- Email: [email protected]
Neonatal myoclonus and neuroblastoma
1410
THE LANCET
EXPECTED CHANGES IN AGE-SPECIFIC INCIDENCE OF CANCER DUETO INCREASED BIOLOGICAL AGEING
Biol = biological. When mortality rate from cancer is calculated on basis of biological age, the increase in biological age associated with acceleration of maturation and growth produces increases m incidence of cancer at any chronological age.
in this calculation are similar to the longer term increases shown in their fig 3. There is another feature of the Adami data that supports the biological ageing hypothesis. They showed that men and women aged about 40 years in 1987 had almost no increase in the incidence of cancer. These people passed through their childhood in the years after World War II, when conditions were not easy for parents and children. If good environmental and social conditions lead to increased growth, health, and maturation, then these 40-year-old individuals may have been deprived as children and their lower incidence of cancer would result from the resulting slower growth and maturation. Data on their growth during childhood and their age at menarche would support or refute that idea. I suggest that biological age does not necessarily parallel chronological age. Although it is usually impracticable, changes in disease incidence might be better based on biological rather than chronological age. Good conditions during childhood are associated with more rapid growth and earlier maturation. The associated increased rate of biological ageing may then lead to earlier malignant disease. If this hypothesis is correct, then we have the paradox that a good social and physical environment may lead to more cancer, whereas a bad environment protects. women
Human Nutrition Research Group, Department of Child Health, University of Queensland, Queensland 4067, Australia
ALAN DUGDALE
1. Eveleth
PB, Tanner JM. Worldwide variation in human growth. Cambridge: Cambridge University Press, 1991. 2. Ljung B-O, Burgsten-Brucefors A, Lindgren G. The secular trend in physical growth in Sweden. Ann Hum Biol 1974; 1: 245-56. 3. WHO Annual Statistical Reports. Geneva: WHO, 1988.
Authors’reply SIR,-Although the
use
of
age-period-cohort models needs
considerable care and awareness of the limitation of the models, the situation is not as bleak as Dr Bonneux and colleagues indicate.1-5 Indeed, such models are important in the analysis of incidence or mortality trends. In our original report we had full discussion of the methodological limitations inherent in these models, but a few additional comments seem warranted. Age-period-cohort modelling can be divided into two steps. The first involves testing of various submodels against the full ageperiod-cohort model. This part of the analysis is fairly straightforward, since traditional statistical considerations usually allow the choice of the best model. In our data, we found that the age-cohort model was sufficient for women. For men, the full model was an improvement on both the age-period and the age-cohort model, but cohort seemed to be a more important factor than
period.
The second step of the modelling consists of an estimation of the relative risks from the model identified in the first step. In our data, this scheme caused no difficulties for women, since we could use the age-cohort model, which has uniquely defined variables (although they could still be biased if there was a true period effect). For men the situation was more difficult, since the linear parts of a full model are not uniquely identified. As we noted in our report, all choices of further restrictions to force unique index estimates have drawbacks. Our choice of a linear period effect with slope zero was based on the arguments of Holford.3-5 This assumption has the further advantage that it is almost equivalent to the method suggested by Clayton and Schifilers.1,2 A possible result of this choice is that the cohort effect is somewhat exaggerated, because it may include part of the period effect. The only real alternative is to report no estimates at all, or to report several different sets of estimates, each based on specific assumptions. Such an approach needs to be considered when it is unclear which of the factors are more important, a situation we were able to avoid for men because of the clear importance of the cohort effects. Thus it is incorrect to deny the presence of the strong cohort effects in our data. To what