Pre- and post-diagnosis leisure time physical activity and survival following diagnosis with ovarian cancer

Abstract

Little is known about the influence of pre- and post-diagnosis physical activity on ovarian cancer survival. We investigated this association in two large cohorts, the Nurses’ Health Study (NHS) and NHSII. Analyses included 1461 women with confirmed invasive, epithelial ovarian cancer and data on physical activity. Cox regression models were used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for ovarian cancer-specific mortality. Ovarian cancer-specific mortality was not associated with physical activity reported 1–8 years before diagnosis overall (≥7.5 vs. <1.5 MET-hours/week, HR=0.96), for high-grade serous/ poorly differentiated tumors, or non-serous/ low-grade serous tumors (P-heterogeneity=0.45). An inverse association was observed for activity 1–4 years after diagnosis (≥7.5 vs. <1.5 MET-hours/week, HR=0.67, 95%CI: 0.48–0.94), with similar results by histotype (P-heterogeneity=0.53). Women who decreased their activity from ≥7.5 MET-hours/week 1–8 years before diagnosis to <7.5 MET-hours/week 1–4 years after diagnosis, compared with those with <7.5 MET-hours/week across periods, had a 49% increased risk of death (HR=1.49, 95%CI: 1.07–2.08). Physical activity after, but not before, ovarian cancer diagnosis was associated with better prognosis.

BACKGROUND

Based on research conducted over the past two decades, cancer survivors are recommended to undertake 150–300 minutes of moderate or 75–150 minutes vigorous intensity aerobic activity per week ^{1, 2}. However, these recommendations have been largely based on research involving breast and prostate cancer survivors. Little is known regarding appropriate exercise recommendations for individuals with less common cancers, such as ovarian cancer ³, which is the most fatal gynecological malignancy in the US ⁴. Also, fewer ovarian cancer patients achieved activity guidelines than breast and prostate cancer patients, possibly due to the aggressive treatment typically required for ovarian cancer ⁵. Therefore, research to tailor physical activity guidelines for ovarian cancer survivors is crucial to improve outcomes ⁶.

Previous studies of ovarian cancer survival considered both pre- and post-diagnosis exposure, as pre-diagnosis factors may alter the aggressiveness of the tumor, while post-diagnosis factors may impact the efficacy of cancer treatments and recurrence ⁷. For example, we previously observed that pre- and post-diagnosis smoking were associated with ovarian cancer survival, with stronger associations for post-diagnosis smoking ⁸, while only post-diagnosis aspirin use was associated with survival ⁹. Therefore, it is critical to consider the timing of modifiable risk factors assessment when evaluating cancer outcomes.

In the current study, we investigated the associations of physical activity before and after ovarian cancer diagnosis and change in activity from pre- to post-diagnosis period with mortality in two large prospective cohort studies. We considered the associations among all patients and by tumor histotype and stage, and the associations with vigorous and moderate activity separately.

METHODS

Study participants

The Nurses’ Health Study (NHS) was established in 1976 among 121700 US female registered nurses aged 30–55 years from 11 states ¹⁰. NHSII began in 1989 among 116429 US female registered nurses aged 25–42 years from 14 states ¹¹. In both cohorts, participants reported detailed information about their lifestyle and medical history at study entry and were followed prospectively by biennial questionnaires to update this information.

Ovarian cancer, primary peritoneal and fallopian tube cancer are typically grouped together because of common histotypes and origins, and are subsequently referred as ovarian cancer herein ¹². Eligible participants included women with confirmed invasive, epithelial ovarian cancer who provided information on physical activity either before or after ovarian cancer diagnosis (Figure 1). Ovarian cancer cases were identified by self-report on the biennial questionnaires and deaths were identified via the National Death Index, family members, or the US Postal Service. All ovarian cancer diagnoses were confirmed via medical record review by a gynecological pathologist or by linkage with the relevant cancer registry. The gynecological pathologist abstracted data on tumor stage, histology, grade, and morphology ¹³. Cause and date of death was obtained from related medical records and death certificates. We excluded patients who died within one month of diagnosis (n=33) or had missing data on physical activity at all timepoints (n=97).

Figure 1.

