Treatment plan consistency with guidelines for older adults with cancer

The population of older adult cancer survivors in the US is rapidly expanding. Today, more than half of all cancer survivors are over 65 years of age; by 2040, this number is expected to rise to 73% [1]. Currently, treatment of the older adult with cancer often poses a dilemma: whether to attempt cure and risk toxicity or to design a treatment plan that may compromise survival but preserve quality of life [2].

Currently little is known about how oncologists make decisions about treatment planning for older adults with cancer. One study used data from a state cancer registry to evaluate adherence to National Comprehensive Cancer Network guidelines for older adults with stage II or III colon cancer [3]. Approximately 50% of patients between ages 66 and 80 received guideline-concordant treatment. Guideline-concordant treatment was associated with better survival outcomes, however adherence to guidelines was less likely for single, male older individuals (≥ age 65) with comorbid conditions. A cohort study of stage I to III breast cancer patients showed that while overall chemotherapy adherence was approximately 70%, the number of guideline violations increased proportionally with age [4].

Our research group conducted a pilot study to evaluate recommended treatment plans for older adults with cancer in one large urban oncology practice in the United States prior to initiation of therapy. We also sought to identify factors that influenced recommendations for treatment of older adults with cancer. The team conducted a retrospective medical records review as part of a larger, prospective study evaluating clinical outcomes for older adults with cancer. Patients were eligible if they were ≥ age 65 with a new or recurrent cancer diagnosis, and were seen in a general oncology practice group in the cancer center. The Institutional Review Board approved this study.

Information derived from the electronic medical record included demographics (i.e., age, gender, race, ethnicity, insurance status), clinical (ECOG performance status (PS), Charlson comorbidity index (CCI)) [5,6], and cancer data (i.e., diagnosis, stage, grade, recurrence or new diagnosis, history of previous cancer). The CCI was dichotomized as cancer alone versus cancer plus comorbidity. The primary outcome assessed was concordance of treatment plans with NCCN guidelines.

For each case, AM (from the research team) reviewed NCCN treatment guidelines [7] for all treatment modalities, chemotherapy drugs and doses, and length of planned regimen. Non-concordant treatment plans were cases in which any treatment modality was omitted (e.g., surgery, radiation or chemotherapy), mono-chemotherapy was used when poly-chemotherapy was recommended, a non-recommended chemotherapeutic agent was substituted, prescribed drug dosage was reduced, or an abbreviated course of treatment was recommended. If the case met all NCCN criteria, the plan was considered concordant with guidelines. For cases in which treatment plans were determined to be non-concordant, we conducted a further review of the medical record to determine why. Reasons for a non-concordant plan were recorded verbatim, and were classified into three categories: patient preference, medical contraindication, or frailty.

The data were summarized with frequencies and percentages. Fisher’s exact test was used to evaluate unadjusted means for significant differences. Following univariable analyses, logistic regression was used to model consistency with treatment guidelines (concordant versus non-concordant) using patient demographic, clinical, and disease characteristics as predictors. The most promising predictors from the univariable analyses were candidates for inclusion in the model. The final model was selected using a backward stepwise variable selection procedure. Due to the small sample size, we tuned the procedure to select no more than two predictor variables with p < .20. The significance level for each test was set in advance at α = 0.05. SAS version 9.4 (SAS Institute, Cary, NC, USA) was used to conduct all statistical analyses.

A total of 82 cases were identified as receiving a treatment plan for a new or recurrent cancer. Eleven cases were eliminated due to treatment by a study investigator with geriatric training and two were eliminated because at follow up, a probable cancer diagnosis was not confirmed. A final sample of 69 cases was analyzed.

Slightly >50% of patients included in the study were ≥ 75 years of age (Table 1). Women (58%) outnumbered men. Sixty-one percent of patients were white and 26% were African American. Most patients had a limited performance status at diagnosis (PS ≥ 1 = 64%). Sixty percent had co-morbid conditions in addition to cancer. The most common cancer type was gastrointestinal cancer, followed by breast, lung, genitourinary, and hematological malignancy. Most patients (77%) were newly diagnosed. About one third of the sample (32%) had stage 0-I disease, 32% had stage II-III disease, and 36% had stage IV disease.

Table 1.

Demographics stratified by guideline-consistency.

