bundi children in a mission school in Papua New Guinea⁽ Reference Malcolm ^{163 –} Reference Malcolm ^{165 )} included the testing of energy deficiency. These children and the adults were very severely stunted consuming a nutrient-poor, low-protein-quality diet of mainly sweet potato and taro. Given the bulk and low energy density of this low-fat diet, an energy insufficiency was possible, especially for children in the mission school, who had a restricted three meals/d schedule. They were more stunted than village children who had an unrestricted access to food throughout the day. There was a small but significant increase in height growth with 60 % more energy and protein when the children ate five rather than three meals per d, but no stimulation of height growth with extra energy as margarine, although there was a marked increase in skinfold thickness. This suggests that energy intake was not the limiting factor for height growth in these children. As discussed below (Animal-source foods and linear growth section), a skimmed milk supplement had the greatest effect on height. Malcolm argued that there were few signs of overt deficiency or morbidity other than stunting, with malaria rates low at 3–7 %. However, diarrhoea and dysentery did occur, accounting for a quarter of the admissions to the mission hospital, and given this evidence of infection, Malcolm’s assertion that ‘It is unlikely that disease is a significant factor in the slow linear-growth rates in Bundi’ has been challenged⁽ Reference Garn ^{166 )}.

Taken together, these studies do indicate that inadequate energy intakes can occur and that in some circumstances the energy deficiency is associated with stunting.

Protein and amino acid intake

Contrary to popular belief, the dietary protein requirement of young children is low in terms of the protein:energy ratio of the requirement: the safe level is about 5 % energy for healthy preschool children^{( 167 ,} Reference Millward and Jackson ^{168 )}. This means that with the exception of some very low-protein starchy-root staples like plantain, cassava, taro and sweet potato, most diets and especially cereal-based diets provide more than adequate amounts of protein if sufficient is consumed to meet the energy requirement. The challenge is to define the lower limits of the range of protein:energy ratios of intakes which ensures that the genotype in relation to stature can be expressed.

During exclusive breast feeding, length growth is generally considered to be optimal with protein intakes much lower (about 6–7 % protein:energy) than after weaning and at later stages of childhood. Furthermore, the linear growth of healthy, breast-fed or formula-fed infants in the USA during the first year of life does not differ and is independent of protein intake for both groups⁽ Reference Heinig, Nommsen and Peerson ^{169 )}. No difference in length growth in infants fed either breast milk or experimental formulae was observed with protein intakes either higher than or similar to breast milk⁽ Reference Räiha, Fazzolari-Nesci and Cajozzo ^{170 )}. Whether this means that during infancy linear growth is insensitive to dietary signals or that the threshold for dietary proteins’ influence is below the lowest level of intake observed in healthy breast- or formula-fed infants is not known. Current regulatory advice for infant formula in Europe^{( 171 )} is for a minimum protein content of 1·8 g/100 kcal (4·184 kJ) (protein:energy ratio=0·072) for cows’ and goats’ milk-based formulae (of which protein quality and digestibility are assumed to match breast milk (which is the case)⁽ Reference Macé, Steenhout and Klassen ^{172 )}), and 2·25 g/100 kcal (4·184 kJ) (protein:energy ratio=0·09) for formula based on isolated soya protein because of a slightly lower quality (lower sulfur amino acid content) and digestibility.

In developed societies the change to a diet based on family foods at weaning involves a dramatic increase in protein intake to levels considerably in excess of requirements, for example, 13 % protein energy in Danish infants at 9 months⁽ Reference Hoppe, Molgaard and Thomsen ^{173 )} with even higher values maintained in many Nordic countries up to puberty⁽ Reference Hörnell, Lagström and Lande ^{174 )}. The consequences of this increase in protein intakes for linear growth and whether any influence is desirable is a complicated issue. A systematic review of protein intakes and growth of children in the Nordic countries up to puberty⁽ Reference Hörnell, Lagström and Lande ^{174 )} concluded that a higher protein intake in infancy and early childhood is convincingly associated with increased growth and higher BMI in childhood and possibly related to an undesirable earlier puberty. However, some of the studies quoted in support of this indicated no relationship between protein intake at early childhood and body fatness at 10 years expressed as body fat percentage or BMI⁽ Reference Hoppe, Molgaard and Thomsen ^{173 )}. Also, in healthy Danish preschool children, height was positively associated with protein intake from milk but not from meat⁽ Reference Hoppe, Udam and Lauritzen ^{175 )}, a possible example of the ‘milk effect’ on growth as discussed below. Thus the issue of excess protein intakes in early childhood and obesity in late childhood and adult life requires further clarification.

As far as protein quality is concerned, it is not clear how well linear growth can occur in the best circumstances for cereal-based diets with minimal ASF, given the likely dietary limitations of micronutrients and minerals, as well as the suboptimal protein quality. The high protein:energy ratio of cereals, especially wheat, means that their lysine limitation is to some extent offset by the higher protein supply, although digestibility may limit the protein quality of cereals like millet⁽ Reference Millward and Jackson ^{168 )}. In fact, the largely cereal-based diets reported for the rural, urban, slum and tribal communities in India, which exhibit high prevalence rates of stunting, appear adequate in terms of protein and lysine intakes⁽ Reference Swaminathan, Vaz and Kurpad ^{176 )}. Although countries with the lowest protein and lysine intakes on the basis of FAO food balance sheets have the highest prevalence rates of stunting⁽ Reference Ghosh, Suri and Uauy ^{177 )}, they also have the lowest gross domestic product per capita and highest infant mortality rates, indicating a poor environment. This means that it is unwise to suggest, as these authors did, a causal relationship between stunting and protein adequacy and quality of the adult diets.

