Document G58B28xLE8x9ke8qexvgnNRQn

X U N I V ER SI T Y OF M A N CH EST ER M A N C H E S T E R 13 'T DEPARTM ENT OF OCCUPATION AL HEALTH S 5COTT m o . m 0 c * EK / JM M r. Don Fow l er, L ead I ndustri es A ssoci ati on, 292 M adi son A venue, NEW Y ORK 10017, U. S. A. CLIN ICAL SCIENCES BUILDING YORK PLACE M ANCHESTER 13 Telephone : ARDwick 4441 [OccupciMr-nal Hygiene Service] 25th October, 1968. D ear Don, Y ou w i l l have recei v ed M organ'is l etter bef ore thi s, gi vi ng the basi c probl em . I t is: -- Z i el hui s has ex tracted f rom the l i ter atur e al l group data f or H b/ PbB . He has cal cul ated the two r egressi on l i nes, and then thei r bi sector, argui ng that the l atter gi ves the " real " rel ati onshi p. T hus, f rom 25 groups, he gets a high correl ati on (-0. 80) betw een them, and a bi sector of Hb = 15. 43 - 0. 033 PbB (Hb i n gram s, PbB i n ug/ 100 m i s.). The techni que of col l ecti ng di f f eri ng data i s arguabl e, as i s the use of the bi sector to give the ' ' r eal " rel ati onshi p. T hi s ex erci se sti m ul ated me to do l i k ew i se w ith the grouped V auxhal l data (1962, 540 men) and another f actory group (1967, 157 men). I encl ose a copy of the anal y ses of these data, f or PbU, PbB and Hb. A s you w i l l see, the H b/ PbB data ar e so cl ose to Z i el hui s that any argument that the rel ati onshi p i s f al se has to be f orgotten. F urther, f rom my data, the di f f erence betw een men of 0 - 29 ug/ 100 m i s. and those of 30 - 49 ug/ 100 m i s, i s stati sti cal l y sound. T hese data show two thi ngs: - (a) T here i s a r eal H b/ PbB rel ati onshi p, and (b) I t star ts at a v ery low PbB value. The only outstandi ng data ar e M ike W i l l i am s. We have di scussed these endl essl y . They f al l down on two counts: - (a) The accuracy of hi s Hb m easurem ents. I have had tri pl i cates done by oursel v es and M anchester Royal I nf i rm ary (we agree with each other) and M ike, who shows about 2 g hi gher on av erage of 50 sam pl es, with consi derabl e scatter. D ennis now accepts an er r o r i n Chl ori de (M anchester) data, and w hile argui ng that i t i s a constant, cannot OCT 261968 HE 0009024 Fow l er, ORK 10017. -2- 25th October, 1968. substanti ate thi s. He f urther show s a " ranki ng" di f f erence of Hb w ith PbB on enormous groups, giving qual i tati v e support to the rel ati onshi ps. (b) The popul ati ons are sel ected - the best excuse av ai l abl e f or thei r omi ssi on. A ll of thi s l eav es us i n a posi ti on of having to agree w ith C ram er and company, that there i s a f al l of Hb with PbB , starti ng at_a low l evel . T he rem ai ni ng questi on i s " how m uch?" and " i s i t soci al l y si gni f i cant?" . I f we accept the use of the " bi sectors" - I have seen stati sti ci ans and they si m pl y w i l l not commi t them sel v es on rel ati onshi ps betw een two i nterdependant v ari abl es as opposed to the