	All		Inconsistent with Guidelines (IG)		Consistent with Guidelines (CG)		p-value
	N	%	N	%	N	%
All	69	100.0	12	17.4	57	82.6
Age							0.229
65–69	14	20.3	1	7.1	13	92.9
70–74	34	49.3	9	26.5	25	73.5
75+	21	30.4	2	9.5	19	90.5
Gender							0.749
Male	29	42.0	4	13.8	25	86.2
Female	40	58.0	8	20.0	32	80.0
Racea							0.394
White	42	60.9	6	14.3	36	85.7
African-American	18	26.1	5	27.8	13	72.2
Other	7	10.1	1	14.3	6	85.7
ECOG performance statusa							>0.999
Fully Active (0)	24	34.8	4	16.7	20	83.3
Limited [1–4]b	44	63.8	8	18.2	36	81.8
Charlson Scorec							0.104
Cancer only	28	40.6	2	7.1	26	92.9
Additional comorbidities	41	59.4	10	24.4	31	75.6
Stage							0.004
0-I	22	31.9	1	4.5	21	95.5
II-III	22	31.9	9	40.9	13	59.1
IV	25	36.2	2	8.0	23	92.0
Cancer Gradea							0.499
Well differentiated	10	14.5	1	10.0	9	90.0
Moderately differentiated	15	21.7	3	20.0	12	80.0
Poorly differentiated	24	34.8	7	29.2	17	70.8
Recurrent cancer diagnosis	16	23.2	4	25.0	12	75.0	0.453
Previous cancer	11	15.9	3	27.3	8	72.7	0.390
Referred for CGA	3	4.3	1	33.3	2	66.7	0.442

^aData were not available for all subjects: race was missing for 2; ECOG status for 1; Cancer grade for 20.

^bincludes 36 restricted [1]; 6-ambulatory [2]; 1 limited self-care [1]; 1 disabled [4]

^cCharlson Comorbidity Index computed without the two cancer components–any cancer, metastatic cancer.

Fifty-seven patients (83%) were treated according to NCCN guidelines. Among the 12 patients with non-concordant plans, four received non-concordant care due to one or more medical contraindications, three were due to patient preference, and one was due to patient frailty. In four cases, the reason could not be obtained. In univariable analyses we found that being recommended a non-concordant plan was significantly associated with disease stage (p = .004). Of patients recommended a non-concordant plan, 41% had stage II-III disease - compared to only 5% with stage 0-I and 8% with stage IV disease. In this small sample, a higher prevalence of patients with a co-morbid condition were recommended a non-concordant plan than those without a co-morbid condition (24% and 7%, respectively), although the results were not significant (p = .104). Findings from logistic regression model analyses suggested that the odds of being prescribed a non-concordant plan were most strongly predicted by stage of cancer. Because rates of non-concordant plans in patients were comparable for patients with the lowest (0-I) and highest (IV) disease staging (5% and 8% respectively), these groups were collapsed for modeling purposes. We found that the odds of being recommended a non-concordant plan were significantly higher in patients with stage II-III disease versus those with early or late stage disease (OR = 9.41, 95% CI: 2.14–41.28, p = .003). While not statistically significant, the presence of comorbid conditions was an important adjustment term for the selected model (OR = 3.93, 95% CI: 0.71–21.74, p = .177).

In conclusion, the aims of this pilot research were to determine whether recommended treatment plans for older adults with cancer were concordant or non-concordant with NCCN guidelines and identify factors associated with plans that were non-concordant with guidelines. The results showed that while oncologists were more likely to recommend plans that were concordant for individuals with early or late stage disease, patients with stage II or III potentially curable cancers were more likely to be recommended a treatment plan that was non-concordant with guidelines. This result is consistent with previous research showing physicians may not recommend guideline-concordant care for potentially curable cancers due to factors other than age [4,8]. This situation highlights the need for treatment guidelines designed for older patients, demonstrates the importance of educating physicians on how to optimize clinical care for older adults with cancer (i.e. through continuing medical education and formal fellowship training), and reinforces the value of conducting more research on predictors of non-concordant care in this segment of the patient population.

Guidelines are created with the best evidence available at the time of development. Until recently, the evidence was based on literature that excluded older patients from treatment trials. Trials now routinely do not have an upper age cut off, and literature is starting to reflect the population with the disease, namely older adults. Newer, age-focused guidelines have been created recently by the American Society of Clinical Oncology that aim to improve the assessment and treatment planning for older adults [9].

National guidelines are developed to assist oncologists in creating treatment plans for their patients. However, it is not known how many clinicians are using and/or aware of guidelines that take older adult issues into account. It may be that oncologists recognize warning signals in their older patients, and therefore, choose a less aggressive treatment approach, thus leading to non-concordant recommendations. The development of guidelines that take the needs of older adult patients into account may help to provide guidance in the management of these situations. Furthermore, it has been reported that clinical fellows are undertrained in geriatric oncology concepts, but are interested in enriching this part of their fellowship [10].

Limitations of this pilot study include the small sample size and retrospective single site design.

Therefore, this research was hypothesis-generating regarding the practice patterns of oncologists.