The validity of cancer information on death certificates in Norway and the impact of death certificate initiated cases on cancer incidence and survival

BackgroundDeath certificates are an important source of information for cancer registries. The aim of this study was to validate the cancer information on death certificates, and to investigate the effect of including death certificate initiated (DCI) cases in the Cancer Registry of Norway when estimating cancer incidence and survival.MethodsAll deaths in Norway in the period 2011–2015 with cancer mentioned on the death certificates were linked to the cancer registry. Notifications not registered from other sources were labelled death certificate notifications (DCNs), and considered as either cancer or not, based on available information in the registry or from trace-back to another source.ResultsFrom the total of 65 091 cancers mentioned on death certificates in the period 2011–2015, 58,425 (89.8%) were already in the registry. Of the remaining 6 666 notifications, 2 636 (2 129 with cancer as underlying cause) were not regarded to be new cancers, which constitutes 4.0% of all cancers mentioned on death certificates and 39.5% of the DCNs. Inclusion of the DCI cases increased the incidence of all cancers combined by 2.6%, with largest differences for cancers with poorer prognosis and for older age groups. Without validation, including the 2 129 disregarded death certificates would over-estimate the incidence by 1.3%. Including DCI cases decreased the five-year relative survival estimate for all cancer sites combined with 0.5% points.ConclusionIn this study, almost 40% of the DCNs were regarded not to be a new cancer case, indicating unreliability of death certificate information for cancers that are not already registered from other sources. The majority of the DCNs where, however, registered as new cases that would have been missed without death certificates. Both including and excluding the DCI cases will potentially bias the survival estimates, but in different directions. This biases were shown to be small in the Cancer Registry of Norway.     

Impact of using multiple causes of death codes to compute site-specific, death certificate-based cancer mortality statistics in the United States

Background: Cancer mortality statistics, an important indicator for monitoring cancer burden, are traditionally restricted to instances when cancer is determined to be the underlying cause of death (UCD) based on information recorded on standard certificates of death. This study's objective was to determine the impact of using multiple causes of death codes to compute site-specific cancer mortality statistics. Methods: The state cancer registries of California, Colorado and Idaho provided linked cancer registry and death certificate data for individuals who died between 2002 and 2004, had at least one cancer listed on their death certificate and were diagnosed with cancer between 1993 and 2004. These linked data were used to calculate the site-specific proportion of cancers not selected as the UCD (non-UCD) among all cancer-related deaths (any mention on the death certificate). In addition, the retrospective concordance between the death certificate and the population-based cancer registry, measured as confirmations rates, was calculated for deaths with cancer as the UCD, as a non-UCD, and for any mention. Results: Overall, non-UCD deaths comprised 9.5 percent of total deaths; 11 of the 79 cancer sites had proportions greater than 3 standard deviations from 9.5 percent. The confirmation rates for UCD and for any mention did not differ significantly for any of the cancer sites. Conclusion and impact: The site-specific variation in proportions and rates suggests that for a few cancer sites, death rates might be computed for both UCD and any mention of the cancer site on the death certificate. Nevertheless, this study provides evidence that, in general, restricting to UCD deaths will not under report cancer mortality statistics.     

Reliability of recording uterine cancer in death certification in France and age-specific proportions of deaths from cervix and corpus uteri