Study participant flow diagram in the Nurses’ Health Study (NHS) and NHSII

Assessment of physical activity and other covariates

Physical activity was assessed every 2–4 years since 1986 in NHS and 1989 in NHSII with good reproducibility and validity ¹⁴. Participants were asked about average time spent weekly doing each of eight common leisure-time activities, including walking, jogging, running, bicycling, swimming, tennis, squash/racquet ball, and calisthenics/aerobics/rowing machine. A metabolic equivalent task (MET) score was assigned to each type of activity to quantify its energy expenditure, then MET-hours per week were calculated by multiplying the corresponding MET score and hours per week spent in each activity ¹⁵. Total leisure time physical activity was calculated by summing MET-hours per week over all eight activities. Activities with a MET score ≥6 were defined as vigorous, while other activities were defined as moderate. We used 1.5 and 7.5 MET-hours/week as the cutoff values, which equate to 30- (low level of physical activity) and 150-minute (recommended level of physical activity) walking with average pace (2–2.9 mph), respectively ^{1, 2}.

Primary analyses assessed activity at one period before ovarian cancer diagnosis and one after diagnosis, excluding the one year before diagnosis, when women may be experiencing symptoms of ovarian cancer and the one year after diagnosis, when most patients were undergoing treatment ¹⁶. Pre-diagnosis activity was assessed on the questionnaire with the return date 1–8 years (median assessment of 4.6, interquartile range [IQR]: 4.0–5.2 years; average was used if multiple questionnaires were available) prior to ovarian cancer diagnosis. This period was chosen as previous studies identified a window of seven years between development of a serous tubal intraepithelial carcinomas (precursor of ovarian carcinoma) and initiation of ovarian carcinoma ¹⁷. Post-diagnosis activity was obtained from the questionnaire with the earliest return date 1–4 years (median 1.8, IQR: 1.4–2.5 years) after diagnosis.

Body mass index (BMI), smoking status and non-steroidal anti-inflammatory drug (NSAID) use were queried on biennial questionnaires. Briefly, height, weight and smoking status were queried at baseline, and weight and smoking status were updated on each follow-up questionnaire. In NHS, the use of aspirin and non-aspirin NSAIDs was queried in 1980, 1990, and every two years thereafter. In NHSII, aspirin and non-aspirin NSAID use was collected in 1989, 1993, and every two years thereafter. Pre- and post-diagnosis values of these covariates were defined using the same timeframe as the relevant activity assessment.

Outcomes

The primary and secondary outcomes were ovarian cancer-specific mortality and all-cause mortality, respectively.

Statistical methods

Cox regression models were used to estimate hazard ratios (HR) and 95% confidence intervals (CI). For pre-diagnosis exposures, the analytic time scale was measured in months from diagnosis to death or the end of follow-up (June 2016 for NHS and June 2017 for NHSII). In analyses of post-diagnosis activity and change from pre- to post-diagnosis, the analytic time scale was measured in months from the return of the post-diagnosis questionnaire containing activity information to death or the end of follow-up. Primary pre- and post-diagnosis exposure models and models analyzing change in activity levels were adjusted for age and calendar year of diagnosis, tumor stage (I-IV or unknown), histology (high-grade serous/poorly differentiated, low-grade serous/non-serous, other histology/unknown), cohort (NHS, NHSII), BMI (<25, 25-<30, ≥30 kg/m², unknown), and smoking status (never smoker, ever smoker, unknown). We adjusted for post-diagnosis NSAID use (current user, non-current user, unknown) only for post-diagnosis and change analyses, as we previously observed no association of pre-diagnosis NSAID use with survival ⁹. Tumor grade and additional factors, including parity, oral contraceptive use, postmenopausal hormone therapy, family history of ovarian cancer or breast cancer, alcohol intake, diet, depression, change of BMI, and self-reported physical limitations ¹⁸, did not substantially alter risk estimates and thus were not included in the final models. To better evaluate the influence of unmeasured confounders, we calculated the E-value, which is defined as the minimum strength of association on the risk ratio scale that an unmeasured confounder would need to have with both exposure and outcome to fully explain an observed association ¹⁹. We conducted analyses stratified by histotype and stage and tested potential heterogeneity in effect estimates using a likelihood ratio test comparing models with versus without interaction terms.

