Abstract
To determine the demographic and systemic parameters in children with solid malignancies and to ascertain which of them affected the delay in diagnosis, a retrospective study was performed on 315 children diagnosed with a solid tumor at our hospital, including epidemiological, social, and medical issues concerning the family, the child, the medical system, and the tumor. Lag time, defined as the interval between onset of symptoms and final diagnosis, including parent delay and physician delay, was estimated for each child. Mean lag time: 15.75 weeks (w), median: 7 w, range: 0–208 w. Lowest mean values appeared in kidney tumors, highest in epithelial, brain and soft tissue sarcomas. Mean parent delay: 4.42 w, median: 1 w, range: 0–130 w. Mean physician delay: 11.17 w, median: 4 w, range: 0–206 w. Among the demographic and personal parameters, the best predictors for diagnosis delay were age of child and father's ethnic origin. Several factors influenced diagnosis delay of childhood solid tumors. Recognizing these factors could minimize the delay, thereby improving the child's chances of survival.
Cancer in children is a rare disease, with one child out of 600 developing cancer by the age of 16 years. Still, malignant disesese are the main cause of death due to disease at 1–15 years of age Citation[1–3]. During the last 20 years, the survival of children with cancer has improved significantly, due to advances in therapeutic management, especially combination therapy, including chemotherapy, surgery, and radiotherapy, and to the more specifically adapted laboratory and imaging investigations. More than 70% of children with cancer are cured and most efforts are now concentrated on decreasing the late effects of treatment in long-term survivors Citation[1–4]. However, children diagnosed with advanced solid tumors still need more aggressive and adapted therapies in order to improve their cure rate. Diagnosis delay in childhood cancer is one of the predominant causes of disseminated malignancies.
Cancer in children can be difficult to diagnose in the primary setting: the index of suspicion tends to be low because of the relative rarity of the disese in children, the presenting features of malignancy in childhood usually are non specific and mimic those of common pediatric illnesses (fever, pain, headache, and vomiting) and therefore, the diagnosis may be delayed Citation[3, 4].
Lag time—the interval between the start of the symptoms and diagnosis—has been evaluated in a few studies in the UK Citation[5], the USA Citation[6], and Sweden Citation[7]. The main conclusions drawn concerning the causes of late diagnosis were related to the child's age, with older children presenting later, and the greatest lag time for brain tumors and Ewing's sarcoma.
The objective of this study was to investigate the determinants of symptom interval in children with malignancies [leukemia excluded] diagnosed in northern Israel. Our study included epidemiological, socioeconomic, medical and paramedical parameters, including the medical system. Doctor delay and parent delay were specifically analyzed to reach practical conclusions and to better train and educate parents and physicians.
PATIENTS AND METHODS
A retrospective analysis was performed on 347 children (aged 0–20 years) diagnosed with cancer, excluding leukemia, at Rambam Medical Center in Haifa, Israel, from 1993 to 2001. A total of 315 children were eligible for the study, as 32 files could not be fully completed. Informed consent was obtained from a parent. A questionnaire was fullfilled and 72 parameters were analyzed concerning the family's and the child's socioeconomic and epidemiological status, the medical system and the tumor. Three physicians reviewed each file according to the description of the disease reported by the parents and the primary care physician.
Lag times (‘overall delay’) were defined by the number of weeks between the onset of first symptoms and the date of diagnosis. Parent and doctor delays were defined respectively by the onset of first symptom until the visit to the first physician, and by the first visit to the physician until the final diagnosis. Lag time was divided into four time intervals: 1–3 weeks [w], 3–7 w, 7–15 w, and over 15 w. These time intervals were chosen according to the distribution of 25% of the children in each. The different parameters were described and statistically analyzed regarding to these time intervals.
Statistical Analysis
Statistical analysis was performed with the SPSS statistical program. Correlation between the categorical parameters were examined with the chi-square test. In cases where the sample was small, the Fisher exact test was used. The Spearman or Pierson adapter was used to examine the correlation between two parameters. The t test was used to determine the difference between two averages of two independent samples. When the number of independent samples was greater than two, the F test (ANOVA) was used. If the variables did not come from normal range or the sample was small, aparametric tests, Mann-Whitney or Kruskal-Wallis, were used. Finally, we examined the correlation between the dependent variables (continuous) and the ‘explanatory’ (dependent) variables (which were continuous or categorical), by using multivariate analysis of variance Citation[7, 8].
