Cancer survival in a southern African urban population



This paper provides the first comprehensive population based cancer survival estimates from the African continent. Five-year absolute and relative survival estimates are presented for black and white Zimbabwean patients diagnosed with cancer in Harare, Zimbabwe between the years 1993 and 1997. The survival of black Zimbabwean cancer patients are among the lowest ever reported from population based cancer registries. For most cancer sites, white Zimbabwean patients have much higher survival than black Zimbabweans, except for lung and colorectal cancer, for which the estimates are similarly poor. Race specific comparisons to cancer patients in the United States show that Zimbabwean patients have much lower survival than American cancer patients and that the gap between black Zimbabwean patients and black American patients is broader than between white Zimbabwean and white American patients. Access to and the ability to pay for medical care may be a very important barrier to better survival for the majority of black Zimbabwean patients and the most important cause for the very low cancer survival in this population. © 2004 Wiley-Liss, Inc.

Comprehensive population-based cancer survival estimates from southern Africa were not available until now. Several years ago, a project was started by the International Agency for Research on Cancer (IARC, Lyon) to assess cancer occurrence and management in developing countries. Cooperation with many cancer registries in developing countries was initiated and the first population-based cancer survival estimates from various developing countries were published in 1998.1 However, no data from Africa was available at that time. This paper provides the first population-based estimates of cancer patient survival from an urban environment in southern Africa. Five-year absolute and relative survival estimates are presented for the most common cancers of the black and white populations of Harare, the capital of Zimbabwe. We compared these with those of the black and white American populations included in the Surveillance, Epidemiology and End Results (SEER) program in the United States.2

During the last decade, the burden of cancer in the black Zimbabwean population has risen dramatically in Harare, and there has been a major shift in the cancer spectrum, mainly due to the AIDS epidemic. Around one third of the registered black cancer patients suffer from AIDS associated Kaposi sarcoma, the cancer with the highest incidence in Harare today.3 A considerable rise in the incidence of eye cancer, particularly the AIDS associated squamous cell carcinoma of the conjunctiva, was also recorded. Cancer incidence among the white population and that of non-AIDS associated cancers has remained largely unchanged.4



Our analysis is based on data collected by the Zimbabwe National Cancer Registry (ZNCR) in Harare, Zimbabwe. The city is a large, fast growing African metropolis, with an estimated population of 1.5 million in 1997, comprising 95% black, 3% white and 2% Asian or mixed ethnicity. The ZNCR was established in 1985 and achieved adequate population coverage for the population of Harare by 1990.5 History, methods, and quality and completeness of cancer registration have been described in detail elsewhere.3, 5, 6 Briefly, the registry performs active case finding, based on hospital in- and out-patient records, histology reports and death certificates.5 However, owing to the lack of complete and reliable routine vital statistics and limited resources, the follow-up of cases, done by active individual tracing of patients, is not carried out routinely but is restricted to special studies.

For this survival study, a subsample of Harare resident cases was selected from the patients diagnosed with cancer and reported to the Harare cancer registry. The sample includes randomly selected patients with one of the 15 most common cancers diagnosed between 1993–1997. For cervical cancer, only patients diagnosed during 1995–1997 were included. For each cancer site, the study intended to include 150 patients, except for cervix and breast cancer, for which 300 patients were planned. For breast cancer, 150 black and 150 white patients were to be included. For other cancer sites, races other than African (i.e., white, coloured and Asian) were only included by chance. Where the death occurred in Harare, the date of death was retrieved from the death certificates, an important source of information for the registry. For those cases not known to have died, an active follow-up was carried out. This involved the retrieval of patient records of the 2 government central hospitals (Harare and Parirenyatwa). If the records did not provide sufficient follow-up information, home visits were conducted to ascertain the vital status of the patients.

The follow-up was carried out from January 2001 to June 2002. The effective cut-off date for the follow-up of the study population was 31 December 1999, i.e., the patient's vital status was considered on that date. Epi Info was used for data entry and data sets were converted to the Statistical Software System (SAS) for analysis.

Statistical analysis

First, a careful check of the quality and completeness of the data was done. Records with missing or implausible data on the critical variables were excluded, as were records of patients reported by death certificate only (DCO) and records of patients who were lost to follow-up. The final number of patients included in the analysis was influenced by many factors: the number of patients in the registry data set (for some cancer sites, considerably fewer than 150 cases were registered during the recruitment period), DCO and lost to follow-up cases, missing data on critical variables and racial composition of the patients with a specific cancer. As a result, for several cancer sites, the number of patients included in the analysis was lower than 150. For black Zimbabweans, only cancer sites with more then 35 cases were considered for the analysis. This limit was lowered to 20 for white Zimbabweans, due to the smaller overall number of these patients in the sample. Other races (Asian and Coloured) were excluded. Kaposi sarcoma cases were analysed together regardless of the site of disease.