To test the proportional hazards assumption, we computed multiplicative interaction terms between physical activity and the analytic time scale and compared models with versus without interaction terms using likelihood ratio test. To test for heterogeneity by cohort, HRs were calculated separately in each cohort and pooled using random-effects meta-analysis. We conducted a sensitivity analysis excluding participants missing tumor stage. Considering that activity was assessed every 2–4 years in our study, and the similar associations of activity 1–2.5 years and 2.6–4 years after diagnosis with survival, we only show the results of activity 1–4 years after diagnosis. Considering the small number of women with high total activity, and the similar associations of 7.5–14.9 and ≥15 MET-hour/week with survival, these two levels were merged as the top category of activity. A secondary analysis using inverse probability weighting (IPW) was applied to reduce the potential bias due to the availability of questionnaire data ^{20, 21} (see Supplementary Methods). Additionally, a sensitivity analysis was conducted among single population who had provided activity data both before and after diagnosis.

Statistical analyses were conducted with SAS statistical software version 9.4 (SAS Institute Inc., Cary, North Carolina, US). All statistical tests were two-sided and P values <0.05 were considered statistically significant.

RESULTS

In total, 1591 confirmed epithelial ovarian cancer cases were identified (1261 from NHS; 330 from NHSII; Figure 1). After exclusions, 1461 (91.8%) were included in the analyses (1158 from NHS; 303 from NHSII). Physical activity data reported 1–8 years before diagnosis and 1–4 years after diagnosis were available for 1431 and 669 women with ovarian cancer, respectively.

Among all cases with activity reported in at least one period during the study period (1986–2016 in NHS, and 1989–2017 in NHSII), the median age at ovarian cancer diagnosis was 62 years (IQR: 53–70), 73.7% of patients died and median survival time after ovarian cancer diagnosis was 3.8 years (IQR: 1.7–8.8; Table 1). Most patients were diagnosed with high-grade serous and stage III cancer and were current NSAID users and ever smokers. More than 40% of women had BMI ≥25 kg/m².

Table 1.

Characteristics of women with ovarian cancer included in each analytic dataset, NHS and NHSII

	1–8 years BEFORE diagnosis (median 4.6, IQR 4.0–5.2 years)	1–4 years AFTER diagnosis (median 1.8, IQR 1.4–2.5 years)	Total study population
Total cases	1431	669	1461
Deaths from ovarian cancer, n (%)	901 (63.0)	294 (44.0)	911 (62.4)
Deaths from all causes, n (%)	1061 (74.1)	403 (60.2)	1076 (73.7)
Age at diagnosis, years, median (IQR)	62 (54,70)	60 (51,68)	62 (53,70)
Calendar year of diagnosis, median (IQR)	2000 (1992,2007)	1999 (1990,2006)	2000 (1992,2007)
Overall survival time since ovarian cancer diagnosis, years, median (IQR)	3.7 (1.6,8.7)	7.1 (4.3,15.0)	3.8 (1.7,8.8)
Tumor histology subtype, n (%)
High-grade serous or poorly differentiated	943 (65.9)	424 (63.4)	957 (65.5)
Mucinous	50 (3.5)	27 (4.0)	52 (3.6)
Endometrioid	178 (12.4)	107 (16.0)	183 (12.5)
Clear cell	84 (5.9)	48 (7.2)	87 (6.0)
Other or unknown	176 (12.3)	63 (9.4)	182 (12.5)
Stage, n (%)
I	284 (19.9)	207 (30.9)	297 (20.3)
II	108 (7.6)	58 (8.7)	111 (7.6)
III	810 (56.6)	348 (52.0)	821 (56.2)
IV	139 (9.7)	34 (5.1)	141 (9.7)
Missing	90 (6.3)	22 (3.3)	91 (6.2)
BMI, kg/m2, n (%)
<25	738 (51.6)	350 (52.3)	743 (50.9)
25–29.9	376 (26.3)	163 (24.4)	381 (26.1)
≥30	283 (19.8)	128 (19.1)	285 (19.5)
Missing	34 (2.4)	28 (4.2)	52 (3.6)
NSAID use, n (%)
Non-current user	462 (32.3)	192 (28.7)	464 (31.8)
Current user	703 (49.1)	315 (47.0)	708 (48.5)
Missing	266 (18.6)	162 (24.2)	289 (19.8)
Smoking, n (%)
Never smoker	672 (47.0)	321 (48.0)	675 (46.2)
Ever smoker	751 (52.5)	331 (49.5)	762 (52.2)
Missing	8 (0.6)	17 (2.5)	24 (1.6)

Abbreviations: BMI, body mass index; IQR, interquartile range; MET, metabolic equivalent task; NHS, Nurses’ Health Study; NHSII, Nurses’ Health Study II; NSAID, non-steroidal anti-inflammatory drug.

