Do older individuals who are diagnosed with cancer have worse physical performance prior to diagnosis compared to matched controls? A longitudinal cohort study

The study sample was selected from the Longitudinal Aging Study Amsterdam (LASA), a longitudinal study on a nationally representative sample of the Dutch older population among 3107 community dwelling individuals aged 55–84 years (first cohort, 1992–1993) and 1002 individuals aged 55–64 years (second cohort, 2002–2003) [29]. Individuals were followed about every 3 years since these baseline measurements. Thus after the baseline measurement of cohort one in 1992–1993, a second cohort was added in 2002-2003 and the latest follow-up included in this study took place in 2011-2012. Detailed information on data procedures and collection has been published elsewhere [29, 30]. The study has been approved by the Medical Ethics Committee of the VU University Medical Center in Amsterdam.

Individuals were divided into a cancer group and control group based on self-report of having cancer during any of the follow-up measurements. To determine the physical performance of cancer patients prior to their diagnosis, the measurement cycle before they reported to have cancer for the first time was used for analysis. After the baseline and follow-up interviews of 2001–2002, 2005–2006 and 2008–2009, information on presence of cancer was also obtained from general practitioners by use of a questionnaire. As self-reports of cancer were fairly accurate compared to reports of general practitioners (1992–1993 kappa = 0.64 (0.58–0.70) and 2008–2009 kappa = 0.64 (0.57–0.70), the cancer group in this study was based on self-report of cancer [31, 32]. Individuals were included in the cancer group if they reported cancer for the first time at any of the follow up measurements. Individuals were excluded if 1) they reported cancer at baseline (n = 372); 2) they did not participate in the most recent interview before reporting cancer (n = 9); 3) no physical performance tests were conducted during the interview before reporting cancer (n = 2). The number of individuals included in the cancer group was N = 414.

The control group consisted of individuals who reported no cancer at baseline nor during the complete 10 (cohort 2) to 20 (cohort 1) years of follow-up. The measurement cycle of the year 2005–2006 was chosen as the physical performance measurement for the control group, since the median age of individuals in this cycle was nearest to the median age of the individuals diagnosed with cancer. Individuals were excluded from the control group if: 1) they died before the interview in 2005–2006, did not participate in the interview or data of the interview or physical performance measurements were missing in 2005–2006 (n = 1865); 2) they did not report presence of cancer at any of the interviews but the general practitioner did (n = 116), to maximize the probability that the controls were cancer free. The number of individuals included in the control group was N = 1321. Figure 1 illustrates the selection of the study sample from the cohort.

Data on age, gender, lifestyle factors, anthropometry and health characteristics were included as covariates. Lifestyle variables included current smoking status and alcohol use. Current smoking status (never, former and current smoker) was dichotomized into yes (current smoker) or no (never or former smoker). Alcohol consumption was measured with an adaptation of the alcohol consumption index by Garretsen [33]. This index was categorized into (very) excessive alcohol use, light/moderate alcohol use and no alcohol use. Anthropometric measurements included height, weight and body mass index (BMI). Health characteristics included cognitive functioning, number of chronic diseases and number of medicines used per day. Cognitive functioning was measured by the Mini Mental State Examination (MMSE) [34]. Number of chronic diseases was obtained by asking explicitly about the presence of six prevalent somatic chronic diseases (chronic non-specific lung disease including asthma and chronic obstructive pulmonary disease, cardiac disease, peripheral arterial disease, diabetes mellitus, cerebrovascular accident, and osteoarthritis or rheumatoid arthritis [35, 36]. The medicines taken by the respondent were recorded from the containers by the interviewer.

Physical performance was assessed at respondents’ homes and included handgrip strength (HGS), six-meter walking (6MWT), chair stand (CST), tandem balance and putting on and taking off a cardigan. HGS was measured using a grip strength dynamometer that was adjusted for hand size and recorded the grip strength to the nearest kilogram (Takei TKK 5001, Takei Scientific Instruments Co. Ltd., Tokyo, Japan). HGS was measured twice for both the left and right hand. Maximum HGS was determined by the highest grip strength. Gait speed was assessed by time in seconds doing the 6MWT which consists of walking three meters, turn around and walk three meters back as quickly as they can. Gait speed was measured in meters per second by dividing six meters by the time to do the test. The CST was used to assess the ability of individuals to rise from a chair with their arms crossed over their chest, stand up to a straight position and sit down again. This test was performed five times. The ability and the time in seconds needed to perform the CST were used for analysis. The tandem balance test is performed by placing the heel of 1 foot directly in front of the other foot, making sure that the toes of the back foot are touching the heel of the front foot. The ability to remain in tandem stand for 10 s was used for analysis. The cardigan test measured the time in seconds required to put on and take off a cardigan.

The median age of the total study sample was 68.7 [IQR 63.3–76.4] years. Table 1 shows the characteristics of the group prior to diagnosis of cancer and the control group. None of the characteristics (age, gender, smoking, alcohol, anthropometric measurements or health characteristics) differed significantly between the cancer and the control group.

Table 2 shows the mean physical performance and Table 3 shows the proportions of individuals prior to cancer diagnosis and controls on the physical performance measurements. Both tables also show the association between cancer diagnosis and physical performance in models adjusted for gender and age (model 1) and for additional potential confounders (model 2). HGS, gait speed, ability and time to perform the CST and the cardigan test did not differ between individuals prior to cancer diagnosis and controls. Individuals prior to cancer diagnosis were more likely to be able to remain in tandem position for 10 s than controls, also after adjustment for possible confounders. Results did not differ when individuals of whom the general practitioner reported cancer while the individuals themselves did not, were included in the control group.

A limitation of this study is that actual cancer diagnosis could have occurred anytime between the two measurements varying from 1 day to 4 years. Unfortunately, it was not possible to adjust for the effect of the time interval between measurement of physical performance and cancer diagnosis because the actual date of cancer diagnosis is unknown, there is only information on the date that individuals reported to have cancer during any of the measurement cycles. Furthermore, presence of cancer was based on self-report of individuals. Although there was substantial accuracy in the self-report of individuals [31, 32], self-report could still have led to under- or over reporting of cancer. Based on previous studies, underreport of cancer is most likely [32, 54] and underreport is more common in individuals without mobility limitations [32]. Thus, if there was substantial underreport of cancer, it will not have changed the results and conclusions of this study as it will have led to more individuals with higher physical performance in the cancer group. Moreover, controls were matched based on the median age of the cancer group and not per individual. However, the groups proved to be comparable in major sociodemographic characteristics. Lastly, the cancer group included all types and severity of cancer and it may be argued that some types of tumor would have a larger influence on physical performance and that more severe stages of cancer will also have a bigger impact on physical performance. This argument is supported by the study of Petrick et al. which showed that self-reported physical performance was only significantly lower in the group with lung cancer in the 3 years leading up to cancer diagnosis [22].