T. Dang-Tan is recipient of a CIHR doctoral scholarship.
Diagnosis delays in childhood cancer
Article first published online: 9 JUL 2007
Copyright © 2007 American Cancer Society
Volume 110, Issue 4, pages 703–713, 15 August 2007
How to Cite
Dang-Tan, T. and Franco, E. L. (2007), Diagnosis delays in childhood cancer. Cancer, 110: 703–713. doi: 10.1002/cncr.22849
- Issue published online: 2 AUG 2007
- Article first published online: 9 JUL 2007
- Manuscript Accepted: 30 APR 2007
- Manuscript Revised: 29 APR 2007
- Manuscript Received: 26 FEB 2007
- Canadian Institutes of Health Research (CIHR). Grant Number: MOP-64455
Timely access to quality healthcare has become an increasingly important public health concern over the years. Early diagnosis of cancer is a fundamental goal in oncology because it allows an opportunity for timely treatment while disease burden is still in its earliest stages. Consequently, prognosis may improve, and a cure can be attained with minimal side or late effects. This review examined delays present in diagnosis of childhood cancers and factors that influence these delays. An extensive search of the literature published before April 15, 2007 was conducted for studies that evaluated any type of delay along the cancer-care continuum. Twenty-three studies were included. Diagnosis delay varied across studies. Physician delays were generally longer than those consequent to parents' or patients' recognition of underlying disease. Causes of delays can be grouped into 3 categories: patient and/or parent, disease, and healthcare. The main factors related to diagnosis delay were the child's age at diagnosis, parent level of education, type of cancer, presentation of symptoms, tumor site, cancer stage, and first medical specialty consulted. Greater understanding of factors that influence delays and the individual impact of patient and provider delays on disease severity and prognosis would be useful to form effective policies and programs aimed at ensuring timely access to healthcare for children with cancer. Cancer 2007. © 2007 American Cancer Society.
Cancer is the leading cause of disease-related death in children in developed countries and remains an important public health concern because of its great physical and psychological impact on the affected children and their families.1 Many possible risk factors for development of cancer in children and adolescents have been investigated.1 However, the causes of childhood cancer are mostly unknown. Currently, timely diagnosis immediately followed by effective treatment is an essential approach for control of the public health burden due to childhood cancers. Appropriate early diagnosis and treatment require primary care physicians and parents to be aware of early symptoms of childhood malignancies. Public and professional education can be effective in eliminating disparities in cancer survival.2 Despite these suggestions, the study of diagnosis delays in children's malignancies has not received as much attention as cancers in adults.
Research on this topic is complicated by methodological difficulties as well as problems inherent to the biological properties and clinical behavior of childhood cancers. Childhood cancers tend to have short latency periods and often grow rapidly. Tumors in children are very invasive but are more responsive to treatment than adult tumors. Factors related to perception of severity of signs and symptoms are also different. Children are usually under the care of their parents, which underscores the importance of parents' knowledge, attitudes, and behavior in the cancer diagnosis pathway. Conversely, rapid and early cancer diagnosis is not solely a function of actions taken by patients or parents. A complex chain of events is triggered once a patient with cancer is seen by a healthcare professional. The combination of various factors related to the healthcare provider and the complexity of the patient's disease may also lead to a delay attributable to the healthcare system.
Few studies have been published on determinants and impacts of diagnosis delays in childhood cancer. Development of effective new strategies to shorten delays in childhood cancer diagnosis requires an understanding of these delays and their effect on cancer prognosis. To our knowledge, no review has assessed research on this topic. We present, herein, a review of current knowledge on diagnosis delay in childhood cancer and discuss methodological issues and challenges faced in this area of research.