extent are incidence rates influenced by factors other than changes in carcinogenic exposure and disease occurrence? It is possible that screening, new diagnostic technologies, increased diagnostic intensity, drifts in histopathological criteria, and changing necropsy rates might result in earlier diagnosis or even overdiagnosis of cancer; these factors could perturb incidence trends, but only overdiagnosis can lead to sustained increases in incidence in the absence of an increase in true cancer incidence. Nevertheless, there is virtually no empirical evidence to support an important role for any of these possibilities. Hence, there is a substantial burden of proof for those who claim that they are
quantitatively important. All quantitative methods used to assess the burden of cancer have advantages and disadvantages.6 Comparison of the incidence, and mortality and survival rates from the same setting are by definition mutually dependent. In fact, the decrease in case fatality estimated by Bonneux et al (and dismissed by them) agrees fairly well with our estimates for Sweden during the period 1960-84.’’ However, Bonneux and colleagues are incorrect in stating that we have claimed improved treatment accounts for the decreased cancer case fatality in Sweden. In fact, we have repeatedly stated that we believe other factors probably play the more important part.6,7 Epidemiology Unit, University Hospital, 75185 Uppsala, Sweden
HANS-OLOV ADAMI PÄR SPARÉN JOHN BARON
Department of Statistics, Uppsala University, Uppsala
REINHOLD BERGSTRÖM
Cancer
Clayton D, Schifflers E. Models for temporal variation in cancer rates, I: age-period and age-cohort models. Stat Med 1987; 6: 449-67. 2. Clayton D, Schifflers E. Models for temporal variation in cancer rates, I: age-penod and age-cohort models. Stat Med 1987; 6: 469-81. 3. Holford TR. Understanding the effects of age, period, and cohort on incidence and mortality rates. Annu Rev Public Health 1991; 12: 425-57. 4. Holford TR, Roush GC, McKay LA. Trends in female breast cancer in Connecticut and the United States. J Clin Epidemiol 1991; 44: 29-39. 5. Dubrow R, Bernstein J, Holford TR. Age-period-cohort modelling of large-bowelcancer incidence by anatomic sub-site and sex in Connecticut. Int J Cancer 1993; 1.
53: 907-13. 6. Pontén J, Adami HO, Sparén P. Trends in cancer survival and mortality rates. Med Oncol Tumor Pharmacother 1991; 8: 147-53. 7. Adami HO, Sparén P, Bergström R, Holmberg L, Krusemo UB, Pontén J. Increasing survival trend after cancer diagnosis in Sweden 1960-1984. J Natl Cancer Inst 1989; 81: 1640-47.
Neonatal myoclonus and neuroblastoma SIR,-Myoclonus occurs in various disorders in neonates but, to our knowledge, neonatal myoclonus has never been reported in association with neoplasms, in contrast to myoclonic encephalopathy of infancy.3 We report a newborn baby with myoclonus who was found to have a neuroblastoma. A 3740 g baby girl, who was born uneventfully at 39 weeks’ gestation, developed myoclonic jerks in her sleep on the third day after birth. The cluster of jerks consisted of symmetrical rapid movements of the arms and/or legs that lasted for several minutes.
1411
THE LANCET
While awake, she looked healthy and no jerks were seen. There was laboratory evidence of meningitis, lacticacidosis, aminoaciduria, hypogylcaemia, hypocalcaemia, or intrauterine infection. Neither cranial computed tomography (CT) nor electroencephalography showed any abnormality. The jerking movements continued despite treatment with several anticonvulsants. At 1 month of age, the bilaterally synchronous jerks were replaced by more asynchronous myoclonic jerks, which affected the arms, legs, head, and trunk erratically. We suspected myoclonic encephalopathy of infancy, despite the absence of opsoclonus.3 Ultrasonography showed a round mass (18 x 18 mm) in the left adrenal gland. Abdominal CT also showed a mass with fine calcification, and the presumptive diagnosis was neuroblastoma. There was no evidence of metastasis. The myoclonus disappeared at 3 months of age, and over the next 6 months she was observed without any therapeutic intervention. Abdominal ultrasonography done every month showed no change in tumour size. Laparotomy at 10 months revealed rosette-fibrillary types of neuroblastoma involving the left adrenal gland and lymph nodes at the left renal hilus. These were completely removed. Her postoperative course was uneventful, but her motor development was slightly