French uterine cancer recordings in death certificates include 60% of "uterine cancer, Not Otherwise Specified (NOS)"; this hampers the estimation of mortalities from cervix and corpus uteri cancers. The aims of this work were to study the reliability of uterine cancer recordings in death certificates using a case matching with cancer registries and estimate age-specific proportions of deaths from cervix and corpus uteri cancers among all uterine cancer deaths by a statistical approach that uses incidence and survival data. Deaths from uterine cancer between 1989 and 2001 were extracted from the French National database of causes of death and case-to-case matched to women diagnosed with uterine cancer between 1989 and 1997 in 8 cancer registries. Registry data were considered as "gold-standard". Among the 1825 matched deaths, cancer registries recorded 830 cervix and 995 corpus uteri cancers. In death certificates, 5% and 40% of "true" cervix cancers were respectively coded "corpus" and "uterus, NOS" and 5% and 59% of "true" corpus cancers respectively coded "cervix" and "uterus, NOS". Miscoding cervix cancers was more frequent at advanced ages at death and in deaths at home or in small urban areas. Miscoding corpus cancers was more frequent in deaths at home or in small urban areas. From the statistical method, the estimated proportion of deaths from cervix cancer among all uterine cancer deaths was higher than 95% in women aged 30-40 years old but declined to 35% in women older than 70 years. The study clarifies the reason for poor encoding of uterus cancer mortality and refines the estimation of mortalities from cervix and corpus uteri cancers allowing future studies on the efficacy of cervical cancer screening.     

Causes of death in men with prostate cancer: Results from the Danish Prostate Cancer Registry (DAPROCAdata)

BackgroundCurrent knowledge of the validity of registry data on prostate cancer-specific death is limited. We aimed to determine the underlying cause of death among Danish men with prostate cancer, to estimate the level of misattribution of prostate cancer death, and to examine the risk of death from prostate cancer when accounting for competing risk of death.Material and methodsWe investigated a nationwide cohort of 15,878 prostate cancer patients diagnosed in 2010–2014; with 3343 deaths occurring through 2016. Blinded medical chart review was carried out for 670 deaths and compared to the national cause of death registry. Five death categories were defined: 1) prostate cancer-specific death, 2) other unspecified urological cancer death, 3) other cancer death 4) cardiovascular disease death, and 5) other causes of death. Competing risk analyses compared Cox cause-specific and Fine-Gray regression models.ResultsChart review attributed 51.2% of deaths to prostate cancer, 17.0% to cardiovascular disease, and 16.7% to other causes. The Danish Register of Causes of Death attributed 71.7% of deaths to prostate cancer when including all registered contributing causes of death, and 57.0% of deaths when including only the primary registered cause of death. The probability of death by prostate cancer was 10% at 2-year survival.ConclusionsMore than half of the deceased men in our study cohort died of their prostate cancer disease within a mean of 2.4 years of follow up. Data from the death registry is prone to misclassification, potentially overestimating the proportion of deaths from prostate cancer.     

Effects of the length of central cancer registry operations on identification of subsequent cancers and on survival estimates

BackgroundPopulation-based cancer survival analyses have traditionally been based on the first primary cancer. Recent studies have brought this practice into question, arguing that varying registry reference dates affect the ability to identify earlier cancers, resulting in selection bias. We used a theoretical approach to evaluate the extent to which the length of registry operations affects the classification of first versus subsequent cancers and consequently survival estimates.MethodsSequence number central was used to classify tumors from the New York State Cancer Registry, diagnosed 2001–2010, as either first primaries (value = 0 or 1) or subsequent primaries (≥2). A set of three sequence numbers, each based on an assumed reference year (1976, 1986 or 1996), was assigned to each tumor. Percent of subsequent cancers was evaluated by reference year, cancer site and age. 5-year relative survival estimates were compared under four different selection scenarios.ResultsThe percent of cancer cases classified as subsequent primaries was 15.3%, 14.3% and 11.2% for reference years 1976, 1986 and 1996, respectively; and varied by cancer site and age. When only the first primary was included, shorter registry operation time was associated with slightly lower 5-year survival estimates. When all primary cancers were included, survival estimates decreased, with the largest decreases seen for the earliest reference year.ConclusionsRegistry operation length affected the identification of subsequent cancers, but the overall effect of this misclassification on survival estimates was small. Survival estimates based on all primary cancers were slightly lower, but might be more comparable across registries.     