For each cancer site, absolute and relative survival were calculated for the 2 groups of the study population. The relative survival was calculated as the ratio of observed (absolute) survival and expected survival.7 The latter were determined using the age- and sex-specific survival estimate values for the general population from the WHO life-tables, which, in the case of Zimbabwe, are based on model calculations as complete vital statistics are not available for the country.8 The survival table represents the survival experience of the entire Zimbabwean population and takes the AIDS epidemic into account. The calculations of relative survival were done according to Hakulinen's method, using the SAS macro periodh.9, 10 Standard errors of survival proportions were calculated according to Greenwood's method.11

Absolute and relative 5-year race-specific survival of Zimbabwean patients were compared with those of patients diagnosed in the same time interval in the United States. For the American patients, data from the cancer registries included in the Surveillance, Epidemiology and End Result (SEER) program were used (only patients with a first primary tumour were included).2 Owing to the large age differences between the Zimbabwean and the SEER cancer populations, the survival estimates of the latter were adjusted to the age distributions of the former using 3 broad age groups (<30, 30–59 and ≥60). The adjustment of relative survival estimates was done as described by Brenner and Hakulinen.12


Study subjects

Table I describes the populations of black and white Zimbabwean cancer patients that were included in this analysis. For most forms of cancer, more than 80% of the patients could be included. The percentage of cases lost to follow-up ranged from 0 to 14% and from 0 to 10% among black and white cancer patients, respectively. Proportions of DCO cases were between 0 and 10% among black Zimbabweans. For most cancer sites, similar proportions of DCO cases were observed for black and white cancer patients. However, the DCO rates were quite high for white patients with lung cancer and colorectal cancer.

Table I. Zimbabwean Cancer Patient Population
Cancer siteNumber of patients selected for survival studyExcluded from analysisIncluded in the analysis
Lost to follow-upDCOMissing data on critical variablesExcluded overall %NMedian ageHistologically verified (%)
a) Description of the black Zimbabwean cancer patient population included in the survival study, Harare, Zimbabwe, 1993–1997
 Colon and rectum7894117.9645473.4
 Skin melanoma4950316.3415682.9
 Kaposi sarcoma1511011.31492669.8
b) Description of the white Zimbabwean cancer patient population included in the survival study, Harare, Zimbabwe, 1993–1997
 Colon and rectum632112244867.587.5
 Skin melanoma553017515098.0

Black cancer patients were on average younger than white patients for all cancer sites. Histological verification ranged from 10.9% (liver cancer) to 92.4% (lymphomas) among black and from 62.5% (lung cancer) to 100% (bladder cancer) among white patients. The proportions of cases with histological verification among black Zimbabweans compare well to those of other developing countries,13, 14, 15, 16 but they were lower than the corresponding proportions for white Zimbabwean patients.

Table II compares the median ages at diagnosis for the sample and the comparison populations. For most cancer sites, the median age was substantially higher among black American patients from the United States than among black Zimbabwean patients, whereas median ages were quite similar for most cancer sites among white patients from Zimbabwe and the United States. The differences in median age among black patients with cancers of the eye are explained by the different nature of the diseases most common among Zimbabwean and American patients (HIV/AIDS associated squamous cell carcinoma of the conjunctiva and retinoblastoma, respectively).

Table II. Comparison of Patients' Median Age at Diagnosis, Zimbabwean Cancer Patient Populations and Seer Cancer Patient Populations, 1993–1997
Cancer siteZimbabwe (Harare)USA (SEER)
Black patientsWhite patientsBlack AmericansWhite Americans
Skin melanoma56506457
Kaposi sarcoma263739

Survival analysis

In Tables III and IV, the 5-year absolute and relative survival of Zimbabwean cancer patients are presented and compared with those of black and white cancer patients from the United States. Black Zimbabwean patients with cancer in the gastrointestinal tract and lung cancer had very low survival, as well as patients with Kaposi sarcoma, which among these young patients is certainly the result of HIV infection. Patients with cancers of the eye and skin melanoma had the highest survival (absolute and relative survival between 40 and 55%). For none of the cancers included into this analysis was the relative survival above 55%. The survival of white Zimbabwean patients was substantially higher than that of black Zimbabwean patients, except for colorectal and lung cancer, for which very poor prognosis was found even among white patients. The highest relative survival, 97.7%, was found for white patients with skin melanoma. We do not provide the survival comparison for patients with cancers of the eye, for the already mentioned different nature of disease would make the comparison between the 2 populations misleading.

