The efficacy of zidovudine for the prevention of HIV transmission from mother to child has been clearly demonstrated and remains one of the greatest successes of antiretroviral therapy . Tolerance of this treatment in infants is generally considered good in the short term [2–4], but remains to be demonstrated in the long term [5–7]. The first placebo-controlled protocol identified a significant reduction in hemoglobin concentration in newborns during the 6 weeks of treatment, but which normalized from the twelfth week of life onwards . Two subsequent analyses in the same protocol at 18 months and 3 years, involving a smaller number of infants, found no significant differences in hematological variables, including levels of CD4+ and CD8+ lymphocytes [4,5]. In contrast, a recent study reported that levels of CD4+ lymphocytes in 20 uninfected newborns born to mothers infected with HIV, 19 of whom were exposed to antiretroviral drugs, were significantly lower than those in infants born to mothers not infected with HIV. This difference mostly involved naive CD4 fractions (CD4+ lymphocytes, CD45RA+) and was accompanied by a decrease in thymic output in fetal thymic organ culture. The number of colony forming units (CFU) was also lower, as was the cloning efficiency of CD34+ progenitor cells . It was thus suggested that zidovudine is toxic to stem cells but this could not be demonstrated in the absence of a control group consisting of infants born to mothers infected with HIV but not receiving antiretroviral drugs.
The French Perinatal Study is appropriate for observational studies of the tolerance of treatments administered during the perinatal period: since 1986, it has included a large number of infants all followed in a uniform manner whether or not exposed to antiretroviral drugs. Although most of the non-exposed infants were born before 1994, hematological and immunological data has been collected in an identical manner for all infants, with unchanged time intervals. The large number of infants included allows powerful statistical analysis of measures with high levels of inter- and intra-individual variability. Here, we report an analysis of all the measures of hemoglobin, platelets, polynuclear neutrophils, total lymphocytes and the CD4+ and CD8+ lymphocytes subpopulations, for more than 4000 HIV-1-uninfected infants, almost two-thirds of whom were exposed to antiretroviral treatment during the perinatal period. Data were modelized to take into account non-linear changes in the variables with age as well as important intra-and inter-individual variability.
Patients and methods
The French Perinatal Study
The French Perinatal Study was established in 1986 and is a prospective national epidemiological study. It was designed to investigate the risk of HIV transmission from mother to child and its prevention, and to study the progression of the disease in infected infants. The protocol has been described in detail elsewhere . More than 7000 mother–child pairs from more than 90 obstetric and pediatric centers have now been included in the study. Each mother–child pair is included no later than delivery and the follow-up of the infants included is strictly prospective from birth. Clinical and biological data are collected at regular intervals. The duration of follow-up for non-infected infants was initially 36 months and was reduced to 18 months from 1993 onwards. For hematological and immunological follow-up, the protocol includes the planned collection of the standard data (levels of polynuclear neutrophils, total lymphocytes, platelets and subpopulations of lymphocytes) at birth (during the first 4 days of life) and then at the ages of 1 and 3 months, and then every 3 months until the age of 18 months. The follow-up protocol for these variables remained unchanged after the implementation of preventive zidovudine treatment in 1994. All the measurements are made locally, in real time, at each participating center, according to standard methods. The results are transmitted to the coordinating center prospectively according to the schedule, together with the clinical data and biological information required as part of the protocol.
The analyses presented here concern infants not infected with HIV born to mothers infected with HIV-1. A child was considered non-infected if serological tests for HIV were negative at the age of 18 months. Infants below the age of 18 months were considered to be uninfected if at least two viral isolation attempts by culture, or by RNA or DNA polymerase chain reaction after 1 month were negative.