RESULTS
Mean Overall lag time was 15.75 w (0–208 w), mean parent delay 4.42 w (0–130 w), and mean doctor delay 11.17 w (0–206 w) ().
The Characteristics of the ‘Overall-Delay’ (Lag Time), Parent's Delay, and Doctor's Delay (Results in Weeks)
In disseminated diseases—the mean lag time and doctor's delay—were significantly longer than in localized disease: Overall (mean) lag time for 53/317 children who died during the study was 18.8 w, doctor delay 14.7 w. Overall lag time for 50 children with disseminated disease was 17.3 w, doctor delay 11.2 w. Overall lag time for 200 children diagnosed with localized disease was 10.2 w, doctor delay 4.2 w. Twelve children with localized disease had a recurrence and found to be with active disease: lag time 14.8 w, doctor delay 10.6 w.
Parameters Concerning the Family and Child
33% of non-Jews were diagnosed at 0–3 w, comparing to only 20% for Jewish families (p <. 01). (). When we examined this according to the specific religions—the mean Lag time of the Jews was the longest, but the median lag time of the Jews was longer than that of the Christians and Moslems but shorter than that of the Druses (, ).
Relationship between Religion and Lag Time
Relationship between Specific Religion Groups and Lag Time
1 Mean lag times according to religion.
Children of Isreali, Ashkenazi, or Arabic fathers had a shorter lag time (median 5–7 w) than children of Sephardic fathers (median 10 w) (p <. 05). The mean and median doctor's delay were similarly longer for children of Sephardic fathers compared to the other origins of the fathers.
Mean and median lag times for children who live in towns was the highest as compared to children who live in villages, cities, or kibbutzim.
There was a significant association between the father's age and lag time: children of younger fathers had an earlier diagnosis (p <. 01); the diagnosis was also reached earlier in children of younger mothers (p <. 01). The mean and median lag times were longer as the mother's age was higher ().
Relationship between Age of Mother and Lag Time
We did not find any correlation between the father's profession and lag time, but lag time for mothers with a ‘blue collar’ profession was greater than for housewives or mothers with academic professions (8.5–13.5 w vs 5.5–6 w).
In families with other chronic diseases, the overall delay was longer (median 25 w compared to 14 w for families without chronic diseases). If the child had other diseases, the lag time was higher (47% were diagnosed after 15 w), but the incidence in this group was low (11% of all children).
There was found a significant association between the order of the child in the family and the parent's delay—the higher it was—the shorter the parent's delay was.
and show the significant (positive) correlation between the age of the child at diagnosis and the lag time: as the age of the child increases, so does the lag time. There was no significant correlation between the child's gender and lag time.
Relation between Age of Child and Lag Time
Relation between the Age of the Child, Lag Time, Parent's Delay, and Doctor's Delay
2 Corelation between age of child and lag time.
Parameters Related to the Medical System
When the first consultant was a private physician or a hospital practitioner, the lag time was shorter; when the child was examined at a public clinic, the lag time was longer. Kruskal-Wallis tests revealed a significant association between the specialty of the first doctor and the lag time and doctor's delay: mean and median lag times and doctor's delays—were shorter for children examined by pediatricians in comparison to family physicians or other specialists (; ).
Relation between Type of Physician First Approached and Lag Time
3 Mean parent's delay, doctor's delay, and lag time according to the specialty of the first doctor.
As the number of consulted doctors increased, the lag time was longer; when only one doctor examined the child before the diagnosis, 38% were diagnosed within three w, compared to only 11% when three or more doctors were involved in the same time period (p <. 01). According to Pearson analysis, there was found to be a positive correlation between the number of times that the child visited the first doctor and the lag time: the more times the child was examined by the first doctor, the longer the lag time.