MATERIALS AND METHODS
For this review, we identified epidemiological studies on diagnosis delays in childhood cancer listed in Medline and PubMed before April 15, 2007. Specifically, we performed a literature search by using the index terms children, cancer, diagnosis, delay, prognosis, risk factor, epidemiology, cohort, case-control, and alternate synonyms in various combinations. The reference lists of articles were also examined for any additional publications that were not identified by the bibliographic search. Twenty-three published studies, written in English, were identified, and relevant data were abstracted for this review. These studies calculated delays from data extracted retrospectively from patient medical charts, tumor registries, and, in some studies, interviews with parents. Delay times were converted to weeks for this review.3–15 Table 1 summarizes the main characteristics of these 23 investigations.
|Study||Year||Country||Patient population||Sample size||Age, y||Study type||Cancer type||Data source|
|Gjerris15||1976||Denmark||Hospital-based||299||0–14||Retrospective cohort||Brain tumor||Medical chart|
|Pratt8||1978||USA||Hospital-based||46||4 mo–20 mo||Descriptive||Head and neck rhabdomyosarcoma||Medical charts|
|Haik21||1985||USA||Hospital-based||254||1 mo–3.5 mo||Descriptive||Retinoblastoma||Medical charts|
|Flores25||1986||USA||Hospital-based||79||0–19||Retrospective cohort||Brain tumors vs Wilms and acute leukemia||Medical charts|
|DerKinderen*12||1989||Netherlands||Population||130||0–20||Retrospective cohort||Retinoblastoma||Medical charts/Parent interview|
|Pollock7||1991||USA||POG Protocol||2665||0–29||Cross-sectional||Lymphoma and solid tumors||POG Protocol Database|
|Saha19||1993||England||Hospital-based||236||0–15||Retrospective cohort||All cancers||Medical charts|
|Edgeworth24||1995||England||Hospital-based||74||0–16||Descriptive||Brain tumor||Medical charts/Parent interview|
|Goddard23||1999||England||Hospital-based||100||0–8||Retrospective cohort||Retinoblastoma||Parent interview|
|Chantada22||1999||Argentina||Population||95||0–9||Prospective cohort||Retinoblastoma||Medical charts/Parent interview|
|Thulesius20||2000||Sweden||Population||68||0–16||Descriptive||All cancers||Medical charts|
|Halperin17||2001||USA||Hospital-based||122||0–20||Retrospective cohort||Medulloblastoma||Medical charts|
|Butros10||2002||USA||Hospital-based||57||1.5 mo–9.69 mo||Retrospective cohort||Retinoblastoma||Medical charts|
|Mehta6||2002||Canada||Population||104||0–17||Retrospective cohort||Brain tumors||Brain tumor database|
|Dobrovoljac3||2002||Switzerland||Hospital-based||252||0–17||Retrospective cohort||Brain tumor||Medical charts|
|Fajardo-Gutierrez4||2002||Mexico||Hospital-based||4940||0–14||Retrospective cohort||All cancers||Medical charts|
|Haimi18||2004||Israel||Hospital-based||315||0–20||Retrospective cohort||All cancers excluding leukemia||Medical charts/Questionnaire|
|Goyal13||2004||England||Hospital-based||103||4–22||Retrospective cohort||Bone cancer||Medical charts|
|Rodrigues14||2004||Brazil||Hospital-based||327||2 mo–12 mo||Retrospective cohort||Retinoblastoma||Medical charts|
|Klein-Geltink5||2005||Canada||Population||2316||0–14||Retrospective cohort||All cancers||CCCSCP database|
|Wallach*11||2006||Switzerland||Population||139||0–20||Retrospective cohort||Retinoblastoma||Medical charts|
Figure 1 shows the cancer-care pathway milestones that we used to define diagnosis delay and its components. Along this continuum, events may occur that can negatively interfere with cancer care. Studies included in this review have focused on the time between a patient's first symptom recognition to a diagnosis of cancer. This time period, called diagnosis delay in Figure 1, has also been designated as prediagnosis symptomatic interval,3 symptom duration and/or interval,11, 13, 15–17 time to diagnosis,4, 6 lag time,7, 14, 18–20 or wait time5 by different authors. Some studies made a distinction between patient and physician delays. The former was defined as the length of time between the onset of signs and symptoms and the patient's first visit to the healthcare system, whereas the length of delay between the first healthcare visit and the diagnosis was designated physician delay.