NSAID use includes aspirin and non-aspirin NSAID use.

Percentages may not add up to 100% due to rounding.

Ovarian cancer-specific mortality was not associated with activity reported 1–8 years before diagnosis (≥7.5 vs. <1.5 MET-hours/week, HR=0.96, 95%CI: 0.72–1.27; Table 2). However, physical activity 1–4 years after diagnosis was associated with a lower risk of mortality (≥7.5 vs. <1.5 MET-hours/week, HR=0.67, 95%CI: 0.48–0.94; P_trend=0.01). This inverse association could only be explained away by an unmeasured confounder that was associated with both physical activity and mortality with a risk ratio (E-value) ≥2.35-fold each. Similar results were observed for the associations with vigorous and moderate activity, although for post-diagnosis activity, a statistically significant association was only found for moderate (HR=0.76 and 0.70 for 1.5–7.4 and ≥7.5 MET-hours/week), but not vigorous activity (comparable HR=0.90 and 0.76; Supplementary Table 1).

Table 2.

Total physical activity and mortality, NHS and NHSII

	Total cases, n	Epithelial ovarian cancer-specific mortality		All-cause mortality
		Deaths, n	HR (95%CI)	Deaths, n	HR (95%CI)
Total physical activity, 1–8 years BEFORE diagnosis
Total	1431	901	-	1061	-
<1.5 MET-hours/week	93	59	1.00 (ref)	69	1.00 (ref)
1.5 to <7.5 MET-hours/week	587	383	0.91 (0.68,1.22)	470	0.89 (0.68,1.16)
≥7.5 MET-hours/week	751	459	0.96 (0.72,1.27)	522	0.92 (0.71,1.20)
Ptrend	-	-	0.89	-	0.88
Total physical activity, 1–4 years AFTER diagnosis
Total	669	294	-	403	-
<1.5 MET-hours/week	107	59	1.00 (ref)	79	1.00 (ref)
1.5 to <7.5 MET-hours/week	272	130	0.95 (0.69,1.31)	180	0.87 (0.65,1.15)
≥7.5 MET-hours/week	290	105	0.67 (0.48,0.94)	144	0.71 (0.53,0.95)
Ptrend	-	-	0.01	-	0.02

Abbreviations: CI, confidence interval; HR, hazard ratio; MET, metabolic equivalent task; NHS, Nurses’ Health Study; NHSII, Nurses’ Health Study II.

Models adjusted for age and calendar year at diagnosis, tumor histotype, stage, body mass index (categorical), smoking status, and cohort. Post-diagnosis models additionally adjusted for non-steroidal anti-inflammatory drug usage (NSAID), which included aspirin and non-aspirin NSAID use.

In the analysis of change in activity level from 1–8 years before diagnosis to 1–4 years after diagnosis, compared with women with activity <7.5 MET-hours/week at both timepoints, those whose activity decreased from ≥7.5 to <7.5 MET-hours/week had a 49% elevated risk of ovarian cancer-specific mortality (HR=1.49, 95%CI: 1.07–2.08; Table 3). A non-significant inverse relationship was observed for those with ≥7.5 MET-hours/week in both periods (HR=0.83, 95%CI: 0.59–1.17), or those who increased their level of activity from pre- to post-diagnosis (HR=0.88, 95%CI: 0.58–1.35).

Table 3.