The older literature does include some lysine supplementation studies, in one case examining child growth and bone density⁽ Reference Mack, Vose and Kinard ^{178 )}. Lysine supplementation of the diets of sub-adolescent orphanage children for 5 months increased growth and bone density compared with supplemented controls. However, total lysine intakes were >5 times their requirement. Thus the changes were most probably pharmacological effects of lysine which is known to increase Ca absorption⁽ Reference Civitelli, Villareal and Agnusdei ^{179 )}.

There is evidence that improving the protein quality of maize can increase height growth. Zein, the main storage protein of maize, is low in lysine and tryptophan, hence its association with pellagra (niacin deficiency). Maize varieties with good agronomic properties have been developed by selective breeding with much higher levels of lysine and tryptophan. One variety is quality protein maize (QPM) in which the protein content is unchanged but the concentrations of lysine, tryptophan, the sulfur amino acids and threonine are increased by 50, 75, 90 and 40 %, respectively. The adequacy of QPM protein content and quality for child linear growth was demonstrated in rapidly growing infants recovering from malnutrition⁽ Reference Graham, Lembcke and Morales ^{180 )}. When fed as the sole source of dietary protein, fat and energy (but with micronutrient and mineral supplementation), it enabled height growth comparable with that observed with a cows’ milk formula⁽ Reference Graham, Lembcke and Morales ^{180 )}. Although the introduction of QPM has been relatively slow, a recent meta-analysis of community-based studies in maize-eating populations⁽ Reference Nilupa, Gunaratna and De Groote ^{181 )} showed modest positive effects of QPM on both weight gain (12 % increase) and height gain (9 % increase) compared with conventional maize in infants and young children with mild to moderate undernutrition.

Serum concentrations of most (sixteen out of nineteen) amino acids were low in stunted preschool children from rural Malawi with no evidence of severe acute malnutrition, compared with non-stunted children⁽ Reference Semba, Shardell and Ashour ^{80 )}. Leucine and the other essential amino acids were particularly low⁽ Reference Semba, Shardell and Ashour ^{80 )}. While this suggests that dietary protein intake was low, without any indication of the timing of the blood sampling in relation to their last meal it would be unwise to over-interpret the findings.

Taken together, these studies provide some suggestion that height growth in children is sensitive to dietary protein and specifically dietary indispensable amino acids intakes as in the animal models, but with the exception of studies with milk protein (discussed below) any influences are small and some of the evidence is indirect. On the other hand, sufficient or even excess dietary protein intake may not be able to prevent the stunting observed in poor communities. Thus the Nutrition Collaborative Research Support Programme (CRSP) found that linear growth faltering was prevalent in preschool children in Egypt, Kenya and Mexico, even though their intakes of protein and essential amino acids were adequate⁽ Reference Beaton, Calloway and Murphy ^{182 )}.

Zinc intake

The prevalence of Zn deficiency is widely believed to be widespread in developing countries because of low intakes of Zn-rich animal products, diets high in phytates, which inhibit Zn absorption, and Zn losses due to diarrhoea. However, few nationally representative surveys of Zn status⁽ Reference de Benoist, Darnton-Hill and Davidsson ^{183 )} or intakes have been conducted in low-income countries, and clear evidence of its role in the aetiology of stunting has been difficult to demonstrate. Recently, Zn deficiency prevalence has been modelled from both country-specific absorbable Zn content of the national food supply (from national food balance sheet data), and estimates of physiological requirements for absorbed Zn⁽ Reference Wessells and Brown ^{184 )}, finding 17·3 % of the world’s population to be at risk of inadequate Zn intake, with estimates ranging from 7·5 % in high-income regions to 30 % in South Asia. Overall, the prevalence of inadequate Zn intake correlated with the prevalence of stunting in children under 5 years of age, explaining almost a quarter of the between-country variation in stunting.

As for Zn supplementation trials, evidence for their efficacy in terms of pregnancy and birth weight is very mixed, with the majority of studies identifying no influence⁽ Reference Allen and Gillespie ^{153 )}, and Zn supplementation during pregnancy is not included in the most recent Lancet series on Maternal and Child Undernutrition⁽ Reference Bhutta, Das and Rizvi ^{185 )}. For children, benefits of Zn supplementation include reductions in the duration and severity of both diarrhoea and dysentery and acute lower respiratory infections⁽ Reference Allen and Gillespie ^{153 ,} Reference Yakoob, Theodoratou and Jabeen ^{186 )}, all of which may indirectly influence height growth. As for linear growth, results of randomised controlled trials of Zn supplementation for children have been mixed, as might be expected given the limitations of Zn supplementation trials: i.e. difficulty of discerning baseline status in supplemented populations, low bioavailability of supplements, low adherence and side effects, etc. Thus although a 1997 meta-analysis indicated a small, but highly significant, impact of Zn supplements in stunted children⁽ Reference Brown, Peerson and Allen ^{187 )}, more recent meta-analyses have found no influence⁽ Reference Ramakrishnan, Nguyen and Martorell ^{118 )}, a significant positive effect in children under 5 years of age⁽ Reference Imdad and Bhutta ^{188 )}, and, in the case of a Cochrane review⁽ Reference Mayo-Wilson, Junior and Imdad ^{189 )}, moderate-quality evidence of a very small improvement in height in children aged 6 months to 12 years of age which was unlikely to be clinically important.