retarded: she sat at 8 months and walked at 18 months of age. Cranial magnetic resonance imaging at the age of 18 months showed incomplete myelination of subcortical white matter. Neonatal sleep myoclonus and neuroblastoma are rare, but they can coexist. Furthermore, the myoclonus disappeared before the tumour was removed. Despite these findings, a possible relation between myoclonus and neuroblastoma cannot be ruled out. The tumour was clinically silent and remained unchanged for 6 months, which suggests that it might have been present at birth. Without ultrasonography, it might have remained undetected. Hence, it is possible that neuroblastoma is unrecognised in some neonates with clusters of myoclonic jerks during sleep. In addition, neurological manifestations sometimes result from the remote action of neoplasms. In children, the most well-known paraneoplastic syndrome is myoclonic encephalopathy, which often complicates neuroblastoma.3 However, our patient did not have opsoclonus or clinical deterioration. Her myoclonic jerks were similar to the polymyoclonia of myoclonic encephalopathy of infancy, but they disappeared spontaneously without steroid therapy. Although it is uncertain whether our patient has a newly identified paraneoplastic syndrome, the possible association of neonatal myoclonus with an occult neoplasm should be carefully investigated. no
Department of Paediatrics, Nagoya University School of Medicine, Showa-ku, Nagoya 466, Japan Department of Paediatric Surgery, Nagoya University School of Medicine Department of Paediatrics, Nagoya First Red Cross Hospital
KOSABURO ASO KAZUYOSHI WATANABE TAMIKO NEGORO KEIZO HORIBE YOSHIO WATANABE TAKASHI UMEDA SUNAO FURUNE
1. Commission on Classification and Terminology of the International League against Epilepsy. Proposal for revised classification of epilepsies and epileptic syndromes. Epilepsia 1989; 30: 389-99. 2. Daoust-Roy J, Seshia SS. Benign neonatal sleep myoclonus: a differential diagnosis of neonatal seizures. Am J Dis Child 1992; 146: 1236-41. 3. Lott I, Kinsborne M. Myoclonic encephalopathy of infants. In: Fahn S, et al, eds. Advances in neurology: myoclonus, vol 43. New York: Raven Press, 1986: 127-36, 665.
Sensitisation to peanut and vitamin D oily
preparations SiR,—In a recent case of anaphylactic shock after ingestion of butter, the history indicated that peanut oil in vitamin D preparations taken during the first two years of life was the only possible method of sensitisation. Peanut sensitisation by peanut oil in infant formulae has also been reported.’ We have done a prospective analysis of peanut cutaneous reactivity according to the type of vitamin D supplementation in early infancy. A skin prick test with peanut antigen was done in 122 children aged between 7 and 60 months (mean [SD] 37[14-4]) referred to the paediatric allergy unit. During the first two years of life 40 of peanut
these children received appropriate doses of vitamin D free from peanut oil (group A), and 72 had vitamin D in preparations that contained peanut oil: 28 a daily dose (group B), and 44 a monthly dose (group C). A positive skin prick test, defined as a weal greater than 3 mm, was found in 5% in group A, 32-1% in group B, and 25% in group C: Children tested Allergic children Positive peanut skin
Group A
Group B
Group C
40 27
28 20
44 30
2(5%)
test
9(32-1%)
11(25%)
The odds ratio of B versus A was 9 00 (p < 0003) and of C versus A No significant difference was found between was 6-33 (p < 001). groups B and C. Sensitisation to peanut is widely acknowledged to play a role in atopic dermatitis and is a major factor in life-threatening anaphylactic reactions.2 Traces of proteins in the peanut oil of some infant formulae are thought to be the sensitising factor.1 The same sensitising process is suggested in children treated with vitamin D in preparations made partly of peanut oil. Although cutaneous sensitisation is not synonymous with clinical reactivity to peanut but merely indicates a risk, our data suggest the need to avoid all medications containing peanut oil in infancy. GUY DE MONTIS
DOMINIQUE GENDREL MARTINE CHEMILLIER-TRUONG CHRISTOPHE DUPONT
Hôpital Saint-Vincent-de-Paul, 75674 Paris, France
1. Moneret-Vautrin DA, Hatahet R, Kanny G, Ait-Djafer Z. Allergenic peanut oil in milk formulas. Lancet 1991; 338: 1149. 2. Yunginger JW, Squillace DL, Jones RT, Helm RM. Fatal anaphylactic reactions induced by peanuts. Allergy Proc 1989; 10: 249-54.
