Ugburo A.O.1, Oyeneyin J.O.1, Atuk T.A.2, Desalu I.S.2, Sowemimo G.O.A.3

1 Department of Surgery, College of Medicine, University of Lagos, Lagos, Nigeria
2 Lagos University Teaching Hospital, Lagos
3 Plastic Surgery Unit, Department of Surgery, College of Medicine, University of Lagos, Lagos

SUMMARY. This is a study of 94 patients in Nigeria who sustained flame burns resulting from explosions of kerosene contaminated with petrol between 10 October and 25 November 2001. Fifty-eight families were involved, with multiple family members affected. The incident was later discovered to have been due to contamination of kerosene from a storage tank at a fuel depot. Most of the accidents occurred while people were trying to refuel a lighted lantern. The ages of the 94 patients seen in hospital ranged from 3 weeks to 55 yr (mean age, 21.88 ± 1.41 yr). Many of the patients suffered severe burns with a mean percentage of burned body surface area (BSA) of 37.05 ± 3.22%. A significant percentage of the burns were full-thickness injury (mean percentage burned BSA, 19.95 ± 2.02). The right upper limb holding the lantern was usually more severely affected than the left in the person refuelling the lighted lantern. The incidence of inhalation injury resulting from inhaled kerosene and smoke due to the impossibility of rescuing the victims promptly - because of “burglar-proof security” - was 37% of all the patients. The majority of these patients died of inhalation injury, which was the single commonest cause of death (40% of fatal cases). Such disasters due to human error in the use of petrol-contaminated kerosene could be minimized by stringent control. The custom of refuelling a lantern while it is lighted should be discouraged by public education campaigns. There is also a need to improve the design of the hurricane lantern commonly used in rural areas and in developing countries.


Kerosene, a petroleum product with a relatively high flash point of 35 °C and low flammability, is commonly used for domestic purposes in rural areas and in developing countries.1 It is very popular in Nigeria among the low and middle socio-economic classes as a fuel for lighting and cooking. Its importance is more pertinent because of the poor electricity supply in rural areas and frequent power cuts.

Epidemic flame burns resulting from kerosene contaminated with petrol has been reported from Nigeria before.2 This situation can rightly be described as an epidemic because of the sudden increase in the number of patients, which overwhelms local resources and necessitates regional and national assistance. This type of burns disaster has the following common features: 1. complete preventability of the cause; 2. continuation of the disaster over a long period.1 The aims of this report are to review all the cases of patients admitted to the Lagos University Teaching Hospital with burns resulting from kerosene appliance explosions occurring between 10 October and 25 November 2001, when the epidemic ended, to investigate the cause of the epidemic, and to compare the pattern of burns in the epidemic with that of previous reports from this centre2 and of another report found in a search of the literature.

Materials and methods

This is a prospective study of patients seen and discharged or admitted to the adult and children’s emergency wards between 10 October and 25 November 2001 suffering from kerosene explosion burns. This was fortuitous, as we had commenced an epidemiological study of burns in April 2001, using a modified version of the WHO burns data form.3 All patient data were recorded on this pro forma. Major burns were defined as any burn greater than 15% body surface area (BSA) in adults or greater than 10% BSA in children. Any patient presenting with burns of any size plus inhalation injury was also classified as a major burn and admitted. All other patients were classified as minor burns and treated on an out-patient basis. Inhalation injury was diagnosed by a combination of any finding of facial burn and singeing of the nasal vibrissae or burns in an enclosed space, in combination with any of the following signs: brassy cough, carbonaceous sputum, dyspnoea, wheezing, and the finding by the anaesthetist of erythema or soot in the larynx and vocal cords. We do not have the facilities for fibre-optic bronchoscopy or ventilation perfusion scans. The cause of death in the fatal cases was confirmed by our morbid anatomy department. We also closely monitored the investigation of the source of the contamination carried out by an expert committee appointed by the National Petroleum Corporation. The data were analysed with SPSS for Windows, standard version, release 10.0.1, October 1999 (commercial statistical package). The confidence interval was set at 95% and all values were reported as mean ± SEM. The 0.05 level of significance was used for statistical difference. A descriptive statistical analysis was applied and the paired Student’s t-test was used for the comparison of means.


The common finding in the case histories was that the accident usually occurred after a power cut. The family would have gathered around a hurricane lantern (Fig. 1a), which was usually the only source of light.