The small-quantity lipid-based nutrient supplements (discussed below) contain Zn and these have not been shown to be effective in Malawi⁽ Reference Phuka, Maleta and Thakwalakwa ^{190 –} Reference Ashorn, Alho and Ashorn ^{192 )} or in Ghana except as part of a Nutributter supplement⁽ Reference Adu-Afarwuah, Lartey and Brown ^{193 )}. Overall, it seems that for Zn, the initial indications that this could easily be remedied with Zn supplements⁽ Reference Brown, Peerson and Allen ^{187 )} have yet to be achieved. Nevertheless, it is probably appropriate that in the most recent Lancet Series on Maternal and Child Undernutrition⁽ Reference Bhutta, Das and Rizvi ^{185 )}, preventative Zn supplementation in children is included as an intervention in the Lives Saved Tool modelling, acting both on diarrhoea incidence and stunting to reduce mortality.

Iodine intake

Iodine deficiency is seldom included in discussions of the nutritional aetiology of stunting. Yet even after the widespread introduction of salt-iodination programmes, 2 billion individuals globally may have an insufficient iodine intake, especially in South Asia and sub-Saharan Africa, and 241 million (30 %) of school-age children globally are estimated to have insufficient iodine intakes⁽ Reference Zimmermann ^{123 ,} Reference Andersson, Karumbunathan and Zimmermann ^{125 )}.

The health consequence of severe iodine deficiency for human growth and development is clear in terms of delayed physical development through growth arrest and delayed bone maturation, and impaired mental function⁽ Reference Zimmermann and Boelaert ^{124 ,} Reference Rohner, Zimmermann and Jooste ^{126 )}. An insufficient iodine supply can be aggravated by goitrogens (glucosinolates, cyanogenic glucosides and some flavonoids) in many plant foods and unclean drinking water, which interfere with thyroid metabolism⁽ Reference Zimmermann ^{123 ,} Reference Rohner, Zimmermann and Jooste ^{126 )}. Deficiencies of Se, Fe and vitamin A exacerbate the effects of iodine deficiency. Given the importance of the T₄/T₃ axis for growth and development, with hypothyroidism decreasing circulating IGF-1 and IGFBP-3 levels, and thyroid hormone replacement increasing them⁽ Reference Miell, Zini and Quin ^{194 ,} Reference Iglesias, Bayon and Mendez ^{195 )} and the fact that severe iodine deficiency results in severe linear-growth retardation, it might be expected that growth faltering would occur in otherwise asymptomatic iodine deficiency. In fact, the evidence for this is scarce. Adaptive changes in iodine and thyroid hormone metabolism occur with iodine deficiency, initially conserving iodine within thyroid tissue in association with increased TSH, and maintaining T₄ and T₃ levels (subclinical hypothyroidism), eventually followed by reductions in T₄ (overt hypothyroidism). Thus, in both severely iodine-deficient and moderately iodine-deficient children, as indicated by their urinary iodine concentration, median T₄ concentrations can be in the low–normal range, with only a minority identifiable as hypothyroxinaemic⁽ Reference Zimmermann, Jooste and Mabapa ^{128 )}. Serum total and free T₃ concentrations may not decrease until disease is far advanced, because increased TSH concentrations stimulate T₃ release from the thyroid⁽ Reference Rohner, Zimmermann and Jooste ^{126 )}. Thus, the enlargement of the thyroid may precede any fall in T₃ and any impairment of linear growth. In fact, most randomised controlled trials of iodine supplementation have involved pregnancy, aimed to prevent irreversible brain damage associated with cretinism⁽ Reference Zimmermann ^{127 )}, and, of these, a recent systematic review identified only two studies which reported on child linear growth, observing no improvements⁽ Reference Zhou, Anderson and Gibson ^{196 )}. However, hypothyroid iodine-deficient Columbian children with goitre, short stature and retarded bone age increased in stature when given T₄ ⁽ Reference Hernandez-Cassis, Cure-Cure and Lopez-Jaramillo ^{197 )}. Iodine repletion in school-age children who were severely iodine deficient (7- to 10-year-old Moroccan children), moderately iodine deficient (10- to 12-year-old Albanian children) or mildly iodine deficient (5- to 14-year-old South African children) increased IGF-1 levels in each case, and also increased total T₄, IGF-1, IGFBP-3 and both WA and HA Z scores in the Moroccan and Albanian children⁽ Reference Zimmermann, Jooste and Mabapa ^{128 )}. This suggests that in communities of low iodine availability where use of iodised salt is limited, if stunting is prevalent, iodine deficiency should be included as a potential part of any nutritional aetiology.

Multiple micronutrient intakes

Given that multiple deficiencies of both type 1 and type 2 nutrients are likely to be common in the diets of many mothers and stunted children, poor growth may reflect interactions of their deficiencies. The recognition of this has resulted in the introduction of multiple micronutrient supplements.