Primary central
nervous
and brain
system lymphoma
biopsy
in AIDS
SIR,-In an editorial1 on brain biopsy for intracranial mass lesions, you say that inability to treat a central nervous system (CNS) lymphoma argues against regular use of brain biopsy in patients with late-stage AID S. At our centre, a prospective study on the role of brain biopsy in the diagnosis and treatment of intracranial mass lesions in patients with AIDS is ongoing. We enrolled, from Oct 1, 1991 to March 31, 1993, 56 patients with brain mass lesions. All were treated with an antitoxoplasma empirical regimen for 3-4 weeks. 30 were considered as having toxoplasmic encephalitis because of response to therapy. The other 26 were selected for potential brain biopsy; indication was in all cases a poor clinical and/or radiological response to antitoxoplasma therapy. In 13 (50%) biopsy could not be done because of contraindicatons or refusal. 13 had brain biopsy (8 stereotactic and 5 ultrasound guided), and 12 (92%) had a final diagnosis (7 primary CNS lymphoma, 3 progressive multifocal leukoencephalopathy, 1 mycotic abscess, 1 tuberculous abscess). Only minimum bleeding confined to the tumour bed occurred in 1 patient. Of the 7 patients with primary CNS lymphoma (3 immunoblastic, 3 large cell, and 1 lymphoblastic according to WF) 6 received whole-brain irradiation (cobalt-60, opposed lateral portals, 40-44 Gy). 5 completed a full course and were clinically stabilised. 3 showed a partial radiological response, and 2 a stable lesion (computed tomography). We observed clinical and radiological progression in only 1 patient, with death occurring 28 days after biopsy and before completion of radiotherapy. We retrospectively compared the 7 patients with biopsy-proven CNS lymphoma to the 7 patients with a necropsy-diagnosed cerebral lymphoma observed at our centre during 1986-90. The two groups were similar for age, sex, CD4 count, previous AIDS diagnosis, time from onset to presumptive diagnosis, severity of neurological status, white cell count, haemoglobin, and time of treatment with antiretroviral drugs. Patients with a biopsyconfirmed diagnosis had significantly longer survival from onset of neurological symptoms, with an estimated median survival time (EMST) of 233 days compared with 73 days in necropsy-proven patients (p 0-0002, logrank test). This result was confirmed by calculating survival from presumptive diagnosis (221 vs 9 days, =
THE LANCET
EXPECTED CHANGES IN AGE-SPECIFIC INCIDENCE OF CANCER DUETO INCREASED BIOLOGICAL AGEING
Biol = biological. When mortality rate from cancer is calculated on basis of biological age, the increase in biological age associated with acceleration of maturation and growth produces increases m incidence of cancer at any chronological age.
in this calculation are similar to the longer term increases shown in their fig 3. There is another feature of the Adami data that supports the biological ageing hypothesis. They showed that men and women aged about 40 years in 1987 had almost no increase in the incidence of cancer. These people passed through their childhood in the years after World War II, when conditions were not easy for parents and children. If good environmental and social conditions lead to increased growth, health, and maturation, then these 40-year-old individuals may have been deprived as children and their lower incidence of cancer would result from the resulting slower growth and maturation. Data on their growth during childhood and their age at menarche would support or refute that idea. I suggest that biological age does not necessarily parallel chronological age. Although it is usually impracticable, changes in disease incidence might be better based on biological rather than chronological age. Good conditions during childhood are associated with more rapid growth and earlier maturation. The associated increased rate of biological ageing may then lead to earlier malignant disease. If this hypothesis is correct, then we have the paradox that a good social and physical environment may lead to more cancer, whereas a bad environment protects. women
Human Nutrition Research Group, Department of Child Health, University of Queensland, Queensland 4067, Australia
ALAN DUGDALE
1. Eveleth
PB, Tanner JM. Worldwide variation in human growth. Cambridge: Cambridge University Press, 1991. 2. Ljung B-O, Burgsten-Brucefors A, Lindgren G. The secular trend in physical growth in Sweden. Ann Hum Biol 1974; 1: 245-56. 3. WHO Annual Statistical Reports. Geneva: WHO, 1988.