<% immagine "Fig. 1a","gr0000001.jpg","The common hurricane lantern used in Nigeria and the usual manoeuvre for refilling the lantern by holding it in the left hand.",230 %>

After a period ranging from a few minutes to some hours, it became clear that the flames were dying out because of exhaustion of the fuel. The event occurred while the lantern was being refuelled. There was usually an explosion, with the spraying of kerosene from the lantern and the storage can over the patient and other family members. If the family members were not in the vicinity, the patient sometimes threw the lantern and kerosene container towards them. This sequence of events resulted in burns involving multiple family members and sometimes house fires. Escape from the building and rescue were usually delayed because of “burglar proofs” (Fig. 1b).

<% immagine "Fig. 1b","gr0000002.jpg"," Typical building with “burglar-proof” grilles.",230 %>

These “burglar proofs” - steel grilles placed over the windows and doors - prevented escape as the keys could not be easily located in the immediate post-explosion chaos. Many victims thus suffered inhalation injury. Patients refuelling lanterns also sometimes suffered chemical inhalation injury due to inhalation of kerosene.

<% immagine "Fig. 2a","gr0000003.jpg"," Incidence of flame burns due to kerosene lantern explosions (April 2001-March 2002)",230 %>

Fig. 2a shows all admissions from kerosene explosion burns between April 2001 and March 2002. This figure shows that burns from kerosene explosions are an uncommon event. However, from 10 October to 30 November 2001, the incidence of burns resulting from kerosene appliances increased to 74 cases in October and 17 in November. The bar chart in Fig. 2b shows the weekly trend in the number of patients seen. There was a sudden rise to 43 patients in the incidence of burns due to kerosene lantern explosions in the first week. By the second week the incidence had dropped steeply to 19 cases, and this progressively reduced to one case each by the seventh and eighth weeks of the epidemic. When the epidemic was declared over by the ninth week, 91 patients had been affected. Three cases - one seen in September 2001, one in January 2002, and another in February 2002 - were added to the analysis to give an overall total of 94 patients.

<% immagine "Fig. 2b","gr0000004.jpg"," Bar graph showing number of patients seen per week.",230 %>

A study of the patients’ addresses indicates that most of the local councils in the state of Lagos were affected, the majority of cases coming from local council areas that we class as low-income areas, e.g. 29% from Mushin, 14% from Ajeromi-Ifelodun, 6% from Oshodi-Isolo, and 34% from Surulere. The other eight council areas produced 27% of the cases. Most of the injuries occurred between 6 p.m. and 6 a.m. (Fig. 2c). During this time interval the day is usually dark, and in an associated power cut the patient was usually attempting to refuel a lantern for use in the night, or trying to refuel it in the early hours of the morning for use next day. The six patients who were burned between 10 and 12 a.m. were members of a family who sustained burns while trying to fill a kerosene container in the presence of a naked stove light. The time interval between the onset of injury and the moment of hospital presentation in the 88 patients who presented within 24 h was 4.39 ± 0.44 h, with a range of 30 min to 17 h. The rest of the patients were referred from other hospitals after several days because of complications.

<% immagine "Fig. 2c","gr0000005.jpg","Time of injury on 24-hour clock",230 %>

With regard to age, the patients ranged from a 3-week-old neonate to a 55-yr-old man. The patients’ mean age was 21.88 ± 1.41 yr. Table I shows the sex distribution of the adult and children patients. There were 35 children (17 males and 18 females; male/female ratio, 0.94:1) and 59 adults (24 males and 35 females; male/female ratio, 0.68:1). The overall male to female ratio was 0.77:1. Thirteen per cent of the patients were assessed as having minor burns and treated as out-patients. Seventy-five patients were admitted as fresh patients, while 12% were transferred with sepsis from other hospitals.

<% createTable "Table I ","Distribution of patients by sex","§1,4§Sex@; Female;Male;Total@;Children;18;17;35@;Adult;35;24;59@;Total;53;41;94","",4,300,true %>

The burn injuries tended to involve multiple family members. Fig. 3 shows the distribution of the number of family members affected per incident. Altogether, there were 63 fire incidents. One was due to a spontaneous lantern explosion involving a family of 10, of whom only two children survived. Other incidents involved the refuelling of a lighted kerosene stove. The rest of the incidents were due to lantern explosions occurring while attempts were made to refuel a lighted kerosene lantern. One person only was affected in 38 fire incidents, two persons in 20 incidents, three in seven incidents, six in one incident, and ten in one other incident. However, not all members of the families involved in all cases were treated at our centre, as some of the less severe burns were treated in general hospitals.