These programmes have developed from an initial focus on Fe, vitamin A, iodine, folate and Zn^{( 198 )} to multiple-micronutrient supplementations (MMS) of pregnant women in developing countries^{( 199 )}, recommending a single daily tablet containing the ADA (or higher for folic acid) of fifteen micronutrients (vitamins A, B₁, B₂, B₆, B₁₂, C, D, E, niacin, folic acid, Fe, Zn, Cu, I and Se). It has often been given as a powder. A Cochrane review of MMS during pregnancy indicated that it was efficacious for maternal, fetal and infant health outcomes⁽ Reference Haider and Bhutta ^{200 )}: i.e. a reduced frequency of low birth weight, small for gestational age and stillbirths compared with supplements of only Fe, with or without folic acid. However, for postnatal linear growth, while an early MMS 12-month trial in Mexican infants aged 8–14 months showed benefit for the younger children (<12 months)⁽ Reference Rivera, Gonzalez-Cossio and Flores ^{201 )}, a 2009 meta-analysis of multiple micronutrient supplementations of at least three micronutrients identified only a very small significant change in height⁽ Reference Ramakrishnan, Nguyen and Martorell ^{118 )}. Reviews of trials of micronutrient powders given to children under 2 years of age⁽ Reference De-Regil, Suchdev and Vist ^{202 )} and in children mainly aged 6 months to 6 years of age⁽ Reference Salam, MacPhail and Das ^{203 )} concluded that micronutrient powders did not increase linear growth. Since these reviews, as discussed below⁽ Reference Krebs, Mazariegos and Chomba ^{162 )}, a multi-micronutrient-fortified cereal supplement given as a control for a multi-country meat intervention trial in stunted infants and toddlers was unable to prevent the worsening of stunting during the trial, even though both Zn and especially Fe status benefitted from the intervention.

The most recent approach is to provide MMS as part of a small-quantity lipid-based nutrient supplements (LNS), for home fortification during complimentary feeding⁽ Reference Arimond, Zeilani and Jungjohann ^{204 )}. This provides about 20 g/120 kcal (502 kJ) of fat (73 % energy of which essential fatty acids are about 50 %), protein at 9 % energy, and aims to be nutritionally complete in terms of both type 1 and type 2 nutrients. Whilst this International Lipid-Based Nutrient Supplements project is ongoing^{( 205 )}, to date they have achieved only very limited success, with a modest linear growth increase in Ghana, but no growth effects in three trials in Malawi. Thus, LNS appear to promote linear growth in some but not all infant populations. The project team suggests that lack of an intervention effect could reflect asymptomatic infections, environmental enteropathy and an unfavourable composition of intestinal bacterial microbiota, restricting linear growth through a multitude of inflammation-related or other pathways in children with adequate dietary intakes⁽ Reference Ashorn, Alho and Ashorn ^{192 )}. As a result, trials of LNS are now underway combined with interventions aimed at water quality, sanitation and hygiene (WASH)⁽ Reference Arnold, Null and Luby ^{206 , 207 )} in which the nutrition intervention will use a LNS comprising what is described as a next-generation version of Nutributter⁽ Reference Arnold, Null and Luby ^{206 )}.

Animal-source foods and linear growth

ASF are nutrient dense in terms of many highly bioavailable micronutrients and are often more energy dense than plant-based foods through their fat content; thus, they are a good source of fat-soluble vitamins and essential fatty acids. They are also the only source of vitamin B₁₂. It is the case that for many population groups of children, ASF often comprise only a very small part of their diet, if any at all. As discussed above, while diets free of animal products can support near-normal growth, they need to contain a wide variety of plant food sources and to be supplemented with key micronutrients. However, in poor communities in much of the developing world, diets are much less varied and often contain few ASF or energy-dense foods, with ASF culturally unacceptable in some communities.

Several observational studies, for example The Nutrition Collaborative Research Support Program (N/CRSP) in Egypt, Kenya and Mexico, show associations between intakes of ASF and better growth which persisted even after controlling covariates and confounders such as socio-economic status, morbidity, parental literacy and nutritional status⁽ Reference Neumann, Harris and Rogers ^{208 )}. In these studies the greatest deficits in linear growth were found in those with little or no ASF in their diet. On the basis of these studies, Allen & Dror⁽ Reference Allen and Dror ^{209 )} have widely promoted the strategy of increasing consumption of ASF, i.e. meat, fish, eggs and milk, for improving the amount and bioavailability of micronutrients available to children in the developing world.

As far as meat per se is concerned, few of the interventions associated with better growth identified by Allen & Gillespie⁽ Reference Allen and Gillespie ^{153 )} involved meat as such and these generally failed to increase height growth or reduce stunting. In a Kenyan trial that supplied school children daily snacks of a local dish (maize, beans and greens) for 2 years with additions of meat, milk or equivalent energy, while all children increased weight compared with no snacks, there was little effect on height, although meat did improve muscle mass and cognitive function more than milk⁽ Reference Neumann, Bwibo and Murphy ^{210 ,} Reference Neumann, Murphy and Gewa ^{211 )}. A more recent multicentre 12-month intervention at 6 months of age compared the effect of lyophilised beef with an equi-energetic multiple micronutrient-fortified cereal. This involved infants with an estimated prevalence of stunting of ≥20 %, in rural communities in the Democratic Republic of Congo and Zambia, in semirural communities in the Western Highlands of Guatemala, and in urban communities in Karachi, Pakistan⁽ Reference Krebs, Mazariegos and Chomba ^{162 )}. The micronutrient content of the rice–soya cereal supplement was based on guidelines for multiple micronutrients in complementary foods. Consequently, apart from vitamin B₁₂, dietary levels of minerals and vitamins were higher in the cereal compared with beef supplement, four times higher for Fe, and Fe status at 18 months was better in the cereal group. The design was surprising given that higher indices of Fe status from the meat was a secondary hypothesis. However, the extent of stunting worsened from a length-for-age Z score of –1·4 at baseline to –1·9 at 18 m with no difference between the two groups. In fact, only maternal height and education emerged as (positive) influences on length growth and the authors commented that this may be a surrogate for better socio-economic status and favourable practices such as hygiene and medical care⁽ Reference Krebs, Mazariegos and Chomba ^{162 )}. Importantly, handwashing and use of boiled water for food preparation were emphasised in the trial.