Diagnosis Delays in Childhood Cancer
Table 2 shows the total mean or median diagnosis delay reported in all studies. Mean delay times varied by cancer type from a low of 2.5 weeks (nephroblastoma) to a high of 29.3 weeks (brain tumor). Haimi et al. reported a mean diagnosis delay of 15.8 weeks with a range of 0 to 208 weeks for all cancers in Israel.18 In the case of brain tumors, a Swedish study by Thulesius et al.20 found a median delay of 9 weeks with a range of 1 to 199 weeks, whereas in a study in eastern Canada, Mehta et al.6 found a mean delay of more than 7 months with a 95% confidence interval of 5 to 10 months.
|Cancer type||Mehta6 Canada||Pollock7 USA*||Halperin16 USA||Pratt8 USA*||Flores25 USA||Butros10 USA*||Saha19 England*||Edgeworth24 England||Goyal13 England*||Goddard23 England*||Dobrovoljac3 Switzerland*||Wallach11 Switzerland†||Gjerris15 Denmark*||Thulesius20 Sweden*||Haimi18 Israel||Fajardo-Gutierrez4 Mexico*||Rodrigues14 Brazil|
|Head and neck rhabdomyosarcoma||NA||NA||NA||8||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA|
Ten studies separated the total average diagnosis delay into its components, as described above3, 5, 9, 10, 12, 18, 21–24 (Table 3). Generally, the tendency was for physician delays to be longer than patient delays. Variations were observed in patient delay (range, 2–12.8 weeks) and physician delay (range, 2–15 weeks) for studies that investigated retinoblastoma.
|Patient delay||Physician delay‡||Patient delay||Physician delay||Patient delay||Physician delay||Patient delay||Physician delay||Patient delay||Physician delay||Patient delay||Physician delay||Patient delay||Physician delay‡||Patient delay§||Physician delay||Patient delay||Physician delay‡||Patient delay||Physician delay‡|
|Soft tissue tumor||1.7||1||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA|
|Germ cell tumor||0.6||1||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA||NA|
Patient Factors Associated with Diagnosis Delay
Eleven studies investigated the relation between patient and parental factors and diagnosis delays (Table 4). A positive association between the patient's age at diagnosis and diagnosis delay was observed in 7 of 11 studies. Most studies supported the hypothesis that older patients are at higher risk of delayed diagnosis than younger patients. In Mexico, Fajardo-Gutierrez et al. found that risk of increased diagnosis delay for children between the ages of 10 years and 14 years is 1.8 times that of infants younger than 1 year of age (odds ratio [OR] for 10–14 years of age: 1.8; 95% confidence interval [CI], 1.4–2.3).4 It was also reported that diagnosis delay was shortest for children aged 0–2 years despite no significant differences in histopathology, grade or location of tumors, or parental persistence (number of consultations before diagnosis) across age groups.22 Dobrovoljac et al. analyzed patient and physician delays separately and found a positive correlation between age and patient delay, but not physician delay.3 The negative correlation with age observed by Klein-Geltink5 contrasts with the above findings.