Parameters Related to the Tumor
There was a significant correlation between tumor type and lag time (). The greatest mean and median lag times were in brain tumors, especially astrocytoma, epithelial tumors, and in bone tumors, especially Ewing's sarcoma; the shortest was in children with Wilm's tumor.
Relation between Type of Tumor and Lag Time (Weeks)
When comparing the results for brain tumors: For medulloblastoma, there are lower mean and median lag times (8.9/4.5 w), compared to astrocytoma (30.9/15 w). When comparing the results for bone tumors, the mean and median lag times are lower for osteosarcoma (8.1/8 w), compared to Ewing sarcoma (24.2/12 w). When comparing the results for lymphomas, the mean and median lag times are lower for Hodgkin's lymphoma (9.1/5 w), compared to NHL (12.4/6 w).
When we used Kruskal Wallis aparameric test, there were significant differences in lag times according to the site of the primary tumor: when the tumor was located in the back and spine, 87% of the children were diagnosed after 7 w, in the gonads 72%, pharynx 64%, pelvis 58%, and extremities 58% (longest lag times); when the tumor was located in the brain, 53% of the children were diagnosed after 7 w, in the neck 52%, skin 50%; when the tumor was located in the abdomen, 66% were diagnosed within 7 w, skull 64%, eye 56%, and chest 55% (shortest lag times) (p <. 01).
The main presenting symptom was pain, and the overall delay and doctor delay were significantly higher in those cases. The number of presenting symptoms did not influence the lag time, but there was significantly lower parent's delay when the presenting symptom was rare as compared to cases with common presenting symptoms.
Multivariate Analysis
In order to estimate which were the most influencing parameters on the lag time, a multivariate analysis was performed using the GLM procedure, when only parameters concerning the parents and the child were included.
Four parameters were found significantly influencing the lag time: age of the child, ethnic origin of the father, religion, and place of residence.
A regression procedure was performed on transformation of the lag time and a significant model was received (p <. 001) with the predictors explaining 20% of the difference in lag time; only personal and demographic factors were examined. The two parameters with the greatest influence on the overall delay, were the child's age and the ethnic origin of the father (these results were relevant in only 16% of the changes): when the father's ethnic origin was non-Ashkenazi and as the child's age was getting higher, the lag time was getting longer.
Goodness of Fit
The Chi-Square Test procedure tabulates a variable into categories and computes a chi-square statistic. This goodness-of-fit test compares the observed and expected frequencies in each category to test either that all categories contain the same proportion of values or that each category contains a user-specified proportion of values:
The results demonstrate that there is no significant difference between the observed freqency and the expected freqency in any category of patients (according to the lag times intervals). Probably, the observed freqency is not much different from the expected freqency.
DISCUSSION
Few studies on delay in diagnosis in pediatric malignancies have been reported, most reporting the influence of age and type of tumor Citation[5, 6]. The rarity and the non-specific clinical presentation of the disease influenced parent delay in seeking medical advice and doctor delay in reaching the diagnosis. Parent delay was reported in the study of Thulesius et al. Citation[9] regarding diagnosis (type of tumor) only. Social factors were analyzed by Fajardo-Gutierrez et al. Citation[10] with a risk of delay reported to be higher in the low educational level group.
In our study, the main parent parameters studied were ethnicity, religion, profession, educational level, socio-economic level, age, place of residence, number of children in the family, and diseases in the family. Although the ethnicity of the father was a significant factor in delay of diagnosis in multivariate analysis, the other results must be considered and larger studies may better explain our findings. A possible explanation to the observation that younger parents had a shorter lag time is that these parents tend to consult quicker the doctors than older parents (shorter parent's delay), and by the fact that usually they have younger children.
Differences according to race were reported only by Pollock et al. Citation[6], where white children with osteogenic sarcomas had a longer lag time. In our study, Arabic children had a shorter lag time than Jewish children. A possible explanation is that the doctor's awareness is higher among the Arab population because of the greater consanguinity rate (between the parents) in this population, contributing to shorter lag times.
An issue that is more reported is correlation between the age of the child and delay in diagnosis. Saha et al. Citation[5], Pollock et al. Citation[6], Pratt et al. Citation[11], and Flores et al. Citation[12] confirmed that delay in young patients is less than in older children. Our results confirm these findings.