Authors’reply SIR,-Although the
use
of
age-period-cohort models needs
considerable care and awareness of the limitation of the models, the situation is not as bleak as Dr Bonneux and colleagues indicate.1-5 Indeed, such models are important in the analysis of incidence or mortality trends. In our original report we had full discussion of the methodological limitations inherent in these models, but a few additional comments seem warranted. Age-period-cohort modelling can be divided into two steps. The first involves testing of various submodels against the full ageperiod-cohort model. This part of the analysis is fairly straightforward, since traditional statistical considerations usually allow the choice of the best model. In our data, we found that the age-cohort model was sufficient for women. For men, the full model was an improvement on both the age-period and the age-cohort model, but cohort seemed to be a more important factor than
period.
The second step of the modelling consists of an estimation of the relative risks from the model identified in the first step. In our data, this scheme caused no difficulties for women, since we could use the age-cohort model, which has uniquely defined variables (although they could still be biased if there was a true period effect). For men the situation was more difficult, since the linear parts of a full model are not uniquely identified. As we noted in our report, all choices of further restrictions to force unique index estimates have drawbacks. Our choice of a linear period effect with slope zero was based on the arguments of Holford.3-5 This assumption has the further advantage that it is almost equivalent to the method suggested by Clayton and Schifilers.1,2 A possible result of this choice is that the cohort effect is somewhat exaggerated, because it may include part of the period effect. The only real alternative is to report no estimates at all, or to report several different sets of estimates, each based on specific assumptions. Such an approach needs to be considered when it is unclear which of the factors are more important, a situation we were able to avoid for men because of the clear importance of the cohort effects. Thus it is incorrect to deny the presence of the strong cohort effects in our data. To what extent are incidence rates influenced by factors other than changes in carcinogenic exposure and disease occurrence? It is possible that screening, new diagnostic technologies, increased diagnostic intensity, drifts in histopathological criteria, and changing necropsy rates might result in earlier diagnosis or even overdiagnosis of cancer; these factors could perturb incidence trends, but only overdiagnosis can lead to sustained increases in incidence in the absence of an increase in true cancer incidence. Nevertheless, there is virtually no empirical evidence to support an important role for any of these possibilities. Hence, there is a substantial burden of proof for those who claim that they are
quantitatively important. All quantitative methods used to assess the burden of cancer have advantages and disadvantages.6 Comparison of the incidence, and mortality and survival rates from the same setting are by definition mutually dependent. In fact, the decrease in case fatality estimated by Bonneux et al (and dismissed by them) agrees fairly well with our estimates for Sweden during the period 1960-84.’’ However, Bonneux and colleagues are incorrect in stating that we have claimed improved treatment accounts for the decreased cancer case fatality in Sweden. In fact, we have repeatedly stated that we believe other factors probably play the more important part.6,7 Epidemiology Unit, University Hospital, 75185 Uppsala, Sweden
HANS-OLOV ADAMI PÄR SPARÉN JOHN BARON
Department of Statistics, Uppsala University, Uppsala
REINHOLD BERGSTRÖM
Cancer
Clayton D, Schifflers E. Models for temporal variation in cancer rates, I: age-period and age-cohort models. Stat Med 1987; 6: 449-67. 2. Clayton D, Schifflers E. Models for temporal variation in cancer rates, I: age-penod and age-cohort models. Stat Med 1987; 6: 469-81. 3. Holford TR. Understanding the effects of age, period, and cohort on incidence and mortality rates. Annu Rev Public Health 1991; 12: 425-57. 4. Holford TR, Roush GC, McKay LA. Trends in female breast cancer in Connecticut and the United States. J Clin Epidemiol 1991; 44: 29-39. 5. Dubrow R, Bernstein J, Holford TR. Age-period-cohort modelling of large-bowelcancer incidence by anatomic sub-site and sex in Connecticut. Int J Cancer 1993; 1.
53: 907-13. 6. Pontén J, Adami HO, Sparén P. Trends in cancer survival and mortality rates. Med Oncol Tumor Pharmacother 1991; 8: 147-53. 7. Adami HO, Sparén P, Bergström R, Holmberg L, Krusemo UB, Pontén J. Increasing survival trend after cancer diagnosis in Sweden 1960-1984. J Natl Cancer Inst 1989; 81: 1640-47.