<% immagine "Fig. 3","gr0000006.jpg","Number of persons affected per incident",230 %>

The mean BSA burned ranged from 2 to 100%. The mean total BSA burned was 37.05 ± 3.22%. Forty-one patients sustained full-thickness burns ranging from a minimum of 2% BSA to a maximum of 45% BSA. The mean full-thickness area of burns in this subset was 19.95 ± 2.02% BSA. We noted that the left upper limb that held the lantern (Fig. 1a) in 61 patients was more severely affected in terms of surface area and depth of burns than the right upper limb. In most patients the anterior trunk was also more severely affected than the posterior trunk. The mean percentage BSA affected for the left upper limb was 6.67 ± 0.27% and, for the right, 3.76 ± 0.32%. Using the paired sample Student’s t-test to compare the difference, surface area burns in the left upper and right upper limbs showed a highly significant difference (p = 0.000). Comparing the percentage BSA affected in the anterior and the posterior trunk, we found the mean anterior trunk surface area affected to be 7.10 ± 0.68; the mean posterior trunk was 4.86 ± 0.68. Using the paired Student’s t-test to compare the difference between the mean values of the anterior trunk surface area with the posterior trunk surface area, we found a highly significant difference (p = 0.000).

The number of fatalities among all the patients was 35, equal to 37% (Table II). This was largely contributed to by the deaths of children (20 patients out of 35, i.e. 57%) compared with 15 out of 59 patients among adults, i.e. 25%. Table II also shows that the majority of the adults suffered less severe injuries (i.e. in less than 40% BSA) - 43 adults compared with four children. More children suffered more severe burns, i.e. greater than 40% BSA - 18 children (51%) compared with 16 adults (27%). Children are more likely to be trapped in house fires and to be unable to escape, suffering significant surface burns and inhalation injury (Table II).

<% createTable "Table II ","Extent of burns and mortality",";Extent;Number of children and deaths (%);Number of adults and deaths (%);Total number of cases and deaths (%)@;< 20%;8 1 (12.5%);30 2 (7%);38 3 (8%)@;20 - 40%;9 3 (33%);13 0 (0%);22 3 (14%)@;4 - 60%;7 5 (71%);6 4 (66%);13 9 (70%)@;6 - 80%;3 3 (100%);3 2 (66%);6 5 (83%)@;> 80%;8 8 (100%);7 7 (100%);15 15 (100%)@;Total;35 20 (57%);59 15 (25%);94 35 (37%)","",4,300,true %>

In this series, 35 (37%) of the patients were found to have suffered various degrees of inhalation injury. Of this number, only six patients (17%) were admitted to the available bed space in the intensive care unit for naso-tracheal intubation and ventilation (Table III). The rest were treated with oxygen by means of face masks, but early mortality in this group was high. This is demonstrated by the mean length of survival before death (3.81 ± 0.77 days).

Among survivors, the mean length of hospitalization was 34.35 ± 5.79 days. As said, there were 35 (37%) deaths in this study (Table II). The commonest cause of death was inhalation injury, i.e. in 13 of the fatal cases (37.1%). This was followed by septicaemia (12 cases, or 34.3% of all the fatal cases) (Fig. 4).

<% immagine "Fig. 4","gr0000007.jpg","Cause of death",230 %> <% createTable "Table III ","Distribution of patients with inhalation injury who were intubated","; Total;Deaths (%)@;Intubated and ventilated;6;3 (50%)@;Oxygen therapy only;29;10 (35%)@;Total;35;13 (37%)","",4,300,true %>

The commonest surgical procedure was escharotomy, which was carried out in 29 patients in 78 limbs (48 upper and 30 lower). Of these patients, fourteen underwent the procedure in all four limbs, six patients in both upper limbs, eight patients in only the left upper limb, and one patient in both lower limbs. There was no isolated right upper limb involvement. Eight patients underwent skin grafting and one patient each underwent release of contracture and incision and drainage for a subperichondrial abscess.

<% createTable "Table IV ","Distribution of chronic complications in survivors",";Type of complication;Number of patients;Proportion@;Dyschromia ;22;27%@;Hypertrophic scars ;19;23%@;Ear chondritis;5;6%@;Contractures;4;6%@;No complication;33;40%","",4,300,true %>

The 59 survivors had a combination of complications. The commonest of these was burn dyschromia, which was found in 22 patients. This was followed by hypertrophic scars (19 patients). Five patients developed auricular chondritis, leading to loss of portions of the ear and deformity. Four patients who refused skin grafting developed contractures.