Table III. Absolute Survival (in %) of Black and White Patients with Cancer, Harare, Zimbabwe and of Black and White American Patients with Cancer in the United States, 1993–1997
Cancer siteZimbabwe (Harare)USA (SEER)
Black patientsWhite patientsBlack Americans, adjusted survival1White Americans, adjusted survival2
5 years absolute survivalSE5 years absolute survivalSE5 years absolute survivalSE5 years absolute survivalSE
  • 1

    Adjusted to age distribution of black Zimbabwean patients.

  • 2

    Adjusted to age distribution of white Zimbabwean patients.

  • 3

    3 years survival.

Skin melanoma41.48.380.
Kaposi sarcoma3.91.714.64.3
Table IV. Relative Survival (in %) of Black and White Patients with Cancer, Harare, Zimbabwe and of Black and White American Patients with Cancer in the United States, 1993–1997
Cancer siteZimbabwe (Harare)USA (SEER)
Black patientsWhite patientsBlack Americans, adjusted survival1White Americans, adjusted survival2
5 yrs relative survivalSE5 yrs relative survivalSE5 yrs relative survivalSE5 yrs relative survivalSE
  • 1

    Adjusted to age distribution of Zimbabwean African patients.

  • 2

    Adjusted to age distribution of Zimbabwean European patients.

  • 3

    3 years survival.

Skin melanoma49.910.
Kaposi sarcoma4.41.814.94.4

Comparing the survival of black Zimbabwean with black American patients, the absolute and the relative survival of American patients were substantially higher for nearly all cancer sites. Only for patients with oesophagus, liver and lung cancer and Kaposi sarcoma, who have very low survival even in the United States, were the differences less pronounced. Among black Zimbabwean patients, the survival of patients with larynx and prostate cancer is strikingly lower than that of black American patients. The differences seen in the survival of patients with lymphomas may partly be explained by a proportional difference in Hodgkin and Non-Hodgkin lymphoma patients in the 2 patient groups. Unfortunately, stratification was not possible due to the many unspecified lymphomas among Zimbabwean patients. For white Zimbabwean patients with colorectal cancer, survival was very low in comparison to white American patients (relative survival 19% vs. 61%). For the other cancer sites, the differences in absolute and relative survival were somewhat less pronounced than those between black patients in Zimbabwe and the United States, but they were still rather substantial, except for skin melanoma.


In this pioneering study on cancer patient survival from a population-based cancer registry from southern Africa, we found that cancer survival among black Zimbabweans was low, with no survival estimate exceeding 55%. The 5-year survival estimates for black Zimbabwean cancer patients are in fact among the lowest ever values reported from a population based cancer registry. White Zimbabweans experienced a much higher survival than black Zimbabwean cancer patients. Nevertheless, both groups of Zimbabwean cancer patients had much lower 5-year survival than cancer patients living in the United States. These patterns suggest that for most cancer sites differential access to effective medical care rather than ethnic differences account for most of the observed differences in survival.

The survival estimates for black Zimbabweans, as the comparison in Table V shows, are lower than or at the low end of the corresponding survival range reported in a recent study from other developing countries by the International Agency for Research on Cancer.18 Compared with other countries, survival is particularly low for patients with cancer of the colon and rectum, larynx, breast, cervix, prostate and the urinary bladder, and for lymphomas, suggesting, for most cancer patients in Zimbabwe, a lack of access to effective measures of early detection and treatment available for those cancers in other countries. Although direct measures of access to and receipt of medical care were not available in this dataset, the much lower proportion of histologically verified cancers among black patients than among white patients may be an indirect indication of the particularly low access to diagnostic and therapeutic procedures among the black patients.

Table V. Comparison of Cancer Survival (in %) between Zimbabwe and Other Developing Countries
Cancer site5-year relative survival, ZimbabweOther developing countries1 range
Black patientsWhite patients
  • 1

    Range of age standardized relative survival for the cancer patient populations included into the IARC Cancer Survival in Developing Countries monograph17.

  • 2

    3 years survival for Zimbabwean patients.

Skin melanoma49.997.739.2–47.0

Another important consideration for the interpretation of our results is the effect of HIV/AIDS on cancer and cancer survival in Zimbabwe. The rise in incidence of several different types of cancers within the past 2 decades is clearly associated with the evolution of the AIDS epidemic.19 In particular, it has been shown that the explosion in the number of patients with Kaposi sarcoma and eye cancer, the 2 youngest groups of cancer patients in our study, are directly linked to the AIDS epidemic in Africa. With respect to lymphomas, the incidence of NHL in the black population of Harare was not high in 1990–1992,6 but the rates appear to have increased somewhat in the period 1993–1995, at least in adult females, possibly due to the maturation of the HIV epidemic.4 In Uganda, the most recent data from the cancer registry in Kampala suggest that the incidence of NHL increased significantly between 1991 and 1994 and 1995 and 1997, especially in children and young adults.20 It remains unclear as to what extent patients with non-AIDS associated cancers are independently affected by HIV/AIDS.