Estimating bias in causes of death ascertainment in the Finnish Randomized Study of Screening for Prostate Cancer

BackgroundPrecise cause of death (CoD) ascertainment is crucial in any cancer screening trial to avoid bias from misclassification due to excessive recording of diagnosed cancer as a CoD in death certificates instead of non-cancer disease that actually caused death. We estimated whether there was bias in CoD determination between screening (SA) and control arms (CA) in a population-based prostate cancer (PCa) screening trial.MethodsOur trial is the largest component of the European Randomized Study of Screening for Prostate Cancer with more than 80,000 men. Randomly selected deaths in men with PCa (N = 442/2568 cases, 17.2%) were reviewed by an independent CoD committee. Median follow-up was 16.8 years in both arms.ResultsOverdiagnosis of PCa was present in the SA as the risk ratio for PCa incidence was 1.19 (95% confidence interval (CI) 1.14–1.24). The hazard ratio (HR) for PCa mortality was 0.94 (95%CI 0.82–1.08) in favor of the SA. Agreement with official CoD registry was 94.6% (κ = 0.88) in the SA and 95.4% (κ = 0.91) in the CA. Altogether 14 PCa deaths were estimated as false-positive in both arms and exclusion of these resulted in HR 0.92 (95% CI 0.80–1.06).ConclusionsA small differential misclassification bias in ascertainment of CoD was present, most likely due to attribution bias (overdiagnosis in the SA). Maximum precision in CoD ascertainment can only be achieved with independent review of all deaths in the diseased population. However, this is cumbersome and expensive and may provide little benefit compared to random sampling.     

Problems in determining the incidence of cervical cancer.

Since cancer registries have different recording practices, the incidence rates that they report must be compared with caution. Indexes of reliability of recording indicated that in 1971 the reported incidence of cervical cancer in Ontario was too high. In 1971 Ontario used a method of passive reporting of cancer cases: the Ontario Cancer Registry linked hospital reports, death certificates and reports from the Ontario Cancer Treatment and Research Foundation''s treatment centres to produce a single record for each case. Pathological confirmation was requested for cases thus recorded by the registry. In 26% of cases a diagnosis other than cervical cancer was indicated. With these cases omitted, the incidence rate became 15.1/100 000, as opposed to the 20.5/100 000 reported by the registry.     

The impact of National Death Index linkages on population-based cancer survival rates in the United States

Background: In order to ensure accurate survival estimates, population-based cancer registries must ascertain all, or nearly all, patients diagnosed with cancer in their catchment area, and obtain complete follow-up information on all deaths that occurred among registered cancer patients. In the US, linkage with state death records may not be sufficient to ascertain all deaths. Since 1979, all state vital statistics offices have reported their death certificate information to the National Death Index (NDI). Objective: This study was designed to measure the impact of linkage with the NDI on population-based relative and cancer cause-specific survival rates in the US. Methods: Central cancer registry records for patients diagnosed 1993–1995 from California, Colorado, and Idaho were linked with death certificate information (deaths 1993–2004) from their individual state vital statistics offices and with the NDI. Two databases were created: one contained incident records with deceased patients linked only to state death records and the second database contained incident records with deceased patients linked to both state death records and the NDI. Survival estimates and 95% confidence intervals from each database were compared by state and primary site category. Results: At 60 months follow-up, 42.1–48.1% of incident records linked with state death records and an additional 0.7–3.4% of records linked with the NDI. Survival point estimates from the analysis without NDI were not contained within the corresponding 95% CIs from the NDI augmented analysis for all sites combined and colorectal, pancreas, lung and bronchus, breast, prostate, non-Hodgkin lymphoma, and Kaposi sarcoma cases in all 3 states using relative survival methods. Additional combinations of state and primary site had significant survival estimate differences, which differed by method (relative versus cause-specific survival). Conclusion: To ensure accurate population-based cancer survival rates, linkage with the National Death Index to ascertain out of state and late registered deaths is a necessary process for US central cancer registries.     

The effect of diagnostic misclassification on non-cancer and cancer mortality dose response in A-bomb survivors.