The results of this report indicate that adults were more affected than children (63% of cases). This differs from the 1984 report from Lagos,2 according to which children formed the bulk of the patients (63%). We could not find any explanation for the disparity. However, most of the burn injuries in the adults in this report were not extensive in their surface area and resulted from attempts to rescue the patient. In house fires, children are more likely to be trapped and to suffer inhalation injury and severe surface burns. The distribution of burns in children in this series is similar to that reported from Lagos in 1984;1 there was a large number of small burns in less than 20% BSA and greater than 60%. This pattern was also found in adults in the 1984 disaster. However, while there was no mortality in burns in less than 20% in the 1984 report, there was a total mortality of 8%, and 12.5% of children and 7% of adults with less than 20% burned BSA died of inhalation injury.

In contrast with the 1984 report, not all patients with burns greater than 60% BSA died. In the disaster we describe, only 66% of the victims with burns in more than 60% BSA died. This may have been due to better intensive care facilities for the monitoring of burn patients. Inhalation injury was the commonest cause of death (40%). One patient had acute renal failure in addition to inhalation injury. Septicaemia was the cause of death in 34.3% of the patients. Septicaemia was the commonest cause of death in the 1984 report from our centre.1 The reason for this was that escape from the building and rescue were usually delayed because of “burglar proofs”, i.e. steel grilles constructed over windows and doors that prevented escape. These were installed in many buildings from the mid-1980s onwards following an increase in burglaries and armed robbery attacks in the home.

The body sites most commonly affected in family members were the face, the anterior trunk, and both upper limbs. However, in cases of clothing ignition and house fires, other sites were usually affected. It was commonly found that the left upper limb holding the exploding lantern was more significantly affected than the right, while in one left-handed patient who held the lantern in the right hand we found the opposite: the left upper limb was more severely affected in terms of surface area burned.

In most major disasters the casualty figures are known or can be easily estimated, and logistics can be planned in terms of manpower, space, equipment, drugs, and other consumables. In this special type of burn disaster, which has a continuous quality, it is difficult to predict the total number of victims who will eventually be affected. It is therefore difficult to plan and manage the disaster situation. Sharpe and Foo4 have provided a mnemonic for the hospital management of major burn disasters, in which the provision of adequate resources precedes the updating of casualty figures.

In this type of disaster situation, it is difficult to predict the required logistics in terms of manpower, medication, and hospital facilities. It is difficult to know whether to call on national or international assistance. In the management of such a situation, it is essential that a daily situation report should be drawn up and projections made for further casualties in order to determine when to call for outside assistance. In this type of disaster situation, the updating of casualty figures goes hand in hand with the making of requests for further resources.

The extent of this disaster was limited by the urgent intervention. Within the first week, it was reported to the hospital management by the surgery department. Press conferences were called and people were warned via the electronic and printed media to discontinue the use of kerosene. The national government was informed and by the second week the Ministry of Health had brought in supplies of drugs, intravenous fluids, antibiotics, and other medical consumables. The national petroleum company and the State government also offered to exchange all available kerosene for safe kerosene. This kerosene exchange must have assisted in the early termination of the epidemic. In the report by Gupta et al.,1 the burn epidemic lasted for three months because people had no other alternative to kerosene for lighting, as no alternative source of uncontaminated kerosene was provided.

Kerosene does not usually explode and many of the victims had previously been accustomed to refuelling lighted lanterns in the past. It was therefore suspected that there must have been major contamination by a flammable substance. Investigations later carried out by the national oil company traced the source of the contamination to a major marketer at Apapa, in the Ojo local council in the suburbs of Lagos, one of the most severely affected areas. The depot tank had initially been used for storage of petrol but had been abandoned for years. It was later used to store kerosene before distribution, leading to the present disaster. The previous disaster reported from Lagos1 also resulted from a filling-station petrol storage tank being used for storage of kerosene. In the report from India1 the accident was due to the human error of emptying kerosene from a train wagon into a truck that had been used to convey petrol. This type of burn disaster thus has the following common features:1 1. complete preventability of the cause; 2. continuation of the disaster over a long period. We would add a third feature: 3. such a disaster is commonly caused by human error.

