Iodine and Adult IQ meta-analysis

Iodine improves IQ in fetuses; adults as well? A meta-analysis of relevant studies says no.
nootropics, psychology, statistics, meta-analysis, charity, R, power-analysis, IQ, bibliography
2012-02-292017-07-18 in progress certainty: highly likely importance: 10


Iodiza­tion is one of the great suc­cess sto­ries of pub­lic health inter­ven­tion: iodiz­ing salt costs pen­nies per ton, but as demon­strated in ran­dom­ized & nat­ural exper­i­ments, pre­vents goi­ters, cre­tinism, and can boost pop­u­la­tion IQs by a frac­tion of a stan­dard devi­a­tion in the most iodine-d­e­fi­cient pop­u­la­tions.

These exper­i­ments are typ­i­cally done on preg­nant wom­en, and results sug­gest that the ben­e­fits of iodiza­tion dimin­ish through­out the trimesters of a preg­nan­cy. So does iodiza­tion ben­e­fit nor­mal healthy adults, poten­tially even ones in rel­a­tively iodine-suffi­cient West­ern coun­tries?

Com­pil­ing exist­ing post-na­tal iodiza­tion stud­ies which use cog­ni­tive tests, I find that—out­liers aside—the ben­e­fit appears to be nearly zero, and so likely it does not help nor­mal healthy adults, par­tic­u­larly in West­ern adults.

is inter­est­ing from the eth­i­cal stand­point as one of the most cost-effec­tive—yet obscure— mea­sures ever devised which deserves a name like : a few pen­nies of iodine added to salt elim­i­nates many cases of men­tal retar­da­tion & goi­ters.

Background

Sup­ple­men­ta­tion of in salt, water, or oil increases body iodine lev­els and reduces iodine defi­ciency dis­or­ders (Wu et al 2002). Sup­ple­ment­ing, dur­ing preg­nancy or infan­cy, can raise aver­age IQs in the worst-off regions by <13 IQ points1 (close to a full stan­dard devi­a­tion); such an increase is of con­sid­er­able eco­nomic val­ue, even in devel­oped coun­tries with iodiza­tion pro­grams (see & appen­dices for a cost-ben­e­fit analy­sis). In an addi­tional bonus for our post-fem­i­nist soci­ety, females ben­e­fit more from iodiza­tion than males2. Because salt pro­duc­tion is gen­er­ally so cen­tral­ized as a bulk com­mod­ity extracted from a very few areas, iodiza­tion is almost triv­ial to imple­ment. (Although humans being humans, there are obsta­cles even to suc­cess­ful iodiza­tion pro­grams3.)

Cre­tinism is only the most extreme form of iodine defi­cien­cy, although a major & worth­while human­i­tar­ian task; iodine cor­re­lates with IQ in non-d­e­fi­cient chil­dren, eg. Japan is simul­ta­ne­ously one of the longest-lived & high­est IQ coun­tries in the world and one of the great­est con­sumers of iodine4 (from sea­weed, prin­ci­pal­ly, lev­els so high they sug­gest that cur­rent rec­om­men­da­tions are overly con­ser­v­a­tive), and even weak hypothy­roidism impairs men­tal per­for­mance (in the old). School is directly impacted in ran­dom­ized tri­als; from a review of Poor Eco­nom­ics (Baner­jee & Duflo 2011):

Pro­vid­ing iodine cap­sules to preg­nant moth­ers is an inter­ven­tion that helps brain devel­op­ment in fetus­es. It costs around 51 cents per dose—and leads to kids who stay in school about five months longer because they are cog­ni­tively bet­ter able to learn.

The orig­i­nal waves of iodiza­tion caused large-s­cale changes in: num­bers of peo­ple going to school5, work­ing at all6, their occu­pa­tion7, how they voted, or even how many recruits from a region are accepted to selec­tive flight schools8.

Of course, iodine can be a dou­ble-edged sword. Feyrer et al 2008 men­tioned a wave of dis­or­ders after iodiza­tion of salt, as long-d­e­fi­cient thy­roids were shocked with ade­quate lev­els of iodine, and nat­ural iodine lev­els can be so high as to begin to inversely cor­re­late with IQ in Chi­na.9

More wor­ri­some is recent trends in the devel­oped world such as the War on Salt. Un-iodized salt or low-io­dide salt like sea salt is ever more pop­u­lar. Exist­ing iodized table salt often has far less iodide than rec­om­mend­ed, or even what the man­u­fac­turer claims it has10. Iodized salt used in cook­ing—as opposed to a straight table-side condi­men­t—loses large chunks of its iodine con­tent11. Small sam­ples of ordi­nary peo­ple turn in severe or mild iodine defi­ciency rates of 67%12 to as much as 73.7%13. This is plau­si­ble given a steady sec­u­lar trend of iodine reduc­tion in the US (although one that seems to have paused in the 2000s)14. I was unsur­prised to read Morse 2012:

Numer­ous pop­u­la­tion stud­ies from a vari­ety of coun­tries includ­ing Chi­na, Hong Kong, Iran, India, Kyr­gyzs­tan and Eng­land have reported iodine defi­ciency in girls of child bear­ing age 76, in preg­nant 154,155,156,157, and in preg­nant and lac­ta­tion women 158,159. Some of these stud­ies included regions where salt iodiza­tion is prac­ticed, yet a [sub­stan­tial] pro­por­tion of preg­nant and lac­tat­ing women were still defi­cient 155,156,157,158,159,160,161. A few exam­ples of recent stud­ies fol­low…

This has led to an observ­able impact on the IQ of the chil­dren of Eng­lish women (Bath et al 2013), and there is no rea­son to expect the effect to be con­fined to them.

Meta-analysis

Given all this, one nat­u­rally won­ders what the effect might be in older humans: ele­men­tary school age and above. If iodine before birth can be respon­si­ble for increas­ing IQ by a full stan­dard devi­a­tion or more in com­bi­na­tion with iron, what about iodine post-birth? Iron sup­ple­men­ta­tion treats and there is evi­dence it also treats the cog­ni­tive prob­lems as well, so what about iodine?

Liu et al 2009, and tan­gen­tial results like Bonger­s-Schokking et al 2005 (where there was a dis­cernible IQ differ­ence between TSH treat­ment of infants with con­gen­i­tal hypothy­roidism before & after 13 days post-birth) sug­gest that the win­dow for iodine inter­ven­tion may close rapidly dur­ing preg­nancy and be closed post-birth. While iodine has been exten­sively stud­ied in infants and other unusual pop­u­la­tions, the nar­row ques­tion of iodine’s effect on IQ in healthy adult First World pop­u­la­tions has not been (a com­mon prob­lem in nootrop­ic­s); we are inter­ested in cases only where some­one is men­tally tested before and after iodine sup­ple­men­ta­tion, or where 1 cohort receives sup­ple­men­ta­tion after birth when com­pared against a sim­i­lar cohort who receive no sup­ple­men­ta­tion. Most stud­ies turn out to be either cor­re­la­tional (eg. strat­i­fy­ing by blood lev­els of thy­roid hor­mone) or com­par­ing fetal sup­ple­men­ta­tion against a non-sup­ple­mented con­trol group. Unfor­tu­nate­ly, no study is so large and high­-qual­ity that it defin­i­tively resolves our ques­tion. So we resort to of what is avail­able: we pool many stud­ies together to derive a sum­mary aver­age of the over­all results, weighted by how many sub­jects each study had (since more is bet­ter) ver­sus how strong a result they yield­ed. An exam­ple of this is the meta-analy­sis & review, which is the clos­est exist­ing study to what I want, “Iodine and Men­tal Devel­op­ment of Chil­dren 5 Years Old and Under”:

Sev­eral reviews and meta-analy­ses have exam­ined the effects of iodine on men­tal devel­op­ment. None focused on young chil­dren, so they were incom­plete in sum­ma­riz­ing the effects on this impor­tant age group. The cur­rent sys­tem­atic review there­fore exam­ined the rela­tion­ship between iodine and men­tal devel­op­ment of chil­dren 5 years old and under. A sys­tem­atic review of arti­cles using MEDLINE (1980-No­vem­ber 2011) was car­ried out. We orga­nized stud­ies accord­ing to four designs: (1) ran­dom­ized con­trolled trial with iodine sup­ple­men­ta­tion of moth­ers; (2) non-ran­dom­ized trial with iodine sup­ple­men­ta­tion of moth­ers and/or infants; (3) prospec­tive cohort study strat­i­fied by preg­nant wom­en’s iodine sta­tus; (4) prospec­tive cohort study strat­i­fied by new­born iodine sta­tus. Aver­age effect sizes for these four designs were 0.68 (2 RCT stud­ies), 0.46 (8 non-RCT stud­ies), 0.52 (9 cohort strat­i­fied by moth­ers’ iodine sta­tus), and 0.54 (4 cohort strat­i­fied by infants’ iodine sta­tus). This trans­lates into 6.9 to 10.2 IQ points lower in iodine defi­cient chil­dren com­pared with iodine replete chil­dren. Thus, regard­less of study design, iodine defi­ciency had a sub­stan­tial impact on men­tal devel­op­ment. Method­olog­i­cal con­cerns included weak study designs, the omis­sion of impor­tant con­founders, small sam­ple sizes, the lack of clus­ter analy­ses, and the lack of sep­a­rate analy­ses of ver­bal and non-ver­bal sub­tests. Quan­ti­fy­ing more pre­cisely the con­tri­bu­tion of iodine defi­ciency to delayed men­tal devel­op­ment in young chil­dren requires more well-de­signed ran­dom­ized con­trolled tri­als, includ­ing ones on the role of iodized salt.