Change in total physical activity and mortality, NHS and NHSII

Physical activity			Epithelial ovarian cancer-specific mortality		All-cause mortality
			n Deaths/Total cases	HR (95%CI)	n Deaths/Total cases	HR (95%CI)
Group	1–8 years BEFORE diagnosis	1–4 years AFTER diagnosis
Low-stable	<7.5 MET-hours/week	<7.5 MET-hours/week	105/230	1.00 (ref)	157/230	1.00 (ref)
Low to high increase	<7.5 MET-hours/week	≥7.5 MET-hours/week	28/72	0.88 (0.58,1.35)	44/72	0.95 (0.68,1.34)
High to low decrease	≥7.5 MET-hours/week	<7.5 MET-hours/week	75/127	1.49 (1.07,2.08)	88/127	1.31 (0.97,1.76)
High-stable	≥7.5 MET-hours/week	≥7.5 MET-hours/week	76/210	0.83 (0.59,1.17)	99/210	0.85 (0.63,1.14)

Abbreviations: CI, confidence interval; HR, hazard ratio; MET, metabolic equivalent task; NHS, Nurses’ Health Study; NHSII, Nurses’ Health Study II.

Models adjusted for age and calendar year at diagnosis, tumor histotype, stage, body mass index (categorical), smoking status, non-steroidal anti-inflammatory drug usage (NSAID, including aspirin and non-aspirin NSAID use), and cohort.

An inverse association was noted for activity 1–4 years after diagnosis among women with high-grade serous or poorly differentiated tumors (≥7.5 vs. <1.5 MET-hours/week, HR=0.68, 95%CI: 0.47–0.98), and suggestively with non-serous or low-grade serous tumors (comparable HR=0.55, 95%CI: 0.21–1.44; P-heterogeneity=0.53; Table 4). In analyses by tumor stage, the inverse association was found for post-diagnosis activity regardless of tumor stage, although the association appeared suggestively stronger for stage I/II (HR=0.25) versus stage III/IV (HR=0.68; P-heterogeneity=0.21; Supplementary Table 2). Ovarian cancer-specific mortality was not associated with physical activity 1–8 years before diagnosis or change in activity from 1–8 years before diagnosis to 1–4 years after diagnosis for either histotype or stage.

Table 4.

Total physical activity and mortality by tumor histotype, NHS and NHSII

	Epithelial ovarian cancer-specific mortality			All-cause mortality
	High-grade serous or poorly differentiated histology	Non-serous or low-grade serous histology	P-heterogeneity	High-grade serous or poorly differentiated histology	Non-serous or low-grade serous histology	P-heterogeneity
	HR (95%CI)	HR (95%CI)		HR (95%CI)	HR (95%CI)
Total physical activity, 1–8 years BEFORE diagnosis
n, deaths/cases	668/943	179/424	0.45	753/943	250/424	0.15
<1.5 MET-hours/week	1.00 (ref)	1.00 (ref)		1.00 (ref)	1.00 (ref)
1.5 to <7.5 MET-hours/week	0.82 (0.59,1.15)	1.45 (0.67,3.15)		0.81 (0.60,1.11)	1.26 (0.65,2.44)
≥7.5 MET-hours/week	0.96 (0.70,1.33)	1.11 (0.52,2.36)		0.93 (0.69,1.25)	0.98 (0.51,1.88)
Ptrend	0.45	0.44		0.60	0.33
Total physical activity, 1–4 years AFTER diagnosis
n, deaths/cases	241/423	42/227	0.53	293/423	96/227	0.49
<1.5 MET-hours/week	1.00 (ref)	1.00 (ref)		1.00 (ref)	1.00 (ref)
1.5 to <7.5 MET-hours/week	0.96 (0.67,1.37)	0.90 (0.37,2.19)		0.92 (0.67,1.27)	0.75 (0.38,1.47)
≥7.5 MET-hours/week	0.68 (0.47,0.98)	0.55 (0.21,1.44)		0.65 (0.47,0.91)	0.92 (0.47,1.81)
Ptrend	0.02	0.18		<0.01	0.95
Change in total physical activity from 4–8 years BEFORE diagnosis to 1–4 years AFTER diagnosis
n, deaths/cases	231/409	42/212	0.50	281/409	94/212	0.02
Low-stable (<7.5 MET-hours/week)	1.00 (ref)	1.00 (ref)		1.00 (ref)	1.00 (ref)
Low to high increase (from <7.5 to ≥7.5 MET-hours/week)	0.79 (0.48,1.30)	1.09 (0.39,3.03)		0.66 (0.41,1.05)	1.71 (0.96,3.03)
High to low decrease (from ≥7.5 to <7.5 MET-hours/week)	1.45 (0.99,2.10)	0.93 (0.37,2.37)		1.28 (0.91,1.80)	0.75 (0.35,1.60)
High-stable (≥7.5 MET-hours/week)	0.85 (0.59,1.24)	0.37 (0.12,1.09)		0.82 (0.58,1.14)	0.57 (0.26,1.23)

Abbreviations: CI, confidence interval; HR, hazard ratio; MET, metabolic equivalent task; NHS, Nurses’ Health Study; NHSII, Nurses’ Health Study II.Non-serous histology subtype refers to mucinous, endometrioid, clear cell, transitional/Brenner, carcinosarcoma, or mixed subtypes of ovarian cancer.