|Patient factor||Study||Statistical analysis||Summary of findings||Conclusions|
|Age||Mehta6||Chi-square||P = .8||No significant difference|
|Klein-Geltink5||Logistic regression||Multiple results||Negative correlation; reduced risk of patient and physician delay for older patients|
|Pollock7||Pearson correlation||P < .001||Positive correlation; diagnosis delay longer for older children|
|Halperin16||Kruskal-Wallis||P = .11||No significant difference|
|Saha19||F||P < .001||Significant difference; diagnosis delay longer for older children|
|Dobrovoljac3||Pearson correlation||r = 0.32; P < .0001||Positive correlation; diagnosis delay longer for older children|
|Haimi18||ANOVA||P < .01||Positive correlation; diagnosis delay longer for older children|
|Goyal13||Mann-Whitney U||P = .05||Significant difference; diagnosis delay longer for older children; patient delay longer only for Ewing sarcoma patients|
|Goddard23||Kruskal-Wallis||Result not shown||No significant difference in patient delay; significant difference in physician delay; physician delay inversely related to patient age|
|Fajardo-Gutierrez4||Logistic regression||OR for age < 1 y vs 10–14 y, 1.8; 95% CI, 1.4–2.3||Positive correlation; older children at higher risk of diagnosis delay|
|Rodrigues14||Mann-Whitney U||P = .001||Positive correlation; diagnosis delay longer for children >24 mo of age than children < 24 mo|
|Sex||Mehta6||Chi-square||P = .131||No significant difference|
|Pollock7||Pooled Student t||P = .18||No significant difference overall; longer delays for girls than boys for non-Hodgkin lymphoma only; longer delays for boys than girls for Ewing sarcoma|
|Halperin16||Kruskal-Wallis||P = .08||No significant difference|
|Saha19||2 × 2 Table||1.2 (0.9–1.6)||No association|
|Dobrovoljac3||Wilcoxon rank sum||Result not shown||No significant difference|
|Haimi18||Student t||Result not shown||No significant difference|
|Rodrigues14||Mann-Whitney U||P = .949||No significant difference|
|Fajardo-Gutierrez4||Logistic regression||OR, 1.1; 95% CI, 1–1.3||Small increased risk of delays for boys|
|Gjerris15||Chi-square||Result not shown||No significant difference|
|Race||Pollock7||Pooled Student t||P = 0.23||No significant difference; longer delay for white children than nonwhite children for osteosarcoma only|
|Halperin16||Kruskal-Wallis||P = .89||No significant difference|
|Rodrigues14||Mann-Whitney U||P = .5333||No significant difference|
|Haimi18||Wilcoxon rank sum||P < .05||Shorter delays for children of Israeli, Ashkenazi, or Arabic origin than children of Sephardic fathers|
|Parental age||Haimi18||F||P < .01||Shorter delays for children of younger parents than for children of older parents|
|Parental education||Fajardo-Gutierrez4||Logistic regression||OR (fathers), 1.4; 95% CI, 1.1–1.8||Higher risk of long delays for children of parents with 0 to 5 y of education than for children of parents with >12 y of education|
|OR (mothers), 1.5; 95% CI, 1.2–2.1|
|Chantada22||Logistic regression||OR, 6.34; 95% CI, 3.7–29.3||Greater risk of longer patient delay with parental elementary education or lower|
Statistically significant differences in diagnosis delay between male and female patients was observed in only 2 of 9 studies. Fajardo-Gutierrez et al. reported a slight increase in risk of diagnosis delay in male patients (OR = 1.1; 95% CI, 1–1.3).4 In an analysis of sex and delay across diagnostic groups, Pollock et al. found that female patients had significantly longer diagnosis delay only for non-Hodgkin lymphoma.7 This association remained even after adjusting for age and date of diagnosis. In multivariate regression analyses, they also found that male patients with Ewing sarcoma had a significantly longer delay in diagnosis than female patients with the same disease. Of the 4 studies that examined the effect of ethnicity on diagnosis delay, 2 studies did not find any significant difference in delay time among patients of different ethnicities.14, 17 Pollock et al. reported a significant association between ethnicity and diagnosis delay only for osteosarcoma, for which white children had longer delay times.7 Similarly, a study on all cancers examined differences in lag time of children of different racial origins (based on the father) and found that Arabic children had shorter lag time than Jewish children.18
Because children are usually under the care of their parents, parental characteristics and behavior are also important factors in recognizing symptoms and signs of cancer. In Argentina, retinoblastoma patients whose parents had an elementary education or lower had a greater risk of longer patient delay (OR = 6.34).22 This was also observed in Mexico, where children whose parents had the lowest level of education had longer delays in diagnosis than children with parents with the highest level of education (OR = 1.4 for fathers and 1.5 for mothers).4 Haimi et al. looked at various parental factors and found that the parents' age, mother's profession, father's ethnicity, and religion were all significantly associated with diagnosis delay.18 In the case of parental age, children of younger parents had significantly shorter delay times than children of older parents. Although no relation between father's profession and delay was found, diagnosis delays were shorter for housewives or mothers with academic professions than for mothers in a “blue collar” profession. However, when parent-related and child-related factors were included in a multivariate analysis, religion and father's ethnicity were the only parental factors to remain significantly associated with increased diagnosis delay.