But its sec­ond design con­flates sup­ple­men­ta­tion of moth­ers with that of infants & chil­dren, and so the d = 0.46 (fig­ure 2) is not directly mean­ing­ful (the authors note that the stud­ies are het­ero­ge­neous but do not attempt strat­i­fy­ing by fetal vs infan­cy). More rel­e­vant is “Effect of iodine sup­ple­men­ta­tion in preg­nancy on child devel­op­ment and other clin­i­cal out­comes: a sys­tem­atic review of ran­dom­ized con­trolled tri­als”, Zhou et al 2013:

…Four­teen pub­li­ca­tions that involved 8 tri­als met the inclu­sion cri­te­ria. Only 2 included tri­als reported the growth and devel­op­ment of chil­dren and clin­i­cal out­comes. Iodine sup­ple­men­ta­tion dur­ing preg­nancy or the peri­con­cep­tional period in regions of severe iodine defi­ciency reduced risk of cre­tinism, but there were no improve­ments in child­hood intel­li­gence, gross devel­op­ment, growth, or preg­nancy out­comes, although there was an improve­ment in some motor func­tions. None of the remain­ing 6 RCTs con­ducted in regions of mild to mod­er­ate iodine defi­ciency reported child­hood devel­op­ment or growth or preg­nancy out­comes. Effects of iodine sup­ple­men­ta­tion on the thy­roid func­tion of moth­ers and their chil­dren were incon­sis­tent.

They cor­rectly observe that the avail­able stud­ies are not very method­olog­i­cally rig­or­ous and most do not allow of any real analy­sis, but I think it may be worth doing a more per­mis­sive sum­mary and see what it says. Tay­lor et al 2014’s “Impact of iodine sup­ple­men­ta­tion in mild-to-mod­er­ate iodine defi­cien­cy: sys­tem­atic review and meta-analy­sis” looks at Gor­don & Zim­mer­man, com­bin­ing the avail­able tests and find­ing:

For the meta-analy­sis cog­ni­tive tests were cat­e­gorised into the fol­low­ing domains: (i) per­cep­tual rea­son­ing; (ii) pro­cess­ing speed; (iii) work­ing mem­o­ry; and (iv) global cog­ni­tive index. The global cog­ni­tive index was derived from the aver­age of the scores in each of the domains. Unad­justed SMDs of the change in cog­ni­tive scores from base­line were com­puted from the raw scores reported by the authors, while adjusted SMDs were derived from the reported mean-ad­justed treat­ment effects. s.e.m.-ad­justed treat­ment effects were cal­cu­lated using the rec­om­mended for­mula in the Cochrane hand­book (44). The results of the analy­sis for indi­vid­ual domains are pre­sented in Table 3, while Fig. 3 shows the for­est plots for the global cog­ni­tive index. Ben­e­fi­cial effects of iodine sup­ple­men­ta­tion were seen for both adjusted and unad­justed global indices with mild het­ero­gene­ity observed between the stud­ies. For indi­vid­ual unad­justed domain scores, ben­e­fits were seen for pro­cess­ing speed but not for per­cep­tual rea­son­ing or work­ing mem­o­ry, while for the adjusted domains iodine was ben­e­fi­cial for per­cep­tual rea­son­ing although this was asso­ci­ated with sig­nifi­cant het­ero­gene­ity.

Possible studies

I have looked at the fol­low­ing stud­ies:

  1. “Neu­ro­log­i­cal dam­age to the fetus result­ing from severe iodine defi­ciency dur­ing preg­nancy”, Pharoah et al 1971; “The effect of iodine pro­phy­laxis on the inci­dence of endemic cre­tinism”, Pharoah et al 1972; “Fetal iodine defi­ciency and motor per­for­mance dur­ing child­hood”, Con­nolly et al 1979; “A con­trolled trial of iod­i­nated oil for the pre­ven­tion of endemic cre­tinism: a long-term fol­low-up”, Pharoah & Con­nolly 1987; a 1966 Papua New Guinea trial which in 1972 began sup­ple­ment­ing the con­trol group as well (who were now 0–6 years old); at the 1982 fol­lowup, the con­trol group still suffered cre­tinism and deficits com­pared to the exper­i­men­tal—no effect. These stud­ies turn out to not test the adults or chil­dren born before iodiza­tion and so are not use­ful for our pur­pose.

  2. “Pro­phy­laxis of endemic goi­ter with iodized oil in rural Peru”, Pretell et al 1972; “Iodine defi­ciency and the mater­nal-fe­tal rela­tion­ship”, Pretell et al 1974 (see also the ret­ro­spec­tive review Pretell & Cac­eres 1994): a 1966 Peru tri­al; females (<45 years old) and males (<18 years old); exper­i­men­tal infants did not out­per­form con­trol infants—no effect. These stud­ies turn out to not test the adults or chil­dren born before iodiza­tion and so are not use­ful for our pur­pose.

  3. “The effects of oral iodized oil on intel­li­gence, thy­roid sta­tus, and somatic growth in school-age chil­dren from an area of endemic goi­ter”, Bautista et al. A 1982 Bolivia tri­al; 200 school­child­ren mea­sur­ing IQ—no effect

  4. “Sup­ple­men­tary iodine fails to reverse hypothy­roidism in ado­les­cents and adults with endemic cre­tinism”, Boy­ages et al 1990; 28 Chi­nese aged 14–52, severely defi­cient, with no gains on the “Hiskey Nebraska Test of Learn­ing Apti­tude” or the “Griffiths Men­tal Devel­op­ment Scales”—no effect

  5. “Con­trolled trial of vit­a­m­in-min­eral sup­ple­men­ta­tion: Effects of intel­li­gence and per­for­mance”, Schoen­thaler et al 1991; mul­ti­vi­t­a­min with 0.075mg iodine, IQ gain in the 8th and 10th grader­s—ben­e­fi­cial effect

  6. “Tim­ing of Vul­ner­a­bil­ity of the Brain to Iodine Defi­ciency in Endemic Cre­tinism”, Cao et al 1994: 689 chil­dren 0–3 years of age, min­i­mal effect in the chil­dren

  7. “Dietary intake and micronu­tri­ent sta­tus of ado­les­cents: effect of vit­a­min and trace ele­ment sup­ple­men­ta­tion on indices of sta­tus and per­for­mance in tests of ver­bal and non-ver­bal intel­li­gence”, Southon et al 1994: 13–15 year olds with a mul­ti­vi­t­a­min of things includ­ing iodine (0.15mg); no effect on IQ test

  8. “Effect of iodine and iron sup­ple­men­ta­tion on phys­i­cal, psy­chomo­tor and men­tal devel­op­ment in pri­mary school chil­dren in Malawi” (mir­ror), Shrestha 1994

  9. “Effect of iodized oil sup­ple­men­ta­tion on thy­roid hor­mone lev­els and men­tal per­for­mance among Orang Asli school­child­ren and preg­nant moth­ers in an endemic goitre area in Penin­su­lar Malaysia”, Isa et al 2000

  10. “Improved iodine sta­tus is asso­ci­ated with improved men­tal per­for­mance of school­child­ren in Benin”, van den Briel et al 2000; orig­i­nally a dou­ble-blind RCT, but sub­jects were con­t­a­m­i­nated by impor­ta­tion of iodized salt (there was an appar­ent smal­l­-medium effect on the matrix sub­test)

  11. “Cog­ni­tive and motor func­tions of iodine-d­e­fi­cient but euthy­roid chil­dren in Bangladesh do not ben­e­fit from iodized poppy seed oil (Li­p­i­odol)”, Huda et al 2001: 1st and 2nd grade chil­dren, ~1mg? no effect

  12. “Influ­ence of sup­ple­men­tary vit­a­mins, min­er­als and essen­tial fatty acids on the anti­so­cial behav­iour of young adult pris­on­ers”, Gesch et al 2002: over 18 pris­on­ers, 0.14mg; ben­e­fi­cial effect on mea­sures of vio­lence, but they did not do an intel­li­gence retest and so are not use­ful for our pur­pose.

  13. “Effects of iodine sup­ple­men­ta­tion dur­ing preg­nancy on child growth and devel­op­ment at school age”, O’Don­nell 2002; included chil­dren who were sup­ple­mented at age 2 but only tested with Raven’s at age 7, which ren­ders their data not use­ful. How­ev­er, the gen­eral results con­firm that tim­ing of sup­ple­men­ta­tion even within preg­nancy fol­lows the ‘ear­lier is bet­ter’ rule.