Models adjusted for age and calendar year at diagnosis, stage, body mass index (categorical), smoking status, and cohort. Post-diagnosis and change in total physical activity models additionally adjusted for non-steroidal anti-inflammatory drug usage (NSAID), which included aspirin and non-aspirin NSAID use.

For all analyses, results for all-cause mortality were similar to those for ovarian cancer-specific mortality. Findings were similar across cohorts (all P-heterogeneity>0.05; Supplementary Table 3), or by BMI (all P-heterogeneity>0.05; data not shown). No different result was observed in the sensitivity analysis among single population who provided physical activity data both before and after diagnosis. Risk estimates from a reduced model excluding BMI, smoking status, and NSIADs use were similar to those in the primary analyses. Findings did not differ when excluding cases with unknown stage, or using IPW-adjusted multivariable Cox regression model (data not shown).

DISCUSSION

This is the largest prospective study of the relationship of physical activity before and after diagnosis with survival among women with ovarian cancer. We observed that physical activity after ovarian cancer diagnosis, but not before, was associated with better prognosis. Further, women whose activity decreased substantially from 1–8 years before diagnosis to 1–4 years after diagnosis had a worse outcome compared to those with low activity at both times.

Among the few studies of pre-diagnosis physical activity and ovarian cancer survival, inconsistent results have been reported. For example, women reporting no regular moderate or vigorous recreational physical activity had a 22% higher risk of mortality in a pooled analysis of 12 case-control studies ²², and participation in >0 versus 0 MET-hours/week of vigorous activity was associated with a 26% lower risk of mortality in a prospective study ²³. In contrast, no association was observed in two population-based case-control studies ^{24, 25}. In our prospective study we did not find an association between total activity 1–8 years before diagnosis and mortality. Differences in results across studies may be due to variation in how activity or inactivity was defined and assessed (prospectively or retrospectively) as well as potentially different ovarian cancer histotype distributions across studies and varying timing of exposure assessment prior to diagnosis ⁶. Further, most previous studies were limited by short follow-up and use of a single assessment of activity lacking details on the type and intensity of activity ²⁶.

Hypothesized mechanisms regarding a putative association between pre-diagnosis physical activity and risk of ovarian cancer survival are complex. Pre-diagnosis physical activity may affect carcinogenesis or prognosis through effects on adiposity, changes in hormones levels, cytokines, blood insulin levels, and insulin resistance ^{27, 28}. However, exercise may increase tumor-generated lactic acid secretion and result in an acidified tumor microenvironment, leading to the development of acid-resistant phenotypes with invasive potential from precursor lesions, suppressed local anticancer immune response, and enhanced angiogenesis ^29–31. Further research should assess these potential mechanisms to understand the impact of pre-diagnosis physical activity on ovarian cancer development and prognosis.

High physical activity after cancer diagnosis has been associated with reduced risk of mortality in breast, colorectal and prostate cancers ^{32, 33}. Most ovarian cancer patients decreased activity from pre- to post-diagnosis and remained insufficiently active following diagnosis ^{16, 34, 35}. Most studies of post-diagnosis activity and ovarian cancer survival utilized data collected at or shortly after diagnosis ³⁶. One study of 264 African-American women with ovarian cancer that investigated activity one year after diagnosis observed no association with mortality ²⁴. Conversely, our study of predominantly women of European ancestry observed an inverse association between post-diagnosis activity assessed 1–4 years after diagnosis and ovarian-cancer specific mortality. Notably, we observed a suggestively stronger association for women with lower stage disease, suggesting that activity may play a larger role in less aggressive tumor types. African-American women are more likely to develop aggressive ovarian cancer ³⁷, potentially explaining the difference between these two studies. One cohort study consisting of 93% white women supported our finding of a potential improved survival among habitually active patients among 151 newly diagnosed ovarian cancer patients ³⁸.