The majority of ASF interventions have involved milk, starting with studies in the 1920–1930s in the UK and India and in the 1950s in USA⁽ Reference Pollock ^{212 )}. Whilst all the studies leave much to be desired in terms of their design and analysis, they resulted in the identification of milk as an important determinant of height growth in children, and they were instrumental in the subsequent legislation in the UK in 1934, 1940 and in 1945 when free school milk was introduced⁽ Reference Atkins ^{213 )}. Since these early studies the influence of cows’ milk intake on height growth post-weaning has been repeatedly confirmed in a large variety of cross-sectional, observational and intervention studies in healthy children in the UK, Denmark, USA and Japan⁽ Reference Hoppe, Mølgaard and Michaelsen ^{214 –} Reference Wiley ^{219 )} and in stunted children in developing societies.

In healthy Danish preschool children, all with protein intakes well above current protein requirements, with >60 % from ASF, height was positively associated with intakes of total animal protein and milk, but not with intakes of vegetable protein or meat⁽ Reference Hoppe, Udam and Lauritzen ^{175 )}. In New Zealand, long-term avoidance of cows’ milk is associated with small stature in pre-pubertal school children which was argued to reflect the simple chronic avoidance of milk, rather than health problems associated with milk allergy or intolerance⁽ Reference Black, Williams and Jones ^{220 )}. In the USA, analysis of National Health and Nutrition Examination Survey (NHANES) 1999–2002 indicated that adult height was positively associated with milk consumption at the ages of 5–12 years and 13–17 years, after controlling for sex, education and ethnicity⁽ Reference Wiley ^{219 )}. However, given the complexity of height growth regulation, it would be very surprising if milk intake explained all the variation and Wiley⁽ Reference Wiley ^{219 )} reports on five intervention studies in the UK, Switzerland, New Zealand and the USA which report no significant effect on height in mainly adolescent girls. Nevertheless, it is undeniable that milk, which has the function of supporting rapid postnatal growth, can also influence growth throughout childhood and is a determinant of adult height.

At the outset of these studies milk was viewed simply as a nutrient-rich food providing for growth in stature, with elaborations of this view into a ‘Milk hypothesis’ predicting that a greater consumption of milk during infancy and childhood will result in greater stature in adult life⁽ Reference Bogin ^{218 )}. It is the case that milk consumption and lactose digestion after weaning are exclusively human mammalian traits made possible by lactase persistence (LP), the continued production of the enzyme in the post-weaning period into adulthood. LP post-weaning is not the ancestral condition in humans but is made possible by a relatively recent SNP of the LPH (lactase–phlorizin hydrolase) gene⁽ Reference Romero, Mallick and Liebert ^{221 )}. Whilst the emergence of LP was advantageous amongst early pastoralists with a constant source of fresh milk, consumption in industrialised urban societies was very limited until the pasteurisation of the milk supply to cities in the late 19th and early 20th centuries and the widespread use of refrigeration. This means that the marked increase in milk consumption by urban populations in northern Europe and North America is a 20th-century phenomenon.

It has certainly become clear in recent years that because the specific biological function of milk is to enable rapid post-natal growth, it differs in many ways from all other human foods in having growth-promoting properties not found in meat or other ASF, although these properties are by no means understood. Also the widely assumed beneficial role of milk in the human diet after weaning is being questioned⁽ Reference Hoppe, Mølgaard and Michaelsen ^{214 )}. Melnik et al. ⁽ Reference Melnik, John and Schmitz ^{222 )} identifies the presence of microRNA in milk exosomes which could possibly act as anabolic signals for the stimulation of cellular growth and proliferation. Also, milk has uniquely high levels of amino acids known to be involved in anabolic signaling, especially tryptophan with concentrations in milk proteins uniquely high, 40 % higher than egg and double that of meat. Trytophan is said to activate the GH–IGF-1–mTORC1 pathway either directly or through gastrointestinal glucose-dependent insulinotropic peptide production. The caseins and lactoglobulins contain high concentrations of leucine, although this is not unique to milk, with higher levels in maize and sorghum. As indicated above, leucine is an upstream activator of the mTORC1 signalling pathway, important for cell replication and growth⁽ Reference Laplante and Sabatini ^{77 )}. These features may explain why Hoppe et al. ⁽ Reference Hoppe, Udam and Lauritzen ^{175 )} showed, in a cross-sectional study of Danish preschool children, that serum IGF-1 levels and overall height were significantly related to milk but not meat intake. They also showed in an intervention study with a high protein supplement of either milk or beef in 8-year-olds that milk, and not meat, increased concentrations of serum IGF-1 and the serum IGF-1:serum IGFBP-3 ratio significantly⁽ Reference Hoppe, Mølgaard and Juul ^{223 )} and that fasting insulin levels and consequently insulin resistance (calculated with the homeostasis model assessment) were higher in the milk- but not meat-supplemented children⁽ Reference Hoppe, Mølgaard and Vaag ^{224 )}. This combination of increased growth and insulin resistance raises the question of whether these effects of milk in these otherwise healthy children are of benefit or not⁽ Reference Hoppe, Mølgaard and Michaelsen ^{214 )}. The benefit of tallness is said to include a decreased risk of some diseases (cardiovascular and respiratory diseases) but an increased risk of colorectal, postmenopausal breast and ovarian cancers, and possibly pancreatic, prostate and premenopausal breast cancers^{( 3 )}. Furthermore, it has been known for some time that societies with high rates of milk consumption and with tall adults generally exhibit poor bone health in old age with higher fracture rates compared with less developed countries with much less milk consumption⁽ Reference Hegsted ^{225 )}. Why this should occur is not known but the possibility that the more rapid growth of childhood and earlier puberty results in a bone architecture which becomes more fragile in old age deserves investigation.