  14. “Iodine sup­ple­men­ta­tion improves cog­ni­tion in iodine-d­e­fi­cient school­child­ren in Alba­nia”, Zim­mer­man et al 2006

  15. “Effect of mul­ti­vi­t­a­min and mul­ti­min­eral sup­ple­men­ta­tion on cog­ni­tive func­tion in men and women aged 65 years and over: a ran­domised con­trolled trial”, McNeill et al 2007; old adults, 0.15mg; no effect

  16. “A micronu­tri­en­t-for­ti­fied sea­son­ing pow­der reduces mor­bid­ity and improves short­-term cog­ni­tive func­tion, but has no effect on anthro­po­met­ric mea­sures in pri­mary school chil­dren in north­east Thai­land”, Manger et al 2008

  17. “Delayed Neu­robe­hav­ioral Devel­op­ment in Chil­dren Born to Preg­nant Women with Mild Hypothy­rox­ine­mia Dur­ing the First Month of Ges­ta­tion: The Impor­tance of Early Iodine Sup­ple­men­ta­tion”, Berbel et al 2009; turns out that sup­ple­men­ta­tion was admin­is­tered to all women dur­ing their preg­nan­cy, so the devel­op­ment scores are not use­ful here.

  18. “Iodine sup­ple­men­ta­tion improves cog­ni­tion in mildly iodine-d­e­fi­cient chil­dren”, Gor­don et al 2009

  19. think2 pro­ject:

  20. Dewi et al 2012, “Iodine sup­ple­men­ta­tion into drink­ing water improved intel­li­gence of preschool chil­dren aged 25–59 months in Ngar­goyoso sub­-dis­trict, Cen­tral Java, Indone­sia: A ran­dom­ized con­trol trial”; sub­jects selected for low uri­nary iodine excre­tion, region has >50% goi­ter rates

  21. van Stu­i­jven­berg et al 1999, “Effect of iron-, iodine-, and β-carotene -for­ti­fied bis­cuits on the micronu­tri­ent sta­tus of pri­mary school chil­dren: a ran­dom­ized con­trolled trial”

  22. “Tim­ing of the Effect of Iodine Sup­ple­men­ta­tion on Intel­li­gence Quo­tients of School­child­ren”, Salarkia et al 2004; fol­lowup in 2 Iran­ian vil­lages after the national insti­tu­tion of iodiza­tion in 1989

  23. Untoro 1999, Use of Oral Iodized Oil to Con­trol Iodine Defi­ciency in Indone­sia, chap­ter 4 “Effect of stunt­ing, iodine defi­ciency and oral iodized oil sup­ple­men­ta­tion on cog­ni­tive per­for­mance of school chil­dren in an endemic iodine defi­cient area of Indone­sia”, pg69–87

  24. Fier­ro-Benítez et al 1972 & Fier­ro-Benítez et al 1974, which are fol­lowups to the inter­ven­tion reported in Fier­ro-Benítez et al 1968 & (for a review of the Ecuado­rian exper­i­ments, see Greene 1994); a ground­break­ing series of stud­ies begin­ning in the 1960s in which 8 goi­ter­ous vil­lages in Ecuador were treated with iodine injec­tions and sub­se­quent long-term out­comes mea­sured through to the 1980s; for the most part, stud­ied chil­dren were affected pre­na­tally

  25. Gli­noer et al 1995, “A ran­dom­ized trial for the treat­ment of mild iodine defi­ciency dur­ing preg­nan­cy: mater­nal and neona­tal effects”: preg­nant women only, no cog­ni­tive mea­sures

  26. Liesenkot­ter et al 1996, “Ear­li­est pre­ven­tion of endemic goi­ter by iodine sup­ple­men­ta­tion dur­ing preg­nancy”: ditto

  27. Nøhr et al 2000, “Post­par­tum thy­roid dys­func­tion in preg­nant thy­roid per­ox­i­dase anti­body-pos­i­tive women liv­ing in an area with mild to mod­er­ate iodine defi­cien­cy: is iodine sup­ple­men­ta­tion safe?”: ditto

  28. Ped­er­sen et al 1993, “Ame­lio­ra­tion of some preg­nan­cy-as­so­ci­ated vari­a­tions in thy­roid func­tion by iodine sup­ple­men­ta­tion”: ditto

  29. Silva & Silva 1981, “Inter­re­la­tion­ships among serum thy­rox­ine, tri­iodothy­ronine, reverse tri­iodothy­ronine, and thy­roid-s­tim­u­lat­ing hor­mone in iodine-d­e­fi­cient preg­nant women and their off­spring: effects of iodine sup­ple­men­ta­tion”: ditto

  30. Man­gani et al 2013, “Pro­vid­ing lipid-based nutri­ent sup­ple­ments does not affect devel­op­men­tal mile­stones among Malaw­ian chil­dren”: inter­ven­tion included iodine (90mcg per 150g con­tain­er), but the devel­op­ment mile­stones are too differ­ent from IQ or cog­ni­tion to use for this

  31. San­ti­ago et al 2013, “Infant neu­rocog­ni­tive devel­op­ment is inde­pen­dent of the use of iodised salt or iodine sup­ple­ments given dur­ing preg­nancy”: preg­nancy only

  32. Lavy et al 2015, “Out of Africa: Human Cap­i­tal Con­se­quences of In Utero Con­di­tions”: an analy­sis of an intrigu­ing , “”, in which 14k Ethiopi­ans (the entire Ethiopian Jew­ish com­mu­ni­ty) were over 2 days air­lifted from Ethiopia to Israel and set­tled there. Since a fair num­ber of the women were at vary­ing stages of preg­nan­cy, the abrupt­ness & com­pre­hen­sive­ness of the relo­ca­tion means that the causal effect of bet­ter Israeli pre­na­tal care can be inferred with­out any of the usual con­founders which would defeat an attempt to com­pare Ethiopian school or test scores with Israeli. They find large effects, espe­cially for those air­lifted dur­ing the first trimester of preg­nancy & for females (which is con­sis­tent with the run­ning theme of iodine stud­ies: that the ben­e­fits are largest when done early dur­ing preg­nancy and larger for females than males). Unfor­tu­nate­ly, it can’t be included as the design is too differ­ent, noth­ing sim­i­lar to IQ tests were admin­is­tered, and the effects reflect all the fac­tors entailed in mov­ing to Israel, which highly likely goes well beyond just a mul­ti­vi­t­a­min.

  33. Sangün et al 2015, “The Effects of Six-Month L-Thy­rox­ine Treat­ment on Cog­ni­tive Func­tions and Even­t-Re­lated Brain Poten­tials in Chil­dren with Sub­clin­i­cal Hypothy­roidism”: non-ran­dom­ized

  34. Prado et al 2016, “Effects of mater­nal and child lipid-based nutri­ent sup­ple­ments on infant devel­op­ment: a ran­dom­ized trial in Malawi”: ran­dom­ized, and includes chil­dren 6–18­mo, but assessed only on devel­op­ment at 18mo, which can’t be con­sid­ered a use­ful IQ test

  35. Li et al 2015, “Pre­na­tal Micronu­tri­ent Sup­ple­men­ta­tion Is Not Asso­ci­ated with Intel­lec­tual Devel­op­ment of Young School-Aged Chil­dren”: 3 exper­i­men­tal groups: folate, folate+iron, folate+iron+others+150.0mg/d iodine, allow­ing a com­par­i­son of folate with folate+­mul­ti­vi­t­a­m­in; but only pre­na­tal inter­ven­tion was done.

  36. Zahrou et al 2016, “For­ti­fied Iodine Milk Improves Iodine Sta­tus and Cog­ni­tive Abil­i­ties in School­child­ren Aged 7–9 Years Liv­ing in a Rural Moun­tain­ous Area of Morocco” (fol­lowup to Zahrou et al 2015): rel­e­vant

  37. “A clus­ter-ran­dom­ized, con­trolled trial of nutri­tional sup­ple­men­ta­tion and pro­mo­tion of respon­sive par­ent­ing in Mada­gas­car: the MAHAY study design and ratio­nale”, Fer­nald et al 2016: pre­reg­is­tra­tion of a study pro­to­col for a Mada­gas­car study; daily 90μg iodine in the lipid-based T2/T3 arms vs T0/T1/T4 for chil­dren aged 6–18 months, devel­op­men­tal out­comes “Ages and Stages Ques­tion­naire: Inven­tory” & “Bay­ley Scale” but no IQ test-ana­logue.