We used the EM algorithm in the context of a joint Poisson regression analysis of cancer and non-cancer mortality in the Radiation Effects Research Foundation (RERF) Life Span Study (LSS) to assess whether the observed increased risk of non-cancer death due to radiation exposure (Shimizu et al., RERF Technical Report 02-91, 1991) can be attributed solely to misclassification of cancer as non-cancer on death certificates. We show that greater levels of dose-independent misclassification than are indicated by a series of autopsies conducted on a subset of LSS members would be required to explain the non-cancer dose response, but that a relatively small amount of dose-dependence in the misclassification of cancer would explain the result. The adjustment for misclassification also results in higher risk estimates for cancer mortality. We review applications of similar statistical methods in other contexts and discuss extensions of the methods to more than two causes of death.     

On estimation in relative survival

Estimation of relative survival has become the first and the most basic step when reporting cancer survival statistics. Standard estimators are in routine use by all cancer registries. However, it has been recently noted that these estimators do not provide information on cancer mortality that is independent of the national general population mortality. Thus they are not suitable for comparison between countries. Furthermore, the commonly used interpretation of the relative survival curve is vague and misleading. The present article attempts to remedy these basic problems. The population quantities of the traditional estimators are carefully described and their interpretation discussed. We then propose a new estimator of net survival probability that enables the desired comparability between countries. The new estimator requires no modeling and is accompanied with a straightforward variance estimate. The methods are described on real as well as simulated data.     

Is neoadjuvant chemotherapy for pT2 bladder cancer associated with a survival benefit in a population-based analysis?

BackgroundPatients with organ confined muscle-invasive bladder cancer (MIBC) who are candidates for radical cystectomy (RC) should receive neoadjuvant chemotherapy (CHT). However, the most contemporary CHT use rates indicate low adherence to these guidelines. We tested contemporary neoadjuvant CHT rates and associated cancer-specific mortality (CSM) and overall mortality (OM) in pT2N0 MIBC patients treated with RC.Materials and methodsWithin the SEER database (2004–2015), we identified patients with pT2N0 MIBC patients who underwent RC. CHT administration rates were evaluated using estimated annual percentage changes (EAPCs) analyses. After inverse probability of treatment weighting (IPTW), Kaplan–Meier (KM) analyses and Cox regression models (CRMs) were used to test the effect of CHT vs no CHT on survival. Landmark analyses tested for immortal time bias.ResultsOf 3978 RC patients, 38.2% of patients received CHT. Between 2004 and 2015, CHT rates increased from 15.9% to 66.2% (EAPC: +14.2%; p < 0.001). IPTW-adjusted KM showed 10-year CSM-free survival rates of 78.9% for CHT vs 76.7% for no CHT patients (p = 0.6). Similarly, IPTW-adjusted KM showed 10-year OM-free survival rates of 54.6% for CHT vs 57.9% for no CHT patients (p = 0.8). In IPTW-adjusted MCRMs, CHT was not significantly associated with lower CSM (HR 0.97, CI 0.82–1.14; p = 0.7) or OM (HR 1.02, CI 0.90–1.16; p = 0.7). Virtually the same CSM and OM rates were recorded after landmark analyses.ConclusionsCHT use in pT2N0 MIBC RC patients sharply increased over the study span. However, neoadjuvant CHT was not associated with better survival in this patient group.     

Accuracy of industry and occupation on death certificates of electric utility workers: implications for epidemiologic studies of magnetic fields and cancer