predi cti on of a v ari abl e agai nst a f ixed absci ssa - we are l ef t w ith a f al l of 3 g Hb f or 100 ug/ 100 m i s. PbB , or 20%, and pro r ata. T hi s does not make practi cal sense to me, but then I have nev er looked at i t thi s way. C ertai nl y , 10% drop (50 ug/ 100 m i s. ) w?ould be the upper l i m i t of acceptabi l i ty soci al l y . If, how ever, we can say that PbB i s the def i ni ti v e v ari abl e, and our concern i s the predi cti on of the f al l i n Hb f rom PbB , we use the H b/ PbB regr essi on l i ne, giving a f al l of about 6% at 100 ug Pb/ 100 mi s. and pro r ata, f ul l y soci al l y acceptabl e at, say , 80 ug Pb/ 100 m i s. - A ll of these excl ude conf i dence l i m i ts, etc. - I am argui ng pri nci pl e at the moment. A ll of these data come up f or di scussi on i n A m sterdam . I tol d M organ that I would send you photostats of Z i el hui s's paper, but thi s i s i m possi bl e f or me to do today, so I have sum m ari sed the argum ents i n thi s l etter. Frank l y , I do not know what approach to take. R ef erri ng to the above, and juggling my rem ai ni ng m ental f acul ti es, 20% at 100 ug Pb/ 100 mi s. i s too hi gh, but 6% at 100 ug/ 100 m i s. i s too low. M y stati sti ci ans cannot hel p me ref ute the bi sector approach, and the regr essi on l i ne i s m erel y a predi cti on of Hb f rom PbB , and not necessari l y the " real " rel ati onshi p. Have you any i d eas? I have l ef t my f orm al contri buti on to A m sterdam pendi ng a " v i si on" , but hav en't had one y et, and ti m e i s runni ng out. K ind r egards, K F 0009025 E. K ing D i rector X 560 , If XT PbU Hb Hb PbU Croupe Group s X I Sx Range Range Sy r p X Y Regressi on A 4 9 75 5.5 -0 .2 1 0 .0 5 40-280 76-10 8 Hb a 96 - 0 . 0l 6PbU 9 4 5.5 75. It tt 76-10 8 40-280 PbU a 382 - 2 . 8211b B i secto Hb a 104 -0.09 PbU a 1140 - H ft PbB PbU PbU PbB 5 14 16 5 30 95 95 30 0.43 0 .0 1 0 -130 0 -50 0 PbU 43 + 1.37FbB n it 0 -50 0 0- 130 PbB = 0.13FbU + 44 PbU a 2.7PbB PbB a 0.37PWJ + A" PbB Hb m Hb PbB 5 9 9 5 30 5.5 5.5 30 -0 .2 7 11 0 .0 1 ft 0-130 76-10 8 76-10 8 0 -130 Hb a 9 7.3 - 0 . 052PbB Hb a 10 6 PbB a 195 -1.42H b PbB = 541 - 0.19 5.1H B 157 PbB Hb h - Hb PbB 4 9 9 21 7.1 .0 .2 2 0 .0 5 0 -10 0 76-10 8 Hb = 95.8 - 0.065PbB Hb a 105 - 0.24 4 7.1 21 11 tt 76-10 8 0 190 PbB 5=127 - O.OTSfib PbBa 440 - 4.2K PbU = u g / l i t r e Hb * % PbB ug/ 100mls. 666- Tram Bf "A " (PbB-IO, N .16) U r eat er **" Kb(PbB .10, N=51> " If tt ft It It '" `(PbB J.O, >fc65) <t= ^ 53, p 0,02) k(PhBei*0 Bb B=-I1O9 2p;i (tJs2#5f Pp 00*#0052)) g. (PbBe=70j N *51) (f _-r 20 rjO 010 " (PbB= 100, H58) U ~5*29* P0 *0 10 When Hb100# * Hb l 4.8g , B i sect o n become : Sel f Z j el hui a S el f , from other two v ar i ab l es PbU * 2.7PbB - 43 PbU a 3.12FbB - 45 PbU = 2.2PbB - 2Bt ( A) Hb - 15.6 - 0.029PbB (B ) Hb * 15.0 - 0.035FbB Hb 15.4 - O.OJJPbB Hb = 16.0 - 0.037PbB PbU 1160 - 74Hb PbU = 948 - 59Hb PbU = 1650 - 92Hb*