  38. “Effect of iodine sup­ple­men­ta­tion dur­ing preg­nancy on thy­roid func­tion and cog­ni­tive devel­op­ment of off­spring”, Jaiswal (Ban­ga­lore Indi­a): preg­nant only, no IQ tests

  39. Solon et al 2003, “Effect of a mul­ti­ple-mi­cronu­tri­en­t-for­ti­fied fruit pow­der bev­er­age on the nutri­tion sta­tus, phys­i­cal fit­ness, and cog­ni­tive per­for­mance of school­child­ren in the Philip­pines”: RCT in Philip­pine school-chil­dren (mean age 9.9); 4 groups total (for­ti­fied sup­ple­ment x deworm­ing), 10.8mg total iodine deliv­ered over 16 weeks; Pri­mary Men­tal Abil­i­ties Test for Fil­ipino Chil­dren (PMAT-FC) used to mea­sure cog­ni­tive per­for­mance

  40. Lien et al 2009, “Impact of milk con­sump­tion on per­for­mance and health of pri­mary school chil­dren in rural Viet­nam”: claims cog­ni­tive ben­e­fits, espe­cially to short­-term mem­o­ry, but paper does not report quan­ti­ta­tive results

  41. Prado et al 2016, “Effects of pre- and post-na­tal lipid-based nutri­ent sup­ple­ments on infant devel­op­ment in a ran­dom­ized trial in Ghana” & Adu-A­far­wuah et al 2016, : mea­sured “motor, lan­guage, socio-e­mo­tion­al, and exec­u­tive func­tion at 18 months” but too early for valid IQ tests, and the chil­dren appear to have received sup­ple­men­ta­tion pre-na­tally

  42. Zhou et al, “The effect of iodine sup­ple­men­ta­tion in preg­nancy on early child­hood neu­rode­vel­op­ment and clin­i­cal out­comes: results of an aborted ran­domised place­bo-con­trolled trial”: preg­nancy only

  43. Pearce 2017, “Iodine Sup­ple­men­ta­tion for Pre­ma­ture Infants Does Not Improve IQ”: Bayley-III cog­ni­tive tests at age 2, no sta­tis­ti­cal­ly-sig­nifi­cant differ­ence or trend for the iodine-sup­ple­mented chil­dren (30mcg/kg/day for >3 weeks; pre­ma­ture infants can weigh any­where 0.5—2.5kg) . While the sup­ple­men­ta­tion is post-preg­nan­cy, pre­ma­ture infants are an unusual enough patient pop­u­la­tion (and one which fre­quently ) that it prob­a­bly is not a good idea to meta-an­a­lyze it with RCTs on nor­mal chil­dren & adults.

  44. Haas et al 2014/Wenger et al 2017, “Con­sump­tion of a Dou­ble-For­ti­fied Salt Affects Per­cep­tu­al, Atten­tion­al, and Mnemonic Func­tion­ing in Women in a Ran­dom­ized Con­trolled Trial in India”: RCT of iron+io­dine sup­ple­ment in 18–55yo, but where the con­trol also received iodized salt and the large ben­e­fits were due to reduc­ing ane­mia

  45. , Ser­ena 2019; nat­ural exper­i­ment using the Dan­ish gov­ern­men­t’s legal­iza­tion of and then require­ment to iodize salt, 1998–2001, with a differ­ence-in-d­iffer­ences analy­sis sug­gest­ing d = 0.06 on affected ado­les­cents’ high school GPA. (Not a RCT, and differ­ence-in-d­iffer­ences are one of the weaker nat­ural exper­i­ment design­s.)

My above searches & review turned out to be par­tially redun­dant with a review that was pub­lished online a few months after­wards, in June 2012: Zim­mer­man’s “The Effects of Iodine Defi­ciency in Preg­nancy and Infancy” and also his ear­lier 2009 “Iodine defi­ciency in preg­nancy and the effects of mater­nal iodine sup­ple­men­ta­tion on the off­spring: a review”. Regard­less, the above was still use­ful because Zim­mer­man’s focus was not on any child­hood or adult effects but the preg­nancy & infancy effects. I have also ben­e­fited from . Final­ly, “Impact of iodine sup­ple­men­ta­tion in mild-to-mod­er­ate iodine defi­cien­cy: sys­tem­atic review and meta-analy­sis” (Tay­lor et al 2014) reviews iodine stud­ies and for cog­ni­tive per­for­mance, meta-an­a­lyzes Gor­don et al 2009 & Zim­mer­man et al 2009, find­ing ben­e­fits.

Unfor­tu­nate­ly, a few of the use­ful stud­ies also tested a vari­ety of sub­stances, so any esti­mate of iodine’s effect size (such as it is) would be an over­es­ti­mate: the iodine might be syn­er­giz­ing with one of the other ingre­di­ents (eg. iron or ) or the iodine might not be the respon­si­ble agent at all! We will code that covari­ate in. But what impli­ca­tions can we draw out via a ?

Data

First, the data from the sur­viv­ing stud­ies:

study group year n.e mean.e sd.e n.c mean.c sd.c age dose multi coun­try
Bautista Bautista 1982 100 69.43 10.96 100 70.31 10.96 8.75 475 0 Bolivia
Boy­ages Boy­ages 1990 28 34.4 15.4 24 29.5 8.5 29.5 720 0 China
Schoen­thaler.1 Schoen­thaler 1991 100 10.1 8.9 33.3 8.9 7.3 14 12.68 1 USA
Schoen­thaler.2 Schoen­thaler 1991 105 12.6 7.9 33.3 8.9 7.3 14 25.35 1 USA
Schoen­thaler.3 Schoen­thaler 1991 105 10.4 7.6 33.3 8.9 7.3 14 25.35 1 USA
Southon Southon 1994 22 63.94 2.6 29 64.1 1.87 13.5 16.8 1 UK
Shrestha.1 Shrestha 1994 72 17.1 1.8 36 10.7 2.4 7.1 490 0 Malawi
Shrestha.2 Shrestha 1994 80 18.2 1.5 36 10.7 2.4 7.1 490 0 Malawi
Isa Isa 2000 60 85.25 14.6 100 83.6 16.2 11.39 480 0 Malaysia
Huda Huda 2001 145 14.88 3.28 142 14.60 3.19 9.8 400 0 Bangladesh
Zim­mer­man Zim­mer­man 2006 159 25 6.3 151 20.5 5.6 11 400 0 Alba­nia
McNeill McNeill 2007 398 11.5 2.3 374 11.7 2.1 72 54.75 1 UK
Gor­don Gor­don 2009 84 10.2 3 82 9.6 2.4 11.5 29.4 0 NZ
Dewi Dewi 2012 33 110.27 9.04 34 103.06 9.99 3.1 8.4 0 Indone­sia
Salarkia Salarkia 2004 19 96 10 246 89 13 1.5 1272 0 Iran
Untoro Untoro 1999 121 89 7 43 88 6 9 464 0 Indone­sia
Red­man Red­man 2011 86 20.85 3.43 86 20.59 3.27 21.28 45 0 NZ
Solon Solon 2003 412 1.82 0.22 419 1.66 0.22 9.9 10.752 1 Philip­pines

Com­ments:

  1. “age” vari­able is based on a sim­ple aver­age age in years of sub­jects (un­weight­ed) or the reported mean; “dose” is total admin­is­tered iodine in mil­ligrams (note that stud­ies typ­i­cally report in micrograms/μg per day or week); multi is whether the study used solely iodine (0) or other chem­i­cals such as iron (1)

  2. Bautista notes: Stan­ford-Bi­net IQ, some data re-derived15

  3. Boy­ages: the two groups were not ran­domly cho­sen and may have pre-ex­ist­ing differ­ences in IQ

  4. Schoen­thaler: RAPM IQ

  5. Cao: an ear­lier ver­sion of this meta-analy­sis used its “devel­op­men­tal quo­tient” exclud­ing the preg­nant wom­en’s off­spring; since this does not seem to be IQ, I removed it

  6. Southon: Non-ver­bal IQ scores pooled16

  7. McNeill: used in place of IQ scores (WM cor­re­lates highly with IQ)

  8. Gor­don: score from matrix sub­test (1 of 4 sub­test scores; 2 showed pos­i­tive trend but not sig­nifi­cance)

  9. Manger: ear­lier ver­sion used Manger’s back­wards digit span (back­wards is a WM test which loads on g); how­ev­er, the meta-analy­sis & my own found that WM exer­cises which trans­ferred to digit spans did not also trans­fer to IQ tests, rais­ing ques­tions about using digit span as a proxy for IQ in a causal rather than cor­re­la­tional analy­sis

  10. Stu­i­jven­berg: see Manger

  11. Isa: IQ scores reported in unhelp­ful for­mat; means & devi­a­tions reverse-engi­neered from per­centile dis­tri­b­u­tion17; as usu­al, preg­nan­cy-re­lated scores are omit­ted

  12. Shrestha: ‘Fluid intel­li­gence’ scores report­ed, omit­ting ‘Crys­tal­lized intel­li­gence’ & ‘Per­cep­tual skill’; the con­trol group is split across the iodine inter­ven­tion (“Shrestha.1”) and the iodine+­iron inter­ven­tion (“Shrestha.2”)

  13. Salarkia:

    1. the paper’s orig­i­nal con­trol group using con­tem­po­rary age/sex-matched Tehran chil­dren, does­n’t account for their supe­rior IQ scores and likely supe­rior SES; I have instead used the reported 1989 IQ scores of the pre­vi­ous gen­er­a­tion of chil­dren
    2. the admin­is­tered iodine includes the 480mg from iodized oil but also the 6 years of iodized salt con­sump­tion (40mg/kg, national daily per capita aver­age salt con­sump­tion 9g) or 132mg a year) for a total of 1272mg
  14. Untoro: there were 3 dose groups (200/400/800mg) but Untoro reported sum­mary sta­tis­tics for the iodine group as a whole:

    The uri­nary iodine con­cen­tra­tion and thy­roid vol­ume of all treat­ment groups were [sta­tis­ti­cal­ly-]sig­nifi­cantly improved after the sup­ple­men­ta­tion, but there were no differ­ences among the sup­ple­mented groups in cog­ni­tive per­for­mance. There­fore we com­bined the three iodized oil sup­ple­mented groups into one group.