Younger children are likely to be seen more often by their physician than older children and there is greater parent awareness of the child's condition in younger children, with detection of symptomatic disease in older children dependent greatly on self-reporting.
We should also remember that the age parameter is influenced from additional parameters concerning the type of the tumor, like the anatomical site and symptoms which characterize specific ages and thereby infuencing the lag time Citation[6].
Pain was found to be the main presenting symptom in our study, and a significant factor in children diagnosed with a longer lag time; doctor delay was similarly higher in cases where pain was the presenting symptom, without any other more specific symptomatology.
Anatomic site influences delay in diagnosis but there are few reports on this factor. Pratt et al. Citation[11] reported a shorter lag time for children with rhabdomyosarcoma located in the pharynx and orbit compared with children diagnosed with rhabdomyosarcoma located in the face and neck. Flores et al. Citation[12] reported that children with infratentorial brain tumors have the shortest lag time in comparison with children with supratentorial tumors (similar to our findings).
In our study, the mean and median lag times were short for renal tumors, intermediate for lymphomas, and long for brain, bone, and epithelial tumors. Children with disease located to the abdomen, skull, eyes, and chest had the shortest lag time.
There seems to be a triade of parameters affecting each other: the child's age, the type of the tumor, and the anatomical site: most of the children with abdominal diseases are young children with Wilms' tumor, lymphoma, or neuroblastoma. Children with disease located in the pelvis and extremities are generally older children diagnosed with sarcoma. Most authors Citation[4, 5, 10] concluded that brain tumors and sarcomas have the longest lag time; these tumors are predominant in older children and their presenting symptoms are non specific, especially at the beginning of the disease Citation[5, 13]. Children diagnosed with Wilms' tumor are young and the lag time was the shortest in our study and in other studies Citation[6, 10].
Another example for the contribution of the anatomical site to the lag time is given by Pollock Citation[6], supported by our findings: axial bone tumors (Ewing's sarcoma) are more difficult to observe, therfore having longer lag time compared to tumors in the long peripheral bones.
The lag time is influenced also from the biologic features of the tumors (bone and brain tumors tend to have slower growth rates than other tumors, hence causing longer lag times), and from the presenting features of the tumor (when they are rare in childhood, the tumor will be diagnosed earlier).
When analizing parameters related to the medical system, we have found that the lag time was influenced by the type of medical authority that was consulted (shortest when hospitals approached directly); the specialty of the first doctor (shortest among pediatricians); the number of additional doctors that the child had visited and the number of times that the child visited the first doctor (gets longer as number of doctos/visits increases).
In our hospital, which is a referral center for northern Israel, a few pediatrician specialists are rapidly involved when a child is hospitalized for persistent and progressive symptoms and, therefore, the diagnosis is made earlier. The responsibility of the general physician has been reported in our study and in the study of Saha et al. Citation[5] also, but the education of the general physicians at the primary health care clinics has to be improved.
In our study, lag time was predictive of survival, with a significant correlation with doctor delay. Shortening of the diagnosis period, can probably improve the prognosis. The same conclusions are described by Pratt et al. in the study on survival of children with rhabdomyosarcoma Citation[11], with a significant correlation between stage of disease, lag time, and survival. Flores et al. Citation[12] reported delay diagnosis in brain tumors and correlated the lag time with the prognosis. The relationship between lag time and prognosis is very complex and related to tumor biology as well as to physician or parent delay. In Israel, access to health care is free and does not influence lag time, contrary to other countries. Several authors have reported on delay in breast cancer when access to health institutions is directly correlated to the socioeconomic level of the population Citation[14, 15]. We didn't find any correlation to this parameter.
Like Pollock et al. Citation[6], the regressions performed for each prognostic factor group explained no more than 16% of the variance in lag time; this suggests that the nature, genetic and biological profiles, and epidemiological characteristics of the tumor and individual factors are important determinants for lag time. Nevertheless, further studies need to be preformed to identify the patient characteristics most strongly associated with longer lag times. Early detection interventions in childhood cancer are necessary, including public campaigns aimed at the general population, both adults and children, especially teenagers, along with educational programs for physicians.