Cancer-related Factors Associated with Delays
The timely diagnosis of cancer in children is made difficult because of the rarity of the disease and the nonspecific presentation of symptoms. All studies found that cancer type was an important factor related to diagnosis delay (Table 5). Statistically significant differences were observed in the risk of a delayed time to diagnosis when different groups of cancers were compared with leukemia. Fajardo-Gutierrez et al. reported that renal tumors had a 60% increase in risk compared with leukemia, whereas the risk of delay in diagnosis for Hodgkin disease was 7 times that of leukemia.4 The effect of cancer type on delay remained even after accounting for effects of other covariates, such as age, sex, and race. Likewise, Flores et al. found that patients with brain tumors had significantly longer lag time than patients with either Wilms tumor or acute leukemia.25
|Cancer factor||Study||Statistical analysis||Results||Conclusions|
|Cancer type||Klein-Geltink5||Logistic regression||Multiple results||Cancer type influenced risk of patient and physician delay in diagnosis. Compared with children diagnosed with leukemia, children with bone tumors had the highest risk of increased patient delay, whereas children with CNS tumors had the highest risk of increased physician delay.|
|Pollock7||ANOVA||P = .0001||Significant difference in diagnosis delay among tumor types; difference remained after adjustment for age, sex, and race.|
|Flores25||Chi-Square, test not specified||P < .0001 at 4 wk||Brain tumors had significant delay compared with leukemia and Wilms tumor.|
|Saha19||F||P < .001||Significant difference in diagnosis delay among cancer type when compared with leukemia|
|Thulesius20||Chi-Square||P = .002||Diagnosis delay, patient delay, and physician delay were significantly shorter for leukemia patients than for brain tumor patients.|
|Haimi18||F||Result not shown||Significant difference; shortest delay, Wilms tumor (median, 2.5 wk); longest delay, epithelial tumors (median, 13 wk)|
|Fajardo-Gutierrez4||Logistic regression||Multiple results||Compared with leukemia, other cancer types have a higher risk of longer delay (eg, lowest OR renal tumor, 1.6; 95% CI, 1.2–2.1; highest OR Hodgkin disease, 7; 95% CI, 5.3–9.3).|
|Symptoms||Pollock7||Student t||Multiple results||Depending on cancer types, children with shorter lag times are significantly more likely to present certain symptoms. (eg, For neuroblastoma, abdominal masses were more common in patients with shorter diagnosis delay.)|
|Dobrovoljac3||Wilcoxon rank sum||P = .007||Patients with symptoms of increased intracranial pressure had a shorter diagnosis delay and physician delay than other patients.|
|Goddard23||Kruskal-Wallis||Result not shown||No significant difference in patient delay; significant difference in physician delay|
|Haimi18||Kruskal-Wallis||Result not shown||Significant longer overall delay and physician delay when presenting symptom is pain; shorter parent delay when symptom is rare.|
|Rodrigues14||Mann-Whitney U||P = .014||Patients with strabismus had significantly longer diagnosis delay than patients with other symptoms (leukocoria or tumor).|
|Tumor histology||Mehta6||Chi-square||P = .006||Significant difference in diagnosis delay between medulloblastoma vs nonmedulloblastoma|
|Dobrovoljac3||Wilcoxon rank sum||Result not shown||Negative correlation; fast-growing tumors have shorter delays than slow growing tumors.|
|Tumor location||Mehta6||Chi-square||P = .014||Significant difference in delay between tumors located in the brainstem vs nonbrainstem|
|Dobrovoljac3||Kruskal-Wallis||Result not shown||No significant difference|
|Haimi18||Kruskal-Wallis||P < .01||Significant difference in delay according to site of primary tumor|
|Goyal13||Linear regression||P = .002||Physician delay was longer for patients with axial site tumors than limb tumors.|
|Gjerris15||Chi-square||P < .01||Significant difference in delay between supratentorial and infratentorial tumors|
|WBC count||Saha19||2 × 2 Table||OR, 1.1; 95% CI, 0.6–1.4||No association|
|Cancer stage||Halperin16||Student t-||P = .01||Patients with early cancer stage had significantly longer diagnosis delays than patients with advanced cancer stage.|
|Saha19||F||P = .23||No significant difference|
|Chantada22||Logistic regression||Result not shown||Patients with disease stage II, III, and IV had significantly longer patient delay than patients with disease stage I.|
|Gjerris15||Chi-square||P < .05||Patients with a grade 1 malignancy had significantly longer delay than patients with grade 4 malignancy.|
|Wallach11||Logistic regression||OR, 8.09; 95% CI, 1.86–35.23||Patients with >6 mo diagnosis delay had a higher risk of being diagnosed with a higher stage.|