A variable describing the exposure of the child to the treatment was defined. This variable takes into account the treatment of the mother with antiretroviral drugs during the pregnancy and/or of the child during the neonatal period, and the type of treatment used: the first group comprised cases in which neither the mother nor the child was treated; the second group included all the cases in which the mother and/or the child was treated with zidovudine monotherapy; and the third group consisted of all the cases in which the mother and/or the child was treated with a combination of antiretroviral drugs including zidovudine and at least one other nucleoside analog.
For each biological variable (hemoglobin, polynuclear neutrophils, platelets, total lymphocytes and CD4+ and CD8+ lymphocytes), infants with at least one measure of the variable during the first 18 months of life who fell into one of the three treatment groups described above were included in the analysis. Four hundred and sixty five infants were not included in the analysis because the treatment given did not include zidovudine (n = 180) or because the hematological measures or the covariates of the model were not available (n = 285).
Exploratory analyses were performed by non- parametric smoothing techniques . To model the non-linear evolution of the variables studied, between-child variability and correlations between measurements for the same child, we used mixed effects models based on natural cubic B-spline curves [11,12] estimated by the maximum likelihood method. Natural cubic B-splines are very flexible smooth curves for modeling the mean profiles of these variables [13,14]. A cubic spline is a series of cubic functions joined together smoothly at a series of specified time points or knots in the follow-up period. A natural spline is one in which the spline is constrained such that it is linear beyond the first and last knots. The user sets the number and locations of the knots. Models including 5, 7 and 9 knots within the follow-up period were tested, with the knots at equally spaced fixed percentiles of the age distribution expressed in days . In addition, random effects were introduced to model individual deviations from the mean profile allowing the between-child variability to be taken into account. Two random effects models for each number of knots were tested. The first model included a random intercept at 3 months and a random slope. The second model used a different random slope before and after 3 months. These two models were compared in all cases, using likelihood ratio tests. The choice between models with different numbers of knots was based on Aikaike's and other similar criteria .
For each biological variable studied, the model incorporated both the terms of the B-splines, and also the variables sex, geographic origin of the mother (African or Caribbean versus other origins), prematurity (born before 37 weeks of gestation), maternal drug use during pregnancy (recorded if the child suffered withdrawal syndrome at birth) and the treatment administered during the pregnancy and/or during the neonatal period. Initially, the overall effect of these variables was modeled as constant over time and without distinguishing between the two types of treatment (zidovudine in monotherapy or combination). Differences were then studied from birth until 6 weeks (duration of treatment for infants), from 6 weeks to 15 months and from 15 to 18 months in order to detect any late impact of treatment. All these analyses were then repeated for the subgroup of infants for which CD4 cell counts for the mother at delivery were available.
The same type of modeling was used for the subgroup of treated infants to assess the relationship between the variables studied and the duration of exposure to pre- and postnatal treatment. The model included sex, the geographical origin of the mother, prematurity, maternal drug use and the total duration of exposure to the treatment, defined as the length of time for which the mother was treated during the pregnancy plus the duration of postnatal treatment of the child.
The biological variables were expressed on a fourth-root scale, a transformation widely used because it stabilizes the variance. All the models retained included 7 or 9 knots. The models including a simple random slope were clearly rejected in likelihood ratio tests in which they were compared to models with random sloped differing before and after 3 months. Generally, the values of the coefficients of the various factors of interest were stable, regardless of the number of knots and the random effect models used. The fit of the final models was checked by plotting the residuals. The P-values reported are two-tailed and an alpha level of 0.05 was used to assess statistical significance. SAS software version 8.01 (SAS Institute, Cary, North Carolina, USA) was used for statistical analysis.