A substantial epidemiologic literature has relied on occupation and industry information from death certificates to make inferences about the association of electric and magnetic field exposure with cancer, but the validity of the occupational data on death certificates is questionable. We compared occupation and industry information from death certificates to company work histories for 793 electric utility workers who died from brain cancer (n=143), leukemia (n=156), lung cancer (n=246, randomly sampled), and non-cancer causes (n=248, randomly sampled). Nearly 75% of death certificates correctly indicated utility industry employment and of those, 48% matched the longest held occupation derived from company work histories. Hence, only 36% matched on both industry and occupation. We computed odds ratios relating occupations involving magnetic field exposure to brain cancer and leukemia both for the occupation listed on the death certificate and for the longest-held occupation based on company records in order to examine the impact of exposure misclassification based on reliance on the death certificate information. For brain cancer, the odds ratio was 1.2 based on death certificates and 1.7 based on company work history, suggesting some attenuation due to misclassification. For leukemia, death certificate information yielded an odds ratio of 0.9, whereas company work histories yielded an odds ratio of 1.3. Although work histories are limited to the period of employment in a specific company, these data suggest that there is substantial misclassification in use of death certificate information on industry and occupation of utility workers, as found in other industries. The limited quality of occupation and industry information on death certificates argues against relying on such information to evaluate modest associations with mortality.     

The impact of under and over-recording of cancer on death certificates in a competing risks analysis: A simulation study

Background: With linked register and cause of death data becoming more accessible than ever, competing risks methodology is being increasingly used as a way of obtaining “real world” probabilities of death broken down by specific causes. It is important, in terms of the validity of these studies, to have accurate cause of death information. However, it is well documented that cause of death information taken from death certificates is often lacking in accuracy and completeness. Methods: We assess through use of a simulation study the effect of under and over-recording of cancer on death certificates in a competing risks analysis consisting of three competing causes of death: cancer, heart disease and other causes. Using realistic levels of misclassification, we consider 24 scenarios and examine the bias in the cause-specific hazard ratios and the cumulative incidence function. Results: The bias in the cumulative incidence function was highest in the oldest age group reaching values as high as 2.6 percentage units for the “good” cancer prognosis scenario and 9.7 percentage units for the “poor” prognosis scenario. Conclusion: The bias resulting from the chosen levels of misclassification in this study accentuate concerns that unreliable cause of death information may be providing misleading results. The results of this simulation study convey an important message to applied epidemiological researchers.     

Beta-blocker usage and prostate cancer survival: A nested case–control study in the UK Clinical Practice Research Datalink cohort

BackgroundRecent laboratory and epidemiological evidence suggests that beta-blockers could inhibit prostate cancer progression. Methods: We investigated the effect of beta-blockers on prostate cancer-specific mortality in a cohort of prostate cancer patients. Prostate cancer patients diagnosed between 1998 and 2006 were identified from the UK Clinical Practice Research Database and confirmed by cancer registries. Patients were followed up to 2011 with deaths identified by the Office of National Statistics. A nested case–control analysis compared patients dying from prostate cancer (cases) with up to three controls alive at the time of their death, matched by age and year of diagnosis. Odds ratios (OR) and 95% confidence intervals (CI) were calculated using conditional logistic regression. Results: Post-diagnostic beta-blocker use was identified in 25% of 1184 prostate cancer-specific deaths and 26% of 3531 matched controls. There was little evidence (P = 0.40) of a reduction in the risk of cancer-specific death in beta-blocker users compared with non-users (OR = 0.94 95% CI 0.81, 1.09). Similar results were observed after adjustments for confounders, in analyses by beta-blocker frequency, duration, type and for all-cause mortality. Conclusions: Beta-blocker usage after diagnosis was not associated with cancer-specific or all-cause mortality in prostate cancer patients in this large UK study.     

Adjusting for the proportion of cancer deaths in the general population when using relative survival: a sensitivity analysis