Once diagnosed, most ovarian cancer survivors are relatively inactive ⁶, and low muscle mass may be related to increased treatment toxicity and mortality in cancer survivors ³⁹. In contrast, higher activity is related to reduced symptom burden and less adverse effects of treatment, and increased physical and emotional well-being ^{40, 41}. The mechanisms linking activity with survival include decreased body fat, altered reproductive hormone levels, altered cytokine and growth factor milieu, and changes in immune function ⁴². Specifically, physical activity increases the number and activity of macrophages, natural and lymphokine-activated killer cells, and their regulating cytokines, results in increased mitogen-induced lymphocyte proliferation rates ^{43, 44}, and reduces symptoms of upper respiratory tract infection ⁴⁵. It can also lower the risk of cardiorespiratory disease ⁴⁶, which is common in gynecological cancer survivors ⁴⁷ and associated with worse prognosis ⁴⁸. Also, physical activity was suggested to improve psychological outcomes in small studies ³⁵. However, our findings did not differ when we additionally adjusted for depression.

Our study has several limitations. One potential limitation is reverse causation, which could have occurred if cancer patients had been less physically active due to symptoms, treatment, or tumor burden at the time of activity assessment. To address this, we excluded cancer patients who reported on their physical activity one year before or after cancer diagnosis, and a model additionally adjusted for self-reported physical limitations did not substantially alter risk estimates. Also, our findings may not be explained by the screening bias as most of these women (>99%) have regular check-ups and cancer screenings before ovarian cancer diagnosis. In addition, analyses examining activity reported shortly after diagnosis (1–2.5 years), when many patients may be undergoing treatment, versus later after diagnosis (2.6–4 years) had similar findings. Although calendar year of diagnosis was adjusted to account for possible changes in cancer treatment over time, we were unable to evaluate potential confounding by cancer treatment or debulking status ⁴⁹. Physical activity was measured by self-report, which might increase the likelihood of misclassification ⁶. However, the prospective setting and use of a reliable activity measure in our study increase confidence in the self-reported data ¹⁴. Additionally, while we had well over 1000 cases, there was somewhat limited power to conduct detailed histology-specific analyses. Our results for post-diagnosis physical activity are only generalizable to women who survived at least one year after diagnosis, necessitating future studies in ovarian cancer patient cohorts. Finally, our study population was comprised of registered nurses who may have relatively higher level of occupational physical activity than others and included few non-white women (4%), which might affect the generalizability of our findings.

Strengths of this study include the prospective design, long follow-up for death ascertainment, availability of cause-of-death information, and detailed updated data on physical activity every 2–4 years both before and after ovarian cancer diagnosis. To our knowledge, our study includes the largest number of prospectively ascertained ovarian cancer patients to examine the influence of physical activity from pre- to post-diagnosis on survival overall and by histotype and stage.

Current knowledge gaps regarding lifestyle factors that may improve outcomes for ovarian cancer survivors throughout the entire survivorship spectrum make it challenging to develop tailored guidelines for this patient population. The current large prospective study provides evidence that physical activity after ovarian cancer diagnosis is associated with better prognosis across multiple histotypes. If confirmed, these findings may support inclusion of physical activity in prognostic models to predict outcome of ovarian cancer and inform future studies evaluating the effectiveness and optimal timing of adding structured activity programs in standard clinical management of ovarian cancer patients.

What’s New?

Little is known about the influence of pre- and post-diagnosis physical activity on ovarian cancer survival. In this study, physical activity after, but not before, ovarian cancer diagnosis was found to be associated with better prognosis. Women whose activity decreased substantially from 1–8 years before diagnosis to 1–4 years after diagnosis had a worse outcome compared to those with low activity at both times. Additional investigation is needed to identify factors underlying these associations.

Abbreviations:

BMI

body mass index

CI

confidence interval

HR

hazard ratio

IPW

inverse probability weighting

IQR

interquartile range

MET

metabolic equivalent task

NHS

Nurses’ Health Study

NHSII

Nurses’ Health Study II

NSAID

non-steroidal anti-inflammatory drug