    The listed dose is a weighted aver­age.

  15. Red­man: dose is cal­cu­lated as: 150 μg, 100 pills a bot­tle, 3 bot­tles (1 ini­tial bot­tle, 2 replace­ments in mail) = = 45000μg; scores are from the Matrix Rea­son­ing sub­test

The result of the meta-analy­sis:

Random-Effects Model (k = 17; tau^2 estimator: REML)

tau^2 (estimated amount of total heterogeneity): 1.2100 (SE = 0.4432)
tau (square root of estimated tau^2 value):      1.1000
I^2 (total heterogeneity / total variability):   97.87%
H^2 (total variability / sampling variability):  47.02

Test for Heterogeneity:
Q(df = 16) = 285.8785, p-val < .0001

Model Results:

estimate       se     zval     pval    ci.lb    ci.ub
  0.6204   0.2716   2.2844   0.0223   0.0881   1.1527
Test for Heterogeneity:
Q(df = 16) = 285.8785, p-val < .0001

Model Results:

estimate       se     zval     pval    ci.lb    ci.ub
  0.2565   0.0380   6.7536   <.0001   0.1820   0.3309

So the effect size is, as expect­ed, small: d = 0.2. A far cry from the d > 1 which we might esti­mate from the pre­na­tal stud­ies.

The large differ­ence in d—0.2 vs 0.6—­be­tween the fixed-effects and ran­dom-effects mod­els is con­cern­ing. Given the extremely high het­ero­gene­ity of the i2, which indi­cates that there are large differ­ences between some of the stud­ies, a ran­dom-effects is more appro­pri­ate in prin­ci­ple; but fur­ther analy­sis shows this is being dri­ven by a far out­lier of Shrestha, and so I believe the fixed-effects esti­mate of 0.2 winds up being more accu­rate.

A pretty sum­ma­ry:

A for­est plot of iodine stud­ies

Moderators

Age & dose

We sus­pect­ed, based on the equiv­o­cal results in post-birth stud­ies and the large decline in effect over the dura­tion of a preg­nancy in the Chi­nese stud­ies, that if there was any ben­e­fit, it would be in the young; on the same rea­son­ing, we might expect large doses to do more good than smaller ones. The nec­es­sary data is encoded into the table already, so we run a meta-an­a­lytic regres­sion on them as inde­pen­dent pre­dic­tors:

tau^2 (estimated amount of residual heterogeneity):     1.2521 (SE = 0.4902)
tau (square root of estimated tau^2 value):             1.1190
I^2 (residual heterogeneity / unaccounted variability): 97.49%
H^2 (unaccounted variability / sampling variability):   39.92
R^2 (amount of heterogeneity accounted for):            0.00%

Test for Residual Heterogeneity:
QE(df = 14) = 243.6469, p-val < .0001

Test of Moderators (coefficient(s) 2,3):
QM(df = 2) = 1.6083, p-val = 0.4475

Model Results:

             estimate      se     zval    pval    ci.lb   ci.ub
intrcpt        0.6793  0.5205   1.3053  0.1918  -0.3407  1.6994
iodine$age    -0.0157  0.0183  -0.8568  0.3915  -0.0515  0.0202
iodine$dose    0.0006  0.0009   0.6665  0.5051  -0.0011  0.0023

The coeffi­cients and vari­abil­ity are dis­ap­point­ing: the age and dose mod­er­a­tors explain lit­tle of what is going on.

Some graphs to help us visu­al­ize. Graph­ing by age, we see what might be a slight neg­a­tive rela­tion­ship, as the regres­sion sug­gests (driven by Shrestha 1994):

plot(iodine$age, effects$yi)

Graph­ing by dose, we see—de­spite the cal­cu­lated sig­nifi­cance—no rela­tion­ship at all to my eyes (out­liers are Shrestha 1994, again):

plot(iodine$dose, effects$yi)

The estimate is the impor­tant part: nei­ther of the mod­er­a­tors seem to have a strong rela­tion­ship with the cog­ni­tive ben­e­fits (nor are they at least sta­tis­ti­cal­ly-sig­nifi­can­t). It would seem that any effec­tive­ness of iodine is unre­lated to age and dose. Whether the effec­tive­ness is being dri­ven by a few stud­ies is what we’ll look at next.

Multiple supplements

One method­olog­i­cal con­cern is that by includ­ing stud­ies like Southon which sup­ple­mented many things besides iodine, our results are merely pick­ing up the effi­cacy of other sup­ple­ments (iron is a par­tic­u­lar con­cern). Curi­ous­ly, despite our expec­ta­tion that the mul­ti­-vi­t­a­min stud­ies would have higher effect sizes because any of the ingre­di­ents could be help­ful singly or syn­er­gis­ti­cal­ly, it is strik­ingly the oppo­site:

QM(df = 2) = 6.7027, p-val = 0.0350

Model Results:

                       estimate      se    zval    pval    ci.lb   ci.ub
factor(iodine$multi)0    0.8250  0.3203  2.5756  0.0100   0.1972  1.4529
factor(iodine$multi)1    0.1300  0.4953  0.2625  0.7930  -0.8408  1.1008

Bias checks

We do not have enough to reli­ably check for biases like , but we can still try. The looks pretty bizarre, with almost all stud­ies tightly bunched around the null effect but 2 out­lier­s—the 2 Shrestha 1994 groups we just saw—show­ing shock­ingly high effect sizes of d = 3.1/4.1. Why did Shrestha 1994 observe such large IQ effects? I don’t know, although the Malawi region was cho­sen for its iodine defi­cien­cy.

A test & graph:

test for funnel plot asymmetry: z = 3.3206, p = 0.0009
A fun­nel plot of effect size ver­sus sam­ple size, check­ing for bias in pub­lish­ing only good-look­ing results on iodine

So Shrestha 1994 is dri­ving most of the effect! (Not a sur­prise at this point.) If we omit those points, the fun­nel plot is cleaner

A fun­nel plot with Shrestha removed, show­ing bet­ter fit from the weaker esti­mate.

A trim-and-fill check agrees with us and not the test, by decid­ing not to add in any new stud­ies between Shrestha and the rest; we also notice that the τ2 & i2 are extremely high, which is just telling us what we know—Shrestha 1994 is differ­ent from the other stud­ies:

Estimated number of missing studies on the left side: 0 (SE = NA)

Random-Effects Model (k = 17; tau^2 estimator: REML)

tau^2 (estimated amount of total heterogeneity): 1.2100 (SE = 0.4432)
tau (square root of estimated tau^2 value):      1.1000
I^2 (total heterogeneity / total variability):   97.87%
H^2 (total variability / sampling variability):  47.02

If we were to redo the analy­sis but omit­ting Shrestha 1994, we see a much smaller effec­t-size:

Random-Effects Model (k = 15; tau^2 estimator: REML)

tau^2 (estimated amount of total heterogeneity): 0.0526 (SE = 0.0313)
tau (square root of estimated tau^2 value):      0.2293
I^2 (total heterogeneity / total variability):   68.75%
H^2 (total variability / sampling variability):  3.20

Test for Heterogeneity:
Q(df = 14) = 53.0253, p-val < .0001

Model Results:

estimate       se     zval     pval    ci.lb    ci.ub
  0.2224   0.0751   2.9602   0.0031   0.0752   0.3697

So, it seems a lot of the effect size is being dri­ven by a few stud­ies turn­ing in large or extremely large effect sizes: Shrestha, Zim­mer­man, and Dewi (in chrono­log­i­cal order). One com­mon fac­tor to these stud­ies seems to be that they worked in the most iodine-de­prived areas pos­si­ble: Shrestha chose his region as being the worst he could find, and Dewi cites reports of goi­ter in half the pop­u­la­tion while tar­get­ing the sup­ple­men­ta­tion to the most defi­cient chil­dren.

At most, we can jus­tify an effect esti­mate which is much smaller than would be esti­mated based on the pre­na­tal stud­ies, and the real­ity of this resid­ual effect can be doubt­ed: look­ing at the for­est plot, the more rig­or­ous and West­ern stud­ies seem to have the tini­est and clos­est to zero esti­mates.