We analyzed hematological variables for 4249 infants not infected with HIV-1. The total number of measures was greater than 21 000 for hemoglobin, polynuclear neutrophils, lymphocytes and platelets and more than 15 000 for CD4+ and CD8+ lymphocytes. The median (range) of the number of measures per child was 5 (range, 1–11) for polynuclear neutrophils, lymphocytes and platelets and 4 (1–9) for the lymphocyte subpopulations (Table 1). Of the infants, 2026 (48%) were girls, and 2013 (47%) were born to mothers of sub-Saharan African or Caribbean origin. Five percent of the infants presented withdrawal syndrome at birth, demonstrating active maternal drug use during pregnancy. The median CD4 cell count of the mothers at delivery (available for 3322 women) was 468 × 106 cells/l (range, 36–1688). Ten percent (n = 441) of the infants were premature, defined as birth before 37 weeks of gestation. About one-third (n = 1504) of the infants were not exposed to antiretroviral treatment during the perinatal period; most of these born before 1994. Of the 2745 infants exposed in utero and/or during the postnatal period, 1346 (49%) were exposed to zidovudine monotherapy and 1399 were exposed to treatment with two or more molecules including zidovudine. Multitherapy was a combination of zidovudine and lamivudine (3TC) in 56% of cases (n = 784), a protease inhibitor plus zidovudine and another nucleoside analog in 19% of cases (n = 259) and another antiretroviral association including zidovudine in 25% of cases (n = 356). Ninety-two percent of the infants treated were treated both prenatally and postnatally. Treatment during these two phases was identical in 76% and different (mainly combination for the mother and zidovudine monotherapy for the child) in 24%. The median total duration of treatment (pre- and postnatal) was 171 days (range, 3–333).
General aspects of changes in hematological variables from 0 to 18 months
Exploratory non-parametric smoother curves suggested that hemoglobin levels were transiently lower in exposed than non-exposed children, which is consistent with the findings of the first placebo-controlled study. Unexpectedly, a modest but durable effect of treatment was observed on the three lines (Fig. 1) as well as on CD4+ and CD8+ lymphocytes subpopulations. This led us to conduct a more detailed analysis.
Factors accounting for variation in hematopoiesis between 0 and 18 months
In the first multivariate analysis, the effect of all covariates was modeled as constant over time. Sex and geographical origin were found to be related to the hematological variables (Table 2). The girls had significantly higher counts of neutrophils, lymphocytes and platelets than the boys and infants born to mothers of sub-Saharan or Caribbean origin had significantly lower counts of neutrophils, lymphocytes and platelets than did infants born to mothers of other origins (mostly European). Infants born to mothers who used drugs during pregnancy, recorded if the child had withdrawal syndrome at birth, had significantly higher levels of neutrophils, platelets and, to a lesser extent, lymphocytes, than did other infants. This analysis also indicated that the effect of treatment was significant: treated infants had significantly lower levels of neutrophils, lymphocytes and platelets than did untreated infants (regression coefficients for neutrophils, −0.277, P < 0.0001; lymphocytes, −0.164, P < 0.0001; platelets, −0.042, P < 0.0001). In the absence of treatment, the levels of neutrophils, lymphocytes and platelets did not differ significantly between premature and non-premature infants. Among treated infants, the values of these three variables were significantly lower for premature infants than those for non-premature infants (Table 3).
Persistence of an effect of treatment until the age of 18 months and effect of antiretroviral combinations
The effect of treatment was modeled, distinguishing between the effect of zidovudine alone and the effect of an association of antiretroviral drugs, and considering three periods: 0 to 6 weeks (during treatment), 6 weeks to 15 months and 15 months to 18 months to identify a possible late effect of perinatal drug exposure (Table 4). During the period from 0 to 6 weeks, infants exposed to zidovudine had significantly lower levels of neutrophils and lymphocytes than did untreated infants (neutrophils: −0.442, P < 0.0001; lymphocytes: −0.160, P < 0.0001). A similar, but non-significant trend was observed for platelets (−0.013, P = 0.28). During the other two periods (6 weeks to 15 months and 15 months to 18 months), levels of neutrophils, lymphocytes and platelets were consistently significantly lower in infants treated with zidovudine than in untreated infants.