As for children in developing countries, data on preschool children from seven countries in Central and South America indicate that milk intake was significantly associated with higher HA Z score in all countries, whereas meat and egg/fish/poultry intakes were only associated with height in one of the countries⁽ Reference Ruel ^{226 )}. A comparison of the heights and weights of children of the Ugandan Karamoja, a cattle-herding, milk-consuming people, with those of the Buganda, farmers growing plantain, sweet potatoes and cassava, showed that the Karamoja children exhibited normal height growth but low weight for height, whilst the Bugandan children were stunted but of normal weight for height⁽ Reference Rutishauser and Whitehead ^{227 )}.

Malcolm’s dietary interventions of children of the Bundi people of the highlands of Papua New Guinea with skimmed milk increased their height⁽ Reference Malcolm ^{163 –} Reference Malcolm ^{165 )}. The growth rate of the Bundi children, with their diet of mainly sweet potato and taro providing 3 % dietary protein energy, was slower than that of any other reported population in the world⁽ Reference Malcolm ^{165 )}. At 7 years they were 9·2 and 5·7 cm shorter than Guatemalan boys and girls studied in the INCAP study⁽ Reference Martorell, Schroeder and Rivera ^{228 )}. His 3- to 8-month interventions in preadolescents involved three studies comparing their usual diet with supplemental skimmed milk powder or additional energy as margarine or additional food⁽ Reference Lampl, Johnston and Malcolm ^{164 )}. In all three studies some linear growth was observed with the usual diet which was slightly increased by more of the usual diet. However, the growth rate in length was doubled by the skimmed milk in all three experiments. Although the design of these studies leaves much to be desired compared with best current practice they remain unique in terms of the detailed description of the community in which they were carried out⁽ Reference Malcolm ^{165 )}. Nevertheless, the outcome of these studies is consistent with current knowledge of the role of milk in linear growth regulation. As with all such interventions with milk, although these studies are reported as protein supplementations, they provide increases in minerals, phosphate and other key nutrients. For example, from the published composition of skimmed milk powder, extra Zn in Malcolm’s studies would range from 1·2 to 3 mg/d, not insignificant accounts given that most Zn supplementation studies involve 5 mg/d. Extra iodine would range from 40 to 100 µg/d. The WHO recommends a daily intake of iodine of 120 µg for school children^{( 229 )} so that if these children were iodine deficient, the milk intervention may have met their iodine requirements. Malcolm makes no mention of iodine deficiency in his monograph on the Bundi people⁽ Reference Malcolm ^{165 )}. However, the Highlands of New Guinea were known to be an area of iodine deficiency from the 1960s and were where a classic iodised oil supplementation study occurred⁽ Reference Pharoah and Connolly ^{230 )}, aimed to prevent endemic cretinism. This means that the growth response observed by Malcolm could have reflected an improvement of their protein, Zn and iodine status in addition to any of the other milk effects described above.