Code

Run as R --slave --file=iodine.r:

set.seed(7777) # for reproducible numbers
# TODO: factor out common parts of `png` (& make less square), and `rma` calls
# install.packages("XML") # if not installed
library(XML)
iodine <- readHTMLTable(colClasses = c("character", "integer", rep("numeric", 8)),
                     "https://www.gwern.net/Iodine#data")[[1]]
# install.packages("metafor") # if not installed
library(metafor)

effects <- escalc("SMD", m1i = mean.e, m2i = mean.c, sd1i = sd.e, sd2i = sd.c, n1i = n.e, n2i = n.c,
                  data = iodine)

cat("Basic random-effects meta-analysis of all studies:\n")
res1 <- rma(yi, vi, data = effects); res1

cat("Fixed-effects version:\n")
res2 <- rma(yi, vi, data = effects, method="FE"); res2

png(file="~/wiki/images/iodine/forest.png", width = 480, height = 480)
forest(res1, slab = paste(iodine$study, iodine$year, sep = ", "))
invisible(dev.off())

cat("Random-effects with age & dose moderators:\n")
rma(yi, vi, mods = ~ iodine$age + iodine$dose, data = effects)

png(file="~/wiki/images/iodine/age.png", width = 480, height = 480)
plot(iodine$age, effects$yi)
invisible(dev.off())

png(file="~/wiki/images/iodine/dose.png", width = 480, height = 480)
plot(iodine$dose, effects$yi)
invisible(dev.off())

cat("Check random-effects meta-analysis for whether the multiple-supplement trials are inflating results:\n")
rma(yi, vi, mods = ~ factor(iodine$multi) - 1, data = effects)

png(file="~/wiki/images/iodine/funnel.png", width = 480, height = 480)
funnel(res1)
invisible(dev.off())

cat("The funnel plot looks weird, test it:\n")
regtest(res1, model = "rma", predictor = "sei", ni = NULL)

cat("For kicks, what does a (normal) trim-and-fill say?\n")
tf <- trimfill(res1); tf

cat("Re-run random-effects meta-analysis, omitting Shrestha:\n")
res3 <- rma(yi, vi, data = escalc("SMD", m1i = mean.e, m2i = mean.c, sd1i = sd.e, sd2i = sd.c,
                                  n1i = n.e, n2i = n.c,
                                  data = iodine[!(grepl("Shrestha", iodine$study)),])); res3

# funnel plot with Shrestha omitted
png(file="~/wiki/images/iodine/funnel-moderated.png", width = 480, height = 480)
funnel(res3)
invisible(dev.off())

# optimize the generated graphs by cropping whitespace & losslessly compressing them
system(paste('cd ~/wiki/images/iodine/ &&',
             'for f in *.png; do convert "$f" -crop',
             '`nice convert "$f" -virtual-pixel edge -blur 0x5 -fuzz 10% -trim -format',
             '\'%wx%h%O\' info:` +repage "$f"; done'))
system("optipng -o9 -fix ~/wiki/images/iodine/*.png", ignore.stdout = TRUE)

See Also

  • Iodine sec­tion of Nootrop­ics essay -(mo­ti­va­tion; some sta­tis­ti­cal power cal­cu­la­tions for a self­-ex­per­i­ment look­ing for cog­ni­tive improve­ment; a “value of infor­ma­tion” cal­cu­la­tion on the pos­si­ble ben­e­fit ver­sus the qual­ity of infor­ma­tion from self­-ex­per­i­ment­ing; a silly eye­-color exper­i­ment result­ing in no change from sup­ple­men­ta­tion)

  1. See Scrimshaw 1998; Ble­ichrodt & Born 1994; and Qian et al 2005:

    Thir­ty-seven reported stud­ies, total 12,291 chil­dren, were analysed. The effect size was an increase of 0.83, 0.82, and 0.32 SIQP respec­tive­ly, for the chil­dren liv­ing in IS com­mu­ni­ties com­pared with those liv­ing in ID areas with no iodine sup­ple­men­ta­tion, with inad­e­quate iodine sup­ple­men­ta­tion, or chil­dren who had received iodine dur­ing their moth­ers’ preg­nancy and after birth. These equal to 12.45, 12.3, 4.8 IQ points. Com­pared with that of chil­dren whose moth­ers were per­sis­tently exposed to ID, the total effect size of the 21 entries was an increase of 0.58 SIQP (8.7 IQ points) in the group receiv­ing iodine sup­ple­men­ta­tion dur­ing preg­nan­cy. Fur­ther­more, there was an increase on 1.15 SIQP of Binet or 0.8 SIQP on Raven Scale (17.25 or 12 IQ points) for chil­dren born more than 3.5 years after iodine sup­ple­men­ta­tion pro­gram was intro­duced. The level of iodine nutri­tion plays a cru­cial role in the intel­lec­tual devel­op­ment of chil­dren. The intel­li­gence dam­age of chil­dren exposed to severe ID was pro­found, demon­strated by 12.45 IQ points loss and they recov­ered 8.7 IQ points with iodine sup­ple­men­ta­tion or IS before and dur­ing preg­nancy

    ↩︎
  2. , Jones 2006:

    In a recently pub­lished paper, Erica Field, Omar Rob­les, and Max­imo Torero study the effects of iodine defi­ciency on edu­ca­tional attain­ment in Tan­za­nia (Field, Rob­les and Torero 2009). Theirs is the first pub­lished paper to link eco­nom­i­cally [im­por­tant] out­comes to iodine defi­cien­cy, and it sup­ports the idea of improved school­ing out­comes after iodine sup­ple­men­ta­tion. Inter­est­ing­ly, Field et al. also find big­ger effects for females, which could be due either to bio­log­i­cal differ­ences or differ­ences asso­ci­ated with intra-house­hold resource allo­ca­tion and social responses to increased cog­ni­tion of males and females.

    The orig­i­nal:

    Our find­ings sug­gest that reduc­ing fetal IDD has [sub­stan­tial] ben­e­fits for child cog­ni­tion: Chil­dren who receive iodine in utero attain an aver­age of 0.33 years of edu­ca­tion above sib­lings and older and younger chil­dren in their dis­trict. Fur­ther­more, the effects appear to be sub­stan­tially larger for girls, con­sis­tent with new evi­dence from lab­o­ra­tory stud­ies in ani­mals indi­cat­ing greater cog­ni­tive sen­si­tiv­ity of the female fetus to in utero iodine depri­va­tion, includ­ing sex-spe­cific responses to mater­nal thy­roid hor­mone restric­tion on expres­sion of neural thy­roid hor­mone recep­tors. The results are con­sis­tent across house­hold and dis­trict fixed effects mod­els and pat­terns of vari­a­tion in esti­mated effects are con­sis­tent with pre­dic­tions regard­ing the vul­ner­a­bil­ity of spe­cific sub­pop­u­la­tions to fetal IDD. Cross-coun­try regres­sion esti­mates indi­cate a strong neg­a­tive influ­ence of total goi­ter rate and strong pos­i­tive influ­ence of salt iodiza­tion on female school par­tic­i­pa­tion.

    ↩︎
  3. An exam­ple of the com­mer­cial oppo­si­tion or con­spir­acy the­o­rists is the Kaza­khstan cam­paign; from “In Rais­ing the World’s I.Q., the Secret’s in the Salt”, NYT:

    In fact, Kaza­khstan has become an exam­ple of how even a vast and stil­l-de­vel­op­ing nation like this Cen­tral Asian coun­try can achieve a remark­able pub­lic health suc­cess. In 1999, only 29% of its house­holds were using iodized salt. Now, 94% are. Next year, the United Nations is expected to cer­tify it offi­cially free of iodine defi­ciency dis­or­der­s…Largely out of the pub­lic eye, they made ter­rific pro­gress: 25% of the world’s house­holds con­sumed iodized salt in 1990. Now, about 66% do.

    …In some nations, iodiza­tion becomes tarred as a gov­ern­ment plot to poi­son an essen­tial of life—salt experts com­pare it to the furi­ous oppo­si­tion by 1950s con­ser­v­a­tives to flu­o­ri­da­tion of Amer­i­can water. In oth­ers, civil lib­er­tar­i­ans demand a right to choose plain salt, with the result that the iodized kind rarely reaches the poor. Small salt mak­ers who fear extra expense often lobby against it. So do mak­ers of iodine pills who fear los­ing their mar­ket. Rumors inevitably swirl: iodine has been blamed for AIDS, dia­betes, seizures, impo­tence and peev­ish­ness. Iodized salt, accord­ing to differ­ent national rumor mills, will make pick­led veg­eta­bles explode, ruin caviar or soften hard cheese. Break­ing down that resis­tance takes both money and lead­er­ship…The coun­try’s biggest salt com­pany was ini­tially reluc­tant to coop­er­ate, fear­ing higher costs, a Unicef report said. Car­di­ol­o­gists argued against iodiza­tion, fear­ing it would encour­age peo­ple to use more salt, which can raise blood pres­sure. More insid­i­ous, Dr. Shar­manov said, were pri­vate com­pa­nies that sold iodine pills. “They pro­moted their prod­ucts in the mass media, say­ing iodized salt was dan­ger­ous,” he said, shak­ing his head­…Asked about the Unicef report say­ing that Aral Tuz ini­tially resisted iodiza­tion on the grounds that it would eat up 7% of profits, the com­pa­ny’s pres­i­dent, Onta­lap Akhme­tov, seemed puz­zled. “I’ve only been pres­i­dent three years,” he said. “But that makes no sense.” The expense, he said, was min­i­mal. “Only a few cents a ton.”