For the periods from 0 to 6 weeks and from 6 weeks to 15 months, infants exposed to a combination of drugs had lower counts for all three cell lineages than did infants exposed to zidovudine monotherapy (Table 4). At 12 months, for example, the difference between untreated and monotherapy groups was 192 × 106 cells/l for neutrophils whereas that between untreated and combination therapy groups was 317 × 106 cells/l; the corresponding values for lymphocytes were 316 and 510 × 106 cells/l, and for platelets 10 × 109 and 24 × 109 (Table 5). To take into account other cofactors, these data were obtained from the modelization in the subgroup of non-premature boys born to non-African and non-drug-addicted mothers.
In general, the CD4+ and CD8+ lymphocyte subpopulations followed similar trends. The effects of treatment were greater on CD8+ than on CD4+ lymphocyte levels: the CD4+ lymphocyte levels were significantly decreased only until 15 months. These two variables showed no greater effect due to a combination of antiretroviral drugs than to monotherapy. Consequently, the findings for CD4+ and CD8+ lymphocytes are presented for all treated infants as a single group (Table 6). At 12 months, treated children had 144 × 106 fewer CD4 cells/l and 141 × 106 fewer CD8+ cells/l than untreated children.
Effect of treatment duration
A model including all the treated infants and the factors sex, geographical origin, prematurity, active maternal drug use and total duration of treatment revealed a significant negative relationship between treatment duration and neutrophil levels (−0.00025, P = 0.04), and treatment duration and lymphocyte levels (−0.00051, P < 0.0001). A similar, but non-significant trend was observed for platelets (−0.00008, P = 0.09).
Effect of maternal immunity
We applied the same model, with the same factors, to the subgroup of infants for whom maternal CD4 cell counts at delivery were available (Table 7). The infants were assigned to three groups according to usual CD4 cell count thresholds (CD4 < 250 × 106 cells/l, 250 ≤ CD4 < 500 × 106 cells/l and CD4 ≥ 500 × 106 cells/l, the reference class). Total lymphocyte counts and particularly the CD4+ lymphocyte subset counts were significantly lower in the two groups of infants whose mothers had CD4 counts below 500 × 106 cells/l than in the group of infants whose mothers had CD4 counts of 500 × 106 cells/l or more. A similar trend was also noted for other lineages.
Observational studies of the effects of a drug in a cohort do not have the same rigor as a randomized study. Indeed, only a randomized study allows effects to be definitively attributed to the drug. However, the number of patients in randomized studies may often be too small to detect rare events, and effects on measures with large inter- and intra-individual variability. A large observational cohort may provide a sufficiently large number of subjects, but interpretation biases are possible. There are several factors that may be involved in effects on hematopoesis. Some are known, and can be included in multivariate analyses, but there may well be other as yet unidentified factors. Using mixed effects models to take into account the substantial variability of hematopoïetic variables, we first confirm what is already known, or widely suggested although rarely studied on so large a scale, concerning maternal geographic origin and sex: (1) total lymphocyte, CD4+, CD8+ and polynuclear neutrophil counts in infants with mothers of sub-Saharan African or Caribbean origin are lower than those in infants of European origin ; and (2) girls have higher levels of polynuclear neutrophils, lymphocytes and circulating platelets than boys [17–19].
More surprising is the relationship between the immune status of the mother at delivery and the lymphocyte count in the child. This relationship has not previously been described, to our knowledge, and may result from defective transplacental transfer of hematopoietic cytokines  in women with cellular immune deficiency. Consequences of maternal HIV infection in HIV uninfected babies have been recently evoked in a report of cardiac dysfunction in HIV-exposed but uninfected children .