Environmental enteric dysfunction

In the context of the stunting syndrome, infection includes specific diagnosed infections such as parasitic infections, especially malaria and intestinal helminths, and diarrhoea, particularly in conditions of poor sanitation and hygiene. One study suggested that 25 % of stunting was attributed to five or more episodes of diarrhoea⁽ Reference Checkley, Buckley and Gilman ^{241 )}. However, while all agree that diarrhoea is implicated as a cause of poor growth, there has been a debate about its relative importance⁽ Reference Bhutta, Ahmed and Black ^{242 ,} Reference Solomons, Mazariegos and Brown ^{243 )}. An analysis of seven longitudinal cohort studies conducted in four low-income countries indicated that the average child’s diarrhoea burden between birth and 2 years only accounted for a clinically modest reduction in height growth⁽ Reference Richard, Black and Gilman ^{244 )}. In Gambian infants in which growth faltering in terms of weight and height growth was reported to become apparent in the 2nd year of life, although diarrhoea did occur, its frequency did not explain the growth faltering⁽ Reference Lunn, Northrop-Clewes and Downes ^{245 )}. Direct measurements of an enteropathy associated with increased intestinal permeability, as indicated by the handling of oral lactulose and mannitol, were a much better predictor of impaired weight and height gain. Increased gut permeability implies a failure of normal gut barrier function, which normally prevents translocation of pathogenic organisms and endotoxins, and which can, in extreme cases, be a route to the development of sepsis in patients in intensive care. Subsequent studies in Gambian infants indicated endotoxin translocation in terms of increased plasma concentrations of total IgG and IgG–endotoxin-core antibody (EndoCAb) and these responses were correlated with increased intestinal permeability. In fact, intestinal permeability, plasma IgG concentration and EndoCAb together accounted for 56 % of the growth inhibition⁽ Reference Campbell, Elia and Lunn ^{246 )}. Mucosal biopsies of these children indicated chronic T cell-mediated enteropathy with crypt hyperplasia, villous stunting and high numbers of intra-epithelial lymphocytes in all Gambian children, regardless of nutritional status⁽ Reference Campbell, Murch and Elia ^{247 )}. However, with worsening nutrition, mucosal cytokine production became biased toward a proinflammatory response, i.e. a dominance of interferon-γ over TGF-β expression and with increased TNF-α-producing cells in the mucosa. These authors concluded that in these Gambian children translocation of immunogenic luminal macromolecules across a compromised gut mucosa had resulted in stimulation of systemic immune/inflammatory processes and subsequent growth impairment. On the basis of this and other evidence, Prendergast & Humphrey⁽ Reference Prendergast and Humphrey ^{10 )} argued that subclinical infection with enteric pathogens is common, even in the absence of diarrhoea⁽ Reference Kotloff, Nataro and Blackwelder ^{248 )}. Humphrey⁽ Reference Humphrey ^{249 )} identified this subclinical infection due to enteric pathogens as tropical enteropathy, characterised by villous atrophy, crypt hyperplasia, increased permeability, inflammatory cell infiltrate, and modest malabsorption caused by faecal bacteria ingested in large quantities by young children living in conditions of poor sanitation and hygiene. She suggested it to be the primary causal pathway from poor sanitation and hygiene to undernutrition (and by implication stunting), rather than diarrhoea. Subsequently, Korpe & Petri⁽ Reference Korpe and Petri ^{250 )} reviewed the distinction between this environmental enteropathy and other non-infectious tropical malabsorption syndromes such as tropical sprue, coeliac sprue and Crohn’s disease, and Keusch et al. ⁽ Reference Keusch, Rosenberg and Denno ^{251 )} introduced the term environmental enteric dysfunction. Its relevance to child health has been recently reviewed⁽ Reference Crane, Jones and Berkley ^{252 )}. Importantly, in the context of the present review, Prendergast & Humphrey⁽ Reference Prendergast and Humphrey ^{10 )} wrote: ‘We therefore view stunting as an inflammatory disease arising, in part, from primary gut pathology. Since gut damage also occurs with recurrent (especially persistent) diarrhoea, severe acute malnutrition, HIV infection and micronutrient deficiencies, there are multiple overlapping causes of enteropathy in settings of poverty which may exacerbate the growth failure arising from EED (environmental enteric dysfunction)’. The EED pathway between the poor environment and stunting is illustrated in Fig. 3 ⁽ Reference Humphrey ^{249 –} Reference Crane, Jones and Berkley ^{252 )}. One potential link between WASH and the development of the enteropathy is a detrimental change in the intestinal microbiota⁽ Reference Crane, Jones and Berkley ^{252 )}, with evidence from longitudinal studies of twin cohorts of stunted children from Malawi and Bangladesh⁽ Reference Gough, Stephens and Moodie ^{253 )} in whom the relative abundance of Acidaminococcus sp. was shown to be associated with future linear growth deficits. Because Acidaminococcus sp. can utilise glutamate as their sole source of carbon and energy, with glutamate usually a key source of fuel for the healthy enterocyte, the authors speculate that overgrowth of bacteria that can ferment glutamate may impair barrier function with the consequent deleterious effect on linear child growth. Other recent work points to a specific deleterious role of frequent exposure to mycotoxins⁽ Reference Smith, Prendergast and Turner ^{254 )}. These contaminate a wide range of staple foods, including maize and groundnuts, especially aflatoxin, fumonisin and deoxynivalenol. The authors argue that several billion individuals in predominantly developing countries may be at risk of exposure, with epidemiological evidence from multiple countries suggesting that exposure in pregnant women, infant cord blood, and young children is widespread during early life and associated with impaired infant growth velocity. Although aflatoxins have been most widely studied in relation to liver cancer⁽ Reference Bennett and Klich ^{255 )}, all three mycotoxins can mediate intestinal damage through distinct pathways resulting in increased local and systemic pro-inflammatory cytokines with immunomodulation toward an inflammatory state, potentially interfering with the IGF-1 axis. Some years ago Hendrickse et al. ⁽ Reference Hendrickse, Coulter and Lamplugh ^{256 )} identified a link between aflatoxicosis and kwashiorkor in Sudanese children. Although there was little evidence of aflatoxicosis in malnourished Jamaican children⁽ Reference Golden ^{257 )}, where aflatoxicosis occurs, a link between the induced inflammation and kwashiorkor is consistent with the concept that kwashiorkor involves the catastrophic impact of an external noxious insult on children with inadequate micronutrient protection unable to counter the consequent oxidative stress⁽ Reference Golden ^{257 )}.

Fig. 3 Stunting as a predominantly inflammatory disease resulting from poor hygiene and environmental enteric dysfunction. The pathway between an unsanitary environment, dietary mycotoxins and other potential pathogens, an abnormal intestinal microbiota and stunting includes intestinal inflammation and pathological changes to the GI mucosa. This results in a failure of barrier function allowing translocation of pathogens and endotoxins resulting in a systemic inflammatory response which inhibits bone growth. In addition, nutrient malabsorption occurs exacerbating any dietary insufficiency, worsening malnutrition and increasing susceptibility to infection through its adverse effect on the immune system. This also contributes to the bone growth inhibition. GH, growth hormone; IGF, insulin-like growth factor; IGFBP, insulin-like growth factor binding protein. Adapted from Humphrey⁽ Reference Humphrey ^{249 )}, Korpe & Petri⁽ Reference Korpe and Petri ^{250 )}, Keusch et al. ⁽ Reference Keusch, Rosenberg and Denno ^{251 )} and Crane et al. ⁽ Reference Crane, Jones and Berkley ^{252 )}.