In conclusion, maintaining a high index of suspicion may result in earlier diagnosis; Both parents and doctors can influence the Lag time; increased vigilance on both sides could lead to shortening in the lag time therby improving the prognosis.
The authors thank (1) Mrs. M. Perlmutter for her help in the preparation of this paper; (2) Mrs. R. Govrin for her help in fulfillment of the questionnaires; (3) For the statistical analysis: (Mrs. K. Katz, Diagnostic Statistical Institute, Haifa, Israel) and (Mrs. HF Rennert, Department of Community Medicine and Epidemiology, Carmel Medical Center, Haifa, Israel); and (4) The ‘Israel Cancer Association’ for funding.
REFERENCES
- SEER Cancer Statistics Review, 1973–1996. Feb 7, 2000, http://www-seer.Ims.nci.nih.gov/publications/CSR1973–1996
- Gurney J G, Davis S, Severson R K, Fang J Y, Ross J A, Robinson L L. Trends in cancer incidence among children in the US. Cancer. 1996; 78: 532–541. [CROSSREF], [PUBMED], [INFOTRIEVE]
- Bleyer W A. The impact of childhood cancer on the United States and the world. CA cancer. J Clin.. 1990; 40: 355–367
- Greenwald H P, Becker S W, Nevitt M C. Delay and noncompliance in cancer detection: a behavioral perspective for health planners. Milbank Med Fund Q Health Soc.. 1978; 56: 212–230
- Saha V, Love S, Eden T, Micallef-Eynand P, Mackinlay G. Determinants of symptom interval in childhood cancer. Arch Dis Child. 1993; 68: 771–774. [PUBMED], [INFOTRIEVE]
- Pollock B H, Krischer J P, Vietti T J. Interval between symptom onset and diagnosis of pediatric solid tumors. J Pediatr. 1991; 119: 725–732. [PUBMED], [INFOTRIEVE]
- McGraw K O, Gordy S, Wong S S. How many subject to screen? AS practical procedure for extimating mutivariate normal probabilities for correlated variable. J Consult Clin Psychol.. 1984; 62: 960–964
- Winer B J, Brown D R, Michels K M. Statistical Principles in Experimental Design. New York, McGraw-Hill. 1991
- Thulesius H, Pola J, Hakansson A. Diagnosis delay in pediatric malignancies. A population based study. Acta Oncol.. 2000; 39: 873–876. [CROSSREF], [PUBMED], [INFOTRIEVE]
- Fajardo-Gutierrez A, Sandoval-Mex A M, Mejia-Arangure S M, Rendon-Macias M E, Martinez-Garcia Mdel Cl. Clinical and social factors that affect the time to diagnosis of Mexican children with cancer. Med Ped Oncol.. 2002; 39: 25–31
- Pratt C B, Smith J W, Woerner S. Factors leading to delay in the diagnosis and affecting survival of children with head and neck rhabdomyosarcoma. Pediatrics. 1978; 61: 30–34. [PUBMED], [INFOTRIEVE]
- Flores L E, Williams D L, Bell B A, O'Brien M, Ragab A H. Delay in the diagnosis of pediatric brain tumors. Am J Dis Child. 1986; 140: 684–686. [PUBMED], [INFOTRIEVE]
- Edgeworth J, Bullock P, Bailey A, Gallagher A, Crouchman M. Why are brain tumors still being missed?. Arch Dis Child.. 1996; 74: 148–151. [PUBMED], [INFOTRIEVE]
- Katz S J, Hislop T G, Thomas D B, Larson E B. Delay from symptom to diagnosis and treatment of breast cancer in Washington State and British Columbia. Med Care. 1993; 31: 264–268. [PUBMED], [INFOTRIEVE]
- Wilkinson G G, Edgerton F, Wallace H J, et al, Delay, stage of disease and survival from breast cancer. J Chron Dis.. 1979; 32: 365–373. [CROSSREF], [PUBMED], [INFOTRIEVE]