The initial protocol comparing zidovudine and placebo revealed only an effect on hemoglobin levels that was reversed by the age of 12 weeks [1,4,5]. Our analysis reproduces this finding. The small but significant and persistent impact of perinatal prophylactic antiretroviral treatment on platelets, neutrophils and lymphocytes was not expected. The relatively small number of infants included in the previous analysis and the substantial inter- and intra-individual variability of hematopoietic variables may explain this discrepancy. The difference remained significant in a multivariate analysis, taking into account the factors listed above. Moreover, the greater effect of combination therapy than monotherapy and the link between the duration of exposure and the magnitude of the biological effect strongly implicate the treatment. The effect involves three hematopoiesis lineage and persists until 18 months of age despite the treatment having stopped at 6 weeks of age. An effect on several cell lineages suggests either changes in the medullary stroma or an effect on multipotent stem cells. There is no evidence suggesting that zidovudine has a deleterious effect on the medullary stroma. In contrast, the toxicity of zidovudine to hematopoietic progenitors in vitro has been well established [22,23]. The mechanism underlying this putative toxicity is unknown. The mitochondrial toxicity of antiretroviral nucleoside analogs is currently receiving considerable attention . However, non-mitochondrial mechanisms of toxicity are possible, mediated in particular by toxicity to nuclear DNA. Zidovudine interacts with nuclear DNA in several cell models , and the integration of zidovudine into nuclear DNA has been demonstrated both in an animal model  and in exposed newborns . More specifically, in vitro incorporation of zidovudine into DNA of burst-forming unit-erythroid (BFU.E) and colony-forming unit-erythroid (CFU-E) has been demonstrated  and leads to a reduced expression of the genes for β globin  or granulocyte-monocyte colony stimulating factor receptor (GM CSFR) . Although not addressed in this study, the clinical consequences of these findings are probably minor or non-existent at these ages. The observed deficiencies in lymphocyte, neutrophil and platelets lineages are clinically modest and no subgroup of patients with very low values was identified among the treated children. As suggested by the Danish study , a more detailed analysis of CD4/CD8 lymphocyte subpopulations (i.e naive/memory phenotype and function) could be of value, as would a more long-term evaluation. Indeed, our findings demonstrate that perinatal exposure to zidovudine may result in biological effects, even in a cell system with a high turnover, that persist until the age of 18 months.
This work was supported by Agence Nationale de Recherches sur le Sida (ANRS).
1. Connor EM, Sperling RS, Gelber R, Kiselev P, Scott G, O'Sullivan MJ, et al
. Reduction of maternal-infant transmission of human immunodeficiency virus type-1 with zidovudine treatment. N Engl J Med
2. Taylor GP, Low-Beer N. Antiretroviral therapy in pregnancy. A focus on safety. Drug Safety
3. Lipshultz SE, Easley KA, Orav EJ, Kaplan S, Starc TJ, Bricker JT, et al
. Absence of cardiac toxicity of zidovudine in infants. Pediatric pulmonary and cardiac complications of vertically transmitted HIV infection study group. N Engl J Med
4. Sperling RS, Shapiro DE, McSherry GD, Britto P, Cunningham BE, Culnane M, et al
. Safety of the maternal-infant zidovudine regimen utilized in the Pediatric AIDS Clinical Trial Group 0176 Study. AIDS
5. Culnane M, Fowler M, Lee SS. McSherry G, Brady M, O'Donnell K, et al
. Lack of long-term effects of in utero exposure to zidovudine among uninfected children born to HIV-infected women. Pediatric AIDS Clinical Trials Group Protocol 219/076 Teams. JAMA
6. Blanche S, Tardieu M, Rustin P, Slama A, Barret B, Firtion G, et al
. Persistent mitochondrial dysfunction and perinatal exposure to antiretroviral nucleoside analogues. Lancet
7. Landreau-Mascaro A, Barret B, Mayaux MJ, Tardieu M, Blanche S. Risk of early febrile seizure with perinatal exposure to nucleoside analogues. Lancet
8. Nielsen SD, Jeppesen DL, Kolte L, Clark DR, Sorensen TU, Dreves AM et al
. Impaired progenitor cell function in HIV negative infants of HIV-positive mothers results in decreased thymic output and low CD4 counts. Blood
9. Blanche S, Rouzioux C, Moscato ML, Veber F, Mayaux MJ, Jacome C, et al
. A prospective study of infants born to women seropositive for human immunodeficiency virus type 1. HIV Infection in Newborns French Collaborative Study Group. N Engl J Med
10. Cleveland WS. Robust locally weighted regression and smoothing scatterplots. J Amer Stat Assoc
11. Wold S. Spline functions in data analysis. Technometrics
12. Laird NM, Ware JH. Random-effects models for longitudinal data. Biometrics
13. Hastie TJ, Tibshirani RJ. Generalized Additive Models. Monographs on Statistics and Applied Probability
, London: Chapman & Hall: 1990.