Neonatal myoclonus and neuroblastoma SIR,-Myoclonus occurs in various disorders in neonates but, to our knowledge, neonatal myoclonus has never been reported in association with neoplasms, in contrast to myoclonic encephalopathy of infancy.3 We report a newborn baby with myoclonus who was found to have a neuroblastoma. A 3740 g baby girl, who was born uneventfully at 39 weeks’ gestation, developed myoclonic jerks in her sleep on the third day after birth. The cluster of jerks consisted of symmetrical rapid movements of the arms and/or legs that lasted for several minutes.
1411
THE LANCET
While awake, she looked healthy and no jerks were seen. There was laboratory evidence of meningitis, lacticacidosis, aminoaciduria, hypogylcaemia, hypocalcaemia, or intrauterine infection. Neither cranial computed tomography (CT) nor electroencephalography showed any abnormality. The jerking movements continued despite treatment with several anticonvulsants. At 1 month of age, the bilaterally synchronous jerks were replaced by more asynchronous myoclonic jerks, which affected the arms, legs, head, and trunk erratically. We suspected myoclonic encephalopathy of infancy, despite the absence of opsoclonus.3 Ultrasonography showed a round mass (18 x 18 mm) in the left adrenal gland. Abdominal CT also showed a mass with fine calcification, and the presumptive diagnosis was neuroblastoma. There was no evidence of metastasis. The myoclonus disappeared at 3 months of age, and over the next 6 months she was observed without any therapeutic intervention. Abdominal ultrasonography done every month showed no change in tumour size. Laparotomy at 10 months revealed rosette-fibrillary types of neuroblastoma involving the left adrenal gland and lymph nodes at the left renal hilus. These were completely removed. Her postoperative course was uneventful, but her motor development was slightly retarded: she sat at 8 months and walked at 18 months of age. Cranial magnetic resonance imaging at the age of 18 months showed incomplete myelination of subcortical white matter. Neonatal sleep myoclonus and neuroblastoma are rare, but they can coexist. Furthermore, the myoclonus disappeared before the tumour was removed. Despite these findings, a possible relation between myoclonus and neuroblastoma cannot be ruled out. The tumour was clinically silent and remained unchanged for 6 months, which suggests that it might have been present at birth. Without ultrasonography, it might have remained undetected. Hence, it is possible that neuroblastoma is unrecognised in some neonates with clusters of myoclonic jerks during sleep. In addition, neurological manifestations sometimes result from the remote action of neoplasms. In children, the most well-known paraneoplastic syndrome is myoclonic encephalopathy, which often complicates neuroblastoma.3 However, our patient did not have opsoclonus or clinical deterioration. Her myoclonic jerks were similar to the polymyoclonia of myoclonic encephalopathy of infancy, but they disappeared spontaneously without steroid therapy. Although it is uncertain whether our patient has a newly identified paraneoplastic syndrome, the possible association of neonatal myoclonus with an occult neoplasm should be carefully investigated. no
Department of Paediatrics, Nagoya University School of Medicine, Showa-ku, Nagoya 466, Japan Department of Paediatric Surgery, Nagoya University School of Medicine Department of Paediatrics, Nagoya First Red Cross Hospital
KOSABURO ASO KAZUYOSHI WATANABE TAMIKO NEGORO KEIZO HORIBE YOSHIO WATANABE TAKASHI UMEDA SUNAO FURUNE
1. Commission on Classification and Terminology of the International League against Epilepsy. Proposal for revised classification of epilepsies and epileptic syndromes. Epilepsia 1989; 30: 389-99. 2. Daoust-Roy J, Seshia SS. Benign neonatal sleep myoclonus: a differential diagnosis of neonatal seizures. Am J Dis Child 1992; 146: 1236-41. 3. Lott I, Kinsborne M. Myoclonic encephalopathy of infants. In: Fahn S, et al, eds. Advances in neurology: myoclonus, vol 43. New York: Raven Press, 1986: 127-36, 665.