    …In the 1990s, when the cam­paign for iodiza­tion began, the world’s great­est con­cen­tra­tion of iodine-d­e­fi­cient coun­tries was in the land­locked for­mer Soviet republics of Cen­tral Asia. All of them—Kaza­khstan, Turk­menistan, Tajik­istan, Uzbek­istan, Kyr­gyzs­tan—­saw their economies break down with the col­lapse of the Soviet Union. Across the region, only 28% of all house­holds used iodized salt. “With the col­lapse of the sys­tem, cer­tain babies went out with the bath­wa­ter, and iodiza­tion was one of them,” said Alexan­dre Zouev, chief Unicef rep­re­sen­ta­tive in Kaza­khstan…That com­bi­na­tion is miss­ing in many nations where iodine defi­ciency remains a health cri­sis. In nearby Pak­istan, for instance, where 70% of house­holds have no iodized salt, there are more than 600 small salt pro­duc­ers. [see “Pak­istan takes legal action”, Unicef; for fur­ther read­ing, the WHO’s Assess­ment of iodine defi­ciency dis­or­ders and mon­i­tor­ing their elim­i­na­tion]

    ↩︎
  4. aver­age daily Japan­ese con­sump­tion of iodine ranges from 5.28-13.8mg of iodine; Patrick L. Iodine: defi­ciency and ther­a­peu­tic con­sid­er­a­tions, although Gaby argues this is a mis­in­ter­pre­ta­tion: “the idea that Japan­ese peo­ple con­sume 13.8 mg of iodine per day appears to have arisen from a mis­in­ter­pre­ta­tion of a paper. In that paper, the aver­age intake of sea­weed in Japan was listed as 4.6 g (4,600 mg) per day, and sea­weed was said to con­tain 0.3% iodine. The fig­ure of 13.8 mg comes from mul­ti­ply­ing 4,600 mg by 0.003. How­ev­er, the 4.6 g of sea­weed con­sumed per day was expressed as wet weight, whereas the 0.3%-io­dine fig­ure based on dry weight. Since many veg­eta­bles con­tain at least 90% water, 13.8 mg per day is a [large] over­es­ti­mate of iodine intake. In stud­ies that have specifi­cally looked at iodine intake among Japan­ese peo­ple, the mean dietary intake (es­ti­mated from uri­nary iodine excre­tion) was in the range of 330 to 500 mcg per day, which is at least 2.5-fold lower than 13.8 mg per day.” Abra­ham replies “The uri­nary excre­tions of iodide reported by Ishizuki et al in ref­er­ences 7 and 8 of Gaby’s edi­to­r­ial are obvi­ously from patients who were told to abstain from sea­weed. We will quote other stud­ies by Japan­ese inves­ti­ga­tors who reported uri­nary iodide lev­els 10 to 100 time higher than the val­ues reported by Ishizuki et al .”↩︎

  5. , Politi 2014 (see also Field et al 2009):

    I iden­tify the impact of iodiza­tion in three ways: first, in a differ­ences-in-d­iffer­ences frame­work, I exploit geo­graphic vari­a­tion in iodine defi­cien­cy, as well as the fact that the nation­wide cam­paign to decrease iodine defi­ciency began in 1922. Sec­ond, I use spa­tial and tem­po­ral vari­a­tion in the intro­duc­tion of iodized salt across Swiss can­tons, and exam­ine whether the level of iodized salt sales at the time of one’s birth affected one’s edu­ca­tional attain­ment. Third, I employ a fuzzy regres­sion dis­con­ti­nu­ity design and use jumps in sales of iodized salt across Swiss can­tons to iden­tify the effect of iodiza­tion, by com­par­ing out­comes for those born right before and right after these sud­den changes in the treat­ment envi­ron­ment. These approaches indi­cate that the erad­i­ca­tion of iodine defi­ciency in pre­vi­ously defi­cient areas increased the school­ing of the pop­u­la­tion [sta­tis­ti­cal­ly-]sig­nifi­cant­ly. The effects are larger for females than for males, which is con­sis­tent with med­ical evi­dence show­ing that women are more likely to be affected by iodine defi­ciency dis­or­ders than men.

    ↩︎
  6. “Salt Iodiza­tion and the Enfran­chise­ment of the Amer­i­can Worker”, Adhvaryu et al 2013:

    …We find sub­stan­tial impacts of salt iodiza­tion. High school com­ple­tion rose by 6 per­cent­age points, and labor force par­tic­i­pa­tion went up by 1 point. Analy­sis of income tran­si­tions by quan­tile shows that the new labor force join­ers entered at the bot­tom of the wage dis­tri­b­u­tion and took up blue col­lar labor, pulling down aver­age wage income con­di­tional on employ­ment. Our results inform the ongo­ing debate on salt iodiza­tion in many low-in­come coun­tries. We show that large-s­cale iodized salt dis­tri­b­u­tion had a tar­geted impact, ben­e­fit­ing the worker on the mar­gin of employ­ment, and gen­er­at­ing size­able eco­nomic returns at low cost.

    ↩︎
  7. Politi 2015, ↩︎

  8. “The Eco­nomic Effects of Micronu­tri­ent Defi­cien­cy: Evi­dence from Salt Iodiza­tion in the United States”, Feyrer et al 2008 (quotes are from 2008 draft, not 2013 final ver­sion):

    For instance, Sachs (2003) shows that efforts to elim­i­nate malaria have sub­stan­tial effects on income, through their effect on health, reduced absen­teeism, etc. Ace­moglu and John­son (2007) look at the effect of health inter­ven­tions and find lit­tle effect. Weil (2007) charts a mid­dle course…­For exam­ple, Bleak­ley (2007) exam­ines the effects of hook­worm erad­i­ca­tion in the Amer­i­can South in early 20th cen­tury and finds [sub­stan­tial] effects on edu­ca­tion and future incomes of those cohorts that ben­e­fited from the inter­ven­tion.

    …Fig­ure 1 illus­trates the geo­graphic dis­tri­b­u­tion of goi­ter in the US as mea­sured among World War I recruits (we dis­cuss the data fur­ther below). In 1924 iodized salt was intro­duced in the United States explic­itly to reduce the goi­ter rate. This inter­ven­tion rapidly reduced the inci­dence of iodine defi­cien­cy. Iodiza­tion of salt in the US pro­vides a par­tic­u­larly good nat­ural exper­i­ment due to the geo­graphic dis­tri­b­u­tion of the dis­ease com­bined with a rapid, com­plete treat­ment. Since there are large in utero effects of iodine defi­cien­cy, we should see a [sub­stan­tial] differ­ence between those born before and after the intro­duc­tion of iodized salt in loca­tions with low lev­els of envi­ron­men­tal iodine. Those liv­ing in high iodine regions pro­vide a con­trol group…The prob­a­bil­ity of assign­ment to the Air Corps rises [sub­stan­tial­ly] in low iodine (i.e. high goi­ter) coun­ties in the years after the intro­duc­tion of iodized salt. In the low­est iodine regions, our esti­mates sug­gest a 10–20% increase in the prob­a­bil­ity of a man being assigned to the Air Corps after iodiza­tion. Using infor­ma­tion about aver­age scores of Air and Ground Force recruits we can infer a one-quar­ter to one half stan­dard devi­a­tion increase in aver­age test scores in these regions. The aver­age level of iodine defi­ciency in the US was [sub­stan­tial­ly] lower than in the high­est regions, so the over­all effect in the US was much more mod­est, though iodiza­tion was undoubt­edly extremely cost effec­tive. The increase in cog­ni­tive abil­ity due to salt iodiza­tion may have con­tributed a small amount to trend rise in mea­sured IQ that took place over the course of the twen­ti­eth cen­tu­ry, the so-called Behrman and Rosen­zweig (2004) find that differ­ences in birth weight among iden­ti­cal twins are reflected in differ­ences in school attain­ment and adult earn­ings…­More cites to be added, includ­ing Case & Pax­son.

    Typ­i­cal­ly, non-d­e­fi­cient pop­u­la­tions differ from iodine-d­e­fi­cient pop­u­la­tions by approx­i­mately 10 IQ points, whereas the whole nor­mal IQ dis­tri­b­u­tion of a pop­u­la­tion shifts to the left as a result of iodine defi­ciency7. In endemic areas, cre­tinism can affect up to 15% of the pop­u­la­tion (de Benoist et al., eds 2004). Ble­ichrodt and Born (1994) esti­mate that the aver­age IQ of iodine-d­e­fi­cient groups is 13.5 points lower than the non-d­e­fi­cient groups…Scrimshaw (1998) pro­vides a list of stud­ies and exper­i­ments that have been con­duct­ed, and which have shown the hin­der­ing effects on men­tal devel­op­ment of iodine defi­ciency in utero.

    …It had been doc­u­mented that large amounts of iodine could cause hyper­thy­roidism to develop in some adults and in oth­er­s…thy­ro­tox­i­co­sis might occur as a result of iodiza­tion in those indi­vid­u­als that have suffered from long-term iodine defi­ciency and whose goi­ters have become nodu­lar. In such cas­es, iodine sup­ple­men­ta­tion causes the out­put of hor­mone to jump to toxic lev­el­s…The num­ber of enlisted men started increas­ing in 1942, and it peaked in the sec­ond semes­ter of that year, as well as the first semes­ter of 1943. Peo­ple born after iodiza­tion enlisted in large num­bers start­ing the first semes­ter of 1943. The pro­por­tion of recruits going to the Air Force was par­tic­u­larly low in 1943…How large are these effects? The high­est goi­ter places have goi­ter rates of 30. The coeffi­cients (from Table 4) sug­gest that about 15% more recruits from the high­est goi­ter regions go into the AAF after salt iodiza­tion. From Appen­dix A we know that the Air Force recruits have, on aver­age 9 point higher AGCT scores (al­most a half a stan­dard devi­a­tion). The aver­age increase for the recruits from that sec­tion is there­fore 0.15 time 9 points, or 1.35. So the aver­age cog­ni­tive abil­ity in the sec­tion goes up by greater than one twen­ti­eth of a stan­dard devi­a­tion. Exam­in­ing the high­-low regres­sions yields sim­i­lar fig­ures. In the high goi­ter group, we have a 5–10% higher assign­ment rate to the AAF after salt iodiza­tion. 10% times 8 AGCT points results in 0.8 points higher on aver­age, or about a twen­ti­eth of a stan­dard devi­a­tion. This implies a twen­ti­eth of a stan­dard devi­a­tion increase in cog­ni­tive abil­ity for 25% of the US pop­u­la­tion.