Background: Relative survival is an extensively used method in population based cancer studies as it provides a measure of survival without the need for accurate cause of death information. It gives an estimate for the probability of dying from cancer in the absence of other causes by estimating the excess mortality in the study population when compared to an external group. The external group is usually the general population within a country or state and mortality estimates are taken from national life tables that are broken down by age, sex, calendar year and, where applicable, race/ethnicity. One potential bias when using relative survival that is most often overlooked occurs when there are a high proportion of deaths due to a specific cancer in the external group. Methods: This paper uses data from the Finnish Cancer Registry to illustrate, through the use of a simple sensitivity analysis, the impact that specific cancer deaths in the population mortality figures can have on the estimate of relative survival. Results: We found that when examining specific diseases such as breast cancer and colon cancer, the proportion of deaths due to these specific cancers in the general population is so small in comparison to the total mortality that they make little difference to the relative survival estimates. However, prostate cancer proved to be an exception to this. For all cancer sites combined the sensitivity analysis illustrates a major limitation for this type of analysis, particularly with the older age groups. Conclusion: We recommend that, with a classification of diseases as wide as all cancer sites, relative survival should not be used without appropriate adjustment.     

Using death certificates to estimate the completeness of AIDS case reporting in Ontario in 1985-87.

The completeness of AIDS (acquired immune deficiency syndrome) case reporting in Ontario was assessed by reviewing all AIDS death certificates compiled by the Registrar General between Jan. 1, 1985, and Dec. 31, 1987. Several demographic variables were used to match death certificates with cases reported to the provincial AIDS registry. The completeness of case reporting was then estimated by examining the ratio of reported deaths of patients with AIDS to the total number of deaths reviewed. The estimated completeness of case reporting was 81.1% in 1985, 71.5% in 1986 and 75.4% in 1987; the overall rate for 1985-87 was 75.2%. The difference in the completeness of case reporting from year to year was not statistically significant. There was a significant increase from 1985 to 1986 in the proportion of unreported cases in people who had never been married (p less than 0.02). Reporting was not associated with the patient''s age, sex, occupation or place of residence. The deficiency in AIDS case reporting could adversely affect the long-term planning of health care resources and the development of programs to prevent and control the spread of AIDS.     

Metformin increases cancer specific survival in colorectal cancer patients—National cohort study

PurposeWe aimed to assess oncological outcomes in colorectal cancer patients with type 2 diabetes mellitus (T2DM) using metformin.MethodsPatients with colorectal cancer and T2DM during 2000–2012 period were identified form Lithuanian Cancer Registry and the National Health Insurance Fund database. Colorectal cancer-specific survival (CS) was the primary outcome. It was measured from date of colorectal cancer diagnosis to date of death due to colorectal cancer, or last known date alive.Results15,052 people who met eligibility criteria for this analysis, including 1094 (7.27%) with pre-existing type 2 diabetes (271 metformin never users and 823 metformin users) and 13 958 people without diabetes assessed. During follow-up (mean follow-up time was 4.4 years, with range from 1 day to 17 years) there were 10,927 deaths including 8559 from colorectal cancer. Significantly lower risk in CS between diabetic and non-diabetic people with lower risk of cancer-specific mortality (HR 0.87, 95% CI 0.80–0.94) in diabetic patient population was seen. After adjustment for age, stage at diagnosis and metformin usage, significant difference in colorectal CS between metformin users in diabetic patient population compared to non-diabetics and metformin non-users in diabetic patient population was found (0.80 (0.72–0.89) vs 1.00 and vs 1.05 (0.91–1.23)). Overall survival (OS) was better for diabetic patients with significant difference in diabetic metformin users (HR 0.91, 95% CI 0.79–0.94).ConclusionsColorectal cancer patients with T2DM treated with metformin as part of their diabetic therapy appear to have a superior OS and CS. However, prospective controlled studies are still needed to evaluate the efficacy of metformin as an anti-tumor agent.     