Sensitisation to peanut and vitamin D oily
preparations SiR,—In a recent case of anaphylactic shock after ingestion of butter, the history indicated that peanut oil in vitamin D preparations taken during the first two years of life was the only possible method of sensitisation. Peanut sensitisation by peanut oil in infant formulae has also been reported.’ We have done a prospective analysis of peanut cutaneous reactivity according to the type of vitamin D supplementation in early infancy. A skin prick test with peanut antigen was done in 122 children aged between 7 and 60 months (mean [SD] 37[14-4]) referred to the paediatric allergy unit. During the first two years of life 40 of peanut
these children received appropriate doses of vitamin D free from peanut oil (group A), and 72 had vitamin D in preparations that contained peanut oil: 28 a daily dose (group B), and 44 a monthly dose (group C). A positive skin prick test, defined as a weal greater than 3 mm, was found in 5% in group A, 32-1% in group B, and 25% in group C: Children tested Allergic children Positive peanut skin
Group A
Group B
Group C
40 27
28 20
44 30
2(5%)
test
9(32-1%)
11(25%)
The odds ratio of B versus A was 9 00 (p < 0003) and of C versus A No significant difference was found between was 6-33 (p < 001). groups B and C. Sensitisation to peanut is widely acknowledged to play a role in atopic dermatitis and is a major factor in life-threatening anaphylactic reactions.2 Traces of proteins in the peanut oil of some infant formulae are thought to be the sensitising factor.1 The same sensitising process is suggested in children treated with vitamin D in preparations made partly of peanut oil. Although cutaneous sensitisation is not synonymous with clinical reactivity to peanut but merely indicates a risk, our data suggest the need to avoid all medications containing peanut oil in infancy. GUY DE MONTIS
DOMINIQUE GENDREL MARTINE CHEMILLIER-TRUONG CHRISTOPHE DUPONT
Hôpital Saint-Vincent-de-Paul, 75674 Paris, France
1. Moneret-Vautrin DA, Hatahet R, Kanny G, Ait-Djafer Z. Allergenic peanut oil in milk formulas. Lancet 1991; 338: 1149. 2. Yunginger JW, Squillace DL, Jones RT, Helm RM. Fatal anaphylactic reactions induced by peanuts. Allergy Proc 1989; 10: 249-54.
Primary central
nervous
and brain
system lymphoma
biopsy
in AIDS
SIR,-In an editorial1 on brain biopsy for intracranial mass lesions, you say that inability to treat a central nervous system (CNS) lymphoma argues against regular use of brain biopsy in patients with late-stage AID S. At our centre, a prospective study on the role of brain biopsy in the diagnosis and treatment of intracranial mass lesions in patients with AIDS is ongoing. We enrolled, from Oct 1, 1991 to March 31, 1993, 56 patients with brain mass lesions. All were treated with an antitoxoplasma empirical regimen for 3-4 weeks. 30 were considered as having toxoplasmic encephalitis because of response to therapy. The other 26 were selected for potential brain biopsy; indication was in all cases a poor clinical and/or radiological response to antitoxoplasma therapy. In 13 (50%) biopsy could not be done because of contraindicatons or refusal. 13 had brain biopsy (8 stereotactic and 5 ultrasound guided), and 12 (92%) had a final diagnosis (7 primary CNS lymphoma, 3 progressive multifocal leukoencephalopathy, 1 mycotic abscess, 1 tuberculous abscess). Only minimum bleeding confined to the tumour bed occurred in 1 patient. Of the 7 patients with primary CNS lymphoma (3 immunoblastic, 3 large cell, and 1 lymphoblastic according to WF) 6 received whole-brain irradiation (cobalt-60, opposed lateral portals, 40-44 Gy). 5 completed a full course and were clinically stabilised. 3 showed a partial radiological response, and 2 a stable lesion (computed tomography). We observed clinical and radiological progression in only 1 patient, with death occurring 28 days after biopsy and before completion of radiotherapy. We retrospectively compared the 7 patients with biopsy-proven CNS lymphoma to the 7 patients with a necropsy-diagnosed cerebral lymphoma observed at our centre during 1986-90. The two groups were similar for age, sex, CD4 count, previous AIDS diagnosis, time from onset to presumptive diagnosis, severity of neurological status, white cell count, haemoglobin, and time of treatment with antiretroviral drugs. Patients with a biopsyconfirmed diagnosis had significantly longer survival from onset of neurological symptoms, with an estimated median survival time (EMST) of 233 days compared with 73 days in necropsy-proven patients (p 0-0002, logrank test). This result was confirmed by calculating survival from presumptive diagnosis (221 vs 9 days, =