In the interpretation of our results, the following limitations have to be kept in mind. The fact that in our database the proportion of histologically verified cancers was only 68% implies the potential for misdiagnoses. Furthermore, a nonnegligible proportion of patients were lost to follow-up (6.6% overall) or notified by death certificate only (6.4% overall) and could therefore not be included in the survival analyses. If the prognosis among these patients strongly differed from the prognosis of patients who could be traced, the estimated survival could have become distorted upward or downward to some degree. Exclusion of patients notified by death certificate only can only increase estimates of survival, as all of these patients actually have died. If deceased patients might also have a higher chance of becoming lost to follow-up, even the very low estimates of survival reported in this study may be somewhat too optimistic rather than too pessimistic. Furthermore, the restriction of the database to patients diagnosed during 1993–1997 did not allow the application of the period survival analysis method, which could have provided more up-to-date survival estimates.21, 22

We also have to note that the Zimbabwean life table that we used leads to overestimated relative survival estimates for white Zimbabwean patients, as they experience a lower general mortality than black Zimbabweans.23 Unfortunately, a specific life table for the white Zimbabwean population was not available. We carried out a sensitivity analysis, in which we applied the life table of white Americans to calculate the relative survival of the white Zimbabwean patients. The resulting relative survival values were 1 and 3% units lower for lung and colorectal cancer, and around 10% units lower for the other cancers. We may consider these differences as the upper limit of potential overestimation of the relative survival of white Zimbabwean patients, on the reasonable condition that they do not have a lower general mortality than white Americans.

Finally, for some of the cancer sites, the numbers of cases included in the analyses were rather small, resulting in quite large standard errors (which were provided along with the point estimates in this paper), in particular among white Zimbabwean patients. In this group, proportions of DCO cases were also quite high among patients with colorectal and lung cancer, i.e., those cancers with rather poor prognosis. Nevertheless, the standard errors of the survival estimates were generally small compared with the very large differences in the point estimates between the survival of cancer patients in Zimbabwe and the United States.

It is recognised that at national level, optimal management of the cancer burden is best achieved by establishing and implementing a cancer control programme.24 This attempts to optimise all appropriate mechanisms of prevention, detection, diagnosis, treatment, rehabilitation and palliative care. Capacities for early diagnosis and treatment of established disease are essential in order to maximise the sufferers' chances of survival and to enhance their quality of life post-diagnosis. Information on survival is an important indicator of the success of such activities, and, in this context, it is the data of population-based cancer registries that are most relevant, reflecting several aspects of cancer control, including screening, diagnosis and the organisation of treatment services.25 In recent years, there has been increasing attention to comparative statistics of survival between different populations and time periods, with a view to identifying where improvements in existing management are possible.26

Population-based survival is measured on the totality of cancer patients, including those who receive no treatment whatsoever. Estimates of survival in such groups may be influenced by a range of prognostic and other factors, and the distribution of these determinants of survival must be taken into account in comparing between different populations or over time. Age and sex are important in this context, but stage of disease at diagnosis is generally the most important factor determining survival, and differences in the stage distributions of tumours in populations being compared are of particular concern. In our study, we have not compared stage distribution of cancers between the different ethnic groups in Zimbabwe or with that observed in the USA. It has however been observed that for many cancers, stage at diagnosis is advanced in the African situation.19 This would imply that much improvement could be achieved through better awareness of cancer (education programmes) to encourage earlier diagnosis of disease. But this would be ineffective unless adequate facilities were available for treatment. Although we have no data on survival within stage groups, it is quite likely that it is poor, given the lack of access to treatment facilities in sub Saharan Africa.27 Equitable distribution of resources is an important ethical concern in cancer treatment;24 it is clear that, in Zimbabwe at least, and probably in much of the continent, there is plenty of scope for improvement in this respect.


Our thanks are due to the staff of the Zimbabwe National Cancer Registry (R. Rukainga, M. Mutasa and F. Mandizvidza) who undertook the work of patient follow-up. We also thank Mr. R. Kumar (IARC) for his assistance in managing the data files. The survival analysis was carried out with the support of the German Research Foundation (DFG), within the framework of the Graduiertenkolleg 793. There is no conflict of interest.