|Rodrigues14||Mann-Whitney U||P < .001||Patients with localized disease had a significantly shorter diagnosis delay than patients with advanced or metastatic disease.|
Studies have observed that the initial presentation of symptoms is related to diagnosis delay.3, 18, 20 The rarity and nonspecific clinical presentation of symptoms influenced parent delay in seeking medical advice and physician delay in reaching diagnosis.18, 20 The effect of symptoms on parent delay was inconsistent. A study on all cancers found significantly shorter parent delay when presenting symptoms were rare compared with common symptoms.18 However, a study on brain tumors noted that the presentation of symptoms affected provider delay, but not parent delay.3
Four of 5 studies found that the anatomic site of cancer influences delay in diagnosis. Children with rhabdomyosarcoma located in the pharynx and orbit had a shorter lag time than children who had it in the face or neck.8 Flores reported that children with infratentorial brain tumors have shorter lag time than children with supratentorial tumors.25 Similarly, when differentiating cases into brainstem and nonbrainstem tumors, Mehta et al. found a significant difference in lag time for children diagnosed with brain tumors.6 Only the study by Haimi et al. did not find a correlation between delay and tumor location.18 There is also evidence that tumor histology was correlated with delay. One study reported that aggressive fast-growing tumors had shorter delays than slowly growing tumors.3 However, Mehta et al. found an association between only histological type and diagnosis delay by comparing medulloblastoma versus nonmedulloblastoma tumors.6
Disease stage at diagnosis is an important factor to consider because it is a possible indicator of chronology of disease progression and a determinant of the constellation of signs and symptoms. Halperin and Friedman found that medulloblastoma patients with advanced stage exhibited shorter lag times compared with early stage disease patients.16 This may suggest that patients with shorter diagnosis delays might ultimately have worse prognoses. However, in a study of 64 children with a solid tumor, the difference in lag time between cancer stages was not significant.19 The effect of white blood cell count on delay was also assessed in 65 children with leukemia and was, again, found not to be significant (OR = 1.1; 95% CI, 0.6–1.4).
Healthcare-related Factors Associated with Diagnosis Delays
Table 6 shows the influence of parameters related to the healthcare system. Three studies found that timely intervention by the appropriate specialist may reduce delays. Diagnosis and physician delay was shorter for patients that visited the pediatrician than for patients whose first health contact was a family physician or other specialist.18 Conversely, the risk of patient delay was lower for patients who first contacted the general practitioner, and the risk of physician delay was lower for patients who first contacted the emergency room than patients whose first contact was the pediatrician.5
|Healthcare factor||Study||Statistical analysis||Results||Conclusions|
|Distance||Klein-Geltink5||Logistic regression||Multiple results||No association for patient or physician delay.|
|Fajardo-Gutierrez4||Logistic regression||OR, 1.5; 95% CI, 1.4–1.8||Children who lived near Mexico City had a lower risk of lag time than children who lived far from Mexico City.|
|Chantada22||Logistic regression||Result not shown||No significant difference in patient delay between patients living in urban and rural or between those living in Buenos Aires and the rest of the country.|
|No. of visits||Haimi18||Pearson correlation||Result not shown||Positive correlation between the no. of times a child visits a physician and lag time|
|First health professional contacted||Klein-Geltink5||Logistic regression||Patient delay||Compared with a first pediatrician contact, patients who first visited a general practitioner had a lower risk of patient delay. Patients who first visited the emergency room had a lower risk of physician delay.|
|OR for GP, 0.63; 95% CI, 0.42–0.95;|
|OR for ER, 0.31; 95% CI, 0.20–0.48|
|Chantada22||Logistic regression||Result not shown||No significant difference in patient delay|
|Goddard23||Kruskal-Wallis||Result not shown||No significant difference in patient delay; significant difference in physician delay|
|Goyal13||Linear regression||P = .02||Diagnosis delay and patient delay longer for patients who initially contacted a general practitioner than for patients who contacted the emergency room.|