14. Shi M, Weiss RE, Taylor JMG. An analysis of pediatric CD4 counts for acquired immune deficiency syndrom using flexible random curves.
Appl Stat 2001; 50
15. Harrell FE. Regression Modeling Strategies
. Springer Series in Statistics. Berlin: Springer; 2001.
16. Read WW, Diehl LF. Leukopenia, neutropenia and reduced hemoglobin levels in healthy American blacks. Arch Intern Med
17. Lisse IM, Aaby P, Whittle H, Jensen H, Engelmann M, Christensen LB. T-lymphocyte subsets in West African children: impact of age, sex, and season. J Pediatr
18. Rudy BJ, Crowley-Nowick PA, Douglas SD. Immunology and the REACH study: HIV immunology and preliminary findings. J Adolesc Health
19. Stevens RF, Alexander MK. A sex difference in the platelet count. Br J Haematol
20. Hambleton J. Hematologic complications of HIV infection. Oncology
21. Lipshultz SE, Easley KA, Kaplan S, Starc TJ, Bricker JT, Lai WW, et al
. Cardiovascular status of infants and children of women infected with HIV-1 (p2c2 HIV): a cohort study. Lancet
2002; in press.
22. Gribaldo L, Malerba I, Collotta A, Casati S, Pessina A. Inhibition of CFU-E/BFU-E by 3'-azido-3'-deoxythymidine, chlorpropamide, and protoporphirin IX Zinc (II): A comparison between direct exposure of progenitor cells and long-term exposure of bone marrow cultures. Toxicol Sci
23. Fowler DA, Xi M-Y, Sommadossi J-P. Protection and rescue from 2', 3'-dideoxypyrimidine nucleotide analog toxicity by hemin in human bone marrow progenitor cells. Antimicrob. Agents Chemother
24. Lewis W, Dalakas MC. Mitochondrial toxicity of antiviral drugs. Nature Med
25. Wutzler P, Thust R. Genetic risks of antiviral nucleoside analogues – a survey. Antiviral Res
26. Olivero OA, Fernandez JJ, Antiochos BB, Wagner JL, St Claire ME, Poirier MC. et al
. Transplacental genotoxicity of combined antiretroviral nucleoside analogue therapy in Erythrocebus patas monkeys. J Acquir Immune Defic Syndr
27. Olivero OA, Shearer GM, Chougnet CA, Kovacs AA, Baker R, Stek AM, et al
. Incorporation of zidovudine into leukocyte DNA from HIV-1-positive adults and pregnant woman, and cord blood from infants exposed in utero. AIDS
28. Sommadossi J.P, Carlisle R, Zhou Z. Cellular pharmacology of 3'azido 3'deoxythymidine with evidence of incorporation into DNA of human bone marrow cells. Mol Pharmacol
29. Spiga MG, Weidner DA, Trentesaux C, LeBoeuf RD, Sommadossi JP. Inhibition of β globin gene expression by 3’ azido 3’ deoxythymidine in human erythroïd progenitor cells. Antiviral Res
30. Chitnis S, Mondal D, Agrawal KC. Zidovudine (AZT) treatment suppresses granulocyte-monocyte colony factor receptor type alpha (GM CSFRa) gene expression in murine bone marrow cells. Life Sci
The main investigator at each site and institutions participating in the French Perinatal Cohort Study
Coordinators: M.J. Mayaux and S. Blanche.