Mortality rates in subjects with fetal alcohol spectrum disorders and their siblings

BACKGROUND: Our objective was to estimate the mortality rate in subjects with fetal alcohol spectrum disorders (FASD) and their siblings whose FASD status was unknown. METHODS: We used the state FASD Registry to link subjects with FASD to a North Dakota birth certificate. We were able to link 304 of 486 cases (63%). We used the birth certificates to identify the mother and children born to the mother (siblings). We then searched for death certificates for both the FASD cases and their siblings. We then calculated the annual and age‐adjusted mortality rates for the siblings of the Registry cases and compared them with mortality rates from North Dakota. RESULTS: The FASD case mortality rate was 2.4%, with a 4.5% mortality rate for their sibings, accounting for 14% of all deaths when compared to the North Dakota residents matched by age and year of death. The sibling deaths accounted for 21.5% of all cause mortality matched by age and year of death. The age‐standardized mortality ratios were 4.9 for the FASD cases and 2.6 for their siblings whose FASD status was unknown. CONCLUSIONS: Mortality rates for FASD cases and their siblings were increased and represent a substantial proportion of all cause mortality in North Dakota. Prevention of FASD may be a useful strategy to decrease mortality. Birth Defects Research (Part A), 2008. © 2008 Wiley‐Liss, Inc.     

Probabilities of dying from cancer and other causes in French cancer patients based on an unbiased estimator of net survival: A study of five common cancers

BackgroundNet survival is the survival that would be observed if cancer were the only possible cause of death. Although it is an important epidemiological tool allowing temporal or geographical comparisons, it cannot inform on the “crude” probability of death of cancer patients; i.e., when taking into account other possible causes of deaths.MethodsIn this work, we provide estimates of the crude probabilities of death from cancer and from other causes as well as the probability of being alive up to ten years after cancer diagnosis according to the age and year of diagnosis. Based on a flexible excess hazard model providing unbiased estimates of net survival, our methodology avoids the pitfalls associated with the use of the cause of death. We used data from FRANCIM, the French network of cancer registries, and studied five common cancer sites: head and neck, breast, prostate, lung, and colorectal cancers.ResultsFor breast, prostate, and colorectal cancers, the impact of the other causes on the total probability of death increased with the age at diagnosis whereas it remained negligible for lung and head and neck cancers whatever the age. For breast, prostate, and colorectal cancer, the more recently was the cancer diagnosed, the less was the probability of death from cancer.ConclusionThe crude probability of death is an intuitive concept that may prove particularly useful in choosing an appropriate treatment, or refining the indication of a screening strategy by allowing the clinician to estimate the proportion of cancer patients who will die specifically from cancer.     

Impact of Inclusion of Industry Trial Results Registries as an Information Source for Systematic Reviews

Background

Clinical trial results registries may contain relevant unpublished information. Our main aim was to investigate the potential impact of the inclusion of reports from industry results registries on systematic reviews (SRs).

Methods

We identified a sample of 150 eligible SRs in PubMed via backward selection. Eligible SRs investigated randomized controlled trials of drugs and included at least 2 bibliographic databases (original search date: 11/2009). We checked whether results registries of manufacturers and/or industry associations had also been searched. If not, we searched these registries for additional trials not considered in the SRs, as well as for additional data on trials already considered. We reanalysed the primary outcome and harm outcomes reported in the SRs and determined whether results had changed. A “change” was defined as either a new relevant result or a change in the statistical significance of an existing result. We performed a search update in 8/2013 and identified a sample of 20 eligible SRs to determine whether mandatory results registration from 9/2008 onwards in the public trial and results registry ClinicalTrials.gov had led to its inclusion as a standard information source in SRs, and whether the inclusion rate of industry results registries had changed.

Results

133 of the 150 SRs (89%) in the original analysis did not search industry results registries. For 23 (17%) of these SRs we found 25 additional trials and additional data on 31 trials already included in the SRs. This additional information was found for more than twice as many SRs of drugs approved from 2000 as approved beforehand. The inclusion of the additional trials and data yielded changes in existing results or the addition of new results for 6 of the 23 SRs. Of the 20 SRs retrieved in the search update, 8 considered ClinicalTrials.gov or a meta-registry linking to ClinicalTrials.gov, and 1 considered an industry results registry.

Conclusion

The inclusion of industry and public results registries as an information source in SRs is still insufficient and may result in publication and outcome reporting bias. In addition to an essential search in ClinicalTrials.gov, authors of SRs should consider searching industry results registries.