|Haimi18||Kruskal-Wallis||Result not shown||Diagnosis and physician delay shorter for patients examined by pediatricians compared with family physicians or other specialists|
The relation between diagnosis delays and access to healthcare services showed mixed results. Two studies found that geographical distance was positively associated with diagnosis delays.4, 18 Fajardo-Guitierrez et al. defined distance as “near or far from Mexico City” and found a 1.5 greater risk of delay for people who live farthest from Mexico City.4 However, in a more detailed measure of distance, after adjusting for demographic and disease characteristics, Klein-Geltink et al. did not find that distance to the treating pediatric oncology center affected diagnosis delay in Canada.5
Early diagnosis of cancer is a fundamental goal in oncology because it allows an opportunity for timely treatment, while disease burden is still in its earliest stages. Consequently, prognosis may improve, and a cure can be attained with minimal side effects or late effects. Previous studies have shown that distributions of diagnosis delay were generally wide and skewed toward low values. Other than brain tumors and retinoblastoma, there was little difference in diagnosis delay among studies across each cancer type. However, given the small number of studies available for various cancer categories, no conclusion can be made about this observed consistency of delay across studies or countries. Differences in healthcare systems may account for variation in delay observed in brain tumors and retinoblastoma. For example, the long delay observed by Mehta et al.6 suggests that a single-payer healthcare system, such as that in Canada, might have an over-reliance on family physicians for triaging health complaints, relative to the situation in the US, where insured patients may access specialists directly. This may conceivably introduce an additional provider layer between the initial patient visit and the final cancer diagnosis, a scenario that could potentially create longer diagnosis delays. A related concern has been expressed for England,26 a country in which general practitioners provide the first opportunity for diagnosis.
Most studies in this review have been retrospective cohort studies and are, therefore, subject to certain limitations inherent in their design. The use of pre-existing records makes it difficult to ascertain the reliability of information collected and to obtain information on potential confounding variables. Possible biases may occur if disease status affected patient selection into the study or the collection of exposure information. Parents and patients might have recalled certain events differently when the child was diagnosed with a more severe tumor. However, any diagnosis of cancer would be cause for great worry to all parents, and so there would be little reason to believe that the severity of the disease would lead to differential collection of information. It is also doubtful that patients or physicians would know of any study hypothesis on diagnosis delays when data were collected. Therefore, if present, then these biases were likely nondifferential.
Misinterpretation of ambiguous cancer symptoms by patients, parents, and physicians may lead to diagnosis delay. The relation between symptoms and diagnosis delay is confounded by the association that symptoms have with other factors. Symptom patterns vary by diagnostic groups, and this may partly account for differences observed in diagnosis delay among cancer groups (eg, brain tumors have a slower tumor growth rate than other cancers and, therefore, would have a slower symptom progression). The effect of symptoms on diagnosis delay may also be mediated by the age of the patient. Younger children may experience cancers with more identifiable signs at onset (such as an abdominal mass in Wilms tumor) than older children. However, a positive association between age and delay was still present even after studies controlled for the type of cancer.7, 19, 24 This would suggest that the effect of age on delay may be due to more than differences in tumor-specific characteristics. It is possible that the tendency for providers to screen for tumors in children may be different across age groups. Older children may be more self-reliant concerning their health status than younger children; younger patients may have a closer relationship to their parents than older patients. The influence of increased parent knowledge and awareness of the child's disease on timely diagnosis is supported by the finding of a negative association between parental education and diagnosis delay.4 However, this relation may also be mediated by socioeconomic status or access to healthcare.