Aix-en-Provence, Thevenieau D.; Amiens, Pautard B.; Angers, Chennebault J.M.; Argenteuil, Allizy C.; Basse-Terre, Sibille G.; Bastia, Pincemaille O.; Bayonne, Hernandorena X.; Besançon, Estavoyer J.M.; Bondy, Lachassinne E.; Bordeaux, Douard D.; Boulogne Billancourt, Gilles I.; Bourg La Reine, Gantzer A.; Bullion, Colin-Gorki A.M.; Caen, Brouard J.; Cayenne, Delattre P.; Clamart, Vial M.; Clichy, Mazy F.; Colombes, Floch-Tudal C.; Compiègne, Lagrue A.; Corbeil Essonnes, Devidas A.; Creil, Duval-Arnould M.; Creteil, Touboul C.; Dijon, Guerin M.N.; Dourdan, Ercoli V.; Dreux, Denavit M.F.; Elbeuf, Lahsinat K.; Evreux, Pascal C.; Evry, May A.; Fontainebleau, Dallot M.C.; Fort de France, Cecile W.; Gonesse, Lobut J.B.; Lagny sur Marne, Chalvon Dermesay A.; Le Chesnay, Beal G.; Le Kremlin Bicêtre, Bader-Meunier B.; Le Lamentin, Monlouis M.; Lille, Mazingue F.; Limoges, De Lumley L.; Longjumeau, Seaume H.; Lyon, Kebaili K.; Mantes La Jolie, Botto C.; Marseille, Thuret I.; Meaux, Crumiere C.; Melun, Le Lorier B.; Meulan, Seguy D.; Montfermeil, Talon P.; Montpellier, Nicolas J.; Montreuil, Heller-Roussin B.; Nancy, Hubert C. L.,; Nanterre, De Sarcus B.; Nantes, Mechinaud F.; Neuilly sur Seine, Berterottiere D.; Nice, Monpoux F.; Nîmes, Dendale J.; Orléans, Arsac P.; Orsay, De Gennes C.; Perpignan, Bachelard G.; Pointe-à-Pitre, Bardinet F.; Poissy, Rousset M.C.; Pontoise, Mouchnino G.; Reims, Munzer M.; Rouen, Brossard V.; Saint-Denis, Retbi J.M.; Saint-Etienne, Fresard A.; Saint-Germain en Laye, Narcy P.; Saint-Martin, De Caunes F.; Sèvres, Segard L.; Strasbourg, Partisani M.; Suresnes, Clement; Toulouse, Tricoire J.; Tours, Marchand S.; La Trinité, Hugon N.; Villeneuve Saint Georges, Guillot F.; Villepinte, Broyard A.
Paris: ASE St Vincent de Paul, Commeau A.; Hospitalier Cochin Tarnier Port-Royal, Firtion G.; Groupe Hospitalier Necker, Blanche S., Parat S.; Hôpital Bichat-Claude Bernard, Matheron S.; Hôpital des Métallurgistes, Heller-Roussin B.; Hôpital La Pitié Salpétrière, Noseda G.; Hôpital Lariboisière, Ciraru-Vigneron N.; Hôpital Notre Dame de Bon Secours, Ayral D.; Hôpital Robert Debré, Levine M.; Hôpital Rothschild, Wallet A.; Hôpital Saint-Antoine, Carbonne B.; Hôpital Saint-Michel, Aufrant C.; Hôpital Saint-Vincent de Paul, Boccara J.F.; Hôpital Tenon, Herve F.; Hôpital Trousseau, Dollfus C.; Institut de Puériculture Brune, Dubois M.; Institut Mutualiste Montsouris, Carlus Moncomble C. Cited Here...