    The final ver­sion finds a more sub­stan­tial effect:

    The esti­mated coeffi­cient on the high­-goi­ter dummy in table 5 is in the range 2.5–8.7 per­cent­age points. Sim­i­lar­ly, the logit spec­i­fi­ca­tion in Table 8 implies a mar­ginal effect of iodiza­tion on the prob­a­bil­ity of enter­ing the AAF in the range of 3–7.8 per­cent­age points. Com­par­ing these esti­mates to Table 9, we see that the lower range of these esti­mates are con­sis­tent with iodiza­tion rais­ing IQ by 15 points (that is, X = 15), which is a rea­son­able expec­ta­tion given the work of (Ble­ichrodt and Born 1994). How­ev­er, the higher range of esti­mates is larger than any rea­son­able esti­mate of the increase in IQ that would have resulted from iodiza­tion. Given the uncer­tainty sur­round­ing the selec­tion process and prior lit­er­a­ture on iodiza­tion we do not con­sider the larger results plau­si­ble. In any case, our results are con­sis­tent with a sub­stan­tial effect in line with the exist­ing lit­er­a­ture.

    But unfor­tu­nate­ly, the detailed mech­a­nisms & psy­cho­met­rics are insuffi­cient to sup­port a con­crete esti­mate of IQ gains, as taken lit­er­al­ly, their results indi­cate an IQ gain some­where around 30 points; this is suffi­ciently large as to prove that the analy­sis is not fully cap­tur­ing the dynam­ics and the esti­mates can­not be taken lit­er­ally aside from prov­ing that there was prob­a­bly a sub­stan­tial gain. (Based on the many prob­lems with the iodiza­tion lit­er­a­ture, sug­gest­ing 15 points is prob­a­bly an over­es­ti­mate.)

    Another inter­est­ing paper exploits the sim­i­larly abrupt intro­duc­tion of iron for­ti­fi­ca­tion into the USA ~WWII: “Iron­ing Out Defi­cien­cies: Evi­dence from the United States on the Eco­nomic Effects of Iron Defi­ciency”, Niemesh 2015.↩︎

  9. “Effects of drink­ing water with high iodine con­cen­tra­tion on the intel­li­gence of chil­dren in Tian­jin, China” (slides), Liu et al 2009:

    The results indi­cated that for chil­dren who resided in regions with a very high water iodine con­cen­tra­tion, there was on aver­age a reduc­tion of IQ by nearly nine points (b = −8.7, 95% CI: −15.9 to −1.4) in com­par­i­son with those who resided in a non-high water iodine area. When com­par­ing chil­dren liv­ing in areas with a high water iodine con­cen­tra­tion with those in the non-high areas, a slight reduc­tion of IQ by about two points was observed, how­ev­er, it was not [sta­tis­ti­cal­ly-]sig­nifi­cant. There was also a [sta­tis­ti­cal­ly-]sig­nifi­cant asso­ci­a­tion between year of birth and IQ, which seemed to exhibit a dose-re­sponse rela­tion­ship. There was a pro­gres­sive increase in IQ by about four (p = 0.072) and eight (p = 0.015) points for chil­dren born in 1996 and 1997 when com­pared with those who were born in 1995, the year China first imple­mented the national dietary iodine sup­ple­ment pro­gramme.

    ↩︎
  10. Ana­lyzed 88 sam­ples; 47 fell below USFDA rec­om­mended level and their labeled lev­el: Das­gupta PK, Liu Y, Dyke JV. “Iodine nutri­tion: iodine con­tent of iodized salt in the United States”↩︎

  11. Salt in cook­ing meat loses ~60% of iodine; Wang et al 1999, Effects of stor­age and cook­ing on the iodine con­tent in iodized salt and study on mon­i­tor­ing iodine con­tent in iodized salt↩︎

  12. “Iodine sta­tus of UK school­girls: a cross-sec­tional sur­vey”, Van­der­pump et al 2011; see also “The preva­lence of iodine defi­ciency in women of repro­duc­tive age in the United States of Amer­ica”, Hol­low­ell et al 2007↩︎

  13. Gun­ton JE, Hams G, Fiegert M, McEl­duff A. Iodine defi­ciency in ambu­la­tory par­tic­i­pants at a Syd­ney teach­ing hos­pi­tal: is Aus­tralia truly iodine replete?; see also .↩︎

  14. Hop­tion Cann SA. “Hypoth­e­sis: dietary iodine intake in the eti­ol­ogy of car­dio­vas­cu­lar dis­ease”:

    In the gen­eral United States (US) pop­u­la­tion, a declin­ing trend in uri­nary iodine lev­els has been observed as esti­mated through the large National Health and Nutri­tion Exam­i­na­tion Sur­veys (NHANES I, 1971–74; and NHANES III, 1988–94) [3]. The pro­por­tion of the US pop­u­la­tion with mod­er­ate to severe iodine defi­ciency (<50 μg iodine/L in urine) has more than quadru­pled in the last 20 years, 2.6% in NHANES I vs 11.7% in NHANES III [3]. This decline may be due to a reduced intake of iodized salt. For exam­ple, Enstrom et al. [4] com­pared data from nation­ally rep­re­sen­ta­tive sam­ples of the Amer­i­can pop­u­la­tion from two time peri­ods, 1980–1982 vs 1990–1992. Based on dietary intake stud­ies from over 10,000 indi­vid­u­als, no change was observed in sodium intake from food, while there was a 65% decline in sodium intake from dis­cre­tionary (iodized) salt. Adven­ti­tious sources of iodine have also declined with reduc­tions in the use of iodine in the dairy indus­try and in com­mer­cial bread pro­duc­tion [5]. Com­pa­ra­ble trends have been observed in other coun­tries that iodize salt. For exam­ple, pub­lic health rec­om­men­da­tions to reduce salt intake have been impli­cated in the decreas­ing iodine sta­tus in Aus­tralia and New Zealand [6]. Sim­i­lar­ly, a recent decline in median uri­nary iodine lev­els in Aus­tria may be due to lower salt intake as well as the avail­abil­ity of non­iodized salt in Aus­tria after join­ing the Euro­pean Union in 1995 [7]. In Green­land, as imported foods have replaced con­sump­tion of tra­di­tional foods such as fish and sea mam­mals, there has been a grad­ual decline in iodine sta­tus. Ander­sen et al. [8] found that median uri­nary iodine excre­tion declined with the degree of decrease in the tra­di­tional lifestyle. Evi­dence of iodine defi­ciency was seen in non-I­nuit sub­jects, who had the low­est lev­els. Thus, in coun­tries where there are chang­ing dietary pat­terns, includ­ing those long estab­lished iodiza­tion pro­grams, [sub­stan­tial] declines in iodine intake may occur.

    See also Cald­well et al 2005; Pearce et al 2004; Trumpff et al 2013; and Gahche et al 2013↩︎

  15. Bautista did not report stan­dard devi­a­tions, but did report _n_s, means, and p-val­ues, which per­mits the stan­dard devi­a­tion to be derived, fol­low­ing the Cochrane Hand­book (“7.7.3.3 Obtain­ing stan­dard devi­a­tions from stan­dard errors, con­fi­dence inter­vals, t val­ues and p val­ues for differ­ences in means”): the degrees of free­dom = 198, so with a reported p = 0.57, t = 0.569; then the stan­dard error is the means divided by the t value (); and finally the stan­dard devi­a­tion: .↩︎

  16. Southon reported con­trol & exper­i­men­tal scores split by gen­der, which need to be re-pooled:

    • con­trol

      1. mean =
      2. stan­dard devi­a­tion =
    • active

      1. mean =
      2. stan­dard devi­a­tion =
    ↩︎
  17. See Table 7; 12 month fol­lowup scores are reported grouped in IQ deciles, and if one assumes scores in each decile aver­age to around the median (since they should be fol­low­ing a bell curve, this will over­state stan­dard devi­a­tion­s), one can cal­cu­late out the fig­ures in R as fol­lows:

    # experimental
    sd(c(rep(65,12), rep(75,10), rep(85,17), rep(100,19), rep(115,1), rep(125,1), rep(135,0)))
    [1] 14.59699
    mean(c(rep(65,12), rep(75,10), rep(85,17), rep(100,19), rep(115,1), rep(125,1), rep(135,0)))
    [1] 85.25
    
    # control
    sd(c(rep(65,30), rep(75,13), rep(85,25), rep(100,26), rep(115,5), rep(125,0), rep(135,1)))
    [1] 16.19141
    mean(c(rep(65,30), rep(75,13), rep(85,25), rep(100,26), rep(115,5), rep(125,0), rep(135,1)))
    [1] 83.6
    ↩︎