The reason why the disaster persisted for eight weeks was that the contaminated kerosene did not usually explode with other domestic use. Indeed there was only one case of a kerosene stove exploding and one of a spontaneous lantern explosion. In fact, from the second week on, most of the affected families were aware that contaminated kerosene was on the market, but since the kerosene had been used safely for cooking they did not heed the advice to return all the kerosene and replace it with uncontaminated kerosene. Similar findings where explosions occurred only after the refilling of lighted lanterns have been reported from India1 and Lagos.2

A mechanism for these explosions was suggested by Gupta et al.,1 in which, with usage, the petrol contaminant with its lower specific gravity floats to the top of the kerosene. With the heat generated from the chimney of the lantern, the more volatile petrol forms an explosive vapour just beneath the lantern flame. When an attempt is made to fill the lantern, the explosive petrol fumes are displaced upwards towards the flames, detonating the mixture. It is possible, as suggested by Gupta et al. in the case of the common chimney lamps used in India, that the hurricane lanterns commonly used in Nigeria might benefit from an improvement in their design.


In order to prevent future occurrence of this type of disaster, we suggest that there should be an intense public awareness campaign in the printed and electronic media on the danger of refuelling lighted lanterns. There needs to be legislation requiring that hurricane lanterns should carry warning labels in the major local languages calling attention to the danger of refuelling lighted lanterns. People should be advised to extinguish the flames of the lantern and to refill the lantern by torchlight. There have been other unpublished kerosene burn explosions in Nigeria (Gwagwalada in 2000, and Edo state in 2001), and all have been traced to contamination in storage tanks in petroleum depots. We think it is time for separate depots to be designated for kerosene and for other petroleum products.

RESUME. Les Auteurs ont étudié 94 patients en Nigérie atteints de brûlures par flammes à la suite d’une série d’explosions de kérosène contaminé par de l’essence entre le 10 octobre et le 25 novembre 2001. Cinquante-huit familles ont été intéressées. Ensuite il a été démontré que l’accident a été causé par la contamination de kérosène provenant d’un dépôt de carburant. La plupart des accidents se sont vérifiés quand les personnes tentaient de ravitailler une lanterne illuminée. L’âge des 94 patients observés en hôpital variait de 3 semaines à 55 ans (âge moyen, 21,88 ± 1.41 ans). Beaucoup des patients ont aussi subi des brûlures sévères (pourcentage moyen de la surface corporelle brûlée [SCB], 37,05 ± 3,22%). Un pourcentage significatif des brûlures était constitué de brûlures à toute épaisseur (pourcentage moyen SCB, 19,95 ± 2,02). Le membre droit supérieur qui tenait la lanterne était intéressé plus fréquemment du membre gauche de la personne qui ravitaillait la lanterne illuminée. L’incidence des lésions par inhalation dues au kérosène inhalée et à la fumée à cause de l’impossibilité de sauver les victimes immédiatement - conséquence des grilles de sécurité contre les cambrioleurs - correspondait à 37% de tous les patients. La cause du décès de la plupart de ces patients était la lésion par inhalation (40% des fatalités). Le numéro des désastres de ce type, dus à l’erreur humaine dans l’emploi du kérosène contaminé par l’essence, pourrait être réduit moyennant des contrôles rigoureux. L’habitude de ravitailler une lanterne illuminée doit être découragée moyennant une campagne d’éducation du public. Il faut en outre améliorer le modèle de la lanterne-tempête communément utilisée dans les zones rurales et les pays en voie de développement.


  1. Gupta M., Bansal M., Gupta A., Goil P.: The kerosene tragedy of 1994, an unusual epidemic of burns: Epidemiological aspects and management of patients. Burns, 22: 3-9, 1995.
  2. Grange A.O., Akinsulie A.O., Sowemimo G.O.A.: Flame burn disasters from kerosene appliance explosions in Lagos, Nigeria. Burns, 14: 147-50, 1988.
  3. Computerized burn registry data form. Burns, 21: 234, 1995.
  4. Sharpe D.T., Foo I.T.H.: The management of burns in major disasters. Injury, 21: 41-4, 1990.
<% riquadro "This paper was received on 16 September 2002.

Address correspondence to: Dr A.O. Ugburu, Department of Surgery, College of Medicine, University of Lagos, P.M.B. 12003 Lagos, Nigeria. Tel.: 234 08023124495; e-mail: andyugburo@yahoo.com" %>

<% footer %>