Recent direct evidence for the EED concept comes from studies in pre-school rural Bangladeshi children⁽ Reference Lin, Arnold and Afreen ^{258 )}. Objective measures of lower water quality, poor sanitary/handwashing infrastructure, lower gut permeability and lower IgG EndoCAb titres were related to lower HA Z score.

One important consequence of a widespread occurrence of EED is that it would explain the largely disappointing nature of nutritional intervention trials: i.e. the fact that only in a very few studies has the provision of supplements of either specific nutrients, multi-nutrient mixes or foods been shown to restore growth to normal or to the increased rates which might be expected if catch-up were occurring. Whilst a review by Dewey & Adu-Afarwuah⁽ Reference Dewey and Adu-Afarwuah ^{259 )} of thirty-eight reports of such supplementation trials indicated both weight and height gain, the latter in the range of an increase in 0·0–0·64 length-for-age Z score by 12 to 24 months, Humphrey⁽ Reference Humphrey ^{249 )} commented that none of children studied in these various interventions achieved normal growth. In fact, the height growth achieved in the most successful of these studies was equivalent to only about a third of the average deficit of Asian and African children. Dewey & Adu-Afarwuah⁽ Reference Dewey and Adu-Afarwuah ^{259 )} did argue that although changes in mean HA Z scores were small, nevertheless the impact on the lower tail of the distribution – that is, on stunting rates per se – could be considerably larger (for example, a fall from 15·8 to 4·7 % in the intervention group). However, they concluded that complementary feeding interventions, by themselves, cannot change the underlying conditions of poverty and poor sanitation that contribute to poor child growth. A typical recent example of the failure of a nutritional intervention is the randomised controlled trial of a meat supplement undertaken by Krebs et al. ⁽ Reference Krebs, Mazariegos and Chomba ^{162 )} discussed above. The rural, semi-rural and urban communities involves in that study were all likely to be exposed to poor hygiene and may well have suffered from EED. Also as discussed above, the disappointing outcomes of many of the LNS within the International Lipid-Based Nutrient Supplements programme has also been attributed to asymptomatic infections, EED and/or an unfavourable composition of intestinal bacterial microbiota⁽ Reference Ashorn, Alho and Ashorn ^{192 )}.

On the basis of the existing data, two major randomised studies are now investigating the concepts shown in Fig. 3 in terms of the potential role of EED in childhood stunting. WASH Benefits⁽ Reference Arnold, Null and Luby ^{206 )} is testing, in cluster-randomised trials, improvements in water quality, sanitation, handwashing and child nutrition, alone and in combination, on length growth over the first 12 and 24 months at multiple sites in Bangladesh and Kenya. The Sanitation, Hygiene, Infant Nutrition Efficacy (SHINE) Project^{( 207 ,} Reference Mbuya, Tavengwa and Stoltzfus ^{260 ,} Reference Desai, Smith and Mbuya ^{261 )} is a proof-of-concept cluster-randomised trial underway in Zimbabwe testing the protecting of babies from faecal ingestion with a WASH intervention and optimising nutritional adequacy of the infant diet with an infant and young child feeding intervention on length growth and anaemia. In each trial LNS are used, described as a next-generation version of Nutributter, which are identical except that the SHINE trial includes less Fe (6 mg compared with 9 mg) than in WASH Benefits, to mitigate concerns about the potential effect of supplemental Fe on infections⁽ Reference Desai, Smith and Mbuya ^{261 )}.

The role of nutrition and inflammation on the regulation of endochondrial ossification incorporating what has been discussed in this review is illustrated in Fig. 4 ⁽ Reference Karlberg ^{20 )}.

Fig. 4 Role of nutrition and inflammation in the regulation of endochondral ossification. During the three developmental growth phases of the infancy, childhood and puberty (ICP) model of Karlberg⁽ Reference Karlberg ^{20 )}, the primary endocrine activators of growth are insulin (infancy), growth hormone (childhood) and the sex steroids (puberty) (although current evidence suggests that growth hormone may also be involved in the infancy stage). These act on the paracrine/autocrine system within the growth plate which mediate endochondral ossification. However, a suitable nutritional anabolic drive is necessary for this process to occur involving both type I nutrients such as iodine (to enable adequate thyroid hormone production) and type II nutrients of which amino acids and zinc are particularly important in stimulating the endocrine/paracrine system and in having direct regulatory influences, although this is poorly understood at the molecular level. Milk also appears to have a specific influence not observed with other animal-source foods. Elevated levels of glucocorticoids, inhibitory fibroblast growth factors (for example, FGF21), pro inflammatory cytokines, especially TNFα, IL-1β, IL-6 and other inflammatory mediators associated with infection and inflammation and environmental enteric dysfunction as in Fig. 3, block the nutritional anabolic drive and inhibit endochondral ossification. How these interactions between nutrition and infection occur at the cellular and molecular level is poorly understood. IGF, insulin-like growth factor; IGFBP, insulin-like growth factor binding protein; T₃, triiodothyronine; 1,25(OH)₂D, 1,25-dihydroxyvitamin D; IHH, Indian hedgehog; PTHrP, parathyroid-hormone-related protein; BMP, bone morphogenetic protein; WNT, wingless/integrated protein; VEGF, vascular endothelial growth factor; EED, environmental enteric dysfunction.