Once patients enter the healthcare system, diagnosis delay may be influenced by access to medical care services, knowledge and recognition of the disease by health providers, and availability of appropriate diagnostic capability and instrumentation. It is difficult to interpret physicians' ability to diagnose cancer. It has been suggested that increased vigilance and awareness of cancer on the part of the general practitioner may decrease delay times.3, 18 However, the severity of disease and symptoms on presentation at the physicians' office likely influences this relation. Pediatric oncologists and emergency room physicians will probably see urgent cases in which symptoms are more apparent. Moreover, the added time required for evaluation by the general practitioner, followed by further visits to the pediatric oncologists before making a cancer diagnosis, will add another time segment and, thus, may increase diagnosis delay.
The relation between diagnosis delays and disease stage is complex. Although staging of cancer in children is different from that of adults, it is still a general classification of the extent of disease at presentation, much of which reflects the chronology of disease progression. Because disease stage is determined at diagnosis, studies have been able to examine this relation only cross-sectionally. A key question remains: Do delays in diagnosis worsen the extent of the disease, or does the extent of the disease influence diagnosis delays? Common sense would indicate that longer delays would lead patients to be diagnosed at a more advanced disease stage. However, it can also be concluded that more aggressive, fast-growing tumors may show rapid progression of symptoms even at early stages, which would lead parents to seek medical attention for their child. Alternatively, less aggressive, slow-growing tumors may lead to longer delays due to the incipient nature of disease symptoms, which parents may not notice as quickly. Therefore, the aggressiveness of disease and the attendant severity of symptoms likely play a role in the relation between diagnosis delay and cancer stage. The actual period of disease development is a concern of studies on cancer prognosis. For example, a child may be developing a tumor for some time before seeking a diagnosis. A more advanced disease would lead to a rapid diagnosis, but this short diagnosis delay may not lead to better cancer prognosis.
The impact of diagnosis delays on the prognosis of children with cancer is still unknown. It has generally been believed that long delays would lead to worse prognosis. In the context of childhood cancer, however, few studies have been specifically designed to investigate this assumption, or they have not conducted thorough analyses. Seven studies explored this relation and found conflicting results. Four studies reported negative associations between delays and survival,8, 12, 14, 18 whereas the remaining studies did not.10, 13, 19 DerKinderen et al. found that retinoblastoma patients with a physician delay ≥1 week had a significantly higher death rate (OR = 5.1) than patients with a shorter delay.12 Rodrigues et al. also found that retinoblastoma patients with a delay of < 6 months had a statistically significant higher 5-year survival rate than patients with longer delay.14 Conversely, Goyal et al. did not find any significant differences in event-free survival between patients with < 3 months, 3 months to 6 months, and >6 months diagnosis delay.13 Similarly, Saha et al. did not find that delay times were predictive of event-free survival even after adjusting for possible confounding variables.19 None of the studies examined the possible mediating effect of disease severity at diagnosis on the relation between delay and survival. Research on adult cancers supports the hypothesis that delays in diagnosis adversely affect prognosis.27–29 Conversely, some studies found that longer delays were associated with increased survival.30, 31 It has been reported that the impact of delay on survival is likely mediated through the relation between delay and cancer stage.27 It can be concluded that more aggressive, fast-growing tumors may show rapid progression of symptoms. This would lead patients to present themselves promptly to a medical professional and have their cancer diagnosed and treated quickly. Alternatively, less aggressive, slow-growing tumors may lead to longer delays due to the early nature of disease symptoms. Short delays would, thus, be associated with the poor prognosis intrinsic to aggressive tumors.
Research on diagnosis delays in childhood cancer is still in its early stages. More studies are needed to investigate the potential impact of delays on prognosis outcomes. Information on factors that influence delays independently of each other and the individual impact of patient and provider delays on disease severity and prognosis would be useful to form effective policies and programs aimed at eliminating obstacles in the cancer-care pathway for children with cancer.
- 1IARC Working Group. Epidemiology of Childhood Cancer. IARC Scientific Publications No. 149. Lyon: International Agency for Research on Cancer; 1999.